[llvm] [CostModel][X86] Cleanup unary shuffle cost tests to use poison instead of undef in second operand (PR #222718)
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Thu Sep 10 23:53:21 PDT 2026
https://github.com/RKSimon updated https://github.com/llvm/llvm-project/pull/222718
>From 2c90d6fe9a0ff88cad93446c20a9f882fc72a295 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev at redking.me.uk>
Date: Thu, 10 Sep 2026 18:29:56 +0100
Subject: [PATCH] [CostModel][X86] Cleanup unary shuffle cost tests to use
poison instead of undef in second operand
---
.../CostModel/X86/shuffle-broadcast-fp16.ll | 20 +-
.../CostModel/X86/shuffle-broadcast.ll | 488 +-
.../X86/shuffle-extract_subvector.ll | 4132 ++++++++---------
.../CostModel/X86/shuffle-insert_subvector.ll | 588 +--
.../CostModel/X86/shuffle-non-pow-2.ll | 6 +-
.../CostModel/X86/shuffle-reverse-fp16.ll | 20 +-
.../Analysis/CostModel/X86/shuffle-reverse.ll | 338 +-
.../CostModel/X86/shuffle-single-src-fp16.ll | 16 +-
.../CostModel/X86/shuffle-single-src.ll | 454 +-
.../Analysis/CostModel/X86/shuffle-splat.ll | 488 +-
10 files changed, 3275 insertions(+), 3275 deletions(-)
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-broadcast-fp16.ll b/llvm/test/Analysis/CostModel/X86/shuffle-broadcast-fp16.ll
index c4cf8b79eb492..763b6f2a61482 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-broadcast-fp16.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-broadcast-fp16.ll
@@ -3,17 +3,17 @@
define void @test_vXf16(<2 x half> %src32, <4 x half> %src64, <8 x half> %src128, <16 x half> %src256, <32 x half> %src512) {
; CHECK-LABEL: 'test_vXf16'
-; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
-; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> zeroinitializer
+; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> zeroinitializer
+; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> zeroinitializer
; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
- %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
- %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
- %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
- %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+ %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> zeroinitializer
+ %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> zeroinitializer
+ %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> zeroinitializer
+ %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> zeroinitializer
+ %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> zeroinitializer
ret void
}
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-broadcast.ll b/llvm/test/Analysis/CostModel/X86/shuffle-broadcast.ll
index 79e16225be79a..7bfab69063d55 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-broadcast.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-broadcast.ll
@@ -18,410 +18,410 @@
define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double> %src512) {
; SSE-LABEL: 'test_vXf64'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> zeroinitializer
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXf64'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> zeroinitializer
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf64'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> zeroinitializer
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf64'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> zeroinitializer
- %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> zeroinitializer
- %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> zeroinitializer
+ %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> zeroinitializer
+ %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> zeroinitializer
+ %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> zeroinitializer
ret void
}
define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512) {
; SSE-LABEL: 'test_vXi64'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> zeroinitializer
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi64'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> zeroinitializer
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi64'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> zeroinitializer
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi64'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> zeroinitializer
- %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> zeroinitializer
- %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> zeroinitializer
+ %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> zeroinitializer
+ %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> zeroinitializer
+ %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> zeroinitializer
ret void
}
define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %src256, <16 x float> %src512) {
; SSE-LABEL: 'test_vXf32'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> zeroinitializer
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXf32'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> zeroinitializer
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf32'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> zeroinitializer
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf32'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> zeroinitializer
- %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> zeroinitializer
- %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> zeroinitializer
- %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> zeroinitializer
+ %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> zeroinitializer
+ %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> zeroinitializer
+ %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> zeroinitializer
+ %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> zeroinitializer
ret void
}
define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512) {
; SSE-LABEL: 'test_vXi32'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> zeroinitializer
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi32'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> zeroinitializer
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi32'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> zeroinitializer
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi32'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> zeroinitializer
- %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> zeroinitializer
- %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> zeroinitializer
- %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> zeroinitializer
+ %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> zeroinitializer
+ %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> zeroinitializer
+ %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> zeroinitializer
+ %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> zeroinitializer
ret void
}
define void @test_vXf16(<2 x half> %src32, <4 x half> %src64, <8 x half> %src128, <16 x half> %src256, <32 x half> %src512) {
; SSE2-LABEL: 'test_vXf16'
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> zeroinitializer
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXf16'
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> zeroinitializer
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXf16'
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> zeroinitializer
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf16'
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> zeroinitializer
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf16'
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
- %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
- %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
- %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
- %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+ %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> zeroinitializer
+ %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> zeroinitializer
+ %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> zeroinitializer
+ %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> zeroinitializer
+ %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> zeroinitializer
ret void
}
define void @test_vXbf16(<2 x bfloat> %src32, <4 x bfloat> %src64, <8 x bfloat> %src128, <16 x bfloat> %src256, <32 x bfloat> %src512) {
; SSE-LABEL: 'test_vXbf16'
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> poison, <2 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> poison, <4 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> poison, <8 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> poison, <16 x i32> zeroinitializer
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> poison, <32 x i32> zeroinitializer
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXbf16'
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> poison, <2 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> poison, <32 x i32> zeroinitializer
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXbf16'
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> poison, <2 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> poison, <32 x i32> zeroinitializer
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXbf16'
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> poison, <2 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> poison, <32 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
- %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
- %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
- %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
- %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
+ %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> poison, <2 x i32> zeroinitializer
+ %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> poison, <4 x i32> zeroinitializer
+ %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> poison, <8 x i32> zeroinitializer
+ %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> poison, <16 x i32> zeroinitializer
+ %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> poison, <32 x i32> zeroinitializer
ret void
}
define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src512) {
; SSE2-LABEL: 'test_vXi16'
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> zeroinitializer
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi16'
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> zeroinitializer
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi16'
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> zeroinitializer
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi16'
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> zeroinitializer
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi16'
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> zeroinitializer
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi16'
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
- %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
- %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
- %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
- %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+ %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> zeroinitializer
+ %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> zeroinitializer
+ %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> zeroinitializer
+ %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> zeroinitializer
+ %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> zeroinitializer
ret void
}
define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x i8> %src128, <32 x i8> %src256, <64 x i8> %src512) {
; SSE2-LABEL: 'test_vXi8'
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> zeroinitializer
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi8'
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> zeroinitializer
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi8'
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> zeroinitializer
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi8'
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> zeroinitializer
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi8'
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> zeroinitializer
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi8'
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
- %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
- %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
- %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
- %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
- %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+ %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> zeroinitializer
+ %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> zeroinitializer
+ %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> zeroinitializer
+ %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> zeroinitializer
+ %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> zeroinitializer
+ %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> zeroinitializer
ret void
}
define void @test_vXi1(<2 x i1> %src2, <4 x i1> %src4, <8 x i1> %src8, <16 x i1> %src16, <32 x i1> %src32, <64 x i1> %src64) {
; SSE2-LABEL: 'test_vXi1'
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> zeroinitializer
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> zeroinitializer
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi1'
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> zeroinitializer
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> zeroinitializer
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi1'
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> zeroinitializer
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> zeroinitializer
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi1'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
-; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> zeroinitializer
+; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> zeroinitializer
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi1'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> zeroinitializer
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> zeroinitializer
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi1'
-; AVX512-NEXT: Cost Model: Found costs of 3 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of 5 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of 9 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of 17 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of 33 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
-; AVX512-NEXT: Cost Model: Found costs of 65 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 3 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 5 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 9 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 17 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 33 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> zeroinitializer
+; AVX512-NEXT: Cost Model: Found costs of 65 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> zeroinitializer
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
- %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
- %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
- %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
- %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
- %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+ %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> zeroinitializer
+ %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> zeroinitializer
+ %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> zeroinitializer
+ %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> zeroinitializer
+ %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> zeroinitializer
+ %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> zeroinitializer
ret void
}
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-extract_subvector.ll b/llvm/test/Analysis/CostModel/X86/shuffle-extract_subvector.ll
index af3b627d4edc6..bb44def20808c 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-extract_subvector.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-extract_subvector.ll
@@ -18,2176 +18,2176 @@
define void @test_vXf64(<4 x double> %src256, <8 x double> %src512) {
; SSE-LABEL: 'test_vXf64'
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x double> %src256, <4 x double> undef, <2 x i32> <i32 0, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_23 = shufflevector <4 x double> %src256, <4 x double> undef, <2 x i32> <i32 2, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 0, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_23 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 2, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 4, i32 5>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_67 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 6, i32 7>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_2345 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE-NEXT: Cost Model: Found costs of 2 for: %V512_567u = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 5, i32 6, i32 7, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x double> %src256, <4 x double> poison, <2 x i32> <i32 0, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_23 = shufflevector <4 x double> %src256, <4 x double> poison, <2 x i32> <i32 2, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 0, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_23 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 2, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 4, i32 5>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_67 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 6, i32 7>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_2345 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE-NEXT: Cost Model: Found costs of 2 for: %V512_567u = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 5, i32 6, i32 7, i32 poison>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXf64'
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x double> %src256, <4 x double> undef, <2 x i32> <i32 0, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x double> %src256, <4 x double> undef, <2 x i32> <i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 0, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_2345 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_567u = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 5, i32 6, i32 7, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x double> %src256, <4 x double> poison, <2 x i32> <i32 0, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x double> %src256, <4 x double> poison, <2 x i32> <i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 0, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_2345 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_567u = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 5, i32 6, i32 7, i32 poison>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf64'
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x double> %src256, <4 x double> undef, <2 x i32> <i32 0, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x double> %src256, <4 x double> undef, <2 x i32> <i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 0, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_2345 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_567u = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 5, i32 6, i32 7, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x double> %src256, <4 x double> poison, <2 x i32> <i32 0, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x double> %src256, <4 x double> poison, <2 x i32> <i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 0, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_2345 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_567u = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 5, i32 6, i32 7, i32 poison>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf64'
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x double> %src256, <4 x double> undef, <2 x i32> <i32 0, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x double> %src256, <4 x double> undef, <2 x i32> <i32 2, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 0, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 2, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_45 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 4, i32 5>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 6, i32 7>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512_2345 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_4567 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512_567u = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 5, i32 6, i32 7, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x double> %src256, <4 x double> poison, <2 x i32> <i32 0, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x double> %src256, <4 x double> poison, <2 x i32> <i32 2, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 0, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 2, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_45 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 4, i32 5>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 6, i32 7>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512_2345 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_4567 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512_567u = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 5, i32 6, i32 7, i32 poison>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V256_01 = shufflevector <4 x double> %src256, <4 x double> undef, <2 x i32> <i32 0, i32 1>
- %V256_23 = shufflevector <4 x double> %src256, <4 x double> undef, <2 x i32> <i32 2, i32 3>
- %V512_01 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 0, i32 1>
- %V512_23 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 2, i32 3>
- %V512_45 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 4, i32 5>
- %V512_67 = shufflevector <8 x double> %src512, <8 x double> undef, <2 x i32> <i32 6, i32 7>
- %V512_0123 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %V512_2345 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
- %V512_4567 = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- %V512_567u = shufflevector <8 x double> %src512, <8 x double> undef, <4 x i32> <i32 5, i32 6, i32 7, i32 undef>
+ %V256_01 = shufflevector <4 x double> %src256, <4 x double> poison, <2 x i32> <i32 0, i32 1>
+ %V256_23 = shufflevector <4 x double> %src256, <4 x double> poison, <2 x i32> <i32 2, i32 3>
+ %V512_01 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 0, i32 1>
+ %V512_23 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 2, i32 3>
+ %V512_45 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 4, i32 5>
+ %V512_67 = shufflevector <8 x double> %src512, <8 x double> poison, <2 x i32> <i32 6, i32 7>
+ %V512_0123 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %V512_2345 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+ %V512_4567 = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %V512_567u = shufflevector <8 x double> %src512, <8 x double> poison, <4 x i32> <i32 5, i32 6, i32 7, i32 undef>
ret void
}
define void @test_vXi64(<4 x i64> %src256, <8 x i64> %src512) {
; SSE-LABEL: 'test_vXi64'
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x i64> %src256, <4 x i64> undef, <2 x i32> <i32 0, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_23 = shufflevector <4 x i64> %src256, <4 x i64> undef, <2 x i32> <i32 2, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 0, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_23 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 2, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 4, i32 5>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_67 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 6, i32 7>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_2345 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x i64> %src256, <4 x i64> poison, <2 x i32> <i32 0, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_23 = shufflevector <4 x i64> %src256, <4 x i64> poison, <2 x i32> <i32 2, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 0, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_23 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 2, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 4, i32 5>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_67 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 6, i32 7>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_2345 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX-LABEL: 'test_vXi64'
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x i64> %src256, <4 x i64> undef, <2 x i32> <i32 0, i32 1>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x i64> %src256, <4 x i64> undef, <2 x i32> <i32 2, i32 3>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 0, i32 1>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 2, i32 3>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 4, i32 5>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 6, i32 7>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX-NEXT: Cost Model: Found costs of 2 for: %V512_2345 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x i64> %src256, <4 x i64> poison, <2 x i32> <i32 0, i32 1>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x i64> %src256, <4 x i64> poison, <2 x i32> <i32 2, i32 3>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 0, i32 1>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 2, i32 3>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 4, i32 5>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 6, i32 7>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX-NEXT: Cost Model: Found costs of 2 for: %V512_2345 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi64'
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x i64> %src256, <4 x i64> undef, <2 x i32> <i32 0, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x i64> %src256, <4 x i64> undef, <2 x i32> <i32 2, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 0, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 2, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_45 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 4, i32 5>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 6, i32 7>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512_2345 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_4567 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x i64> %src256, <4 x i64> poison, <2 x i32> <i32 0, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x i64> %src256, <4 x i64> poison, <2 x i32> <i32 2, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 0, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 2, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_45 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 4, i32 5>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 6, i32 7>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512_2345 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_4567 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V256_01 = shufflevector <4 x i64> %src256, <4 x i64> undef, <2 x i32> <i32 0, i32 1>
- %V256_23 = shufflevector <4 x i64> %src256, <4 x i64> undef, <2 x i32> <i32 2, i32 3>
- %V512_01 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 0, i32 1>
- %V512_23 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 2, i32 3>
- %V512_45 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 4, i32 5>
- %V512_67 = shufflevector <8 x i64> %src512, <8 x i64> undef, <2 x i32> <i32 6, i32 7>
- %V512_0123 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %V512_2345 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
- %V512_4567 = shufflevector <8 x i64> %src512, <8 x i64> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %V256_01 = shufflevector <4 x i64> %src256, <4 x i64> poison, <2 x i32> <i32 0, i32 1>
+ %V256_23 = shufflevector <4 x i64> %src256, <4 x i64> poison, <2 x i32> <i32 2, i32 3>
+ %V512_01 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 0, i32 1>
+ %V512_23 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 2, i32 3>
+ %V512_45 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 4, i32 5>
+ %V512_67 = shufflevector <8 x i64> %src512, <8 x i64> poison, <2 x i32> <i32 6, i32 7>
+ %V512_0123 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %V512_2345 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+ %V512_4567 = shufflevector <8 x i64> %src512, <8 x i64> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
ret void
}
define void @test_vXi32(<4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512) {
; SSE-LABEL: 'test_vXi32'
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 0, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 2, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_45 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 4, i32 5>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 6, i32 7>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <8 x i32> %src256, <8 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_4567 = shufflevector <8 x i32> %src256, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 0, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 2, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 4, i32 5>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 6, i32 7>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_89 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 8, i32 9>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512_AB = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 10, i32 11>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_CD = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 12, i32 13>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512_EF = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 14, i32 15>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_89AB = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_CDEF = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_01234567 = shufflevector <16 x i32> %src512, <16 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_89ABCDEF = shufflevector <16 x i32> %src512, <16 x i32> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> poison, <2 x i32> <i32 0, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> poison, <2 x i32> <i32 2, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 0, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 2, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_45 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 4, i32 5>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 6, i32 7>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <8 x i32> %src256, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_4567 = shufflevector <8 x i32> %src256, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 0, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 2, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_45 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 4, i32 5>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512_67 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 6, i32 7>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_89 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 8, i32 9>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512_AB = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 10, i32 11>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_CD = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 12, i32 13>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512_EF = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 14, i32 15>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_4567 = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_89AB = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_CDEF = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_01234567 = shufflevector <16 x i32> %src512, <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_89ABCDEF = shufflevector <16 x i32> %src512, <16 x i32> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX-LABEL: 'test_vXi32'
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 0, i32 1>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 2, i32 3>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 4, i32 5>
-; AVX-NEXT: Cost Model: Found costs of 2 for: %V256_67 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 6, i32 7>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <8 x i32> %src256, <8 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <8 x i32> %src256, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 0, i32 1>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 2, i32 3>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_45 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 4, i32 5>
-; AVX-NEXT: Cost Model: Found costs of 2 for: %V512_67 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 6, i32 7>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_89 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 8, i32 9>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_AB = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 10, i32 11>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_CD = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 12, i32 13>
-; AVX-NEXT: Cost Model: Found costs of 2 for: %V512_EF = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 14, i32 15>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_4567 = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_89AB = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_CDEF = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_01234567 = shufflevector <16 x i32> %src512, <16 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_89ABCDEF = shufflevector <16 x i32> %src512, <16 x i32> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> poison, <2 x i32> <i32 0, i32 1>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> poison, <2 x i32> <i32 2, i32 3>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 0, i32 1>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 2, i32 3>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 4, i32 5>
+; AVX-NEXT: Cost Model: Found costs of 2 for: %V256_67 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 6, i32 7>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <8 x i32> %src256, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <8 x i32> %src256, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 0, i32 1>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 2, i32 3>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_45 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 4, i32 5>
+; AVX-NEXT: Cost Model: Found costs of 2 for: %V512_67 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 6, i32 7>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_89 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 8, i32 9>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_AB = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 10, i32 11>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_CD = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 12, i32 13>
+; AVX-NEXT: Cost Model: Found costs of 2 for: %V512_EF = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 14, i32 15>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_4567 = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_89AB = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_CDEF = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_01234567 = shufflevector <16 x i32> %src512, <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %V512_89ABCDEF = shufflevector <16 x i32> %src512, <16 x i32> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi32'
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 0, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 2, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 4, i32 5>
-; AVX512-NEXT: Cost Model: Found costs of 2 for: %V256_67 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 6, i32 7>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <8 x i32> %src256, <8 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <8 x i32> %src256, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 0, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 2, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_45 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 4, i32 5>
-; AVX512-NEXT: Cost Model: Found costs of 2 for: %V512_67 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 6, i32 7>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_89 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 8, i32 9>
-; AVX512-NEXT: Cost Model: Found costs of 2 for: %V512_AB = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 10, i32 11>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_CD = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 12, i32 13>
-; AVX512-NEXT: Cost Model: Found costs of 2 for: %V512_EF = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 14, i32 15>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_4567 = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_89AB = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_CDEF = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_01234567 = shufflevector <16 x i32> %src512, <16 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_89ABCDEF = shufflevector <16 x i32> %src512, <16 x i32> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> poison, <2 x i32> <i32 0, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> poison, <2 x i32> <i32 2, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 0, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 2, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 4, i32 5>
+; AVX512-NEXT: Cost Model: Found costs of 2 for: %V256_67 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 6, i32 7>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <8 x i32> %src256, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <8 x i32> %src256, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 0, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_23 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 2, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_45 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 4, i32 5>
+; AVX512-NEXT: Cost Model: Found costs of 2 for: %V512_67 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 6, i32 7>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_89 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 8, i32 9>
+; AVX512-NEXT: Cost Model: Found costs of 2 for: %V512_AB = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 10, i32 11>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_CD = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 12, i32 13>
+; AVX512-NEXT: Cost Model: Found costs of 2 for: %V512_EF = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 14, i32 15>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_0123 = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_4567 = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_89AB = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_CDEF = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512_01234567 = shufflevector <16 x i32> %src512, <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_89ABCDEF = shufflevector <16 x i32> %src512, <16 x i32> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
- %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
- %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 0, i32 1>
- %V256_23 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 2, i32 3>
- %V256_45 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 4, i32 5>
- %V256_67 = shufflevector <8 x i32> %src256, <8 x i32> undef, <2 x i32> <i32 6, i32 7>
- %V256_0123 = shufflevector <8 x i32> %src256, <8 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %V256_4567 = shufflevector <8 x i32> %src256, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- %V512_01 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 0, i32 1>
- %V512_23 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 2, i32 3>
- %V512_45 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 4, i32 5>
- %V512_67 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 6, i32 7>
- %V512_89 = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 8, i32 9>
- %V512_AB = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 10, i32 11>
- %V512_CD = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 12, i32 13>
- %V512_EF = shufflevector <16 x i32> %src512, <16 x i32> undef, <2 x i32> <i32 14, i32 15>
- %V512_0123 = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %V512_4567 = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- %V512_89AB = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
- %V512_CDEF = shufflevector <16 x i32> %src512, <16 x i32> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
- %V512_01234567 = shufflevector <16 x i32> %src512, <16 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- %V512_89ABCDEF = shufflevector <16 x i32> %src512, <16 x i32> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> poison, <2 x i32> <i32 0, i32 1>
+ %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> poison, <2 x i32> <i32 2, i32 3>
+ %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 0, i32 1>
+ %V256_23 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 2, i32 3>
+ %V256_45 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 4, i32 5>
+ %V256_67 = shufflevector <8 x i32> %src256, <8 x i32> poison, <2 x i32> <i32 6, i32 7>
+ %V256_0123 = shufflevector <8 x i32> %src256, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %V256_4567 = shufflevector <8 x i32> %src256, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %V512_01 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 0, i32 1>
+ %V512_23 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 2, i32 3>
+ %V512_45 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 4, i32 5>
+ %V512_67 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 6, i32 7>
+ %V512_89 = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 8, i32 9>
+ %V512_AB = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 10, i32 11>
+ %V512_CD = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 12, i32 13>
+ %V512_EF = shufflevector <16 x i32> %src512, <16 x i32> poison, <2 x i32> <i32 14, i32 15>
+ %V512_0123 = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %V512_4567 = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %V512_89AB = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+ %V512_CDEF = shufflevector <16 x i32> %src512, <16 x i32> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+ %V512_01234567 = shufflevector <16 x i32> %src512, <16 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %V512_89ABCDEF = shufflevector <16 x i32> %src512, <16 x i32> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
ret void
}
define void @test_vXi16(<4 x i16> %src64, <8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src512) {
; SSE2-LABEL: 'test_vXi16'
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 8, i32 9>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 10, i32 11>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 12, i32 13>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:5 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 3 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 8, i32 9>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 10, i32 11>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 12, i32 13>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 16, i32 17>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 18, i32 19>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 20, i32 21>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 22, i32 23>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 24, i32 25>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 26, i32 27>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 28, i32 29>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:5 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 3 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 8, i32 9>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 10, i32 11>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 12, i32 13>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:5 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 3 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 8, i32 9>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 10, i32 11>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 12, i32 13>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 16, i32 17>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 18, i32 19>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 20, i32 21>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 22, i32 23>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 24, i32 25>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 26, i32 27>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 28, i32 29>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:5 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 3 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi16'
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 8, i32 9>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 10, i32 11>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 12, i32 13>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 8, i32 9>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 10, i32 11>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 12, i32 13>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 16, i32 17>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 18, i32 19>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 20, i32 21>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 22, i32 23>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 24, i32 25>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 26, i32 27>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 28, i32 29>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 8, i32 9>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 10, i32 11>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 12, i32 13>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 8, i32 9>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 10, i32 11>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 12, i32 13>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 16, i32 17>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 18, i32 19>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 20, i32 21>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 22, i32 23>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 24, i32 25>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 26, i32 27>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 28, i32 29>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi16'
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 8, i32 9>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 10, i32 11>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 12, i32 13>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 8, i32 9>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 10, i32 11>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 12, i32 13>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 16, i32 17>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 18, i32 19>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 20, i32 21>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 22, i32 23>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 24, i32 25>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 26, i32 27>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 28, i32 29>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 8, i32 9>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 10, i32 11>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 12, i32 13>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 8, i32 9>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 10, i32 11>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 12, i32 13>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 16, i32 17>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 18, i32 19>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 20, i32 21>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 22, i32 23>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 24, i32 25>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 26, i32 27>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 28, i32 29>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi16'
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 8, i32 9>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 10, i32 11>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 12, i32 13>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 8 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 8 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 8, i32 9>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 10, i32 11>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 12, i32 13>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 16, i32 17>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 18, i32 19>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 20, i32 21>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 22, i32 23>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 24, i32 25>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 26, i32 27>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 28, i32 29>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 4 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 8 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 8, i32 9>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 10, i32 11>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 12, i32 13>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 8 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 8 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 8, i32 9>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 10, i32 11>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 12, i32 13>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 16, i32 17>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 18, i32 19>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 20, i32 21>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 22, i32 23>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 24, i32 25>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 26, i32 27>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 28, i32 29>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 4 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 8 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi16'
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 8, i32 9>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 10, i32 11>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 12, i32 13>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 4 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 4 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 8, i32 9>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 10, i32 11>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 12, i32 13>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 16, i32 17>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 18, i32 19>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 20, i32 21>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 22, i32 23>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 24, i32 25>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 26, i32 27>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 28, i32 29>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 4 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 8, i32 9>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 10, i32 11>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 12, i32 13>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 4 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 4 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 8, i32 9>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 10, i32 11>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 12, i32 13>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 16, i32 17>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 18, i32 19>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 20, i32 21>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 22, i32 23>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 24, i32 25>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 26, i32 27>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 28, i32 29>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 4 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'test_vXi16'
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 8, i32 9>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 10, i32 11>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 12, i32 13>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 4 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 4 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 8, i32 9>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 10, i32 11>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 12, i32 13>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 16, i32 17>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 18, i32 19>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 20, i32 21>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 22, i32 23>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 24, i32 25>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 26, i32 27>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 28, i32 29>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 14 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 14 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 8, i32 9>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 10, i32 11>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 12, i32 13>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 4 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 4 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 8, i32 9>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 10, i32 11>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 12, i32 13>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 16, i32 17>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 18, i32 19>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 20, i32 21>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 22, i32 23>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 24, i32 25>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 26, i32 27>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 28, i32 29>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 14 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 14 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'test_vXi16'
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 8, i32 9>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 10, i32 11>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 12, i32 13>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 8, i32 9>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 10, i32 11>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 12, i32 13>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 16, i32 17>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 18, i32 19>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 20, i32 21>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 22, i32 23>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 24, i32 25>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 26, i32 27>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 28, i32 29>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 8, i32 9>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 10, i32 11>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 12, i32 13>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 8, i32 9>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 10, i32 11>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 12, i32 13>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 16, i32 17>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 18, i32 19>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 20, i32 21>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 22, i32 23>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 24, i32 25>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 26, i32 27>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 28, i32 29>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI-LABEL: 'test_vXi16'
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 8, i32 9>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 10, i32 11>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 12, i32 13>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 0, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 4, i32 5>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 8, i32 9>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 10, i32 11>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 12, i32 13>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 16, i32 17>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 18, i32 19>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 20, i32 21>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 22, i32 23>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 24, i32 25>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 26, i32 27>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 28, i32 29>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 8, i32 9>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 10, i32 11>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 12, i32 13>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 0, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 4, i32 5>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 8, i32 9>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 10, i32 11>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 12, i32 13>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 16, i32 17>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 18, i32 19>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 20, i32 21>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 22, i32 23>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 24, i32 25>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 26, i32 27>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 28, i32 29>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'test_vXi16'
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 8, i32 9>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 10, i32 11>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 12, i32 13>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 0, i32 1>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 4, i32 5>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 8, i32 9>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 10, i32 11>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 12, i32 13>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 16, i32 17>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 18, i32 19>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 20, i32 21>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 22, i32 23>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 24, i32 25>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 26, i32 27>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 28, i32 29>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 8, i32 9>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 10, i32 11>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 12, i32 13>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 0, i32 1>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 4, i32 5>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 8, i32 9>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 10, i32 11>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 12, i32 13>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 16, i32 17>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 18, i32 19>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 20, i32 21>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 22, i32 23>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 24, i32 25>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 26, i32 27>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 28, i32 29>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'test_vXi16'
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 4, i32 5>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 0, i32 1>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 4, i32 5>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 8, i32 9>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 10, i32 11>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 12, i32 13>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 0, i32 1>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 4, i32 5>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 8, i32 9>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 10, i32 11>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 12, i32 13>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 16, i32 17>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 18, i32 19>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 20, i32 21>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 22, i32 23>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 24, i32 25>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 26, i32 27>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 28, i32 29>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 4, i32 5>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 4, i32 5>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 8, i32 9>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 10, i32 11>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 12, i32 13>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 0, i32 1>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 4, i32 5>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 8, i32 9>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 10, i32 11>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 12, i32 13>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 16, i32 17>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 18, i32 19>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 20, i32 21>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 22, i32 23>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 24, i32 25>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 26, i32 27>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 28, i32 29>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 0, i32 1>
- %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> undef, <2 x i32> <i32 2, i32 3>
- %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
- %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 2, i32 3>
- %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 4, i32 5>
- %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> undef, <2 x i32> <i32 6, i32 7>
- %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 0, i32 1>
- %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 2, i32 3>
- %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 4, i32 5>
- %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 6, i32 7>
- %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 8, i32 9>
- %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 10, i32 11>
- %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 12, i32 13>
- %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> undef, <2 x i32> <i32 14, i32 15>
- %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
- %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
- %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
- %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
- %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
- %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 0, i32 1>
- %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 2, i32 3>
- %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 4, i32 5>
- %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 6, i32 7>
- %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 8, i32 9>
- %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 10, i32 11>
- %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 12, i32 13>
- %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 14, i32 15>
- %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 16, i32 17>
- %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 18, i32 19>
- %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 20, i32 21>
- %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 22, i32 23>
- %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 24, i32 25>
- %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 26, i32 27>
- %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 28, i32 29>
- %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <2 x i32> <i32 30, i32 31>
- %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
- %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
- %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
- %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
- %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
- %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
- %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
- %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
- %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
- %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
- %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
- %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
- %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+ %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 0, i32 1>
+ %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> poison, <2 x i32> <i32 2, i32 3>
+ %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 0, i32 1>
+ %V128_23 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 2, i32 3>
+ %V128_45 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 4, i32 5>
+ %V128_67 = shufflevector <8 x i16> %src128, <8 x i16> poison, <2 x i32> <i32 6, i32 7>
+ %V128_0123 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %V128_4567 = shufflevector <8 x i16> %src128, <8 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %V256_01 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 0, i32 1>
+ %V256_23 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 2, i32 3>
+ %V256_45 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 4, i32 5>
+ %V256_67 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 6, i32 7>
+ %V256_89 = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 8, i32 9>
+ %V256_AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 10, i32 11>
+ %V256_CD = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 12, i32 13>
+ %V256_EF = shufflevector <16 x i16> %src256, <16 x i16> poison, <2 x i32> <i32 14, i32 15>
+ %V256_0123 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %V256_2345 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+ %V256_4567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %V256_6789 = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+ %V256_89AB = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+ %V256_CDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+ %V256_01234567 = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %V512_00_01 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 0, i32 1>
+ %V512_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 2, i32 3>
+ %V512_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 4, i32 5>
+ %V512_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 6, i32 7>
+ %V512_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 8, i32 9>
+ %V512_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 10, i32 11>
+ %V512_0C_0D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 12, i32 13>
+ %V512_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 14, i32 15>
+ %V512_10_11 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 16, i32 17>
+ %V512_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 18, i32 19>
+ %V512_14_15 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 20, i32 21>
+ %V512_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 22, i32 23>
+ %V512_18_19 = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 24, i32 25>
+ %V512_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 26, i32 27>
+ %V512_1C_1D = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 28, i32 29>
+ %V512_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <2 x i32> <i32 30, i32 31>
+ %V512_00_01_02_03 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %V512_02_03_04_05 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+ %V512_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %V512_06_07_08_09 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+ %V512_08_09_0A_0B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+ %V512_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+ %V512_10_11_12_13 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+ %V512_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+ %V512_18_19_1A_1B = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+ %V512_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+ %V512_00_01_02_03_04_05_06_07 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %V512_10_11_12_13_14_15_16_17 = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+ %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+ %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i16> %src512, <32 x i16> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
ret void
}
define void @test_vXi8(<8 x i8> %src64, <16 x i8> %src128, <32 x i8> %src256, <64 x i8> %src512) {
; SSE2-LABEL: 'test_vXi8'
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:10 Lat:10 SizeLat:10 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:10 Lat:10 SizeLat:10 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:10 Lat:10 SizeLat:10 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:10 Lat:10 SizeLat:10 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:10 Lat:10 SizeLat:10 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:10 Lat:10 SizeLat:10 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 16, i32 17>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 18, i32 19>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 20, i32 21>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 22, i32 23>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 24, i32 25>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 26, i32 27>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 28, i32 29>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:10 Lat:10 SizeLat:10 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:10 Lat:10 SizeLat:10 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 16, i32 17>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 18, i32 19>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 20, i32 21>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 22, i32 23>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 24, i32 25>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 26, i32 27>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 28, i32 29>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 32, i32 33>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 34, i32 35>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 36, i32 37>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 38, i32 39>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 40, i32 41>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 42, i32 43>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 44, i32 45>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 46, i32 47>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 48, i32 49>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 50, i32 51>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 52, i32 53>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 54, i32 55>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 56, i32 57>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 58, i32 59>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 60, i32 61>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 62, i32 63>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi8'
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 16, i32 17>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 18, i32 19>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 20, i32 21>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 22, i32 23>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 24, i32 25>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 26, i32 27>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 28, i32 29>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 16, i32 17>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 18, i32 19>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 20, i32 21>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 22, i32 23>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 24, i32 25>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 26, i32 27>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 28, i32 29>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 32, i32 33>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 34, i32 35>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 36, i32 37>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 38, i32 39>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 40, i32 41>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 42, i32 43>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 44, i32 45>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 46, i32 47>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 48, i32 49>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 50, i32 51>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 52, i32 53>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 54, i32 55>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 56, i32 57>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 58, i32 59>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 60, i32 61>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 62, i32 63>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi8'
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 16, i32 17>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 18, i32 19>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 20, i32 21>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 22, i32 23>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 24, i32 25>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 26, i32 27>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 28, i32 29>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 16, i32 17>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 18, i32 19>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 20, i32 21>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 22, i32 23>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 24, i32 25>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 26, i32 27>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 28, i32 29>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 32, i32 33>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 34, i32 35>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 36, i32 37>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 38, i32 39>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 40, i32 41>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 42, i32 43>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 44, i32 45>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 46, i32 47>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 48, i32 49>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 50, i32 51>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 52, i32 53>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 54, i32 55>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 56, i32 57>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 58, i32 59>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 60, i32 61>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 62, i32 63>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi8'
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 8 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 8 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 16, i32 17>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 18, i32 19>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 20, i32 21>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 22, i32 23>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 24, i32 25>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 26, i32 27>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 28, i32 29>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 8 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 8 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 16, i32 17>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 18, i32 19>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 20, i32 21>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 22, i32 23>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 24, i32 25>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 26, i32 27>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 28, i32 29>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 32, i32 33>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 34, i32 35>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 36, i32 37>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 38, i32 39>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 40, i32 41>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 42, i32 43>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 44, i32 45>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 46, i32 47>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 48, i32 49>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 50, i32 51>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 52, i32 53>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 54, i32 55>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 56, i32 57>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 58, i32 59>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 60, i32 61>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 62, i32 63>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi8'
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 4 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 4 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
-; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 16, i32 17>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 18, i32 19>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 20, i32 21>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 22, i32 23>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 24, i32 25>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 26, i32 27>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 28, i32 29>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 4 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 4 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 16, i32 17>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 18, i32 19>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 20, i32 21>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 22, i32 23>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 24, i32 25>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 26, i32 27>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 28, i32 29>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 32, i32 33>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 34, i32 35>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 36, i32 37>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 38, i32 39>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 40, i32 41>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 42, i32 43>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 44, i32 45>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 46, i32 47>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 48, i32 49>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 50, i32 51>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 52, i32 53>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 54, i32 55>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 56, i32 57>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 58, i32 59>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 60, i32 61>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 62, i32 63>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; AVX2-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'test_vXi8'
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 4 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 4 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
-; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 16, i32 17>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 18, i32 19>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 20, i32 21>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 22, i32 23>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 24, i32 25>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 26, i32 27>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 28, i32 29>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 4 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 4 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 16, i32 17>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 18, i32 19>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 20, i32 21>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 22, i32 23>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 24, i32 25>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 26, i32 27>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 28, i32 29>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 32, i32 33>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 34, i32 35>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 36, i32 37>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 38, i32 39>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 40, i32 41>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 42, i32 43>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 44, i32 45>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 46, i32 47>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 48, i32 49>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 50, i32 51>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 52, i32 53>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 54, i32 55>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 56, i32 57>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 58, i32 59>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 60, i32 61>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 62, i32 63>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; AVX512F-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'test_vXi8'
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 4 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 4 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 16, i32 17>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 18, i32 19>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 20, i32 21>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 22, i32 23>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 24, i32 25>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 26, i32 27>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 28, i32 29>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 4 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 4 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 16, i32 17>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 18, i32 19>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 20, i32 21>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 22, i32 23>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 24, i32 25>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 26, i32 27>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 28, i32 29>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 32, i32 33>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 34, i32 35>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 36, i32 37>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 38, i32 39>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 40, i32 41>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 42, i32 43>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 44, i32 45>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 46, i32 47>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 48, i32 49>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 50, i32 51>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 52, i32 53>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 54, i32 55>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 56, i32 57>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 58, i32 59>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 60, i32 61>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 62, i32 63>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI-LABEL: 'test_vXi8'
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 16, i32 17>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 18, i32 19>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 20, i32 21>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 22, i32 23>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 24, i32 25>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 26, i32 27>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 28, i32 29>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 0, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 4, i32 5>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 8, i32 9>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 10, i32 11>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 12, i32 13>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 16, i32 17>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 18, i32 19>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 20, i32 21>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 22, i32 23>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 24, i32 25>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 26, i32 27>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 28, i32 29>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 32, i32 33>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 34, i32 35>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 36, i32 37>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 38, i32 39>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 40, i32 41>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 42, i32 43>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 44, i32 45>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 46, i32 47>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 48, i32 49>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 50, i32 51>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 52, i32 53>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 54, i32 55>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 56, i32 57>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 58, i32 59>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 60, i32 61>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 62, i32 63>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SLM-LABEL: 'test_vXi8'
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
-; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 16, i32 17>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 18, i32 19>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 20, i32 21>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 22, i32 23>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 24, i32 25>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 26, i32 27>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 28, i32 29>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 0, i32 1>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 4, i32 5>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 8, i32 9>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 10, i32 11>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 12, i32 13>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 16, i32 17>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 18, i32 19>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 20, i32 21>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 22, i32 23>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 24, i32 25>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 26, i32 27>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 28, i32 29>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 32, i32 33>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 34, i32 35>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 36, i32 37>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 38, i32 39>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 40, i32 41>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 42, i32 43>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 44, i32 45>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 46, i32 47>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 48, i32 49>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 50, i32 51>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 52, i32 53>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 54, i32 55>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 56, i32 57>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 58, i32 59>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 60, i32 61>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 62, i32 63>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; SLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; SLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
; SLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; GLM-LABEL: 'test_vXi8'
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
-; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
-; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 0, i32 1>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 4, i32 5>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 4, i32 5>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 8, i32 9>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 10, i32 11>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 12, i32 13>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 0, i32 1>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 4, i32 5>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 8, i32 9>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 10, i32 11>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 12, i32 13>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 16, i32 17>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 18, i32 19>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 20, i32 21>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 22, i32 23>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 24, i32 25>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 26, i32 27>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 28, i32 29>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 0, i32 1>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 4, i32 5>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 8, i32 9>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 10, i32 11>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 12, i32 13>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 16, i32 17>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 18, i32 19>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 20, i32 21>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 22, i32 23>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 24, i32 25>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 26, i32 27>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 28, i32 29>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 32, i32 33>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 34, i32 35>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 36, i32 37>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 38, i32 39>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 40, i32 41>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 42, i32 43>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 44, i32 45>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 46, i32 47>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 48, i32 49>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 50, i32 51>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 52, i32 53>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 54, i32 55>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 56, i32 57>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 58, i32 59>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 60, i32 61>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 62, i32 63>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+; GLM-NEXT: Cost Model: Found costs of 1 for: %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+; GLM-NEXT: Cost Model: Found costs of 0 for: %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
; GLM-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 0, i32 1>
- %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 2, i32 3>
- %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 4, i32 5>
- %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> undef, <2 x i32> <i32 6, i32 7>
- %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
- %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 2, i32 3>
- %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 4, i32 5>
- %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 6, i32 7>
- %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 8, i32 9>
- %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 10, i32 11>
- %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 12, i32 13>
- %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> undef, <2 x i32> <i32 14, i32 15>
- %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
- %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
- %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
- %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
- %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
- %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 0, i32 1>
- %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 2, i32 3>
- %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 4, i32 5>
- %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 6, i32 7>
- %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 8, i32 9>
- %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 10, i32 11>
- %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 12, i32 13>
- %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 14, i32 15>
- %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 16, i32 17>
- %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 18, i32 19>
- %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 20, i32 21>
- %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 22, i32 23>
- %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 24, i32 25>
- %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 26, i32 27>
- %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 28, i32 29>
- %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <2 x i32> <i32 30, i32 31>
- %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
- %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
- %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
- %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
- %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
- %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
- %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
- %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
- %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
- %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
- %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
- %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
- %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
- %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 0, i32 1>
- %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 2, i32 3>
- %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 4, i32 5>
- %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 6, i32 7>
- %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 8, i32 9>
- %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 10, i32 11>
- %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 12, i32 13>
- %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 14, i32 15>
- %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 16, i32 17>
- %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 18, i32 19>
- %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 20, i32 21>
- %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 22, i32 23>
- %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 24, i32 25>
- %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 26, i32 27>
- %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 28, i32 29>
- %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 30, i32 31>
- %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 32, i32 33>
- %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 34, i32 35>
- %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 36, i32 37>
- %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 38, i32 39>
- %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 40, i32 41>
- %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 42, i32 43>
- %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 44, i32 45>
- %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 46, i32 47>
- %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 48, i32 49>
- %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 50, i32 51>
- %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 52, i32 53>
- %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 54, i32 55>
- %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 56, i32 57>
- %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 58, i32 59>
- %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 60, i32 61>
- %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <2 x i32> <i32 62, i32 63>
- %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 8, i32 9,i32 10, i32 11>
- %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
- %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
- %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
- %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
- %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
- %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
- %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
- %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
- %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
- %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
- %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
- %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
- %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
- %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
- %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
- %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
- %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
- %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
- %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
- %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
- %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
- %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23,i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
- %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
- %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
- %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23,i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
- %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> undef, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 0, i32 1>
+ %V64_23 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 2, i32 3>
+ %V64_45 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 4, i32 5>
+ %V64_67 = shufflevector <8 x i8> %src64, <8 x i8> poison, <2 x i32> <i32 6, i32 7>
+ %V64_0123 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %V64_4567 = shufflevector <8 x i8> %src64, <8 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %V128_01 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 0, i32 1>
+ %V128_23 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 2, i32 3>
+ %V128_45 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 4, i32 5>
+ %V128_67 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 6, i32 7>
+ %V128_89 = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 8, i32 9>
+ %V128_AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 10, i32 11>
+ %V128_CD = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 12, i32 13>
+ %V128_EF = shufflevector <16 x i8> %src128, <16 x i8> poison, <2 x i32> <i32 14, i32 15>
+ %V128_0123 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %V128_2345 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+ %V128_4567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %V128_6789 = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+ %V128_89AB = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+ %V128_CDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+ %V128_01234567 = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %V256_00_01 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 0, i32 1>
+ %V256_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 2, i32 3>
+ %V256_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 4, i32 5>
+ %V256_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 6, i32 7>
+ %V256_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 8, i32 9>
+ %V256_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 10, i32 11>
+ %V256_0C_0D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 12, i32 13>
+ %V256_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 14, i32 15>
+ %V256_10_11 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 16, i32 17>
+ %V256_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 18, i32 19>
+ %V256_14_15 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 20, i32 21>
+ %V256_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 22, i32 23>
+ %V256_18_19 = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 24, i32 25>
+ %V256_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 26, i32 27>
+ %V256_1C_1D = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 28, i32 29>
+ %V256_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <2 x i32> <i32 30, i32 31>
+ %V256_00_01_02_03 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %V256_02_03_04_05 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
+ %V256_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %V256_06_07_08_09 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 6, i32 7, i32 8, i32 9>
+ %V256_08_09_0A_0B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 8, i32 9, i32 10, i32 11>
+ %V256_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+ %V256_10_11_12_13 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+ %V256_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+ %V256_18_19_1A_1B = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+ %V256_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+ %V256_00_01_02_03_04_05_06_07 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %V256_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %V256_10_11_12_13_14_15_16_17 = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+ %V256_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+ %V256_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %V256_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <32 x i8> %src256, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+ %V512_00_01 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 0, i32 1>
+ %V512_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 2, i32 3>
+ %V512_04_05 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 4, i32 5>
+ %V512_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 6, i32 7>
+ %V512_08_09 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 8, i32 9>
+ %V512_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 10, i32 11>
+ %V512_0C_0D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 12, i32 13>
+ %V512_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 14, i32 15>
+ %V512_10_11 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 16, i32 17>
+ %V512_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 18, i32 19>
+ %V512_14_15 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 20, i32 21>
+ %V512_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 22, i32 23>
+ %V512_18_19 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 24, i32 25>
+ %V512_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 26, i32 27>
+ %V512_1C_1D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 28, i32 29>
+ %V512_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 30, i32 31>
+ %V512_20_21 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 32, i32 33>
+ %V512_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 34, i32 35>
+ %V512_24_25 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 36, i32 37>
+ %V512_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 38, i32 39>
+ %V512_28_29 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 40, i32 41>
+ %V512_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 42, i32 43>
+ %V512_2C_2D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 44, i32 45>
+ %V512_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 46, i32 47>
+ %V512_30_31 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 48, i32 49>
+ %V512_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 50, i32 51>
+ %V512_34_35 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 52, i32 53>
+ %V512_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 54, i32 55>
+ %V512_38_39 = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 56, i32 57>
+ %V512_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 58, i32 59>
+ %V512_3C_3D = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 60, i32 61>
+ %V512_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <2 x i32> <i32 62, i32 63>
+ %V512_00_01_02_03 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %V512_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+ %V512_08_09_0A_0B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 8, i32 9,i32 10, i32 11>
+ %V512_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
+ %V512_10_11_12_13 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 16, i32 17, i32 18, i32 19>
+ %V512_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 20, i32 21, i32 22, i32 23>
+ %V512_18_19_1A_1B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 24, i32 25, i32 26, i32 27>
+ %V512_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 28, i32 29, i32 30, i32 31>
+ %V512_20_21_22_23 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 32, i32 33, i32 34, i32 35>
+ %V512_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 36, i32 37, i32 38, i32 39>
+ %V512_28_29_2A_2B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 40, i32 41, i32 42, i32 43>
+ %V512_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 44, i32 45, i32 46, i32 47>
+ %V512_30_31_32_33 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 48, i32 49, i32 50, i32 51>
+ %V512_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 52, i32 53, i32 54, i32 55>
+ %V512_38_39_3A_3B = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 56, i32 57, i32 58, i32 59>
+ %V512_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <4 x i32> <i32 60, i32 61, i32 62, i32 63>
+ %V512_00_01_02_03_04_05_06_07 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %V512_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %V512_10_11_12_13_14_15_16_17 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+ %V512_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+ %V512_20_21_22_23_24_25_26_27 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39>
+ %V512_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+ %V512_30_31_32_33_34_35_36_37 = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55>
+ %V512_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <8 x i32> <i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %V512_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23,i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+ %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
+ %V512_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <16 x i32> <i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
+ %V512_00_01_02_03_04_05_06_07_08_09_0A_0B_0C_0D_0E_0F_10_11_12_13_14_15_16_17_18_19_1A_1B_1C_1D_1E_1F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23,i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+ %V512_20_21_22_23_24_25_26_27_28_29_2A_2B_2C_2D_2E_2F_30_31_32_33_34_35_36_37_38_39_3A_3B_3C_3D_3E_3F = shufflevector <64 x i8> %src512, <64 x i8> poison, <32 x i32> <i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
ret void
}
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-insert_subvector.ll b/llvm/test/Analysis/CostModel/X86/shuffle-insert_subvector.ll
index fef397cb216e6..770b90e1a364c 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-insert_subvector.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-insert_subvector.ll
@@ -18,9 +18,9 @@
define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double> %src512) {
; SSE-LABEL: 'test_vXf64'
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x double> %src128, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x double> %src128, <2 x double> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x double> %src256, <4 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x double> %src128, <2 x double> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x double> %src128, <2 x double> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x double> %src256, <4 x double> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x double> %src256, <4 x double> %src128_256, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_23 = shufflevector <4 x double> %src256, <4 x double> %src128_256, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x double> %src512, <8 x double> %src128_512, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -32,9 +32,9 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX-LABEL: 'test_vXf64'
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x double> %src128, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x double> %src128, <2 x double> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x double> %src256, <4 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x double> %src128, <2 x double> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x double> %src128, <2 x double> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x double> %src256, <4 x double> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <4 x double> %src256, <4 x double> %src128_256, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x double> %src256, <4 x double> %src128_256, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_01 = shufflevector <8 x double> %src512, <8 x double> %src128_512, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -46,9 +46,9 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf64'
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x double> %src128, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x double> %src128, <2 x double> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x double> %src256, <4 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x double> %src128, <2 x double> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x double> %src128, <2 x double> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x double> %src256, <4 x double> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <4 x double> %src256, <4 x double> %src128_256, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x double> %src256, <4 x double> %src128_256, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_01 = shufflevector <8 x double> %src512, <8 x double> %src128_512, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -59,9 +59,9 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_4567 = shufflevector <8 x double> %src512, <8 x double> %src256_512, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %src128_256 = shufflevector <2 x double> %src128, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
- %src128_512 = shufflevector <2 x double> %src128, <2 x double> undef, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src256_512 = shufflevector <4 x double> %src256, <4 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
+ %src128_256 = shufflevector <2 x double> %src128, <2 x double> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 undef>
+ %src128_512 = shufflevector <2 x double> %src128, <2 x double> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src256_512 = shufflevector <4 x double> %src256, <4 x double> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 undef>
%V256_01 = shufflevector <4 x double> %src256, <4 x double> %src128_256, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
%V256_23 = shufflevector <4 x double> %src256, <4 x double> %src128_256, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
@@ -76,9 +76,9 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512) {
; SSE-LABEL: 'test_vXi64'
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x i64> %src128, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x i64> %src128, <2 x i64> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x i64> %src256, <4 x i64> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x i64> %src128, <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x i64> %src128, <2 x i64> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x i64> %src256, <4 x i64> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_01 = shufflevector <4 x i64> %src256, <4 x i64> %src128_256, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; SSE-NEXT: Cost Model: Found costs of 0 for: %V256_23 = shufflevector <4 x i64> %src256, <4 x i64> %src128_256, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; SSE-NEXT: Cost Model: Found costs of 0 for: %V512_01 = shufflevector <8 x i64> %src512, <8 x i64> %src128_512, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -90,9 +90,9 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512)
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX-LABEL: 'test_vXi64'
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x i64> %src128, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x i64> %src128, <2 x i64> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x i64> %src256, <4 x i64> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x i64> %src128, <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x i64> %src128, <2 x i64> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x i64> %src256, <4 x i64> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <4 x i64> %src256, <4 x i64> %src128_256, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x i64> %src256, <4 x i64> %src128_256, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX-NEXT: Cost Model: Found costs of 1 for: %V512_01 = shufflevector <8 x i64> %src512, <8 x i64> %src128_512, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -104,9 +104,9 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512)
; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi64'
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x i64> %src128, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x i64> %src128, <2 x i64> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x i64> %src256, <4 x i64> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <2 x i64> %src128, <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <2 x i64> %src128, <2 x i64> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <4 x i64> %src256, <4 x i64> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <4 x i64> %src256, <4 x i64> %src128_256, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_23 = shufflevector <4 x i64> %src256, <4 x i64> %src128_256, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_01 = shufflevector <8 x i64> %src512, <8 x i64> %src128_512, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -117,9 +117,9 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512)
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_4567 = shufflevector <8 x i64> %src512, <8 x i64> %src256_512, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %src128_256 = shufflevector <2 x i64> %src128, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
- %src128_512 = shufflevector <2 x i64> %src128, <2 x i64> undef, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src256_512 = shufflevector <4 x i64> %src256, <4 x i64> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
+ %src128_256 = shufflevector <2 x i64> %src128, <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 undef>
+ %src128_512 = shufflevector <2 x i64> %src128, <2 x i64> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src256_512 = shufflevector <4 x i64> %src256, <4 x i64> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 undef>
%V256_01 = shufflevector <4 x i64> %src256, <4 x i64> %src128_256, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
%V256_23 = shufflevector <4 x i64> %src256, <4 x i64> %src128_256, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
@@ -134,12 +134,12 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512)
define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %src256, <16 x float> %src512) {
; SSE-LABEL: 'test_vXf32'
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x float> %src64, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x float> %src64, <2 x float> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x float> %src64, <2 x float> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x float> %src128, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x float> %src128, <4 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x float> %src256, <8 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x float> %src64, <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x float> %src64, <2 x float> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x float> %src64, <2 x float> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x float> %src128, <4 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x float> %src128, <4 x float> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x float> %src256, <8 x float> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; SSE-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <4 x float> %src128, <4 x float> %src64_128, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; SSE-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <4 x float> %src128, <4 x float> %src64_128, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; SSE-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <8 x float> %src256, <8 x float> %src64_256, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -165,12 +165,12 @@ define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %sr
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXf32'
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x float> %src64, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x float> %src64, <2 x float> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x float> %src64, <2 x float> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x float> %src128, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x float> %src128, <4 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x float> %src256, <8 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x float> %src64, <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x float> %src64, <2 x float> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x float> %src64, <2 x float> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x float> %src128, <4 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x float> %src128, <4 x float> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x float> %src256, <8 x float> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <4 x float> %src128, <4 x float> %src64_128, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <4 x float> %src128, <4 x float> %src64_128, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <8 x float> %src256, <8 x float> %src64_256, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -196,12 +196,12 @@ define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %sr
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf32'
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x float> %src64, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x float> %src64, <2 x float> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x float> %src64, <2 x float> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x float> %src128, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x float> %src128, <4 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x float> %src256, <8 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x float> %src64, <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x float> %src64, <2 x float> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x float> %src64, <2 x float> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x float> %src128, <4 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x float> %src128, <4 x float> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x float> %src256, <8 x float> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <4 x float> %src128, <4 x float> %src64_128, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_23 = shufflevector <4 x float> %src128, <4 x float> %src64_128, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <8 x float> %src256, <8 x float> %src64_256, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -227,12 +227,12 @@ define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %sr
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf32'
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x float> %src64, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x float> %src64, <2 x float> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x float> %src64, <2 x float> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x float> %src128, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x float> %src128, <4 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x float> %src256, <8 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x float> %src64, <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x float> %src64, <2 x float> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x float> %src64, <2 x float> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x float> %src128, <4 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x float> %src128, <4 x float> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x float> %src256, <8 x float> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <4 x float> %src128, <4 x float> %src64_128, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_23 = shufflevector <4 x float> %src128, <4 x float> %src64_128, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <8 x float> %src256, <8 x float> %src64_256, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -257,12 +257,12 @@ define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %sr
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_89ABCDEF = shufflevector <16 x float> %src512, <16 x float> %src128_512, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %src64_128 = shufflevector <2 x float> %src64, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
- %src64_256 = shufflevector <2 x float> %src64, <2 x float> undef, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src64_512 = shufflevector <2 x float> %src64, <2 x float> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src128_256 = shufflevector <4 x float> %src128, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
- %src128_512 = shufflevector <4 x float> %src128, <4 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src256_512 = shufflevector <8 x float> %src256, <8 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ %src64_128 = shufflevector <2 x float> %src64, <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 undef>
+ %src64_256 = shufflevector <2 x float> %src64, <2 x float> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src64_512 = shufflevector <2 x float> %src64, <2 x float> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src128_256 = shufflevector <4 x float> %src128, <4 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src128_512 = shufflevector <4 x float> %src128, <4 x float> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src256_512 = shufflevector <8 x float> %src256, <8 x float> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
%V128_01 = shufflevector <4 x float> %src128, <4 x float> %src64_128, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
%V128_23 = shufflevector <4 x float> %src128, <4 x float> %src64_128, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
@@ -293,12 +293,12 @@ define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %sr
define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512) {
; SSE-LABEL: 'test_vXi32'
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x i32> %src64, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x i32> %src64, <2 x i32> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x i32> %src64, <2 x i32> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x i32> %src128, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x i32> %src128, <4 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x i32> %src256, <8 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x i32> %src64, <2 x i32> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x i32> %src64, <2 x i32> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x i32> %src64, <2 x i32> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x i32> %src128, <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x i32> %src128, <4 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x i32> %src256, <8 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; SSE-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> %src64_128, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; SSE-NEXT: Cost Model: Found costs of 2 for: %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> %src64_128, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; SSE-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> %src64_256, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -324,12 +324,12 @@ define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256,
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi32'
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x i32> %src64, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x i32> %src64, <2 x i32> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x i32> %src64, <2 x i32> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x i32> %src128, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x i32> %src128, <4 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x i32> %src256, <8 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x i32> %src64, <2 x i32> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x i32> %src64, <2 x i32> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x i32> %src64, <2 x i32> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x i32> %src128, <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x i32> %src128, <4 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x i32> %src256, <8 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> %src64_128, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX1-NEXT: Cost Model: Found costs of 2 for: %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> %src64_128, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX1-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> %src64_256, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -355,12 +355,12 @@ define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256,
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi32'
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x i32> %src64, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x i32> %src64, <2 x i32> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x i32> %src64, <2 x i32> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x i32> %src128, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x i32> %src128, <4 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x i32> %src256, <8 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x i32> %src64, <2 x i32> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x i32> %src64, <2 x i32> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x i32> %src64, <2 x i32> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x i32> %src128, <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x i32> %src128, <4 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x i32> %src256, <8 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> %src64_128, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX2-NEXT: Cost Model: Found costs of 2 for: %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> %src64_128, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> %src64_256, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -386,12 +386,12 @@ define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256,
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi32'
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x i32> %src64, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x i32> %src64, <2 x i32> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x i32> %src64, <2 x i32> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x i32> %src128, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x i32> %src128, <4 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x i32> %src256, <8 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <2 x i32> %src64, <2 x i32> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <2 x i32> %src64, <2 x i32> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <2 x i32> %src64, <2 x i32> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <4 x i32> %src128, <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <4 x i32> %src128, <4 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <8 x i32> %src256, <8 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <4 x i32> %src128, <4 x i32> %src64_128, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128_23 = shufflevector <4 x i32> %src128, <4 x i32> %src64_128, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V256_01 = shufflevector <8 x i32> %src256, <8 x i32> %src64_256, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -416,12 +416,12 @@ define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256,
; AVX512-NEXT: Cost Model: Found costs of 1 for: %V512_89ABCDEF = shufflevector <16 x i32> %src512, <16 x i32> %src128_512, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %src64_128 = shufflevector <2 x i32> %src64, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
- %src64_256 = shufflevector <2 x i32> %src64, <2 x i32> undef, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src64_512 = shufflevector <2 x i32> %src64, <2 x i32> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src128_256 = shufflevector <4 x i32> %src128, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
- %src128_512 = shufflevector <4 x i32> %src128, <4 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src256_512 = shufflevector <8 x i32> %src256, <8 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ %src64_128 = shufflevector <2 x i32> %src64, <2 x i32> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 undef>
+ %src64_256 = shufflevector <2 x i32> %src64, <2 x i32> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src64_512 = shufflevector <2 x i32> %src64, <2 x i32> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src128_256 = shufflevector <4 x i32> %src128, <4 x i32> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src128_512 = shufflevector <4 x i32> %src128, <4 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src256_512 = shufflevector <8 x i32> %src256, <8 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
%V128_01 = shufflevector <4 x i32> %src128, <4 x i32> %src64_128, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
%V128_23 = shufflevector <4 x i32> %src128, <4 x i32> %src64_128, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
@@ -452,16 +452,16 @@ define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256,
define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <16x i16> %src256, <32 x i16> %src512) {
; SSE2-LABEL: 'test_vXi16'
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> %src32_128, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -487,16 +487,16 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi16'
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; SSSE3-NEXT: Cost Model: Found costs of 3 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> %src32_128, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -522,16 +522,16 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi16'
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; SSE42-NEXT: Cost Model: Found costs of 3 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> %src32_128, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -557,16 +557,16 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi16'
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX1-NEXT: Cost Model: Found costs of 3 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> %src32_128, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -592,16 +592,16 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi16'
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX2-NEXT: Cost Model: Found costs of 3 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> %src32_128, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -627,16 +627,16 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'test_vXi16'
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX512F-NEXT: Cost Model: Found costs of 3 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> %src32_128, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -662,16 +662,16 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'test_vXi16'
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> %src32_128, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -697,16 +697,16 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VMBI-LABEL: 'test_vXi16'
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX512VMBI-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX512VMBI-NEXT: Cost Model: Found costs of 2 for: %V64_23 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX512VMBI-NEXT: Cost Model: Found costs of 1 for: %V128_01 = shufflevector <8 x i16> %src128, <8 x i16> %src32_128, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -731,16 +731,16 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
; AVX512VMBI-NEXT: Cost Model: Found costs of 1 for: %V256_89ABCDEF = shufflevector <16 x i16> %src256, <16 x i16> %src128_256, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
; AVX512VMBI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
- %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> undef, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> undef, <32 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
- %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ %src32_64 = shufflevector <2 x i16> %src32, <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 undef>
+ %src32_128 = shufflevector <2 x i16> %src32, <2 x i16> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src32_256 = shufflevector <2 x i16> %src32, <2 x i16> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src32_512 = shufflevector <2 x i16> %src32, <2 x i16> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src64_128 = shufflevector <4 x i16> %src64, <4 x i16> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src64_256 = shufflevector <4 x i16> %src64, <4 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src64_512 = shufflevector <4 x i16> %src64, <4 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src128_256 = shufflevector <8 x i16> %src128, <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src128_512 = shufflevector <8 x i16> %src128, <8 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src256_512 = shufflevector <16 x i16> %src256, <16 x i16> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
%V64_01 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
%V64_23 = shufflevector <4 x i16> %src64, <4 x i16> %src32_64, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
@@ -772,21 +772,21 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16x i8> %src128, <32 x i8> %src256, <64 x i8> %src512) {
; SSE2-LABEL: 'test_vXi8'
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> undef, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> poison, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; SSE2-NEXT: Cost Model: Found costs of 1 for: %V32_01 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; SSE2-NEXT: Cost Model: Found costs of 4 for: %V32_23 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> %src32_64, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -812,21 +812,21 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16x i
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi8'
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> undef, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> poison, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V32_01 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; SSSE3-NEXT: Cost Model: Found costs of 3 for: %V32_23 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> %src32_64, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -852,21 +852,21 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16x i
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi8'
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> undef, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> poison, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; SSE42-NEXT: Cost Model: Found costs of 1 for: %V32_01 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; SSE42-NEXT: Cost Model: Found costs of 3 for: %V32_23 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> %src32_64, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -892,21 +892,21 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16x i
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX-LABEL: 'test_vXi8'
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> undef, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> poison, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: Cost Model: Found costs of 1 for: %V32_01 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX-NEXT: Cost Model: Found costs of 3 for: %V32_23 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> %src32_64, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -932,21 +932,21 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16x i
; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'test_vXi8'
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> undef, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> poison, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512F-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V32_01 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX512F-NEXT: Cost Model: Found costs of 3 for: %V32_23 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> %src32_64, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -972,21 +972,21 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16x i
; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'test_vXi8'
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> undef, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> poison, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512BW-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V32_01 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX512BW-NEXT: Cost Model: Found costs of 3 for: %V32_23 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> %src32_64, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -1012,21 +1012,21 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16x i
; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VMBI-LABEL: 'test_vXi8'
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> undef, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> undef, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> undef, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> undef, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> poison, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; AVX512VMBI-NEXT: Cost Model: Found costs of 0 for: %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX512VMBI-NEXT: Cost Model: Found costs of 1 for: %V32_01 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
; AVX512VMBI-NEXT: Cost Model: Found costs of 2 for: %V32_23 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
; AVX512VMBI-NEXT: Cost Model: Found costs of 1 for: %V64_01 = shufflevector <8 x i8> %src64, <8 x i8> %src32_64, <8 x i32> <i32 8, i32 9, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
@@ -1051,21 +1051,21 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16x i
; AVX512VMBI-NEXT: Cost Model: Found costs of 2 for: %V128_89ABCDEF = shufflevector <16 x i8> %src128, <16 x i8> %src64_128, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
; AVX512VMBI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
- %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> undef, <8 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> undef, <32 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> undef, <64 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
- %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
- %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> undef, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+ %src16_32 = shufflevector <2 x i8> %src16, <2 x i8> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 undef>
+ %src16_64 = shufflevector <2 x i8> %src16, <2 x i8> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src16_128 = shufflevector <2 x i8> %src16, <2 x i8> poison, <16 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src16_256 = shufflevector <2 x i8> %src16, <2 x i8> poison, <32 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src16_512 = shufflevector <2 x i8> %src16, <2 x i8> poison, <64 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src32_64 = shufflevector <4 x i8> %src32, <4 x i8> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src32_128 = shufflevector <4 x i8> %src32, <4 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src32_256 = shufflevector <4 x i8> %src32, <4 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src32_512 = shufflevector <4 x i8> %src32, <4 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src64_128 = shufflevector <8 x i8> %src64, <8 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src64_256 = shufflevector <8 x i8> %src64, <8 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src64_512 = shufflevector <8 x i8> %src64, <8 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src128_256 = shufflevector <16 x i8> %src128, <16 x i8> poison, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src128_512 = shufflevector <16 x i8> %src128, <16 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
+ %src256_512 = shufflevector <32 x i8> %src256, <32 x i8> poison, <64 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 undef>
%V32_01 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
%V32_23 = shufflevector <4 x i8> %src32, <4 x i8> %src16_32, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-non-pow-2.ll b/llvm/test/Analysis/CostModel/X86/shuffle-non-pow-2.ll
index a4e0ca64e01d7..6fba7ee0d9a35 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-non-pow-2.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-non-pow-2.ll
@@ -4,15 +4,15 @@
define void @test(<9 x double> noundef %src) {
; SSE-LABEL: 'test'
-; SSE-NEXT: Cost Model: Found costs of 4 for: %matins.2.2 = shufflevector <9 x double> %src, <9 x double> undef, <9 x i32> <i32 0, i32 3, i32 6, i32 1, i32 4, i32 7, i32 2, i32 5, i32 8>
+; SSE-NEXT: Cost Model: Found costs of 4 for: %matins.2.2 = shufflevector <9 x double> %src, <9 x double> poison, <9 x i32> <i32 0, i32 3, i32 6, i32 1, i32 4, i32 7, i32 2, i32 5, i32 8>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX-LABEL: 'test'
-; AVX-NEXT: Cost Model: Found costs of 6 for: %matins.2.2 = shufflevector <9 x double> %src, <9 x double> undef, <9 x i32> <i32 0, i32 3, i32 6, i32 1, i32 4, i32 7, i32 2, i32 5, i32 8>
+; AVX-NEXT: Cost Model: Found costs of 6 for: %matins.2.2 = shufflevector <9 x double> %src, <9 x double> poison, <9 x i32> <i32 0, i32 3, i32 6, i32 1, i32 4, i32 7, i32 2, i32 5, i32 8>
; AVX-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
entry:
- %matins.2.2 = shufflevector <9 x double> %src, <9 x double> undef, <9 x i32> <i32 0, i32 3, i32 6, i32 1, i32 4, i32 7, i32 2, i32 5, i32 8>
+ %matins.2.2 = shufflevector <9 x double> %src, <9 x double> poison, <9 x i32> <i32 0, i32 3, i32 6, i32 1, i32 4, i32 7, i32 2, i32 5, i32 8>
ret void
}
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-reverse-fp16.ll b/llvm/test/Analysis/CostModel/X86/shuffle-reverse-fp16.ll
index f47e1a35aa087..2945c05159b1f 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-reverse-fp16.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-reverse-fp16.ll
@@ -3,17 +3,17 @@
define void @test_vXf16(<2 x half> %src32, <4 x half> %src64, <8 x half> %src128, <16 x half> %src256, <32 x half> %src512) {
; CHECK-LABEL: 'test_vXf16'
-; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 0>
-; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:4 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> <i32 1, i32 0>
+; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:4 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 0>
- %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
- %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> <i32 1, i32 0>
+ %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+ %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
ret void
}
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-reverse.ll b/llvm/test/Analysis/CostModel/X86/shuffle-reverse.ll
index aca0ba3ca74e4..b0c761690e5a5 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-reverse.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-reverse.ll
@@ -18,290 +18,290 @@
define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double> %src512) {
; SSE-LABEL: 'test_vXf64'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXf64'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf64'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf64'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:5 SizeLat:3 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:5 SizeLat:3 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 0>
- %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
- %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 0>
+ %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+ %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
ret void
}
define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512) {
; SSE-LABEL: 'test_vXi64'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi64'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi64'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi64'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:5 SizeLat:3 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:5 SizeLat:3 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
- %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
- %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 0>
+ %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+ %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
ret void
}
define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %src256, <16 x float> %src512) {
; SSE-LABEL: 'test_vXf32'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXf32'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf32'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf32'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 0>
- %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
- %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 0>
+ %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+ %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
ret void
}
define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512) {
; SSE-LABEL: 'test_vXi32'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi32'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi32'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi32'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
- %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
- %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 0>
+ %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+ %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
ret void
}
define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src512) {
; SSE2-LABEL: 'test_vXi16'
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:3 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of RThru:4 CodeSize:6 Lat:6 SizeLat:6 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:12 Lat:12 SizeLat:12 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:3 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of RThru:4 CodeSize:6 Lat:6 SizeLat:6 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:12 Lat:12 SizeLat:12 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi16'
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi16'
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi16'
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:5 Lat:9 SizeLat:5 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:10 Lat:18 SizeLat:10 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:5 Lat:9 SizeLat:5 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:10 Lat:18 SizeLat:10 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi16'
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:18 SizeLat:8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:18 SizeLat:8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'test_vXi16'
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of 7 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of 7 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'test_vXi16'
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI-LABEL: 'test_vXi16'
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
- %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
- %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 0>
+ %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+ %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
ret void
}
define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x i8> %src128, <32 x i8> %src256, <64 x i8> %src512) {
; SSE2-LABEL: 'test_vXi8'
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of RThru:5 CodeSize:11 Lat:6 SizeLat:11 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of RThru:5 CodeSize:11 Lat:6 SizeLat:11 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of RThru:10 CodeSize:22 Lat:12 SizeLat:22 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of RThru:20 CodeSize:44 Lat:24 SizeLat:44 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of RThru:5 CodeSize:11 Lat:6 SizeLat:11 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of RThru:5 CodeSize:11 Lat:6 SizeLat:11 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of RThru:10 CodeSize:22 Lat:12 SizeLat:22 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of RThru:20 CodeSize:44 Lat:24 SizeLat:44 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi8'
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi8'
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi8'
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:5 Lat:9 SizeLat:5 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:10 Lat:18 SizeLat:10 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:5 Lat:9 SizeLat:5 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:4 CodeSize:10 Lat:18 SizeLat:10 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi8'
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:18 SizeLat:8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:4 CodeSize:4 Lat:18 SizeLat:8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'test_vXi8'
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of 7 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of 7 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'test_vXi8'
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:3 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:3 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI-LABEL: 'test_vXi8'
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
- %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
- %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 0>
+ %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+ %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
ret void
}
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-single-src-fp16.ll b/llvm/test/Analysis/CostModel/X86/shuffle-single-src-fp16.ll
index 21d576677de0d..22e28f47984f6 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-single-src-fp16.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-single-src-fp16.ll
@@ -3,15 +3,15 @@
define void @test_vXf16(<8 x half> %src128, <16 x half> %src256, <32 x half> %src512, <64 x half> %src1024) {
; CHECK-LABEL: 'test_vXf16'
-; CHECK-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
-; CHECK-NEXT: Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; CHECK-NEXT: Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT: Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x half> %src1024, <64 x half> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; CHECK-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; CHECK-NEXT: Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; CHECK-NEXT: Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT: Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x half> %src1024, <64 x half> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
- %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
- %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V1024 = shufflevector <64 x half> %src1024, <64 x half> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+ %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+ %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+ %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V1024 = shufflevector <64 x half> %src1024, <64 x half> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
ret void
}
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-single-src.ll b/llvm/test/Analysis/CostModel/X86/shuffle-single-src.ll
index accd83bc9444e..abe176f0b4a31 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-single-src.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-single-src.ll
@@ -19,369 +19,369 @@
define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double> %src512, <16 x double> %src1024) {
; SSE-LABEL: 'test_vXf64'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 8 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 8 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> poison, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'test_vXf64'
-; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; XOP-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; XOP-NEXT: Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
-; XOP-NEXT: Cost Model: Found costs of 11 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+; XOP-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 11 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> poison, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXf64'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 11 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 11 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> poison, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf64'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 10 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 10 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> poison, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf64'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> poison, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
- %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
- %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
- %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
+ %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+ %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+ %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+ %V1024 = shufflevector <16 x double> %src1024, <16 x double> poison, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
ret void
}
define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512) {
; SSE-LABEL: 'test_vXi64'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'test_vXi64'
-; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; XOP-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; XOP-NEXT: Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
+; XOP-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi64'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi64'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi64'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
- %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
- %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+ %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
+ %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+ %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
ret void
}
define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %src256, <16 x float> %src512) {
; SSE2-LABEL: 'test_vXf32'
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of 11 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of 11 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXf32'
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of 11 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of 11 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXf32'
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of 10 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of 10 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'test_vXf32'
-; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; XOP-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT: Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXf32'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf32'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 9 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 9 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf32'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
- %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
- %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+ %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+ %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+ %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
ret void
}
define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512, <32 x i32> %src1024) {
; SSE-LABEL: 'test_vXi32'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
-; SSE-NEXT: Cost Model: Found costs of 11 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; SSE-NEXT: Cost Model: Found costs of 18 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+; SSE-NEXT: Cost Model: Found costs of 11 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSE-NEXT: Cost Model: Found costs of 18 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'test_vXi32'
-; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; XOP-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
-; XOP-NEXT: Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; XOP-NEXT: Cost Model: Found costs of 26 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
+; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; XOP-NEXT: Cost Model: Found costs of 26 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi32'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; AVX1-NEXT: Cost Model: Found costs of 26 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX1-NEXT: Cost Model: Found costs of 26 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi32'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 9 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; AVX2-NEXT: Cost Model: Found costs of 20 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 9 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX2-NEXT: Cost Model: Found costs of 20 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi32'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
-; AVX512-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; AVX512-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+; AVX512-NEXT: Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX512-NEXT: Cost Model: Found costs of RThru:4 CodeSize:2 Lat:6 SizeLat:2 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
- %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
- %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
- %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
- %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
+ %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+ %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+ %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+ %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
ret void
}
define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src512, <64 x i16> %src1024) {
; SSE2-LABEL: 'test_vXi16'
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:5 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of RThru:18 CodeSize:24 Lat:24 SizeLat:24 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; SSE2-NEXT: Cost Model: Found costs of RThru:42 CodeSize:56 Lat:56 SizeLat:56 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of RThru:70 CodeSize:94 Lat:92 SizeLat:94 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:5 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of RThru:18 CodeSize:24 Lat:24 SizeLat:24 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSE2-NEXT: Cost Model: Found costs of RThru:42 CodeSize:56 Lat:56 SizeLat:56 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of RThru:70 CodeSize:94 Lat:92 SizeLat:94 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi16'
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; SSSE3-NEXT: Cost Model: Found costs of 21 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:37 CodeSize:39 Lat:37 SizeLat:39 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSSE3-NEXT: Cost Model: Found costs of 21 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:37 CodeSize:39 Lat:37 SizeLat:39 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi16'
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; SSE42-NEXT: Cost Model: Found costs of 21 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of 35 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSE42-NEXT: Cost Model: Found costs of 21 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of 35 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'test_vXi16'
-; XOP-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
-; XOP-NEXT: Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; XOP-NEXT: Cost Model: Found costs of 18 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT: Cost Model: Found costs of 63 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
+; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; XOP-NEXT: Cost Model: Found costs of 18 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 63 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi16'
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 15 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; AVX1-NEXT: Cost Model: Found costs of 30 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 99 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 15 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX1-NEXT: Cost Model: Found costs of 30 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 99 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi16'
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; AVX2-NEXT: Cost Model: Found costs of 14 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 44 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX2-NEXT: Cost Model: Found costs of 14 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 44 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'test_vXi16'
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; AVX512F-NEXT: Cost Model: Found costs of 14 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of 84 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX512F-NEXT: Cost Model: Found costs of 14 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of 84 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'test_vXi16'
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX512BW-NEXT: Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI-LABEL: 'test_vXi16'
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
- %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
- %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
- %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
- %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+ %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+ %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
+ %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+ %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+ %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
ret void
}
define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x i8> %src128, <32 x i8> %src256, <64 x i8> %src512) {
; SSE2-LABEL: 'test_vXi8'
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 5 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:10 Lat:10 SizeLat:10 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of RThru:22 CodeSize:26 Lat:26 SizeLat:26 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT: Cost Model: Found costs of RThru:77 CodeSize:91 Lat:91 SizeLat:91 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 5 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of RThru:8 CodeSize:10 Lat:10 SizeLat:10 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of RThru:22 CodeSize:26 Lat:26 SizeLat:26 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT: Cost Model: Found costs of RThru:77 CodeSize:91 Lat:91 SizeLat:91 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi8'
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of 6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT: Cost Model: Found costs of 21 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of 6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT: Cost Model: Found costs of 21 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi8'
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of 6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT: Cost Model: Found costs of 21 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of 6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT: Cost Model: Found costs of 21 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; XOP-LABEL: 'test_vXi8'
-; XOP-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; XOP-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
-; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
-; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT: Cost Model: Found costs of 18 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; XOP-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
+; XOP-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
+; XOP-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT: Cost Model: Found costs of 18 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; XOP-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi8'
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 8 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT: Cost Model: Found costs of 30 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 8 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT: Cost Model: Found costs of 30 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi8'
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT: Cost Model: Found costs of 14 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT: Cost Model: Found costs of 14 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512F-LABEL: 'test_vXi8'
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
-; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT: Cost Model: Found costs of 14 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
+; AVX512F-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT: Cost Model: Found costs of 14 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; AVX512F-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512BW-LABEL: 'test_vXi8'
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
-; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT: Cost Model: Found costs of 8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
+; AVX512BW-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT: Cost Model: Found costs of 8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; AVX512BW-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512VBMI-LABEL: 'test_vXi8'
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
-; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
+; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
; AVX512VBMI-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
- %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
- %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
- %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
- %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
+ %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+ %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
+ %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
+ %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+ %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
ret void
}
@@ -396,8 +396,8 @@ define void @identity_vXf32(<4 x float> %a128, <4 x float> %b128, <8 x float> %a
; CHECK-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 0, i32 poison, i32 2, i32 poison, i32 4, i32 poison, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 poison>
; CHECK-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V128 = shufflevector <4 x float> %a128, <4 x float> %b128, <4 x i32> <i32 0, i32 undef, i32 2, i32 undef>
- %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 undef, i32 undef, i32 undef, i32 undef>
- %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 0, i32 undef, i32 2, i32 undef, i32 4, i32 undef, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 undef>
+ %V128 = shufflevector <4 x float> %a128, <4 x float> %b128, <4 x i32> <i32 0, i32 poison, i32 2, i32 undef>
+ %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 poison, i32 poison, i32 poison, i32 undef>
+ %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 0, i32 poison, i32 2, i32 poison, i32 4, i32 poison, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 undef>
ret void
}
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-splat.ll b/llvm/test/Analysis/CostModel/X86/shuffle-splat.ll
index c98efcd7fdbb7..b3cc220c0f3c3 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-splat.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-splat.ll
@@ -18,410 +18,410 @@
define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double> %src512) {
; SSE-LABEL: 'test_vXf64'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXf64'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf64'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf64'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
- %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
- %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V128 = shufflevector <2 x double> %src128, <2 x double> poison, <2 x i32> <i32 1, i32 1>
+ %V256 = shufflevector <4 x double> %src256, <4 x double> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+ %V512 = shufflevector <8 x double> %src512, <8 x double> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
ret void
}
define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512) {
; SSE-LABEL: 'test_vXi64'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi64'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi64'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi64'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
- %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
- %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V128 = shufflevector <2 x i64> %src128, <2 x i64> poison, <2 x i32> <i32 1, i32 1>
+ %V256 = shufflevector <4 x i64> %src256, <4 x i64> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+ %V512 = shufflevector <8 x i64> %src512, <8 x i64> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
ret void
}
define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %src256, <16 x float> %src512) {
; SSE-LABEL: 'test_vXf32'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXf32'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf32'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf32'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
- %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
- %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V64 = shufflevector <2 x float> %src64, <2 x float> poison, <2 x i32> <i32 1, i32 1>
+ %V128 = shufflevector <4 x float> %src128, <4 x float> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+ %V256 = shufflevector <8 x float> %src256, <8 x float> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V512 = shufflevector <16 x float> %src512, <16 x float> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
ret void
}
define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512) {
; SSE-LABEL: 'test_vXi32'
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi32'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi32'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi32'
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
- %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
- %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V64 = shufflevector <2 x i32> %src64, <2 x i32> poison, <2 x i32> <i32 1, i32 1>
+ %V128 = shufflevector <4 x i32> %src128, <4 x i32> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+ %V256 = shufflevector <8 x i32> %src256, <8 x i32> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V512 = shufflevector <16 x i32> %src512, <16 x i32> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
ret void
}
define void @test_vXf16(<2 x half> %src32, <4 x half> %src64, <8 x half> %src128, <16 x half> %src256, <32 x half> %src512) {
; SSE2-LABEL: 'test_vXf16'
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> <i32 1, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXf16'
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXf16'
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXf16'
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXf16'
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
- %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
- %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V32 = shufflevector <2 x half> %src32, <2 x half> poison, <2 x i32> <i32 1, i32 1>
+ %V64 = shufflevector <4 x half> %src64, <4 x half> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+ %V128 = shufflevector <8 x half> %src128, <8 x half> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V256 = shufflevector <16 x half> %src256, <16 x half> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V512 = shufflevector <32 x half> %src512, <32 x half> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
ret void
}
define void @test_vXbf16(<2 x bfloat> %src32, <4 x bfloat> %src64, <8 x bfloat> %src128, <16 x bfloat> %src256, <32 x bfloat> %src512) {
; SSE-LABEL: 'test_vXbf16'
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> <i32 1, i32 1>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> poison, <2 x i32> <i32 1, i32 1>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXbf16'
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXbf16'
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXbf16'
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 0 for: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> <i32 1, i32 1>
- %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
- %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> poison, <2 x i32> <i32 1, i32 1>
+ %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+ %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
ret void
}
define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src512) {
; SSE2-LABEL: 'test_vXi16'
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi16'
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi16'
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi16'
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi16'
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi16'
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
- %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
- %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V32 = shufflevector <2 x i16> %src32, <2 x i16> poison, <2 x i32> <i32 1, i32 1>
+ %V64 = shufflevector <4 x i16> %src64, <4 x i16> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+ %V128 = shufflevector <8 x i16> %src128, <8 x i16> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V256 = shufflevector <16 x i16> %src256, <16 x i16> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V512 = shufflevector <32 x i16> %src512, <32 x i16> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
ret void
}
define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x i8> %src128, <32 x i8> %src256, <64 x i8> %src512) {
; SSE2-LABEL: 'test_vXi8'
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi8'
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi8'
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi8'
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi8'
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi8'
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
- %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
- %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V16 = shufflevector <2 x i8> %src16, <2 x i8> poison, <2 x i32> <i32 1, i32 1>
+ %V32 = shufflevector <4 x i8> %src32, <4 x i8> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+ %V64 = shufflevector <8 x i8> %src64, <8 x i8> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V128 = shufflevector <16 x i8> %src128, <16 x i8> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V256 = shufflevector <32 x i8> %src256, <32 x i8> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V512 = shufflevector <64 x i8> %src512, <64 x i8> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
ret void
}
define void @test_vXi1(<2 x i1> %src2, <4 x i1> %src4, <8 x i1> %src8, <16 x i1> %src16, <32 x i1> %src32, <64 x i1> %src64) {
; SSE2-LABEL: 'test_vXi1'
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
-; SSE2-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> <i32 1, i32 1>
+; SSE2-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT: Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSSE3-LABEL: 'test_vXi1'
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
-; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSSE3-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; SSE42-LABEL: 'test_vXi1'
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
-; SSE42-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; SSE42-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX1-LABEL: 'test_vXi1'
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT: Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX1-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX2-LABEL: 'test_vXi1'
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT: Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT: Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX2-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
; AVX512-LABEL: 'test_vXi1'
-; AVX512-NEXT: Cost Model: Found costs of 3 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT: Cost Model: Found costs of 5 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 9 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 17 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 33 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT: Cost Model: Found costs of 65 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 3 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT: Cost Model: Found costs of 5 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 9 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 17 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 33 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT: Cost Model: Found costs of 65 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
; AVX512-NEXT: Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
;
- %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
- %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
- %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
- %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V2 = shufflevector <2 x i1> %src2, <2 x i1> poison, <2 x i32> <i32 1, i32 1>
+ %V4 = shufflevector <4 x i1> %src4, <4 x i1> poison, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+ %V8 = shufflevector <8 x i1> %src8, <8 x i1> poison, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V16 = shufflevector <16 x i1> %src16, <16 x i1> poison, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V32 = shufflevector <32 x i1> %src32, <32 x i1> poison, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+ %V64 = shufflevector <64 x i1> %src64, <64 x i1> poison, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
ret void
}
More information about the llvm-commits
mailing list