[llvm] deaedab - [NFC][AMDGPU][CostModel] Add tests for AMDGPU cost model.
Daniil Fukalov via llvm-commits
llvm-commits at lists.llvm.org
Wed Dec 22 11:33:00 PST 2021
Author: Daniil Fukalov
Date: 2021-12-22T22:32:09+03:00
New Revision: deaedab14a2f91ff0ed26197c22be6b8dd40562f
URL: https://github.com/llvm/llvm-project/commit/deaedab14a2f91ff0ed26197c22be6b8dd40562f
DIFF: https://github.com/llvm/llvm-project/commit/deaedab14a2f91ff0ed26197c22be6b8dd40562f.diff
LOG: [NFC][AMDGPU][CostModel] Add tests for AMDGPU cost model.
Added:
Modified:
llvm/test/Analysis/CostModel/AMDGPU/fdiv.ll
llvm/test/Analysis/CostModel/AMDGPU/fneg.ll
llvm/test/Analysis/CostModel/AMDGPU/insertelement.ll
llvm/test/Analysis/CostModel/AMDGPU/mul.ll
Removed:
################################################################################
diff --git a/llvm/test/Analysis/CostModel/AMDGPU/fdiv.ll b/llvm/test/Analysis/CostModel/AMDGPU/fdiv.ll
index 310598da87d7..fe5b57be5d3b 100644
--- a/llvm/test/Analysis/CostModel/AMDGPU/fdiv.ll
+++ b/llvm/test/Analysis/CostModel/AMDGPU/fdiv.ll
@@ -575,5 +575,107 @@ define amdgpu_kernel void @rcp_ftzdaz() #1 {
ret void
}
+define i32 @frem(i32 %arg) {
+; CIFASTF64-LABEL: 'frem'
+; CIFASTF64-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = frem float undef, undef
+; CIFASTF64-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4F32 = frem <4 x float> undef, undef
+; CIFASTF64-NEXT: Cost Model: Found an estimated cost of 112 for instruction: %V8F32 = frem <8 x float> undef, undef
+; CIFASTF64-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V16F32 = frem <16 x float> undef, undef
+; CIFASTF64-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %F64 = frem double undef, undef
+; CIFASTF64-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V2F64 = frem <2 x double> undef, undef
+; CIFASTF64-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V4F64 = frem <4 x double> undef, undef
+; CIFASTF64-NEXT: Cost Model: Found an estimated cost of 576 for instruction: %V8F64 = frem <8 x double> undef, undef
+; CIFASTF64-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; CISLOWF64-LABEL: 'frem'
+; CISLOWF64-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = frem float undef, undef
+; CISLOWF64-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4F32 = frem <4 x float> undef, undef
+; CISLOWF64-NEXT: Cost Model: Found an estimated cost of 112 for instruction: %V8F32 = frem <8 x float> undef, undef
+; CISLOWF64-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V16F32 = frem <16 x float> undef, undef
+; CISLOWF64-NEXT: Cost Model: Found an estimated cost of 38 for instruction: %F64 = frem double undef, undef
+; CISLOWF64-NEXT: Cost Model: Found an estimated cost of 76 for instruction: %V2F64 = frem <2 x double> undef, undef
+; CISLOWF64-NEXT: Cost Model: Found an estimated cost of 152 for instruction: %V4F64 = frem <4 x double> undef, undef
+; CISLOWF64-NEXT: Cost Model: Found an estimated cost of 912 for instruction: %V8F64 = frem <8 x double> undef, undef
+; CISLOWF64-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; SIFASTF64-LABEL: 'frem'
+; SIFASTF64-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = frem float undef, undef
+; SIFASTF64-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4F32 = frem <4 x float> undef, undef
+; SIFASTF64-NEXT: Cost Model: Found an estimated cost of 112 for instruction: %V8F32 = frem <8 x float> undef, undef
+; SIFASTF64-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V16F32 = frem <16 x float> undef, undef
+; SIFASTF64-NEXT: Cost Model: Found an estimated cost of 27 for instruction: %F64 = frem double undef, undef
+; SIFASTF64-NEXT: Cost Model: Found an estimated cost of 54 for instruction: %V2F64 = frem <2 x double> undef, undef
+; SIFASTF64-NEXT: Cost Model: Found an estimated cost of 108 for instruction: %V4F64 = frem <4 x double> undef, undef
+; SIFASTF64-NEXT: Cost Model: Found an estimated cost of 648 for instruction: %V8F64 = frem <8 x double> undef, undef
+; SIFASTF64-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; SISLOWF64-LABEL: 'frem'
+; SISLOWF64-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = frem float undef, undef
+; SISLOWF64-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4F32 = frem <4 x float> undef, undef
+; SISLOWF64-NEXT: Cost Model: Found an estimated cost of 112 for instruction: %V8F32 = frem <8 x float> undef, undef
+; SISLOWF64-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V16F32 = frem <16 x float> undef, undef
+; SISLOWF64-NEXT: Cost Model: Found an estimated cost of 41 for instruction: %F64 = frem double undef, undef
+; SISLOWF64-NEXT: Cost Model: Found an estimated cost of 82 for instruction: %V2F64 = frem <2 x double> undef, undef
+; SISLOWF64-NEXT: Cost Model: Found an estimated cost of 164 for instruction: %V4F64 = frem <4 x double> undef, undef
+; SISLOWF64-NEXT: Cost Model: Found an estimated cost of 984 for instruction: %V8F64 = frem <8 x double> undef, undef
+; SISLOWF64-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; FP16-LABEL: 'frem'
+; FP16-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %F32 = frem float undef, undef
+; FP16-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4F32 = frem <4 x float> undef, undef
+; FP16-NEXT: Cost Model: Found an estimated cost of 112 for instruction: %V8F32 = frem <8 x float> undef, undef
+; FP16-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V16F32 = frem <16 x float> undef, undef
+; FP16-NEXT: Cost Model: Found an estimated cost of 38 for instruction: %F64 = frem double undef, undef
+; FP16-NEXT: Cost Model: Found an estimated cost of 76 for instruction: %V2F64 = frem <2 x double> undef, undef
+; FP16-NEXT: Cost Model: Found an estimated cost of 152 for instruction: %V4F64 = frem <4 x double> undef, undef
+; FP16-NEXT: Cost Model: Found an estimated cost of 912 for instruction: %V8F64 = frem <8 x double> undef, undef
+; FP16-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; CI-SIZE-LABEL: 'frem'
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %F32 = frem float undef, undef
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4F32 = frem <4 x float> undef, undef
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V8F32 = frem <8 x float> undef, undef
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 576 for instruction: %V16F32 = frem <16 x float> undef, undef
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %F64 = frem double undef, undef
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %V2F64 = frem <2 x double> undef, undef
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 88 for instruction: %V4F64 = frem <4 x double> undef, undef
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 528 for instruction: %V8F64 = frem <8 x double> undef, undef
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+; SI-SIZE-LABEL: 'frem'
+; SI-SIZE-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %F32 = frem float undef, undef
+; SI-SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4F32 = frem <4 x float> undef, undef
+; SI-SIZE-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V8F32 = frem <8 x float> undef, undef
+; SI-SIZE-NEXT: Cost Model: Found an estimated cost of 576 for instruction: %V16F32 = frem <16 x float> undef, undef
+; SI-SIZE-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %F64 = frem double undef, undef
+; SI-SIZE-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %V2F64 = frem <2 x double> undef, undef
+; SI-SIZE-NEXT: Cost Model: Found an estimated cost of 100 for instruction: %V4F64 = frem <4 x double> undef, undef
+; SI-SIZE-NEXT: Cost Model: Found an estimated cost of 600 for instruction: %V8F64 = frem <8 x double> undef, undef
+; SI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+; FP16-SIZE-LABEL: 'frem'
+; FP16-SIZE-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %F32 = frem float undef, undef
+; FP16-SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4F32 = frem <4 x float> undef, undef
+; FP16-SIZE-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V8F32 = frem <8 x float> undef, undef
+; FP16-SIZE-NEXT: Cost Model: Found an estimated cost of 576 for instruction: %V16F32 = frem <16 x float> undef, undef
+; FP16-SIZE-NEXT: Cost Model: Found an estimated cost of 22 for instruction: %F64 = frem double undef, undef
+; FP16-SIZE-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %V2F64 = frem <2 x double> undef, undef
+; FP16-SIZE-NEXT: Cost Model: Found an estimated cost of 88 for instruction: %V4F64 = frem <4 x double> undef, undef
+; FP16-SIZE-NEXT: Cost Model: Found an estimated cost of 528 for instruction: %V8F64 = frem <8 x double> undef, undef
+; FP16-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %F32 = frem float undef, undef
+ %V4F32 = frem <4 x float> undef, undef
+ %V8F32 = frem <8 x float> undef, undef
+ %V16F32 = frem <16 x float> undef, undef
+
+ %F64 = frem double undef, undef
+ %V2F64 = frem <2 x double> undef, undef
+ %V4F64 = frem <4 x double> undef, undef
+ %V8F64 = frem <8 x double> undef, undef
+
+ ret i32 undef
+}
+
attributes #0 = { nounwind "denormal-fp-math-f32"="ieee,ieee" }
attributes #1 = { nounwind "denormal-fp-math-f32"="preserve-sign,preserve-sign" }
diff --git a/llvm/test/Analysis/CostModel/AMDGPU/fneg.ll b/llvm/test/Analysis/CostModel/AMDGPU/fneg.ll
index bf33cdc63553..4125d60f7ecb 100644
--- a/llvm/test/Analysis/CostModel/AMDGPU/fneg.ll
+++ b/llvm/test/Analysis/CostModel/AMDGPU/fneg.ll
@@ -89,3 +89,39 @@ define amdgpu_kernel void @fneg_f16() {
%v17f16 = fneg <17 x half> undef
ret void
}
+
+define i32 @fneg_idiom(i32 %arg) {
+; CHECK-LABEL: 'fneg_idiom'
+; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float -0.000000e+00, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F32 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F32 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V16F32 = fsub <16 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %F64 = fsub double -0.000000e+00, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V2F64 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V4F64 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V8F64 = fsub <8 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
+; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; SIZE-LABEL: 'fneg_idiom'
+; SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float -0.000000e+00, undef
+; SIZE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F32 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
+; SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F32 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
+; SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V16F32 = fsub <16 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, undef
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = fsub double -0.000000e+00, undef
+; SIZE-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, undef
+; SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
+; SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V8F64 = fsub <8 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, undef
+; SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %F32 = fsub float -0.0, undef
+ %V4F32 = fsub <4 x float> <float -0.0, float -0.0, float -0.0, float -0.0>, undef
+ %V8F32 = fsub <8 x float> <float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0>, undef
+ %V16F32 = fsub <16 x float> <float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0, float -0.0>, undef
+
+ %F64 = fsub double -0.0, undef
+ %V2F64 = fsub <2 x double> <double -0.0, double -0.0>, undef
+ %V4F64 = fsub <4 x double> <double -0.0, double -0.0, double -0.0, double -0.0>, undef
+ %V8F64 = fsub <8 x double> <double -0.0, double -0.0, double -0.0, double -0.0, double -0.0, double -0.0, double -0.0, double -0.0>, undef
+
+ ret i32 undef
+}
diff --git a/llvm/test/Analysis/CostModel/AMDGPU/insertelement.ll b/llvm/test/Analysis/CostModel/AMDGPU/insertelement.ll
index 1bdbb5b18920..e587240d8dbc 100644
--- a/llvm/test/Analysis/CostModel/AMDGPU/insertelement.ll
+++ b/llvm/test/Analysis/CostModel/AMDGPU/insertelement.ll
@@ -220,3 +220,562 @@ define amdgpu_kernel void @insertelement_i64(i32 %arg) {
%v5i64_a = insertelement <5 x i64> undef, i64 42, i32 %arg
ret void
}
+
+define i32 @insert_double_poison(i32 %arg) {
+; ALL-LABEL: 'insert_double_poison'
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2f64_a = insertelement <2 x double> poison, double undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2f64_0 = insertelement <2 x double> poison, double undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2f64_1 = insertelement <2 x double> poison, double undef, i32 1
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4f64_a = insertelement <4 x double> poison, double undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4f64_0 = insertelement <4 x double> poison, double undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4f64_3 = insertelement <4 x double> poison, double undef, i32 3
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8f64_a = insertelement <8 x double> poison, double undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f64_0 = insertelement <8 x double> poison, double undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f64_3 = insertelement <8 x double> poison, double undef, i32 3
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f64_4 = insertelement <8 x double> poison, double undef, i32 4
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f64_7 = insertelement <8 x double> poison, double undef, i32 7
+; ALL-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; ALL-SIZE-LABEL: 'insert_double_poison'
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2f64_a = insertelement <2 x double> poison, double undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2f64_0 = insertelement <2 x double> poison, double undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2f64_1 = insertelement <2 x double> poison, double undef, i32 1
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4f64_a = insertelement <4 x double> poison, double undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4f64_0 = insertelement <4 x double> poison, double undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4f64_3 = insertelement <4 x double> poison, double undef, i32 3
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8f64_a = insertelement <8 x double> poison, double undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f64_0 = insertelement <8 x double> poison, double undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f64_3 = insertelement <8 x double> poison, double undef, i32 3
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f64_4 = insertelement <8 x double> poison, double undef, i32 4
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f64_7 = insertelement <8 x double> poison, double undef, i32 7
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %v2f64_a = insertelement <2 x double> poison, double undef, i32 %arg
+ %v2f64_0 = insertelement <2 x double> poison, double undef, i32 0
+ %v2f64_1 = insertelement <2 x double> poison, double undef, i32 1
+
+ %v4f64_a = insertelement <4 x double> poison, double undef, i32 %arg
+ %v4f64_0 = insertelement <4 x double> poison, double undef, i32 0
+ %v4f64_3 = insertelement <4 x double> poison, double undef, i32 3
+
+ %v8f64_a = insertelement <8 x double> poison, double undef, i32 %arg
+ %v8f64_0 = insertelement <8 x double> poison, double undef, i32 0
+ %v8f64_3 = insertelement <8 x double> poison, double undef, i32 3
+ %v8f64_4 = insertelement <8 x double> poison, double undef, i32 4
+ %v8f64_7 = insertelement <8 x double> poison, double undef, i32 7
+
+ ret i32 undef
+}
+
+define i32 @insert_float_poison(i32 %arg) {
+; ALL-LABEL: 'insert_float_poison'
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2f32_a = insertelement <2 x float> poison, float undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2f32_0 = insertelement <2 x float> poison, float undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2f32_1 = insertelement <2 x float> poison, float undef, i32 1
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4f32_a = insertelement <4 x float> poison, float undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4f32_0 = insertelement <4 x float> poison, float undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4f32_3 = insertelement <4 x float> poison, float undef, i32 3
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8f32_a = insertelement <8 x float> poison, float undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f32_0 = insertelement <8 x float> poison, float undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f32_3 = insertelement <8 x float> poison, float undef, i32 3
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f32_4 = insertelement <8 x float> poison, float undef, i32 4
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f32_7 = insertelement <8 x float> poison, float undef, i32 7
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v16f32_a = insertelement <16 x float> poison, float undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16f32_0 = insertelement <16 x float> poison, float undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16f32_3 = insertelement <16 x float> poison, float undef, i32 3
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16f32_8 = insertelement <16 x float> poison, float undef, i32 8
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16f32_15 = insertelement <16 x float> poison, float undef, i32 15
+; ALL-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; ALL-SIZE-LABEL: 'insert_float_poison'
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2f32_a = insertelement <2 x float> poison, float undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2f32_0 = insertelement <2 x float> poison, float undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2f32_1 = insertelement <2 x float> poison, float undef, i32 1
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4f32_a = insertelement <4 x float> poison, float undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4f32_0 = insertelement <4 x float> poison, float undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4f32_3 = insertelement <4 x float> poison, float undef, i32 3
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8f32_a = insertelement <8 x float> poison, float undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f32_0 = insertelement <8 x float> poison, float undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f32_3 = insertelement <8 x float> poison, float undef, i32 3
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f32_4 = insertelement <8 x float> poison, float undef, i32 4
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8f32_7 = insertelement <8 x float> poison, float undef, i32 7
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v16f32_a = insertelement <16 x float> poison, float undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16f32_0 = insertelement <16 x float> poison, float undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16f32_3 = insertelement <16 x float> poison, float undef, i32 3
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16f32_8 = insertelement <16 x float> poison, float undef, i32 8
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16f32_15 = insertelement <16 x float> poison, float undef, i32 15
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %v2f32_a = insertelement <2 x float> poison, float undef, i32 %arg
+ %v2f32_0 = insertelement <2 x float> poison, float undef, i32 0
+ %v2f32_1 = insertelement <2 x float> poison, float undef, i32 1
+
+ %v4f32_a = insertelement <4 x float> poison, float undef, i32 %arg
+ %v4f32_0 = insertelement <4 x float> poison, float undef, i32 0
+ %v4f32_3 = insertelement <4 x float> poison, float undef, i32 3
+
+ %v8f32_a = insertelement <8 x float> poison, float undef, i32 %arg
+ %v8f32_0 = insertelement <8 x float> poison, float undef, i32 0
+ %v8f32_3 = insertelement <8 x float> poison, float undef, i32 3
+ %v8f32_4 = insertelement <8 x float> poison, float undef, i32 4
+ %v8f32_7 = insertelement <8 x float> poison, float undef, i32 7
+
+ %v16f32_a = insertelement <16 x float> poison, float undef, i32 %arg
+ %v16f32_0 = insertelement <16 x float> poison, float undef, i32 0
+ %v16f32_3 = insertelement <16 x float> poison, float undef, i32 3
+ %v16f32_8 = insertelement <16 x float> poison, float undef, i32 8
+ %v16f32_15 = insertelement <16 x float> poison, float undef, i32 15
+
+ ret i32 undef
+}
+
+define i32 @insert_i64_poison(i32 %arg) {
+; ALL-LABEL: 'insert_i64_poison'
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2i64_a = insertelement <2 x i64> poison, i64 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2i64_0 = insertelement <2 x i64> poison, i64 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2i64_1 = insertelement <2 x i64> poison, i64 undef, i32 1
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i64_a = insertelement <4 x i64> poison, i64 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4i64_0 = insertelement <4 x i64> poison, i64 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4i64_3 = insertelement <4 x i64> poison, i64 undef, i32 3
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8i64_a = insertelement <8 x i64> poison, i64 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i64_0 = insertelement <8 x i64> poison, i64 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i64_3 = insertelement <8 x i64> poison, i64 undef, i32 3
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i64_4 = insertelement <8 x i64> poison, i64 undef, i32 4
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i64_7 = insertelement <8 x i64> poison, i64 undef, i32 7
+; ALL-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; ALL-SIZE-LABEL: 'insert_i64_poison'
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2i64_a = insertelement <2 x i64> poison, i64 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2i64_0 = insertelement <2 x i64> poison, i64 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2i64_1 = insertelement <2 x i64> poison, i64 undef, i32 1
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i64_a = insertelement <4 x i64> poison, i64 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4i64_0 = insertelement <4 x i64> poison, i64 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4i64_3 = insertelement <4 x i64> poison, i64 undef, i32 3
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8i64_a = insertelement <8 x i64> poison, i64 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i64_0 = insertelement <8 x i64> poison, i64 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i64_3 = insertelement <8 x i64> poison, i64 undef, i32 3
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i64_4 = insertelement <8 x i64> poison, i64 undef, i32 4
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i64_7 = insertelement <8 x i64> poison, i64 undef, i32 7
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %v2i64_a = insertelement <2 x i64> poison, i64 undef, i32 %arg
+ %v2i64_0 = insertelement <2 x i64> poison, i64 undef, i32 0
+ %v2i64_1 = insertelement <2 x i64> poison, i64 undef, i32 1
+
+ %v4i64_a = insertelement <4 x i64> poison, i64 undef, i32 %arg
+ %v4i64_0 = insertelement <4 x i64> poison, i64 undef, i32 0
+ %v4i64_3 = insertelement <4 x i64> poison, i64 undef, i32 3
+
+ %v8i64_a = insertelement <8 x i64> poison, i64 undef, i32 %arg
+ %v8i64_0 = insertelement <8 x i64> poison, i64 undef, i32 0
+ %v8i64_3 = insertelement <8 x i64> poison, i64 undef, i32 3
+ %v8i64_4 = insertelement <8 x i64> poison, i64 undef, i32 4
+ %v8i64_7 = insertelement <8 x i64> poison, i64 undef, i32 7
+
+ ret i32 undef
+}
+
+define i32 @insert_i32_poison(i32 %arg) {
+; ALL-LABEL: 'insert_i32_poison'
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2i32_a = insertelement <2 x i32> poison, i32 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2i32_0 = insertelement <2 x i32> poison, i32 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2i32_1 = insertelement <2 x i32> poison, i32 undef, i32 1
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i32_a = insertelement <4 x i32> poison, i32 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4i32_0 = insertelement <4 x i32> poison, i32 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4i32_3 = insertelement <4 x i32> poison, i32 undef, i32 3
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8i32_a = insertelement <8 x i32> poison, i32 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i32_0 = insertelement <8 x i32> poison, i32 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i32_3 = insertelement <8 x i32> poison, i32 undef, i32 3
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i32_4 = insertelement <8 x i32> poison, i32 undef, i32 4
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i32_7 = insertelement <8 x i32> poison, i32 undef, i32 7
+; ALL-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v16i32_a = insertelement <16 x i32> poison, i32 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16i32_0 = insertelement <16 x i32> poison, i32 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16i32_3 = insertelement <16 x i32> poison, i32 undef, i32 3
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16i32_8 = insertelement <16 x i32> poison, i32 undef, i32 8
+; ALL-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16i32_15 = insertelement <16 x i32> poison, i32 undef, i32 15
+; ALL-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; ALL-SIZE-LABEL: 'insert_i32_poison'
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2i32_a = insertelement <2 x i32> poison, i32 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2i32_0 = insertelement <2 x i32> poison, i32 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2i32_1 = insertelement <2 x i32> poison, i32 undef, i32 1
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i32_a = insertelement <4 x i32> poison, i32 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4i32_0 = insertelement <4 x i32> poison, i32 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4i32_3 = insertelement <4 x i32> poison, i32 undef, i32 3
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v8i32_a = insertelement <8 x i32> poison, i32 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i32_0 = insertelement <8 x i32> poison, i32 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i32_3 = insertelement <8 x i32> poison, i32 undef, i32 3
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i32_4 = insertelement <8 x i32> poison, i32 undef, i32 4
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i32_7 = insertelement <8 x i32> poison, i32 undef, i32 7
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v16i32_a = insertelement <16 x i32> poison, i32 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16i32_0 = insertelement <16 x i32> poison, i32 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16i32_3 = insertelement <16 x i32> poison, i32 undef, i32 3
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16i32_8 = insertelement <16 x i32> poison, i32 undef, i32 8
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16i32_15 = insertelement <16 x i32> poison, i32 undef, i32 15
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %v2i32_a = insertelement <2 x i32> poison, i32 undef, i32 %arg
+ %v2i32_0 = insertelement <2 x i32> poison, i32 undef, i32 0
+ %v2i32_1 = insertelement <2 x i32> poison, i32 undef, i32 1
+
+ %v4i32_a = insertelement <4 x i32> poison, i32 undef, i32 %arg
+ %v4i32_0 = insertelement <4 x i32> poison, i32 undef, i32 0
+ %v4i32_3 = insertelement <4 x i32> poison, i32 undef, i32 3
+
+ %v8i32_a = insertelement <8 x i32> poison, i32 undef, i32 %arg
+ %v8i32_0 = insertelement <8 x i32> poison, i32 undef, i32 0
+ %v8i32_3 = insertelement <8 x i32> poison, i32 undef, i32 3
+ %v8i32_4 = insertelement <8 x i32> poison, i32 undef, i32 4
+ %v8i32_7 = insertelement <8 x i32> poison, i32 undef, i32 7
+
+ %v16i32_a = insertelement <16 x i32> poison, i32 undef, i32 %arg
+ %v16i32_0 = insertelement <16 x i32> poison, i32 undef, i32 0
+ %v16i32_3 = insertelement <16 x i32> poison, i32 undef, i32 3
+ %v16i32_8 = insertelement <16 x i32> poison, i32 undef, i32 8
+ %v16i32_15 = insertelement <16 x i32> poison, i32 undef, i32 15
+
+ ret i32 undef
+}
+
+define i32 @insert_i16_poison(i32 %arg) {
+; CI-LABEL: 'insert_i16_poison'
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_a = insertelement <2 x i16> poison, i16 undef, i32 %arg
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_0 = insertelement <2 x i16> poison, i16 undef, i32 0
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_1 = insertelement <2 x i16> poison, i16 undef, i32 1
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i16_a = insertelement <4 x i16> poison, i16 undef, i32 %arg
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i16_0 = insertelement <4 x i16> poison, i16 undef, i32 0
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i16_3 = insertelement <4 x i16> poison, i16 undef, i32 3
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i16_a = insertelement <8 x i16> poison, i16 undef, i32 %arg
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i16_0 = insertelement <8 x i16> poison, i16 undef, i32 0
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i16_7 = insertelement <8 x i16> poison, i16 undef, i32 7
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_a = insertelement <16 x i16> poison, i16 undef, i32 %arg
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_0 = insertelement <16 x i16> poison, i16 undef, i32 0
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_7 = insertelement <16 x i16> poison, i16 undef, i32 7
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_8 = insertelement <16 x i16> poison, i16 undef, i32 8
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_15 = insertelement <16 x i16> poison, i16 undef, i32 15
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_a = insertelement <32 x i16> poison, i16 undef, i32 %arg
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_0 = insertelement <32 x i16> poison, i16 undef, i32 0
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_7 = insertelement <32 x i16> poison, i16 undef, i32 7
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_8 = insertelement <32 x i16> poison, i16 undef, i32 8
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_15 = insertelement <32 x i16> poison, i16 undef, i32 15
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_16 = insertelement <32 x i16> poison, i16 undef, i32 16
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_24 = insertelement <32 x i16> poison, i16 undef, i32 24
+; CI-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_31 = insertelement <32 x i16> poison, i16 undef, i32 31
+; CI-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; GFX89-LABEL: 'insert_i16_poison'
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_a = insertelement <2 x i16> poison, i16 undef, i32 %arg
+; GFX89-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2i16_0 = insertelement <2 x i16> poison, i16 undef, i32 0
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_1 = insertelement <2 x i16> poison, i16 undef, i32 1
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i16_a = insertelement <4 x i16> poison, i16 undef, i32 %arg
+; GFX89-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4i16_0 = insertelement <4 x i16> poison, i16 undef, i32 0
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i16_3 = insertelement <4 x i16> poison, i16 undef, i32 3
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i16_a = insertelement <8 x i16> poison, i16 undef, i32 %arg
+; GFX89-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i16_0 = insertelement <8 x i16> poison, i16 undef, i32 0
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i16_7 = insertelement <8 x i16> poison, i16 undef, i32 7
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_a = insertelement <16 x i16> poison, i16 undef, i32 %arg
+; GFX89-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16i16_0 = insertelement <16 x i16> poison, i16 undef, i32 0
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_7 = insertelement <16 x i16> poison, i16 undef, i32 7
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_8 = insertelement <16 x i16> poison, i16 undef, i32 8
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_15 = insertelement <16 x i16> poison, i16 undef, i32 15
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_a = insertelement <32 x i16> poison, i16 undef, i32 %arg
+; GFX89-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v32i16_0 = insertelement <32 x i16> poison, i16 undef, i32 0
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_7 = insertelement <32 x i16> poison, i16 undef, i32 7
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_8 = insertelement <32 x i16> poison, i16 undef, i32 8
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_15 = insertelement <32 x i16> poison, i16 undef, i32 15
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_16 = insertelement <32 x i16> poison, i16 undef, i32 16
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_24 = insertelement <32 x i16> poison, i16 undef, i32 24
+; GFX89-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_31 = insertelement <32 x i16> poison, i16 undef, i32 31
+; GFX89-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; CI-SIZE-LABEL: 'insert_i16_poison'
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_a = insertelement <2 x i16> poison, i16 undef, i32 %arg
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_0 = insertelement <2 x i16> poison, i16 undef, i32 0
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_1 = insertelement <2 x i16> poison, i16 undef, i32 1
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i16_a = insertelement <4 x i16> poison, i16 undef, i32 %arg
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i16_0 = insertelement <4 x i16> poison, i16 undef, i32 0
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i16_3 = insertelement <4 x i16> poison, i16 undef, i32 3
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i16_a = insertelement <8 x i16> poison, i16 undef, i32 %arg
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i16_0 = insertelement <8 x i16> poison, i16 undef, i32 0
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i16_7 = insertelement <8 x i16> poison, i16 undef, i32 7
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_a = insertelement <16 x i16> poison, i16 undef, i32 %arg
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_0 = insertelement <16 x i16> poison, i16 undef, i32 0
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_7 = insertelement <16 x i16> poison, i16 undef, i32 7
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_8 = insertelement <16 x i16> poison, i16 undef, i32 8
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_15 = insertelement <16 x i16> poison, i16 undef, i32 15
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_a = insertelement <32 x i16> poison, i16 undef, i32 %arg
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_0 = insertelement <32 x i16> poison, i16 undef, i32 0
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_7 = insertelement <32 x i16> poison, i16 undef, i32 7
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_8 = insertelement <32 x i16> poison, i16 undef, i32 8
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_15 = insertelement <32 x i16> poison, i16 undef, i32 15
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_16 = insertelement <32 x i16> poison, i16 undef, i32 16
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_24 = insertelement <32 x i16> poison, i16 undef, i32 24
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_31 = insertelement <32 x i16> poison, i16 undef, i32 31
+; CI-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+; GFX89-SIZE-LABEL: 'insert_i16_poison'
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_a = insertelement <2 x i16> poison, i16 undef, i32 %arg
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v2i16_0 = insertelement <2 x i16> poison, i16 undef, i32 0
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i16_1 = insertelement <2 x i16> poison, i16 undef, i32 1
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i16_a = insertelement <4 x i16> poison, i16 undef, i32 %arg
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v4i16_0 = insertelement <4 x i16> poison, i16 undef, i32 0
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i16_3 = insertelement <4 x i16> poison, i16 undef, i32 3
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i16_a = insertelement <8 x i16> poison, i16 undef, i32 %arg
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v8i16_0 = insertelement <8 x i16> poison, i16 undef, i32 0
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i16_7 = insertelement <8 x i16> poison, i16 undef, i32 7
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_a = insertelement <16 x i16> poison, i16 undef, i32 %arg
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v16i16_0 = insertelement <16 x i16> poison, i16 undef, i32 0
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_7 = insertelement <16 x i16> poison, i16 undef, i32 7
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_8 = insertelement <16 x i16> poison, i16 undef, i32 8
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i16_15 = insertelement <16 x i16> poison, i16 undef, i32 15
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_a = insertelement <32 x i16> poison, i16 undef, i32 %arg
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %v32i16_0 = insertelement <32 x i16> poison, i16 undef, i32 0
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_7 = insertelement <32 x i16> poison, i16 undef, i32 7
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_8 = insertelement <32 x i16> poison, i16 undef, i32 8
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_15 = insertelement <32 x i16> poison, i16 undef, i32 15
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_16 = insertelement <32 x i16> poison, i16 undef, i32 16
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_24 = insertelement <32 x i16> poison, i16 undef, i32 24
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i16_31 = insertelement <32 x i16> poison, i16 undef, i32 31
+; GFX89-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %v2i16_a = insertelement <2 x i16> poison, i16 undef, i32 %arg
+ %v2i16_0 = insertelement <2 x i16> poison, i16 undef, i32 0
+ %v2i16_1 = insertelement <2 x i16> poison, i16 undef, i32 1
+
+ %v4i16_a = insertelement <4 x i16> poison, i16 undef, i32 %arg
+ %v4i16_0 = insertelement <4 x i16> poison, i16 undef, i32 0
+ %v4i16_3 = insertelement <4 x i16> poison, i16 undef, i32 3
+
+ %v8i16_a = insertelement <8 x i16> poison, i16 undef, i32 %arg
+ %v8i16_0 = insertelement <8 x i16> poison, i16 undef, i32 0
+ %v8i16_7 = insertelement <8 x i16> poison, i16 undef, i32 7
+
+ %v16i16_a = insertelement <16 x i16> poison, i16 undef, i32 %arg
+ %v16i16_0 = insertelement <16 x i16> poison, i16 undef, i32 0
+ %v16i16_7 = insertelement <16 x i16> poison, i16 undef, i32 7
+ %v16i16_8 = insertelement <16 x i16> poison, i16 undef, i32 8
+ %v16i16_15 = insertelement <16 x i16> poison, i16 undef, i32 15
+
+ %v32i16_a = insertelement <32 x i16> poison, i16 undef, i32 %arg
+ %v32i16_0 = insertelement <32 x i16> poison, i16 undef, i32 0
+ %v32i16_7 = insertelement <32 x i16> poison, i16 undef, i32 7
+ %v32i16_8 = insertelement <32 x i16> poison, i16 undef, i32 8
+ %v32i16_15 = insertelement <32 x i16> poison, i16 undef, i32 15
+ %v32i16_16 = insertelement <32 x i16> poison, i16 undef, i32 16
+ %v32i16_24 = insertelement <32 x i16> poison, i16 undef, i32 24
+ %v32i16_31 = insertelement <32 x i16> poison, i16 undef, i32 31
+
+ ret i32 undef
+}
+
+define i32 @insert_i8_poison(i32 %arg) {
+; ALL-LABEL: 'insert_i8_poison'
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i8_a = insertelement <2 x i8> poison, i8 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i8_0 = insertelement <2 x i8> poison, i8 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i8_3 = insertelement <2 x i8> poison, i8 undef, i32 1
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i8_a = insertelement <4 x i8> poison, i8 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i8_0 = insertelement <4 x i8> poison, i8 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i8_3 = insertelement <4 x i8> poison, i8 undef, i32 3
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i8_a = insertelement <8 x i8> poison, i8 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i8_0 = insertelement <8 x i8> poison, i8 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i8_7 = insertelement <8 x i8> poison, i8 undef, i32 7
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i8_a = insertelement <16 x i8> poison, i8 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i8_0 = insertelement <16 x i8> poison, i8 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i8_8 = insertelement <16 x i8> poison, i8 undef, i32 8
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i8_15 = insertelement <16 x i8> poison, i8 undef, i32 15
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_a = insertelement <32 x i8> poison, i8 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_0 = insertelement <32 x i8> poison, i8 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_7 = insertelement <32 x i8> poison, i8 undef, i32 7
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_8 = insertelement <32 x i8> poison, i8 undef, i32 8
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_15 = insertelement <32 x i8> poison, i8 undef, i32 15
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_24 = insertelement <32 x i8> poison, i8 undef, i32 24
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_31 = insertelement <32 x i8> poison, i8 undef, i32 31
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_a = insertelement <64 x i8> poison, i8 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_0 = insertelement <64 x i8> poison, i8 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_7 = insertelement <64 x i8> poison, i8 undef, i32 7
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_8 = insertelement <64 x i8> poison, i8 undef, i32 8
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_15 = insertelement <64 x i8> poison, i8 undef, i32 15
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_24 = insertelement <64 x i8> poison, i8 undef, i32 24
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_31 = insertelement <64 x i8> poison, i8 undef, i32 31
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_32 = insertelement <64 x i8> poison, i8 undef, i32 32
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_48 = insertelement <64 x i8> poison, i8 undef, i32 48
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_63 = insertelement <64 x i8> poison, i8 undef, i32 63
+; ALL-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; ALL-SIZE-LABEL: 'insert_i8_poison'
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i8_a = insertelement <2 x i8> poison, i8 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i8_0 = insertelement <2 x i8> poison, i8 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i8_3 = insertelement <2 x i8> poison, i8 undef, i32 1
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i8_a = insertelement <4 x i8> poison, i8 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i8_0 = insertelement <4 x i8> poison, i8 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i8_3 = insertelement <4 x i8> poison, i8 undef, i32 3
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i8_a = insertelement <8 x i8> poison, i8 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i8_0 = insertelement <8 x i8> poison, i8 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i8_7 = insertelement <8 x i8> poison, i8 undef, i32 7
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i8_a = insertelement <16 x i8> poison, i8 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i8_0 = insertelement <16 x i8> poison, i8 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i8_8 = insertelement <16 x i8> poison, i8 undef, i32 8
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i8_15 = insertelement <16 x i8> poison, i8 undef, i32 15
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_a = insertelement <32 x i8> poison, i8 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_0 = insertelement <32 x i8> poison, i8 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_7 = insertelement <32 x i8> poison, i8 undef, i32 7
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_8 = insertelement <32 x i8> poison, i8 undef, i32 8
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_15 = insertelement <32 x i8> poison, i8 undef, i32 15
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_24 = insertelement <32 x i8> poison, i8 undef, i32 24
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i8_31 = insertelement <32 x i8> poison, i8 undef, i32 31
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_a = insertelement <64 x i8> poison, i8 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_0 = insertelement <64 x i8> poison, i8 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_7 = insertelement <64 x i8> poison, i8 undef, i32 7
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_8 = insertelement <64 x i8> poison, i8 undef, i32 8
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_15 = insertelement <64 x i8> poison, i8 undef, i32 15
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_24 = insertelement <64 x i8> poison, i8 undef, i32 24
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_31 = insertelement <64 x i8> poison, i8 undef, i32 31
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_32 = insertelement <64 x i8> poison, i8 undef, i32 32
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_48 = insertelement <64 x i8> poison, i8 undef, i32 48
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i8_63 = insertelement <64 x i8> poison, i8 undef, i32 63
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %v2i8_a = insertelement <2 x i8> poison, i8 undef, i32 %arg
+ %v2i8_0 = insertelement <2 x i8> poison, i8 undef, i32 0
+ %v2i8_3 = insertelement <2 x i8> poison, i8 undef, i32 1
+
+ %v4i8_a = insertelement <4 x i8> poison, i8 undef, i32 %arg
+ %v4i8_0 = insertelement <4 x i8> poison, i8 undef, i32 0
+ %v4i8_3 = insertelement <4 x i8> poison, i8 undef, i32 3
+
+ %v8i8_a = insertelement <8 x i8> poison, i8 undef, i32 %arg
+ %v8i8_0 = insertelement <8 x i8> poison, i8 undef, i32 0
+ %v8i8_7 = insertelement <8 x i8> poison, i8 undef, i32 7
+
+ %v16i8_a = insertelement <16 x i8> poison, i8 undef, i32 %arg
+ %v16i8_0 = insertelement <16 x i8> poison, i8 undef, i32 0
+ %v16i8_8 = insertelement <16 x i8> poison, i8 undef, i32 8
+ %v16i8_15 = insertelement <16 x i8> poison, i8 undef, i32 15
+
+ %v32i8_a = insertelement <32 x i8> poison, i8 undef, i32 %arg
+ %v32i8_0 = insertelement <32 x i8> poison, i8 undef, i32 0
+ %v32i8_7 = insertelement <32 x i8> poison, i8 undef, i32 7
+ %v32i8_8 = insertelement <32 x i8> poison, i8 undef, i32 8
+ %v32i8_15 = insertelement <32 x i8> poison, i8 undef, i32 15
+ %v32i8_24 = insertelement <32 x i8> poison, i8 undef, i32 24
+ %v32i8_31 = insertelement <32 x i8> poison, i8 undef, i32 31
+
+ %v64i8_a = insertelement <64 x i8> poison, i8 undef, i32 %arg
+ %v64i8_0 = insertelement <64 x i8> poison, i8 undef, i32 0
+ %v64i8_7 = insertelement <64 x i8> poison, i8 undef, i32 7
+ %v64i8_8 = insertelement <64 x i8> poison, i8 undef, i32 8
+ %v64i8_15 = insertelement <64 x i8> poison, i8 undef, i32 15
+ %v64i8_24 = insertelement <64 x i8> poison, i8 undef, i32 24
+ %v64i8_31 = insertelement <64 x i8> poison, i8 undef, i32 31
+ %v64i8_32 = insertelement <64 x i8> poison, i8 undef, i32 32
+ %v64i8_48 = insertelement <64 x i8> poison, i8 undef, i32 48
+ %v64i8_63 = insertelement <64 x i8> poison, i8 undef, i32 63
+
+ ret i32 undef
+}
+
+define i32 @insert_i1_poison(i32 %arg) {
+; ALL-LABEL: 'insert_i1_poison'
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i1_a = insertelement <2 x i1> poison, i1 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i1_0 = insertelement <2 x i1> poison, i1 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i1_1 = insertelement <2 x i1> poison, i1 undef, i32 1
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i1_a = insertelement <4 x i1> poison, i1 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i1_0 = insertelement <4 x i1> poison, i1 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i1_2 = insertelement <4 x i1> poison, i1 undef, i32 2
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i1_a = insertelement <8 x i1> poison, i1 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i1_0 = insertelement <8 x i1> poison, i1 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i1_4 = insertelement <8 x i1> poison, i1 undef, i32 4
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i1_a = insertelement <16 x i1> poison, i1 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i1_0 = insertelement <16 x i1> poison, i1 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i1_8 = insertelement <16 x i1> poison, i1 undef, i32 8
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i1_15 = insertelement <16 x i1> poison, i1 undef, i32 15
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_a = insertelement <32 x i1> poison, i1 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_0 = insertelement <32 x i1> poison, i1 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_7 = insertelement <32 x i1> poison, i1 undef, i32 7
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_8 = insertelement <32 x i1> poison, i1 undef, i32 8
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_15 = insertelement <32 x i1> poison, i1 undef, i32 15
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_24 = insertelement <32 x i1> poison, i1 undef, i32 24
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_31 = insertelement <32 x i1> poison, i1 undef, i32 31
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_a = insertelement <64 x i1> poison, i1 undef, i32 %arg
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_0 = insertelement <64 x i1> poison, i1 undef, i32 0
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_7 = insertelement <64 x i1> poison, i1 undef, i32 7
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_8 = insertelement <64 x i1> poison, i1 undef, i32 8
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_15 = insertelement <64 x i1> poison, i1 undef, i32 15
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_24 = insertelement <64 x i1> poison, i1 undef, i32 24
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_31 = insertelement <64 x i1> poison, i1 undef, i32 31
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_32 = insertelement <64 x i1> poison, i1 undef, i32 32
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_48 = insertelement <64 x i1> poison, i1 undef, i32 48
+; ALL-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_63 = insertelement <64 x i1> poison, i1 undef, i32 63
+; ALL-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; ALL-SIZE-LABEL: 'insert_i1_poison'
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i1_a = insertelement <2 x i1> poison, i1 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i1_0 = insertelement <2 x i1> poison, i1 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i1_1 = insertelement <2 x i1> poison, i1 undef, i32 1
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i1_a = insertelement <4 x i1> poison, i1 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i1_0 = insertelement <4 x i1> poison, i1 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i1_2 = insertelement <4 x i1> poison, i1 undef, i32 2
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i1_a = insertelement <8 x i1> poison, i1 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i1_0 = insertelement <8 x i1> poison, i1 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v8i1_4 = insertelement <8 x i1> poison, i1 undef, i32 4
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i1_a = insertelement <16 x i1> poison, i1 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i1_0 = insertelement <16 x i1> poison, i1 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i1_8 = insertelement <16 x i1> poison, i1 undef, i32 8
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v16i1_15 = insertelement <16 x i1> poison, i1 undef, i32 15
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_a = insertelement <32 x i1> poison, i1 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_0 = insertelement <32 x i1> poison, i1 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_7 = insertelement <32 x i1> poison, i1 undef, i32 7
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_8 = insertelement <32 x i1> poison, i1 undef, i32 8
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_15 = insertelement <32 x i1> poison, i1 undef, i32 15
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_24 = insertelement <32 x i1> poison, i1 undef, i32 24
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v32i1_31 = insertelement <32 x i1> poison, i1 undef, i32 31
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_a = insertelement <64 x i1> poison, i1 undef, i32 %arg
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_0 = insertelement <64 x i1> poison, i1 undef, i32 0
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_7 = insertelement <64 x i1> poison, i1 undef, i32 7
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_8 = insertelement <64 x i1> poison, i1 undef, i32 8
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_15 = insertelement <64 x i1> poison, i1 undef, i32 15
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_24 = insertelement <64 x i1> poison, i1 undef, i32 24
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_31 = insertelement <64 x i1> poison, i1 undef, i32 31
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_32 = insertelement <64 x i1> poison, i1 undef, i32 32
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_48 = insertelement <64 x i1> poison, i1 undef, i32 48
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v64i1_63 = insertelement <64 x i1> poison, i1 undef, i32 63
+; ALL-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %v2i1_a = insertelement <2 x i1> poison, i1 undef, i32 %arg
+ %v2i1_0 = insertelement <2 x i1> poison, i1 undef, i32 0
+ %v2i1_1 = insertelement <2 x i1> poison, i1 undef, i32 1
+
+ %v4i1_a = insertelement <4 x i1> poison, i1 undef, i32 %arg
+ %v4i1_0 = insertelement <4 x i1> poison, i1 undef, i32 0
+ %v4i1_2 = insertelement <4 x i1> poison, i1 undef, i32 2
+
+ %v8i1_a = insertelement <8 x i1> poison, i1 undef, i32 %arg
+ %v8i1_0 = insertelement <8 x i1> poison, i1 undef, i32 0
+ %v8i1_4 = insertelement <8 x i1> poison, i1 undef, i32 4
+
+ %v16i1_a = insertelement <16 x i1> poison, i1 undef, i32 %arg
+ %v16i1_0 = insertelement <16 x i1> poison, i1 undef, i32 0
+ %v16i1_8 = insertelement <16 x i1> poison, i1 undef, i32 8
+ %v16i1_15 = insertelement <16 x i1> poison, i1 undef, i32 15
+
+ %v32i1_a = insertelement <32 x i1> poison, i1 undef, i32 %arg
+ %v32i1_0 = insertelement <32 x i1> poison, i1 undef, i32 0
+ %v32i1_7 = insertelement <32 x i1> poison, i1 undef, i32 7
+ %v32i1_8 = insertelement <32 x i1> poison, i1 undef, i32 8
+ %v32i1_15 = insertelement <32 x i1> poison, i1 undef, i32 15
+ %v32i1_24 = insertelement <32 x i1> poison, i1 undef, i32 24
+ %v32i1_31 = insertelement <32 x i1> poison, i1 undef, i32 31
+
+ %v64i1_a = insertelement <64 x i1> poison, i1 undef, i32 %arg
+ %v64i1_0 = insertelement <64 x i1> poison, i1 undef, i32 0
+ %v64i1_7 = insertelement <64 x i1> poison, i1 undef, i32 7
+ %v64i1_8 = insertelement <64 x i1> poison, i1 undef, i32 8
+ %v64i1_15 = insertelement <64 x i1> poison, i1 undef, i32 15
+ %v64i1_24 = insertelement <64 x i1> poison, i1 undef, i32 24
+ %v64i1_31 = insertelement <64 x i1> poison, i1 undef, i32 31
+ %v64i1_32 = insertelement <64 x i1> poison, i1 undef, i32 32
+ %v64i1_48 = insertelement <64 x i1> poison, i1 undef, i32 48
+ %v64i1_63 = insertelement <64 x i1> poison, i1 undef, i32 63
+
+ ret i32 undef
+}
diff --git a/llvm/test/Analysis/CostModel/AMDGPU/mul.ll b/llvm/test/Analysis/CostModel/AMDGPU/mul.ll
index 824b7e1ee93d..8fddcbefaf76 100644
--- a/llvm/test/Analysis/CostModel/AMDGPU/mul.ll
+++ b/llvm/test/Analysis/CostModel/AMDGPU/mul.ll
@@ -112,4 +112,404 @@ define amdgpu_kernel void @mul_i16() #0 {
ret void
}
+define i32 @mul_constpow2() {
+; SLOW16-LABEL: 'mul_constpow2'
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %I64 = mul i64 undef, 16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = mul <2 x i64> undef, <i64 8, i64 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = mul <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = mul <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I32 = mul i32 undef, 16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V4i32 = mul <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = mul <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %V16i32 = mul <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I16 = mul i16 undef, 16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i16 = mul <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i16 = mul <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %V32i16 = mul <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I8 = mul i8 undef, 16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i8 = mul <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V32i8 = mul <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 264 for instruction: %V64i8 = mul <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; FAST16-LABEL: 'mul_constpow2'
+; FAST16-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %I64 = mul i64 undef, 16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = mul <2 x i64> undef, <i64 8, i64 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = mul <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = mul <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I32 = mul i32 undef, 16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V4i32 = mul <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = mul <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %V16i32 = mul <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I16 = mul i16 undef, 16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i16 = mul <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i16 = mul <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V32i16 = mul <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I8 = mul i8 undef, 16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i8 = mul <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V32i8 = mul <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 264 for instruction: %V64i8 = mul <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; SLOW16-SIZE-LABEL: 'mul_constpow2'
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %I64 = mul i64 undef, 16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V2i64 = mul <2 x i64> undef, <i64 8, i64 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4i64 = mul <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 288 for instruction: %V8i64 = mul <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, 16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = mul <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, 16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i16 = mul <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i16 = mul <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %V32i16 = mul <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, 16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i8 = mul <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V32i8 = mul <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 132 for instruction: %V64i8 = mul <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+; FAST16-SIZE-LABEL: 'mul_constpow2'
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %I64 = mul i64 undef, 16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V2i64 = mul <2 x i64> undef, <i64 8, i64 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4i64 = mul <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 288 for instruction: %V8i64 = mul <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, 16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = mul <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, 16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = mul <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = mul <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V32i16 = mul <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, 16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i8 = mul <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V32i8 = mul <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 132 for instruction: %V64i8 = mul <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %I64 = mul i64 undef, 16
+ %V2i64 = mul <2 x i64> undef, <i64 8, i64 16>
+ %V4i64 = mul <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
+ %V8i64 = mul <8 x i64> undef, <i64 2, i64 4, i64 8, i64 16, i64 32, i64 64, i64 128, i64 256>
+
+ %I32 = mul i32 undef, 16
+ %V4i32 = mul <4 x i32> undef, <i32 2, i32 4, i32 8, i32 16>
+ %V8i32 = mul <8 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+ %V16i32 = mul <16 x i32> undef, <i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256, i32 2, i32 4, i32 8, i32 16, i32 32, i32 64, i32 128, i32 256>
+
+ %I16 = mul i16 undef, 16
+ %V8i16 = mul <8 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+ %V16i16 = mul <16 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+ %V32i16 = mul <32 x i16> undef, <i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256, i16 2, i16 4, i16 8, i16 16, i16 32, i16 64, i16 128, i16 256>
+
+ %I8 = mul i8 undef, 16
+ %V16i8 = mul <16 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+ %V32i8 = mul <32 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+ %V64i8 = mul <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
+
+ ret i32 undef
+}
+
+define i32 @mul_uniformconstpow2() {
+; SLOW16-LABEL: 'mul_uniformconstpow2'
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %I64 = mul i64 undef, 16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = mul <2 x i64> undef, <i64 16, i64 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = mul <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = mul <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I32 = mul i32 undef, 16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V4i32 = mul <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = mul <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %V16i32 = mul <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I16 = mul i16 undef, 16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i16 = mul <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i16 = mul <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %V32i16 = mul <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I8 = mul i8 undef, 16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i8 = mul <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V32i8 = mul <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 264 for instruction: %V64i8 = mul <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; FAST16-LABEL: 'mul_uniformconstpow2'
+; FAST16-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %I64 = mul i64 undef, 16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = mul <2 x i64> undef, <i64 16, i64 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = mul <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = mul <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I32 = mul i32 undef, 16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V4i32 = mul <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = mul <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %V16i32 = mul <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I16 = mul i16 undef, 16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i16 = mul <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i16 = mul <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V32i16 = mul <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I8 = mul i8 undef, 16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i8 = mul <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V32i8 = mul <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 264 for instruction: %V64i8 = mul <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; SLOW16-SIZE-LABEL: 'mul_uniformconstpow2'
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %I64 = mul i64 undef, 16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V2i64 = mul <2 x i64> undef, <i64 16, i64 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4i64 = mul <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 288 for instruction: %V8i64 = mul <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, 16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = mul <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, 16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i16 = mul <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i16 = mul <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %V32i16 = mul <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, 16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i8 = mul <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V32i8 = mul <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 132 for instruction: %V64i8 = mul <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+; FAST16-SIZE-LABEL: 'mul_uniformconstpow2'
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %I64 = mul i64 undef, 16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V2i64 = mul <2 x i64> undef, <i64 16, i64 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4i64 = mul <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 288 for instruction: %V8i64 = mul <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, 16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = mul <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, 16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = mul <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = mul <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V32i16 = mul <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, 16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i8 = mul <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V32i8 = mul <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 132 for instruction: %V64i8 = mul <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %I64 = mul i64 undef, 16
+ %V2i64 = mul <2 x i64> undef, <i64 16, i64 16>
+ %V4i64 = mul <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
+ %V8i64 = mul <8 x i64> undef, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+
+ %I32 = mul i32 undef, 16
+ %V4i32 = mul <4 x i32> undef, <i32 16, i32 16, i32 16, i32 16>
+ %V8i32 = mul <8 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+ %V16i32 = mul <16 x i32> undef, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+
+ %I16 = mul i16 undef, 16
+ %V8i16 = mul <8 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+ %V16i16 = mul <16 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+ %V32i16 = mul <32 x i16> undef, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+
+ %I8 = mul i8 undef, 16
+ %V16i8 = mul <16 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+ %V32i8 = mul <32 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+ %V64i8 = mul <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
+
+ ret i32 undef
+}
+
+define i32 @mul_constnegpow2() {
+; SLOW16-LABEL: 'mul_constnegpow2'
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %I64 = mul i64 undef, -16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I32 = mul i32 undef, -16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I16 = mul i16 undef, -16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I8 = mul i8 undef, -16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 264 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; FAST16-LABEL: 'mul_constnegpow2'
+; FAST16-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %I64 = mul i64 undef, -16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I32 = mul i32 undef, -16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I16 = mul i16 undef, -16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I8 = mul i8 undef, -16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 264 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; SLOW16-SIZE-LABEL: 'mul_constnegpow2'
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %I64 = mul i64 undef, -16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 288 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 132 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+; FAST16-SIZE-LABEL: 'mul_constnegpow2'
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %I64 = mul i64 undef, -16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 288 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 132 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %I64 = mul i64 undef, -16
+ %V2i64 = mul <2 x i64> undef, <i64 -8, i64 -16>
+ %V4i64 = mul <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
+ %V8i64 = mul <8 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16, i64 -32, i64 -64, i64 -128, i64 -256>
+
+ %I32 = mul i32 undef, -16
+ %V4i32 = mul <4 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16>
+ %V8i32 = mul <8 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+ %V16i32 = mul <16 x i32> undef, <i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256, i32 -2, i32 -4, i32 -8, i32 -16, i32 -32, i32 -64, i32 -128, i32 -256>
+
+ %I16 = mul i16 undef, -16
+ %V8i16 = mul <8 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+ %V16i16 = mul <16 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+ %V32i16 = mul <32 x i16> undef, <i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256, i16 -2, i16 -4, i16 -8, i16 -16, i16 -32, i16 -64, i16 -128, i16 -256>
+
+ %I8 = mul i8 undef, -16
+ %V16i8 = mul <16 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+ %V32i8 = mul <32 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+ %V64i8 = mul <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
+
+ ret i32 undef
+}
+
+define i32 @mul_uniformconstnegpow2() {
+; SLOW16-LABEL: 'mul_uniformconstnegpow2'
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %I64 = mul i64 undef, -16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I32 = mul i32 undef, -16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I16 = mul i16 undef, -16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I8 = mul i8 undef, -16
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 264 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLOW16-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; FAST16-LABEL: 'mul_uniformconstnegpow2'
+; FAST16-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %I64 = mul i64 undef, -16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 480 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I32 = mul i32 undef, -16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I16 = mul i16 undef, -16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %I8 = mul i8 undef, -16
+; FAST16-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 264 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; FAST16-NEXT: Cost Model: Found an estimated cost of 10 for instruction: ret i32 undef
+;
+; SLOW16-SIZE-LABEL: 'mul_uniformconstnegpow2'
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %I64 = mul i64 undef, -16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 288 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 132 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; SLOW16-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+; FAST16-SIZE-LABEL: 'mul_uniformconstnegpow2'
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %I64 = mul i64 undef, -16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 288 for instruction: %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I32 = mul i32 undef, -16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I16 = mul i16 undef, -16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %I8 = mul i8 undef, -16
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 132 for instruction: %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+; FAST16-SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
+;
+ %I64 = mul i64 undef, -16
+ %V2i64 = mul <2 x i64> undef, <i64 -16, i64 -16>
+ %V4i64 = mul <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
+ %V8i64 = mul <8 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16, i64 -16>
+
+ %I32 = mul i32 undef, -16
+ %V4i32 = mul <4 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16>
+ %V8i32 = mul <8 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+ %V16i32 = mul <16 x i32> undef, <i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16, i32 -16>
+
+ %I16 = mul i16 undef, -16
+ %V8i16 = mul <8 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+ %V16i16 = mul <16 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+ %V32i16 = mul <32 x i16> undef, <i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16>
+
+ %I8 = mul i8 undef, -16
+ %V16i8 = mul <16 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+ %V32i8 = mul <32 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+ %V64i8 = mul <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
+
+ ret i32 undef
+}
+
attributes #0 = { nounwind }
More information about the llvm-commits
mailing list