[llvm] 7aa52f3 - [CostModel][X86] Add test coverage for vXi64 multiplies of sign/zero-extended vector
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 13 08:36:25 PDT 2023
Author: Simon Pilgrim
Date: 2023-06-13T16:36:13+01:00
New Revision: 7aa52f30fd5f4fe5a0087fb55c3f34ee8156ee4d
URL: https://github.com/llvm/llvm-project/commit/7aa52f30fd5f4fe5a0087fb55c3f34ee8156ee4d
DIFF: https://github.com/llvm/llvm-project/commit/7aa52f30fd5f4fe5a0087fb55c3f34ee8156ee4d.diff
LOG: [CostModel][X86] Add test coverage for vXi64 multiplies of sign/zero-extended vector
Part of test coverage for Issue #62969
Added:
llvm/test/Analysis/CostModel/X86/mul64.ll
Modified:
Removed:
################################################################################
diff --git a/llvm/test/Analysis/CostModel/X86/mul64.ll b/llvm/test/Analysis/CostModel/X86/mul64.ll
new file mode 100644
index 0000000000000..26530067abb97
--- /dev/null
+++ b/llvm/test/Analysis/CostModel/X86/mul64.ll
@@ -0,0 +1,2116 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mattr=+sse2 | FileCheck %s --check-prefixes=SSE2
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mattr=+ssse3 | FileCheck %s --check-prefixes=SSSE3
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mattr=+sse4.2 | FileCheck %s --check-prefixes=SSE42
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512f | FileCheck %s --check-prefixes=AVX512,AVX512F
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=AVX512,AVX512BW
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512f,+avx512dq | FileCheck %s --check-prefixes=AVX512,AVX512DQ
+;
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=slm | FileCheck %s --check-prefixes=SLM
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX1
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=knl | FileCheck %s --check-prefixes=AVX512,AVX512F
+
+;
+; mul vXi8 -> mXi64
+;
+
+define void @mul_sext_vXi8(<2 x i8> %a2, <2 x i8> %b2, <4 x i8> %a4, <4 x i8> %b4, <8 x i8> %a8, <8 x i8> %b8, <16 x i8> %a16, <16 x i8> %b16, <32 x i8> %a32, <32 x i8> %b32, <64 x i8> %a64, <64 x i8> %b64) {
+; SSE2-LABEL: 'mul_sext_vXi8'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb2 = sext <2 x i8> %b2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb4 = sext <4 x i8> %b4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb8 = sext <8 x i8> %b8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb16 = sext <16 x i8> %b16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb32 = sext <32 x i8> %b32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xb64 = sext <64 x i8> %b64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE2-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSSE3-LABEL: 'mul_sext_vXi8'
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb2 = sext <2 x i8> %b2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb4 = sext <4 x i8> %b4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb8 = sext <8 x i8> %b8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb16 = sext <16 x i8> %b16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb32 = sext <32 x i8> %b32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xb64 = sext <64 x i8> %b64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'mul_sext_vXi8'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i8> %b2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i8> %b4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = sext <8 x i8> %b8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = sext <16 x i8> %b16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = sext <32 x i8> %b32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = sext <64 x i8> %b64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE42-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE42-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE42-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX1-LABEL: 'mul_sext_vXi8'
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i8> %b2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb4 = sext <4 x i8> %b4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb8 = sext <8 x i8> %b8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb16 = sext <16 x i8> %b16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %xb32 = sext <32 x i8> %b32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %xb64 = sext <64 x i8> %b64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX1-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX1-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX1-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX1-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX2-LABEL: 'mul_sext_vXi8'
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i8> %b2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i8> %b4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = sext <8 x i8> %b8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = sext <16 x i8> %b16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %xb32 = sext <32 x i8> %b32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %xb64 = sext <64 x i8> %b64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512F-LABEL: 'mul_sext_vXi8'
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i8> %b2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = sext <4 x i8> %b4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = sext <8 x i8> %b8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = sext <16 x i8> %b16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = sext <32 x i8> %b32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xb64 = sext <64 x i8> %b64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512BW-LABEL: 'mul_sext_vXi8'
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i8> %b2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = sext <4 x i8> %b4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = sext <8 x i8> %b8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = sext <16 x i8> %b16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = sext <32 x i8> %b32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xb64 = sext <64 x i8> %b64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512DQ-LABEL: 'mul_sext_vXi8'
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i8> %b2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = sext <4 x i8> %b4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = sext <8 x i8> %b8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = sext <16 x i8> %b16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = sext <32 x i8> %b32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xb64 = sext <64 x i8> %b64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SLM-LABEL: 'mul_sext_vXi8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i8> %b2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i8> %b4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = sext <8 x i8> %b8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = sext <16 x i8> %b16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = sext <32 x i8> %b32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = sext <64 x i8> %b64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SLM-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SLM-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SLM-NEXT: Cost Model: Found an estimated cost of 272 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SLM-NEXT: Cost Model: Found an estimated cost of 544 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; GLM-LABEL: 'mul_sext_vXi8'
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i8> %b2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i8> %b4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = sext <8 x i8> %b8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = sext <16 x i8> %b16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = sext <32 x i8> %b32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = sext <64 x i8> %b64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; GLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; GLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; GLM-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; GLM-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; GLM-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+ %xa2 = sext <2 x i8> %a2 to <2 x i64>
+ %xb2 = sext <2 x i8> %b2 to <2 x i64>
+ %xa4 = sext <4 x i8> %a4 to <4 x i64>
+ %xb4 = sext <4 x i8> %b4 to <4 x i64>
+ %xa8 = sext <8 x i8> %a8 to <8 x i64>
+ %xb8 = sext <8 x i8> %b8 to <8 x i64>
+ %xa16 = sext <16 x i8> %a16 to <16 x i64>
+ %xb16 = sext <16 x i8> %b16 to <16 x i64>
+ %xa32 = sext <32 x i8> %a32 to <32 x i64>
+ %xb32 = sext <32 x i8> %b32 to <32 x i64>
+ %xa64 = sext <64 x i8> %a64 to <64 x i64>
+ %xb64 = sext <64 x i8> %b64 to <64 x i64>
+ %res2 = mul <2 x i64> %xa2, %xb2
+ %res4 = mul <4 x i64> %xa4, %xb4
+ %res8 = mul <8 x i64> %xa8, %xb8
+ %res16 = mul <16 x i64> %xa16, %xb16
+ %res32 = mul <32 x i64> %xa32, %xb32
+ %res64 = mul <64 x i64> %xa64, %xb64
+ ret void
+}
+
+define void @mul_zext_vXi8(<2 x i8> %a2, <2 x i8> %b2, <4 x i8> %a4, <4 x i8> %b4, <8 x i8> %a8, <8 x i8> %b8, <16 x i8> %a16, <16 x i8> %b16, <32 x i8> %a32, <32 x i8> %b32, <64 x i8> %a64, <64 x i8> %b64) {
+; SSE2-LABEL: 'mul_zext_vXi8'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa2 = zext <2 x i8> %a2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa4 = zext <4 x i8> %a4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa8 = zext <8 x i8> %a8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa16 = zext <16 x i8> %a16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa32 = zext <32 x i8> %a32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xa64 = zext <64 x i8> %a64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE2-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSSE3-LABEL: 'mul_zext_vXi8'
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa2 = zext <2 x i8> %a2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa4 = zext <4 x i8> %a4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa8 = zext <8 x i8> %a8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa16 = zext <16 x i8> %a16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa32 = zext <32 x i8> %a32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xa64 = zext <64 x i8> %a64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'mul_zext_vXi8'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i8> %a2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i8> %a4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i8> %a8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i8> %a16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = zext <32 x i8> %a32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = zext <64 x i8> %a64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE42-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE42-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE42-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX1-LABEL: 'mul_zext_vXi8'
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i8> %a2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa4 = zext <4 x i8> %a4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa8 = zext <8 x i8> %a8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa16 = zext <16 x i8> %a16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %xa32 = zext <32 x i8> %a32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %xa64 = zext <64 x i8> %a64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX1-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX1-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX1-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX1-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX2-LABEL: 'mul_zext_vXi8'
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i8> %a2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i8> %a4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i8> %a8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i8> %a16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %xa32 = zext <32 x i8> %a32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %xa64 = zext <64 x i8> %a64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512F-LABEL: 'mul_zext_vXi8'
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i8> %a2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = zext <4 x i8> %a4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = zext <8 x i8> %a8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = zext <16 x i8> %a16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = zext <32 x i8> %a32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xa64 = zext <64 x i8> %a64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512BW-LABEL: 'mul_zext_vXi8'
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i8> %a2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = zext <4 x i8> %a4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = zext <8 x i8> %a8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = zext <16 x i8> %a16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = zext <32 x i8> %a32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xa64 = zext <64 x i8> %a64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512DQ-LABEL: 'mul_zext_vXi8'
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i8> %a2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = zext <4 x i8> %a4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = zext <8 x i8> %a8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = zext <16 x i8> %a16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = zext <32 x i8> %a32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xa64 = zext <64 x i8> %a64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SLM-LABEL: 'mul_zext_vXi8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i8> %a2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i8> %a4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i8> %a8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i8> %a16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = zext <32 x i8> %a32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = zext <64 x i8> %a64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SLM-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SLM-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SLM-NEXT: Cost Model: Found an estimated cost of 272 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SLM-NEXT: Cost Model: Found an estimated cost of 544 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; GLM-LABEL: 'mul_zext_vXi8'
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i8> %a2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i8> %a4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i8> %a8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i8> %a16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = zext <32 x i8> %a32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = zext <64 x i8> %a64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; GLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; GLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; GLM-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; GLM-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; GLM-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+ %xa2 = zext <2 x i8> %a2 to <2 x i64>
+ %xb2 = zext <2 x i8> %b2 to <2 x i64>
+ %xa4 = zext <4 x i8> %a4 to <4 x i64>
+ %xb4 = zext <4 x i8> %b4 to <4 x i64>
+ %xa8 = zext <8 x i8> %a8 to <8 x i64>
+ %xb8 = zext <8 x i8> %b8 to <8 x i64>
+ %xa16 = zext <16 x i8> %a16 to <16 x i64>
+ %xb16 = zext <16 x i8> %b16 to <16 x i64>
+ %xa32 = zext <32 x i8> %a32 to <32 x i64>
+ %xb32 = zext <32 x i8> %b32 to <32 x i64>
+ %xa64 = zext <64 x i8> %a64 to <64 x i64>
+ %xb64 = zext <64 x i8> %b64 to <64 x i64>
+ %res2 = mul <2 x i64> %xa2, %xb2
+ %res4 = mul <4 x i64> %xa4, %xb4
+ %res8 = mul <8 x i64> %xa8, %xb8
+ %res16 = mul <16 x i64> %xa16, %xb16
+ %res32 = mul <32 x i64> %xa32, %xb32
+ %res64 = mul <64 x i64> %xa64, %xb64
+ ret void
+}
+
+define void @mul_sext_zext_vXi8(<2 x i8> %a2, <2 x i8> %b2, <4 x i8> %a4, <4 x i8> %b4, <8 x i8> %a8, <8 x i8> %b8, <16 x i8> %a16, <16 x i8> %b16, <32 x i8> %a32, <32 x i8> %b32, <64 x i8> %a64, <64 x i8> %b64) {
+; SSE2-LABEL: 'mul_sext_zext_vXi8'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE2-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSSE3-LABEL: 'mul_sext_zext_vXi8'
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'mul_sext_zext_vXi8'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE42-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE42-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE42-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX1-LABEL: 'mul_sext_zext_vXi8'
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 25 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 50 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX1-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX1-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX1-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX1-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX2-LABEL: 'mul_sext_zext_vXi8'
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512F-LABEL: 'mul_sext_zext_vXi8'
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512BW-LABEL: 'mul_sext_zext_vXi8'
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512DQ-LABEL: 'mul_sext_zext_vXi8'
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SLM-LABEL: 'mul_sext_zext_vXi8'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SLM-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SLM-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SLM-NEXT: Cost Model: Found an estimated cost of 272 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SLM-NEXT: Cost Model: Found an estimated cost of 544 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; GLM-LABEL: 'mul_sext_zext_vXi8'
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i8> %a2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i8> %b2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i8> %a4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i8> %b4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i8> %a8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i8> %b8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i8> %a16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i8> %b16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i8> %a32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i8> %b32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i8> %a64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i8> %b64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; GLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; GLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; GLM-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; GLM-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; GLM-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+ %xa2 = sext <2 x i8> %a2 to <2 x i64>
+ %xb2 = zext <2 x i8> %b2 to <2 x i64>
+ %xa4 = sext <4 x i8> %a4 to <4 x i64>
+ %xb4 = zext <4 x i8> %b4 to <4 x i64>
+ %xa8 = sext <8 x i8> %a8 to <8 x i64>
+ %xb8 = zext <8 x i8> %b8 to <8 x i64>
+ %xa16 = sext <16 x i8> %a16 to <16 x i64>
+ %xb16 = zext <16 x i8> %b16 to <16 x i64>
+ %xa32 = sext <32 x i8> %a32 to <32 x i64>
+ %xb32 = zext <32 x i8> %b32 to <32 x i64>
+ %xa64 = sext <64 x i8> %a64 to <64 x i64>
+ %xb64 = zext <64 x i8> %b64 to <64 x i64>
+ %res2 = mul <2 x i64> %xa2, %xb2
+ %res4 = mul <4 x i64> %xa4, %xb4
+ %res8 = mul <8 x i64> %xa8, %xb8
+ %res16 = mul <16 x i64> %xa16, %xb16
+ %res32 = mul <32 x i64> %xa32, %xb32
+ %res64 = mul <64 x i64> %xa64, %xb64
+ ret void
+}
+
+;
+; mul vXi16 -> mXi64
+;
+
+define void @mul_sext_vXi16(<2 x i16> %a2, <2 x i16> %b2, <4 x i16> %a4, <4 x i16> %b4, <8 x i16> %a8, <8 x i16> %b8, <16 x i16> %a16, <16 x i16> %b16, <32 x i16> %a32, <32 x i16> %b32, <64 x i16> %a64, <64 x i16> %b64) {
+; SSE2-LABEL: 'mul_sext_vXi16'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb2 = sext <2 x i16> %b2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb4 = sext <4 x i16> %b4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb8 = sext <8 x i16> %b8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %xb16 = sext <16 x i16> %b16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %xb32 = sext <32 x i16> %b32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %xb64 = sext <64 x i16> %b64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE2-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSSE3-LABEL: 'mul_sext_vXi16'
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb2 = sext <2 x i16> %b2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb4 = sext <4 x i16> %b4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb8 = sext <8 x i16> %b8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %xb16 = sext <16 x i16> %b16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %xb32 = sext <32 x i16> %b32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %xb64 = sext <64 x i16> %b64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'mul_sext_vXi16'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i16> %b2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i16> %b4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = sext <8 x i16> %b8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = sext <16 x i16> %b16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = sext <32 x i16> %b32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = sext <64 x i16> %b64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE42-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE42-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE42-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX1-LABEL: 'mul_sext_vXi16'
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i16> %b2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb4 = sext <4 x i16> %b4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb8 = sext <8 x i16> %b8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %xb16 = sext <16 x i16> %b16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %xb32 = sext <32 x i16> %b32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 52 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 52 for instruction: %xb64 = sext <64 x i16> %b64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX1-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX1-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX1-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX1-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX2-LABEL: 'mul_sext_vXi16'
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i16> %b2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i16> %b4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = sext <8 x i16> %b8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %xb16 = sext <16 x i16> %b16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %xb32 = sext <32 x i16> %b32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 36 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 36 for instruction: %xb64 = sext <64 x i16> %b64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512F-LABEL: 'mul_sext_vXi16'
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i16> %b2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = sext <4 x i16> %b4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = sext <8 x i16> %b8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = sext <16 x i16> %b16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = sext <32 x i16> %b32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb64 = sext <64 x i16> %b64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512BW-LABEL: 'mul_sext_vXi16'
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i16> %b2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = sext <4 x i16> %b4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = sext <8 x i16> %b8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = sext <16 x i16> %b16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = sext <32 x i16> %b32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb64 = sext <64 x i16> %b64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512DQ-LABEL: 'mul_sext_vXi16'
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i16> %b2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = sext <4 x i16> %b4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = sext <8 x i16> %b8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = sext <16 x i16> %b16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = sext <32 x i16> %b32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb64 = sext <64 x i16> %b64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SLM-LABEL: 'mul_sext_vXi16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i16> %b2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i16> %b4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = sext <8 x i16> %b8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = sext <16 x i16> %b16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = sext <32 x i16> %b32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = sext <64 x i16> %b64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SLM-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SLM-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SLM-NEXT: Cost Model: Found an estimated cost of 272 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SLM-NEXT: Cost Model: Found an estimated cost of 544 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; GLM-LABEL: 'mul_sext_vXi16'
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i16> %b2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i16> %b4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = sext <8 x i16> %b8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = sext <16 x i16> %b16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = sext <32 x i16> %b32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = sext <64 x i16> %b64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; GLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; GLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; GLM-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; GLM-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; GLM-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+ %xa2 = sext <2 x i16> %a2 to <2 x i64>
+ %xb2 = sext <2 x i16> %b2 to <2 x i64>
+ %xa4 = sext <4 x i16> %a4 to <4 x i64>
+ %xb4 = sext <4 x i16> %b4 to <4 x i64>
+ %xa8 = sext <8 x i16> %a8 to <8 x i64>
+ %xb8 = sext <8 x i16> %b8 to <8 x i64>
+ %xa16 = sext <16 x i16> %a16 to <16 x i64>
+ %xb16 = sext <16 x i16> %b16 to <16 x i64>
+ %xa32 = sext <32 x i16> %a32 to <32 x i64>
+ %xb32 = sext <32 x i16> %b32 to <32 x i64>
+ %xa64 = sext <64 x i16> %a64 to <64 x i64>
+ %xb64 = sext <64 x i16> %b64 to <64 x i64>
+ %res2 = mul <2 x i64> %xa2, %xb2
+ %res4 = mul <4 x i64> %xa4, %xb4
+ %res8 = mul <8 x i64> %xa8, %xb8
+ %res16 = mul <16 x i64> %xa16, %xb16
+ %res32 = mul <32 x i64> %xa32, %xb32
+ %res64 = mul <64 x i64> %xa64, %xb64
+ ret void
+}
+
+define void @mul_zext_vXi16(<2 x i16> %a2, <2 x i16> %b2, <4 x i16> %a4, <4 x i16> %b4, <8 x i16> %a8, <8 x i16> %b8, <16 x i16> %a16, <16 x i16> %b16, <32 x i16> %a32, <32 x i16> %b32, <64 x i16> %a64, <64 x i16> %b64) {
+; SSE2-LABEL: 'mul_zext_vXi16'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa2 = zext <2 x i16> %a2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa4 = zext <4 x i16> %a4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa8 = zext <8 x i16> %a8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa16 = zext <16 x i16> %a16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa32 = zext <32 x i16> %a32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa64 = zext <64 x i16> %a64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE2-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSSE3-LABEL: 'mul_zext_vXi16'
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa2 = zext <2 x i16> %a2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa4 = zext <4 x i16> %a4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa8 = zext <8 x i16> %a8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa16 = zext <16 x i16> %a16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa32 = zext <32 x i16> %a32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa64 = zext <64 x i16> %a64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'mul_zext_vXi16'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i16> %a2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i16> %a4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i16> %a8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i16> %a16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = zext <32 x i16> %a32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = zext <64 x i16> %a64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE42-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE42-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE42-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX1-LABEL: 'mul_zext_vXi16'
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i16> %a2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa4 = zext <4 x i16> %a4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa8 = zext <8 x i16> %a8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %xa16 = zext <16 x i16> %a16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %xa32 = zext <32 x i16> %a32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 52 for instruction: %xa64 = zext <64 x i16> %a64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 52 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX1-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX1-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX1-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX1-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX2-LABEL: 'mul_zext_vXi16'
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i16> %a2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i16> %a4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i16> %a8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %xa16 = zext <16 x i16> %a16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %xa32 = zext <32 x i16> %a32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 36 for instruction: %xa64 = zext <64 x i16> %a64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 36 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512F-LABEL: 'mul_zext_vXi16'
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i16> %a2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = zext <4 x i16> %a4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = zext <8 x i16> %a8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = zext <16 x i16> %a16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = zext <32 x i16> %a32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa64 = zext <64 x i16> %a64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512BW-LABEL: 'mul_zext_vXi16'
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i16> %a2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = zext <4 x i16> %a4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = zext <8 x i16> %a8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = zext <16 x i16> %a16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = zext <32 x i16> %a32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa64 = zext <64 x i16> %a64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512DQ-LABEL: 'mul_zext_vXi16'
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i16> %a2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = zext <4 x i16> %a4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = zext <8 x i16> %a8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = zext <16 x i16> %a16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = zext <32 x i16> %a32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa64 = zext <64 x i16> %a64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SLM-LABEL: 'mul_zext_vXi16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i16> %a2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i16> %a4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i16> %a8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i16> %a16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = zext <32 x i16> %a32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = zext <64 x i16> %a64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SLM-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SLM-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SLM-NEXT: Cost Model: Found an estimated cost of 272 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SLM-NEXT: Cost Model: Found an estimated cost of 544 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; GLM-LABEL: 'mul_zext_vXi16'
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i16> %a2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i16> %a4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i16> %a8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i16> %a16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = zext <32 x i16> %a32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = zext <64 x i16> %a64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; GLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; GLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; GLM-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; GLM-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; GLM-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+ %xa2 = zext <2 x i16> %a2 to <2 x i64>
+ %xb2 = zext <2 x i16> %b2 to <2 x i64>
+ %xa4 = zext <4 x i16> %a4 to <4 x i64>
+ %xb4 = zext <4 x i16> %b4 to <4 x i64>
+ %xa8 = zext <8 x i16> %a8 to <8 x i64>
+ %xb8 = zext <8 x i16> %b8 to <8 x i64>
+ %xa16 = zext <16 x i16> %a16 to <16 x i64>
+ %xb16 = zext <16 x i16> %b16 to <16 x i64>
+ %xa32 = zext <32 x i16> %a32 to <32 x i64>
+ %xb32 = zext <32 x i16> %b32 to <32 x i64>
+ %xa64 = zext <64 x i16> %a64 to <64 x i64>
+ %xb64 = zext <64 x i16> %b64 to <64 x i64>
+ %res2 = mul <2 x i64> %xa2, %xb2
+ %res4 = mul <4 x i64> %xa4, %xb4
+ %res8 = mul <8 x i64> %xa8, %xb8
+ %res16 = mul <16 x i64> %xa16, %xb16
+ %res32 = mul <32 x i64> %xa32, %xb32
+ %res64 = mul <64 x i64> %xa64, %xb64
+ ret void
+}
+
+define void @mul_sext_zext_vXi16(<2 x i16> %a2, <2 x i16> %b2, <4 x i16> %a4, <4 x i16> %b4, <8 x i16> %a8, <8 x i16> %b8, <16 x i16> %a16, <16 x i16> %b16, <32 x i16> %a32, <32 x i16> %b32, <64 x i16> %a64, <64 x i16> %b64) {
+; SSE2-LABEL: 'mul_sext_zext_vXi16'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE2-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSSE3-LABEL: 'mul_sext_zext_vXi16'
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'mul_sext_zext_vXi16'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE42-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE42-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE42-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX1-LABEL: 'mul_sext_zext_vXi16'
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 13 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 26 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 52 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 52 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX1-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX1-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX1-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX1-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX2-LABEL: 'mul_sext_zext_vXi16'
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 18 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 36 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 36 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512F-LABEL: 'mul_sext_zext_vXi16'
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512BW-LABEL: 'mul_sext_zext_vXi16'
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512DQ-LABEL: 'mul_sext_zext_vXi16'
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SLM-LABEL: 'mul_sext_zext_vXi16'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SLM-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SLM-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SLM-NEXT: Cost Model: Found an estimated cost of 272 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SLM-NEXT: Cost Model: Found an estimated cost of 544 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; GLM-LABEL: 'mul_sext_zext_vXi16'
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i16> %a2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i16> %b2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i16> %a4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i16> %b4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i16> %a8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i16> %b8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i16> %a16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i16> %b16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i16> %a32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i16> %b32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i16> %a64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i16> %b64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; GLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; GLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; GLM-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; GLM-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; GLM-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+ %xa2 = sext <2 x i16> %a2 to <2 x i64>
+ %xb2 = zext <2 x i16> %b2 to <2 x i64>
+ %xa4 = sext <4 x i16> %a4 to <4 x i64>
+ %xb4 = zext <4 x i16> %b4 to <4 x i64>
+ %xa8 = sext <8 x i16> %a8 to <8 x i64>
+ %xb8 = zext <8 x i16> %b8 to <8 x i64>
+ %xa16 = sext <16 x i16> %a16 to <16 x i64>
+ %xb16 = zext <16 x i16> %b16 to <16 x i64>
+ %xa32 = sext <32 x i16> %a32 to <32 x i64>
+ %xb32 = zext <32 x i16> %b32 to <32 x i64>
+ %xa64 = sext <64 x i16> %a64 to <64 x i64>
+ %xb64 = zext <64 x i16> %b64 to <64 x i64>
+ %res2 = mul <2 x i64> %xa2, %xb2
+ %res4 = mul <4 x i64> %xa4, %xb4
+ %res8 = mul <8 x i64> %xa8, %xb8
+ %res16 = mul <16 x i64> %xa16, %xb16
+ %res32 = mul <32 x i64> %xa32, %xb32
+ %res64 = mul <64 x i64> %xa64, %xb64
+ ret void
+}
+
+;
+; mul vXi32 -> mXi64
+;
+
+define void @mul_sext_vXi32(<2 x i32> %a2, <2 x i32> %b2, <4 x i32> %a4, <4 x i32> %b4, <8 x i32> %a8, <8 x i32> %b8, <16 x i32> %a16, <16 x i32> %b16, <32 x i32> %a32, <32 x i32> %b32, <64 x i32> %a64, <64 x i32> %b64) {
+; SSE2-LABEL: 'mul_sext_vXi32'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb2 = sext <2 x i32> %b2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb4 = sext <4 x i32> %b4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb8 = sext <8 x i32> %b8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb16 = sext <16 x i32> %b16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb32 = sext <32 x i32> %b32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb64 = sext <64 x i32> %b64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE2-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSSE3-LABEL: 'mul_sext_vXi32'
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb2 = sext <2 x i32> %b2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb4 = sext <4 x i32> %b4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb8 = sext <8 x i32> %b8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb16 = sext <16 x i32> %b16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb32 = sext <32 x i32> %b32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xb64 = sext <64 x i32> %b64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'mul_sext_vXi32'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i32> %b2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i32> %b4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = sext <8 x i32> %b8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = sext <16 x i32> %b16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = sext <32 x i32> %b32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = sext <64 x i32> %b64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE42-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE42-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE42-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX1-LABEL: 'mul_sext_vXi32'
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i32> %b2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb4 = sext <4 x i32> %b4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb8 = sext <8 x i32> %b8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb16 = sext <16 x i32> %b16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %xb32 = sext <32 x i32> %b32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %xb64 = sext <64 x i32> %b64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX1-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX1-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX1-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX1-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX2-LABEL: 'mul_sext_vXi32'
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i32> %b2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i32> %b4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %xb8 = sext <8 x i32> %b8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %xb16 = sext <16 x i32> %b16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %xb32 = sext <32 x i32> %b32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %xb64 = sext <64 x i32> %b64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512F-LABEL: 'mul_sext_vXi32'
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i32> %b2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = sext <4 x i32> %b4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = sext <8 x i32> %b8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = sext <16 x i32> %b16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb32 = sext <32 x i32> %b32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb64 = sext <64 x i32> %b64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512BW-LABEL: 'mul_sext_vXi32'
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i32> %b2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = sext <4 x i32> %b4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = sext <8 x i32> %b8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = sext <16 x i32> %b16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb32 = sext <32 x i32> %b32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb64 = sext <64 x i32> %b64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512DQ-LABEL: 'mul_sext_vXi32'
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i32> %b2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = sext <4 x i32> %b4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = sext <8 x i32> %b8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = sext <16 x i32> %b16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb32 = sext <32 x i32> %b32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb64 = sext <64 x i32> %b64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SLM-LABEL: 'mul_sext_vXi32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i32> %b2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i32> %b4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = sext <8 x i32> %b8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = sext <16 x i32> %b16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = sext <32 x i32> %b32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = sext <64 x i32> %b64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SLM-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SLM-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SLM-NEXT: Cost Model: Found an estimated cost of 272 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SLM-NEXT: Cost Model: Found an estimated cost of 544 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; GLM-LABEL: 'mul_sext_vXi32'
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = sext <2 x i32> %b2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = sext <4 x i32> %b4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = sext <8 x i32> %b8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = sext <16 x i32> %b16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = sext <32 x i32> %b32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = sext <64 x i32> %b64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; GLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; GLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; GLM-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; GLM-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; GLM-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+ %xa2 = sext <2 x i32> %a2 to <2 x i64>
+ %xb2 = sext <2 x i32> %b2 to <2 x i64>
+ %xa4 = sext <4 x i32> %a4 to <4 x i64>
+ %xb4 = sext <4 x i32> %b4 to <4 x i64>
+ %xa8 = sext <8 x i32> %a8 to <8 x i64>
+ %xb8 = sext <8 x i32> %b8 to <8 x i64>
+ %xa16 = sext <16 x i32> %a16 to <16 x i64>
+ %xb16 = sext <16 x i32> %b16 to <16 x i64>
+ %xa32 = sext <32 x i32> %a32 to <32 x i64>
+ %xb32 = sext <32 x i32> %b32 to <32 x i64>
+ %xa64 = sext <64 x i32> %a64 to <64 x i64>
+ %xb64 = sext <64 x i32> %b64 to <64 x i64>
+ %res2 = mul <2 x i64> %xa2, %xb2
+ %res4 = mul <4 x i64> %xa4, %xb4
+ %res8 = mul <8 x i64> %xa8, %xb8
+ %res16 = mul <16 x i64> %xa16, %xb16
+ %res32 = mul <32 x i64> %xa32, %xb32
+ %res64 = mul <64 x i64> %xa64, %xb64
+ ret void
+}
+
+define void @mul_zext_vXi32(<2 x i32> %a2, <2 x i32> %b2, <4 x i32> %a4, <4 x i32> %b4, <8 x i32> %a8, <8 x i32> %b8, <16 x i32> %a16, <16 x i32> %b16, <32 x i32> %a32, <32 x i32> %b32, <64 x i32> %a64, <64 x i32> %b64) {
+; SSE2-LABEL: 'mul_zext_vXi32'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i32> %a2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i32> %a4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i32> %a8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i32> %a16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = zext <32 x i32> %a32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = zext <64 x i32> %a64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE2-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSSE3-LABEL: 'mul_zext_vXi32'
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i32> %a2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i32> %a4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i32> %a8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i32> %a16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = zext <32 x i32> %a32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = zext <64 x i32> %a64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'mul_zext_vXi32'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i32> %a2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i32> %a4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i32> %a8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i32> %a16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = zext <32 x i32> %a32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = zext <64 x i32> %a64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE42-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE42-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE42-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX1-LABEL: 'mul_zext_vXi32'
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i32> %a2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa4 = zext <4 x i32> %a4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa8 = zext <8 x i32> %a8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa16 = zext <16 x i32> %a16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %xa32 = zext <32 x i32> %a32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %xa64 = zext <64 x i32> %a64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX1-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX1-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX1-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX1-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX2-LABEL: 'mul_zext_vXi32'
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i32> %a2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i32> %a4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %xa8 = zext <8 x i32> %a8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %xa16 = zext <16 x i32> %a16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %xa32 = zext <32 x i32> %a32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %xa64 = zext <64 x i32> %a64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512F-LABEL: 'mul_zext_vXi32'
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i32> %a2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = zext <4 x i32> %a4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = zext <8 x i32> %a8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = zext <16 x i32> %a16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa32 = zext <32 x i32> %a32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa64 = zext <64 x i32> %a64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512BW-LABEL: 'mul_zext_vXi32'
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i32> %a2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = zext <4 x i32> %a4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = zext <8 x i32> %a8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = zext <16 x i32> %a16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa32 = zext <32 x i32> %a32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa64 = zext <64 x i32> %a64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512DQ-LABEL: 'mul_zext_vXi32'
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i32> %a2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = zext <4 x i32> %a4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = zext <8 x i32> %a8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = zext <16 x i32> %a16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa32 = zext <32 x i32> %a32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa64 = zext <64 x i32> %a64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SLM-LABEL: 'mul_zext_vXi32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i32> %a2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i32> %a4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i32> %a8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i32> %a16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = zext <32 x i32> %a32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = zext <64 x i32> %a64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SLM-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SLM-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SLM-NEXT: Cost Model: Found an estimated cost of 272 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SLM-NEXT: Cost Model: Found an estimated cost of 544 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; GLM-LABEL: 'mul_zext_vXi32'
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = zext <2 x i32> %a2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = zext <4 x i32> %a4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = zext <8 x i32> %a8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = zext <16 x i32> %a16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = zext <32 x i32> %a32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = zext <64 x i32> %a64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; GLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; GLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; GLM-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; GLM-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; GLM-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+ %xa2 = zext <2 x i32> %a2 to <2 x i64>
+ %xb2 = zext <2 x i32> %b2 to <2 x i64>
+ %xa4 = zext <4 x i32> %a4 to <4 x i64>
+ %xb4 = zext <4 x i32> %b4 to <4 x i64>
+ %xa8 = zext <8 x i32> %a8 to <8 x i64>
+ %xb8 = zext <8 x i32> %b8 to <8 x i64>
+ %xa16 = zext <16 x i32> %a16 to <16 x i64>
+ %xb16 = zext <16 x i32> %b16 to <16 x i64>
+ %xa32 = zext <32 x i32> %a32 to <32 x i64>
+ %xb32 = zext <32 x i32> %b32 to <32 x i64>
+ %xa64 = zext <64 x i32> %a64 to <64 x i64>
+ %xb64 = zext <64 x i32> %b64 to <64 x i64>
+ %res2 = mul <2 x i64> %xa2, %xb2
+ %res4 = mul <4 x i64> %xa4, %xb4
+ %res8 = mul <8 x i64> %xa8, %xb8
+ %res16 = mul <16 x i64> %xa16, %xb16
+ %res32 = mul <32 x i64> %xa32, %xb32
+ %res64 = mul <64 x i64> %xa64, %xb64
+ ret void
+}
+
+define void @mul_sext_zext_vXi32(<2 x i32> %a2, <2 x i32> %b2, <4 x i32> %a4, <4 x i32> %b4, <8 x i32> %a8, <8 x i32> %b8, <16 x i32> %a16, <16 x i32> %b16, <32 x i32> %a32, <32 x i32> %b32, <64 x i32> %a64, <64 x i32> %b64) {
+; SSE2-LABEL: 'mul_sext_zext_vXi32'
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; SSE2-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE2-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE2-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE2-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE2-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE2-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSSE3-LABEL: 'mul_sext_zext_vXi32'
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 64 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 128 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 256 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SSE42-LABEL: 'mul_sext_zext_vXi32'
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; SSE42-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SSE42-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SSE42-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SSE42-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SSE42-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SSE42-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX1-LABEL: 'mul_sext_zext_vXi32'
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; AVX1-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX1-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX1-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX1-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX1-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX1-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX2-LABEL: 'mul_sext_zext_vXi32'
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX2-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX2-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX2-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX2-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX2-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512F-LABEL: 'mul_sext_zext_vXi32'
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512F-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512BW-LABEL: 'mul_sext_zext_vXi32'
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; AVX512DQ-LABEL: 'mul_sext_zext_vXi32'
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; SLM-LABEL: 'mul_sext_zext_vXi32'
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; SLM-NEXT: Cost Model: Found an estimated cost of 17 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; SLM-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; SLM-NEXT: Cost Model: Found an estimated cost of 68 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; SLM-NEXT: Cost Model: Found an estimated cost of 136 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; SLM-NEXT: Cost Model: Found an estimated cost of 272 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; SLM-NEXT: Cost Model: Found an estimated cost of 544 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; GLM-LABEL: 'mul_sext_zext_vXi32'
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xa2 = sext <2 x i32> %a2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %xb2 = zext <2 x i32> %b2 to <2 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xa4 = sext <4 x i32> %a4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %xb4 = zext <4 x i32> %b4 to <4 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xa8 = sext <8 x i32> %a8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %xb8 = zext <8 x i32> %b8 to <8 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xa16 = sext <16 x i32> %a16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %xb16 = zext <16 x i32> %b16 to <16 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xa32 = sext <32 x i32> %a32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %xb32 = zext <32 x i32> %b32 to <32 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xa64 = sext <64 x i32> %a64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %xb64 = zext <64 x i32> %b64 to <64 x i64>
+; GLM-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res2 = mul <2 x i64> %xa2, %xb2
+; GLM-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %res4 = mul <4 x i64> %xa4, %xb4
+; GLM-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %res8 = mul <8 x i64> %xa8, %xb8
+; GLM-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %res16 = mul <16 x i64> %xa16, %xb16
+; GLM-NEXT: Cost Model: Found an estimated cost of 96 for instruction: %res32 = mul <32 x i64> %xa32, %xb32
+; GLM-NEXT: Cost Model: Found an estimated cost of 192 for instruction: %res64 = mul <64 x i64> %xa64, %xb64
+; GLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+ %xa2 = sext <2 x i32> %a2 to <2 x i64>
+ %xb2 = zext <2 x i32> %b2 to <2 x i64>
+ %xa4 = sext <4 x i32> %a4 to <4 x i64>
+ %xb4 = zext <4 x i32> %b4 to <4 x i64>
+ %xa8 = sext <8 x i32> %a8 to <8 x i64>
+ %xb8 = zext <8 x i32> %b8 to <8 x i64>
+ %xa16 = sext <16 x i32> %a16 to <16 x i64>
+ %xb16 = zext <16 x i32> %b16 to <16 x i64>
+ %xa32 = sext <32 x i32> %a32 to <32 x i64>
+ %xb32 = zext <32 x i32> %b32 to <32 x i64>
+ %xa64 = sext <64 x i32> %a64 to <64 x i64>
+ %xb64 = zext <64 x i32> %b64 to <64 x i64>
+ %res2 = mul <2 x i64> %xa2, %xb2
+ %res4 = mul <4 x i64> %xa4, %xb4
+ %res8 = mul <8 x i64> %xa8, %xb8
+ %res16 = mul <16 x i64> %xa16, %xb16
+ %res32 = mul <32 x i64> %xa32, %xb32
+ %res64 = mul <64 x i64> %xa64, %xb64
+ ret void
+}
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; AVX512: {{.*}}
More information about the llvm-commits
mailing list