From 5c7b975326c08a8b6f73be06ceab6d83b0f0e28e Mon Sep 17 00:00:00 2001 From: Simon Pilgrim Date: Thu, 18 Aug 2022 17:05:30 +0100 Subject: [PATCH] [CostModel][X86] Add explicit cmp/select test coverage for slm/glm targets --- llvm/test/Analysis/CostModel/X86/fcmp-costkinds.ll | 16 +++++ llvm/test/Analysis/CostModel/X86/icmp-costkinds.ll | 8 +++ .../Analysis/CostModel/X86/select-costkinds.ll | 8 +++ llvm/test/Analysis/CostModel/X86/select.ll | 78 ++++++++++++++++++++++ 4 files changed, 110 insertions(+) diff --git a/llvm/test/Analysis/CostModel/X86/fcmp-costkinds.ll b/llvm/test/Analysis/CostModel/X86/fcmp-costkinds.ll index 822f220..da95b93 100644 --- a/llvm/test/Analysis/CostModel/X86/fcmp-costkinds.ll +++ b/llvm/test/Analysis/CostModel/X86/fcmp-costkinds.ll @@ -10,6 +10,22 @@ ; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx2 -cost-kind=latency < %s | FileCheck %s --check-prefixes=LATE ; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx2 -cost-kind=code-size < %s | FileCheck %s --check-prefixes=SIZE ; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx2 -cost-kind=size-latency < %s | FileCheck %s --check-prefixes=SIZE_LATE +; +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512f,+avx512vl -cost-kind=latency < %s | FileCheck %s --check-prefixes=LATE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512f,+avx512vl -cost-kind=code-size < %s | FileCheck %s --check-prefixes=SIZE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512f,+avx512vl -cost-kind=size-latency < %s | FileCheck %s --check-prefixes=SIZE_LATE +; +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512bw,+avx512vl -cost-kind=latency < %s | FileCheck %s --check-prefixes=LATE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512bw,+avx512vl -cost-kind=code-size < %s | FileCheck %s --check-prefixes=SIZE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512bw,+avx512vl -cost-kind=size-latency < %s | FileCheck %s --check-prefixes=SIZE_LATE +; +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=slm -cost-kind=latency < %s | FileCheck %s --check-prefixes=LATE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=slm -cost-kind=code-size < %s | FileCheck %s --check-prefixes=SIZE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=slm -cost-kind=size-latency < %s | FileCheck %s --check-prefixes=SIZE_LATE +; +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=goldmont -cost-kind=latency < %s | FileCheck %s --check-prefixes=LATE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=goldmont -cost-kind=code-size < %s | FileCheck %s --check-prefixes=SIZE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=goldmont -cost-kind=size-latency < %s | FileCheck %s --check-prefixes=SIZE_LATE define i32 @cmp_float_oeq(i32 %arg) { ; LATE-LABEL: 'cmp_float_oeq' diff --git a/llvm/test/Analysis/CostModel/X86/icmp-costkinds.ll b/llvm/test/Analysis/CostModel/X86/icmp-costkinds.ll index 0760157..5a862a9 100644 --- a/llvm/test/Analysis/CostModel/X86/icmp-costkinds.ll +++ b/llvm/test/Analysis/CostModel/X86/icmp-costkinds.ll @@ -18,6 +18,14 @@ ; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512bw,+avx512vl -cost-kind=latency < %s | FileCheck %s --check-prefixes=LATE ; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512bw,+avx512vl -cost-kind=code-size < %s | FileCheck %s --check-prefixes=SIZE ; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512bw,+avx512vl -cost-kind=size-latency < %s | FileCheck %s --check-prefixes=SIZE_LATE +; +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=slm -cost-kind=latency < %s | FileCheck %s --check-prefixes=LATE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=slm -cost-kind=code-size < %s | FileCheck %s --check-prefixes=SIZE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=slm -cost-kind=size-latency < %s | FileCheck %s --check-prefixes=SIZE_LATE +; +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=goldmont -cost-kind=latency < %s | FileCheck %s --check-prefixes=LATE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=goldmont -cost-kind=code-size < %s | FileCheck %s --check-prefixes=SIZE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=goldmont -cost-kind=size-latency < %s | FileCheck %s --check-prefixes=SIZE_LATE define i32 @cmp_int_eq(i32 %arg) { ; LATE-LABEL: 'cmp_int_eq' diff --git a/llvm/test/Analysis/CostModel/X86/select-costkinds.ll b/llvm/test/Analysis/CostModel/X86/select-costkinds.ll index 3d0028d..010ec44 100644 --- a/llvm/test/Analysis/CostModel/X86/select-costkinds.ll +++ b/llvm/test/Analysis/CostModel/X86/select-costkinds.ll @@ -18,6 +18,14 @@ ; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512bw,+avx512vl -cost-kind=latency < %s | FileCheck %s --check-prefixes=LATE ; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512bw,+avx512vl -cost-kind=code-size < %s | FileCheck %s --check-prefixes=SIZE ; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mattr=+avx512bw,+avx512vl -cost-kind=size-latency < %s | FileCheck %s --check-prefixes=SIZE_LATE +; +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=slm -cost-kind=latency < %s | FileCheck %s --check-prefixes=LATE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=slm -cost-kind=code-size < %s | FileCheck %s --check-prefixes=SIZE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=slm -cost-kind=size-latency < %s | FileCheck %s --check-prefixes=SIZE_LATE +; +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=goldmont -cost-kind=latency < %s | FileCheck %s --check-prefixes=LATE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=goldmont -cost-kind=code-size < %s | FileCheck %s --check-prefixes=SIZE +; RUN: opt -mtriple=x86_64-- -passes="print" 2>&1 -disable-output -mcpu=goldmont -cost-kind=size-latency < %s | FileCheck %s --check-prefixes=SIZE_LATE ; Verify the cost of vector select instructions. diff --git a/llvm/test/Analysis/CostModel/X86/select.ll b/llvm/test/Analysis/CostModel/X86/select.ll index e94bf31..b791e81 100644 --- a/llvm/test/Analysis/CostModel/X86/select.ll +++ b/llvm/test/Analysis/CostModel/X86/select.ll @@ -5,6 +5,10 @@ ; RUN: opt < %s -passes="print" 2>&1 -disable-output -mtriple=x86_64-unknown-linux-gnu -mattr=+avx2 | FileCheck %s -check-prefixes=AVX,AVX2 ; RUN: opt < %s -passes="print" 2>&1 -disable-output -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512f,+avx512vl | FileCheck %s -check-prefixes=AVX512,AVX512F ; RUN: opt < %s -passes="print" 2>&1 -disable-output -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512bw,+avx512vl | FileCheck %s -check-prefixes=AVX512,AVX512BW +; +; RUN: opt < %s -passes="print" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=slm | FileCheck %s --check-prefixes=SLM +; RUN: opt < %s -passes="print" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=SSE +; RUN: opt < %s -passes="print" 2>&1 -disable-output -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=AVX,AVX1 ; Verify the cost of vector select instructions. @@ -104,6 +108,25 @@ define i32 @test_select() { ; AVX512BW-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V64I8 = select <64 x i1> undef, <64 x i8> undef, <64 x i8> undef ; AVX512BW-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef ; +; SLM-LABEL: 'test_select' +; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I64 = select i1 undef, i64 undef, i64 undef +; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2I64 = select <2 x i1> undef, <2 x i64> undef, <2 x i64> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4I64 = select <4 x i1> undef, <4 x i64> undef, <4 x i64> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8I64 = select <8 x i1> undef, <8 x i64> undef, <8 x i64> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I32 = select i1 undef, i32 undef, i32 undef +; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4I32 = select <4 x i1> undef, <4 x i32> undef, <4 x i32> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8I32 = select <8 x i1> undef, <8 x i32> undef, <8 x i32> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16I32 = select <16 x i1> undef, <16 x i32> undef, <16 x i32> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I16 = select i1 undef, i16 undef, i16 undef +; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V8I16 = select <8 x i1> undef, <8 x i16> undef, <8 x i16> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = select <16 x i1> undef, <16 x i16> undef, <16 x i16> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = select <32 x i1> undef, <32 x i16> undef, <32 x i16> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I8 = select i1 undef, i8 undef, i8 undef +; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V16I8 = select <16 x i1> undef, <16 x i8> undef, <16 x i8> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = select <32 x i1> undef, <32 x i8> undef, <32 x i8> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = select <64 x i1> undef, <64 x i8> undef, <64 x i8> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef +; %I64 = select i1 undef, i64 undef, i64 undef %V2I64 = select <2 x i1> undef, <2 x i64> undef, <2 x i64> undef %V4I64 = select <4 x i1> undef, <4 x i64> undef, <4 x i64> undef @@ -172,6 +195,17 @@ define i32 @test_select_fp() { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = select <16 x i1> undef, <16 x float> undef, <16 x float> undef ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef ; +; SLM-LABEL: 'test_select_fp' +; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = select i1 undef, double undef, double undef +; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = select <2 x i1> undef, <2 x double> undef, <2 x double> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = select <4 x i1> undef, <4 x double> undef, <4 x double> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = select <8 x i1> undef, <8 x double> undef, <8 x double> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = select i1 undef, float undef, float undef +; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %V4F32 = select <4 x i1> undef, <4 x float> undef, <4 x float> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V8F32 = select <8 x i1> undef, <8 x float> undef, <8 x float> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V16F32 = select <16 x i1> undef, <16 x float> undef, <16 x float> undef +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef +; %F64 = select i1 undef, double undef, double undef %V2F64 = select <2 x i1> undef, <2 x double> undef, <2 x double> undef %V4F64 = select <4 x i1> undef, <4 x double> undef, <4 x double> undef @@ -201,6 +235,10 @@ define <2 x i64> @test_2i64(<2 x i64> %a, <2 x i64> %b) { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sel = select <2 x i1> , <2 x i64> %a, <2 x i64> %b ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %sel ; +; SLM-LABEL: 'test_2i64' +; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %sel = select <2 x i1> , <2 x i64> %a, <2 x i64> %b +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %sel +; %sel = select <2 x i1> , <2 x i64> %a, <2 x i64> %b ret <2 x i64> %sel } @@ -218,6 +256,10 @@ define <2 x double> @test_2double(<2 x double> %a, <2 x double> %b) { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sel = select <2 x i1> , <2 x double> %a, <2 x double> %b ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %sel ; +; SLM-LABEL: 'test_2double' +; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %sel = select <2 x i1> , <2 x double> %a, <2 x double> %b +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %sel +; %sel = select <2 x i1> , <2 x double> %a, <2 x double> %b ret <2 x double> %sel } @@ -235,6 +277,10 @@ define <4 x i32> @test_4i32(<4 x i32> %a, <4 x i32> %b) { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> , <4 x i32> %a, <4 x i32> %b ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %sel ; +; SLM-LABEL: 'test_4i32' +; SLM-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %sel = select <4 x i1> , <4 x i32> %a, <4 x i32> %b +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %sel +; %sel = select <4 x i1> , <4 x i32> %a, <4 x i32> %b ret <4 x i32> %sel } @@ -252,6 +298,10 @@ define <4 x float> @test_4float(<4 x float> %a, <4 x float> %b) { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> , <4 x float> %a, <4 x float> %b ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %sel ; +; SLM-LABEL: 'test_4float' +; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %sel = select <4 x i1> , <4 x float> %a, <4 x float> %b +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %sel +; %sel = select <4 x i1> , <4 x float> %a, <4 x float> %b ret <4 x float> %sel } @@ -269,6 +319,10 @@ define <16 x i8> @test_16i8(<16 x i8> %a, <16 x i8> %b) { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sel = select <16 x i1> , <16 x i8> %a, <16 x i8> %b ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %sel ; +; SLM-LABEL: 'test_16i8' +; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %sel = select <16 x i1> , <16 x i8> %a, <16 x i8> %b +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %sel +; %sel = select <16 x i1> , <16 x i8> %a, <16 x i8> %b ret <16 x i8> %sel } @@ -293,6 +347,10 @@ define <4 x i64> @test_4i64(<4 x i64> %a, <4 x i64> %b) { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> , <4 x i64> %a, <4 x i64> %b ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %sel ; +; SLM-LABEL: 'test_4i64' +; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %sel = select <4 x i1> , <4 x i64> %a, <4 x i64> %b +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %sel +; %sel = select <4 x i1> , <4 x i64> %a, <4 x i64> %b ret <4 x i64> %sel } @@ -314,6 +372,10 @@ define <4 x double> @test_4double(<4 x double> %a, <4 x double> %b) { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sel = select <4 x i1> , <4 x double> %a, <4 x double> %b ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %sel ; +; SLM-LABEL: 'test_4double' +; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %sel = select <4 x i1> , <4 x double> %a, <4 x double> %b +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %sel +; %sel = select <4 x i1> , <4 x double> %a, <4 x double> %b ret <4 x double> %sel } @@ -335,6 +397,10 @@ define <8 x i32> @test_8i32(<8 x i32> %a, <8 x i32> %b) { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sel = select <8 x i1> , <8 x i32> %a, <8 x i32> %b ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %sel ; +; SLM-LABEL: 'test_8i32' +; SLM-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %sel = select <8 x i1> , <8 x i32> %a, <8 x i32> %b +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %sel +; %sel = select <8 x i1> , <8 x i32> %a, <8 x i32> %b ret <8 x i32> %sel } @@ -356,6 +422,10 @@ define <8 x float> @test_8float(<8 x float> %a, <8 x float> %b) { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sel = select <8 x i1> , <8 x float> %a, <8 x float> %b ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %sel ; +; SLM-LABEL: 'test_8float' +; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %sel = select <8 x i1> , <8 x float> %a, <8 x float> %b +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %sel +; %sel = select <8 x i1> , <8 x float> %a, <8 x float> %b ret <8 x float> %sel } @@ -377,6 +447,10 @@ define <16 x i16> @test_16i16(<16 x i16> %a, <16 x i16> %b) { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sel = select <16 x i1> , <16 x i16> %a, <16 x i16> %b ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %sel ; +; SLM-LABEL: 'test_16i16' +; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %sel = select <16 x i1> , <16 x i16> %a, <16 x i16> %b +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %sel +; %sel = select <16 x i1> , <16 x i16> %a, <16 x i16> %b ret <16 x i16> %sel } @@ -398,6 +472,10 @@ define <32 x i8> @test_32i8(<32 x i8> %a, <32 x i8> %b) { ; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %sel = select <32 x i1> , <32 x i8> %a, <32 x i8> %b ; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %sel ; +; SLM-LABEL: 'test_32i8' +; SLM-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %sel = select <32 x i1> , <32 x i8> %a, <32 x i8> %b +; SLM-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %sel +; %sel = select <32 x i1> , <32 x i8> %a, <32 x i8> %b ret <32 x i8> %sel } -- 2.7.4