[llvm] r358552 - Revert "Temporarily Revert "Add basic loop fusion pass.""
Eric Christopher via llvm-commits
llvm-commits at lists.llvm.org
Tue Apr 16 21:53:01 PDT 2019
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-ssat.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-ssat.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-ssat.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-ssat.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,775 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,SSE
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=slm -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,SLM
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=knl -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skx -mattr=-prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skx -mattr=+prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX256BW
+
+ at a64 = common global [8 x i64] zeroinitializer, align 64
+ at b64 = common global [8 x i64] zeroinitializer, align 64
+ at c64 = common global [8 x i64] zeroinitializer, align 64
+ at a32 = common global [16 x i32] zeroinitializer, align 64
+ at b32 = common global [16 x i32] zeroinitializer, align 64
+ at c32 = common global [16 x i32] zeroinitializer, align 64
+ at a16 = common global [32 x i16] zeroinitializer, align 64
+ at b16 = common global [32 x i16] zeroinitializer, align 64
+ at c16 = common global [32 x i16] zeroinitializer, align 64
+ at a8 = common global [64 x i8] zeroinitializer, align 64
+ at b8 = common global [64 x i8] zeroinitializer, align 64
+ at c8 = common global [64 x i8] zeroinitializer, align 64
+
+declare i64 @llvm.ssub.sat.i64(i64, i64)
+declare i32 @llvm.ssub.sat.i32(i32, i32)
+declare i16 @llvm.ssub.sat.i16(i16, i16)
+declare i8 @llvm.ssub.sat.i8 (i8 , i8 )
+
+define void @sub_v8i64() {
+; SSE-LABEL: @sub_v8i64(
+; SSE-NEXT: [[A0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 0), align 8
+; SSE-NEXT: [[A1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 1), align 8
+; SSE-NEXT: [[A2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2), align 8
+; SSE-NEXT: [[A3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 3), align 8
+; SSE-NEXT: [[A4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4), align 8
+; SSE-NEXT: [[A5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 5), align 8
+; SSE-NEXT: [[A6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6), align 8
+; SSE-NEXT: [[A7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 7), align 8
+; SSE-NEXT: [[B0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 0), align 8
+; SSE-NEXT: [[B1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 1), align 8
+; SSE-NEXT: [[B2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2), align 8
+; SSE-NEXT: [[B3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 3), align 8
+; SSE-NEXT: [[B4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4), align 8
+; SSE-NEXT: [[B5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 5), align 8
+; SSE-NEXT: [[B6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6), align 8
+; SSE-NEXT: [[B7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 7), align 8
+; SSE-NEXT: [[R0:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A0]], i64 [[B0]])
+; SSE-NEXT: [[R1:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A1]], i64 [[B1]])
+; SSE-NEXT: [[R2:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A2]], i64 [[B2]])
+; SSE-NEXT: [[R3:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A3]], i64 [[B3]])
+; SSE-NEXT: [[R4:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A4]], i64 [[B4]])
+; SSE-NEXT: [[R5:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A5]], i64 [[B5]])
+; SSE-NEXT: [[R6:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A6]], i64 [[B6]])
+; SSE-NEXT: [[R7:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A7]], i64 [[B7]])
+; SSE-NEXT: store i64 [[R0]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 0), align 8
+; SSE-NEXT: store i64 [[R1]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 1), align 8
+; SSE-NEXT: store i64 [[R2]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2), align 8
+; SSE-NEXT: store i64 [[R3]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 3), align 8
+; SSE-NEXT: store i64 [[R4]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4), align 8
+; SSE-NEXT: store i64 [[R5]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 5), align 8
+; SSE-NEXT: store i64 [[R6]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6), align 8
+; SSE-NEXT: store i64 [[R7]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 7), align 8
+; SSE-NEXT: ret void
+;
+; SLM-LABEL: @sub_v8i64(
+; SLM-NEXT: [[A0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 0), align 8
+; SLM-NEXT: [[A1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 1), align 8
+; SLM-NEXT: [[A2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2), align 8
+; SLM-NEXT: [[A3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 3), align 8
+; SLM-NEXT: [[A4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4), align 8
+; SLM-NEXT: [[A5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 5), align 8
+; SLM-NEXT: [[A6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6), align 8
+; SLM-NEXT: [[A7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 7), align 8
+; SLM-NEXT: [[B0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 0), align 8
+; SLM-NEXT: [[B1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 1), align 8
+; SLM-NEXT: [[B2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2), align 8
+; SLM-NEXT: [[B3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 3), align 8
+; SLM-NEXT: [[B4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4), align 8
+; SLM-NEXT: [[B5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 5), align 8
+; SLM-NEXT: [[B6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6), align 8
+; SLM-NEXT: [[B7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 7), align 8
+; SLM-NEXT: [[R0:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A0]], i64 [[B0]])
+; SLM-NEXT: [[R1:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A1]], i64 [[B1]])
+; SLM-NEXT: [[R2:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A2]], i64 [[B2]])
+; SLM-NEXT: [[R3:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A3]], i64 [[B3]])
+; SLM-NEXT: [[R4:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A4]], i64 [[B4]])
+; SLM-NEXT: [[R5:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A5]], i64 [[B5]])
+; SLM-NEXT: [[R6:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A6]], i64 [[B6]])
+; SLM-NEXT: [[R7:%.*]] = call i64 @llvm.ssub.sat.i64(i64 [[A7]], i64 [[B7]])
+; SLM-NEXT: store i64 [[R0]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 0), align 8
+; SLM-NEXT: store i64 [[R1]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 1), align 8
+; SLM-NEXT: store i64 [[R2]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2), align 8
+; SLM-NEXT: store i64 [[R3]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 3), align 8
+; SLM-NEXT: store i64 [[R4]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4), align 8
+; SLM-NEXT: store i64 [[R5]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 5), align 8
+; SLM-NEXT: store i64 [[R6]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6), align 8
+; SLM-NEXT: store i64 [[R7]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 7), align 8
+; SLM-NEXT: ret void
+;
+; AVX1-LABEL: @sub_v8i64(
+; AVX1-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @a64 to <2 x i64>*), align 8
+; AVX1-NEXT: [[TMP2:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2) to <2 x i64>*), align 8
+; AVX1-NEXT: [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4) to <2 x i64>*), align 8
+; AVX1-NEXT: [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6) to <2 x i64>*), align 8
+; AVX1-NEXT: [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @b64 to <2 x i64>*), align 8
+; AVX1-NEXT: [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2) to <2 x i64>*), align 8
+; AVX1-NEXT: [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4) to <2 x i64>*), align 8
+; AVX1-NEXT: [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6) to <2 x i64>*), align 8
+; AVX1-NEXT: [[TMP9:%.*]] = call <2 x i64> @llvm.ssub.sat.v2i64(<2 x i64> [[TMP1]], <2 x i64> [[TMP5]])
+; AVX1-NEXT: [[TMP10:%.*]] = call <2 x i64> @llvm.ssub.sat.v2i64(<2 x i64> [[TMP2]], <2 x i64> [[TMP6]])
+; AVX1-NEXT: [[TMP11:%.*]] = call <2 x i64> @llvm.ssub.sat.v2i64(<2 x i64> [[TMP3]], <2 x i64> [[TMP7]])
+; AVX1-NEXT: [[TMP12:%.*]] = call <2 x i64> @llvm.ssub.sat.v2i64(<2 x i64> [[TMP4]], <2 x i64> [[TMP8]])
+; AVX1-NEXT: store <2 x i64> [[TMP9]], <2 x i64>* bitcast ([8 x i64]* @c64 to <2 x i64>*), align 8
+; AVX1-NEXT: store <2 x i64> [[TMP10]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2) to <2 x i64>*), align 8
+; AVX1-NEXT: store <2 x i64> [[TMP11]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4) to <2 x i64>*), align 8
+; AVX1-NEXT: store <2 x i64> [[TMP12]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6) to <2 x i64>*), align 8
+; AVX1-NEXT: ret void
+;
+; AVX2-LABEL: @sub_v8i64(
+; AVX2-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @a64 to <4 x i64>*), align 8
+; AVX2-NEXT: [[TMP2:%.*]] = load <4 x i64>, <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX2-NEXT: [[TMP3:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @b64 to <4 x i64>*), align 8
+; AVX2-NEXT: [[TMP4:%.*]] = load <4 x i64>, <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX2-NEXT: [[TMP5:%.*]] = call <4 x i64> @llvm.ssub.sat.v4i64(<4 x i64> [[TMP1]], <4 x i64> [[TMP3]])
+; AVX2-NEXT: [[TMP6:%.*]] = call <4 x i64> @llvm.ssub.sat.v4i64(<4 x i64> [[TMP2]], <4 x i64> [[TMP4]])
+; AVX2-NEXT: store <4 x i64> [[TMP5]], <4 x i64>* bitcast ([8 x i64]* @c64 to <4 x i64>*), align 8
+; AVX2-NEXT: store <4 x i64> [[TMP6]], <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX2-NEXT: ret void
+;
+; AVX512-LABEL: @sub_v8i64(
+; AVX512-NEXT: [[TMP1:%.*]] = load <8 x i64>, <8 x i64>* bitcast ([8 x i64]* @a64 to <8 x i64>*), align 8
+; AVX512-NEXT: [[TMP2:%.*]] = load <8 x i64>, <8 x i64>* bitcast ([8 x i64]* @b64 to <8 x i64>*), align 8
+; AVX512-NEXT: [[TMP3:%.*]] = call <8 x i64> @llvm.ssub.sat.v8i64(<8 x i64> [[TMP1]], <8 x i64> [[TMP2]])
+; AVX512-NEXT: store <8 x i64> [[TMP3]], <8 x i64>* bitcast ([8 x i64]* @c64 to <8 x i64>*), align 8
+; AVX512-NEXT: ret void
+;
+; AVX256BW-LABEL: @sub_v8i64(
+; AVX256BW-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @a64 to <4 x i64>*), align 8
+; AVX256BW-NEXT: [[TMP2:%.*]] = load <4 x i64>, <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX256BW-NEXT: [[TMP3:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @b64 to <4 x i64>*), align 8
+; AVX256BW-NEXT: [[TMP4:%.*]] = load <4 x i64>, <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX256BW-NEXT: [[TMP5:%.*]] = call <4 x i64> @llvm.ssub.sat.v4i64(<4 x i64> [[TMP1]], <4 x i64> [[TMP3]])
+; AVX256BW-NEXT: [[TMP6:%.*]] = call <4 x i64> @llvm.ssub.sat.v4i64(<4 x i64> [[TMP2]], <4 x i64> [[TMP4]])
+; AVX256BW-NEXT: store <4 x i64> [[TMP5]], <4 x i64>* bitcast ([8 x i64]* @c64 to <4 x i64>*), align 8
+; AVX256BW-NEXT: store <4 x i64> [[TMP6]], <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX256BW-NEXT: ret void
+;
+ %a0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 0), align 8
+ %a1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 1), align 8
+ %a2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2), align 8
+ %a3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 3), align 8
+ %a4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4), align 8
+ %a5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 5), align 8
+ %a6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6), align 8
+ %a7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 7), align 8
+ %b0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 0), align 8
+ %b1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 1), align 8
+ %b2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2), align 8
+ %b3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 3), align 8
+ %b4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4), align 8
+ %b5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 5), align 8
+ %b6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6), align 8
+ %b7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 7), align 8
+ %r0 = call i64 @llvm.ssub.sat.i64(i64 %a0, i64 %b0)
+ %r1 = call i64 @llvm.ssub.sat.i64(i64 %a1, i64 %b1)
+ %r2 = call i64 @llvm.ssub.sat.i64(i64 %a2, i64 %b2)
+ %r3 = call i64 @llvm.ssub.sat.i64(i64 %a3, i64 %b3)
+ %r4 = call i64 @llvm.ssub.sat.i64(i64 %a4, i64 %b4)
+ %r5 = call i64 @llvm.ssub.sat.i64(i64 %a5, i64 %b5)
+ %r6 = call i64 @llvm.ssub.sat.i64(i64 %a6, i64 %b6)
+ %r7 = call i64 @llvm.ssub.sat.i64(i64 %a7, i64 %b7)
+ store i64 %r0, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 0), align 8
+ store i64 %r1, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 1), align 8
+ store i64 %r2, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2), align 8
+ store i64 %r3, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 3), align 8
+ store i64 %r4, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4), align 8
+ store i64 %r5, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 5), align 8
+ store i64 %r6, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6), align 8
+ store i64 %r7, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 7), align 8
+ ret void
+}
+
+define void @sub_v16i32() {
+; SSE-LABEL: @sub_v16i32(
+; SSE-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @a32 to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @b32 to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP9:%.*]] = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> [[TMP1]], <4 x i32> [[TMP5]])
+; SSE-NEXT: [[TMP10:%.*]] = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> [[TMP2]], <4 x i32> [[TMP6]])
+; SSE-NEXT: [[TMP11:%.*]] = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> [[TMP3]], <4 x i32> [[TMP7]])
+; SSE-NEXT: [[TMP12:%.*]] = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> [[TMP4]], <4 x i32> [[TMP8]])
+; SSE-NEXT: store <4 x i32> [[TMP9]], <4 x i32>* bitcast ([16 x i32]* @c32 to <4 x i32>*), align 4
+; SSE-NEXT: store <4 x i32> [[TMP10]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4) to <4 x i32>*), align 4
+; SSE-NEXT: store <4 x i32> [[TMP11]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8) to <4 x i32>*), align 4
+; SSE-NEXT: store <4 x i32> [[TMP12]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12) to <4 x i32>*), align 4
+; SSE-NEXT: ret void
+;
+; SLM-LABEL: @sub_v16i32(
+; SLM-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @a32 to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @b32 to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP9:%.*]] = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> [[TMP1]], <4 x i32> [[TMP5]])
+; SLM-NEXT: [[TMP10:%.*]] = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> [[TMP2]], <4 x i32> [[TMP6]])
+; SLM-NEXT: [[TMP11:%.*]] = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> [[TMP3]], <4 x i32> [[TMP7]])
+; SLM-NEXT: [[TMP12:%.*]] = call <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32> [[TMP4]], <4 x i32> [[TMP8]])
+; SLM-NEXT: store <4 x i32> [[TMP9]], <4 x i32>* bitcast ([16 x i32]* @c32 to <4 x i32>*), align 4
+; SLM-NEXT: store <4 x i32> [[TMP10]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4) to <4 x i32>*), align 4
+; SLM-NEXT: store <4 x i32> [[TMP11]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8) to <4 x i32>*), align 4
+; SLM-NEXT: store <4 x i32> [[TMP12]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12) to <4 x i32>*), align 4
+; SLM-NEXT: ret void
+;
+; AVX-LABEL: @sub_v16i32(
+; AVX-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([16 x i32]* @a32 to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8) to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP3:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([16 x i32]* @b32 to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP4:%.*]] = load <8 x i32>, <8 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8) to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP5:%.*]] = call <8 x i32> @llvm.ssub.sat.v8i32(<8 x i32> [[TMP1]], <8 x i32> [[TMP3]])
+; AVX-NEXT: [[TMP6:%.*]] = call <8 x i32> @llvm.ssub.sat.v8i32(<8 x i32> [[TMP2]], <8 x i32> [[TMP4]])
+; AVX-NEXT: store <8 x i32> [[TMP5]], <8 x i32>* bitcast ([16 x i32]* @c32 to <8 x i32>*), align 4
+; AVX-NEXT: store <8 x i32> [[TMP6]], <8 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8) to <8 x i32>*), align 4
+; AVX-NEXT: ret void
+;
+; AVX512-LABEL: @sub_v16i32(
+; AVX512-NEXT: [[TMP1:%.*]] = load <16 x i32>, <16 x i32>* bitcast ([16 x i32]* @a32 to <16 x i32>*), align 4
+; AVX512-NEXT: [[TMP2:%.*]] = load <16 x i32>, <16 x i32>* bitcast ([16 x i32]* @b32 to <16 x i32>*), align 4
+; AVX512-NEXT: [[TMP3:%.*]] = call <16 x i32> @llvm.ssub.sat.v16i32(<16 x i32> [[TMP1]], <16 x i32> [[TMP2]])
+; AVX512-NEXT: store <16 x i32> [[TMP3]], <16 x i32>* bitcast ([16 x i32]* @c32 to <16 x i32>*), align 4
+; AVX512-NEXT: ret void
+;
+ %a0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 0 ), align 4
+ %a1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 1 ), align 4
+ %a2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 2 ), align 4
+ %a3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 3 ), align 4
+ %a4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4 ), align 4
+ %a5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 5 ), align 4
+ %a6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 6 ), align 4
+ %a7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 7 ), align 4
+ %a8 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8 ), align 4
+ %a9 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 9 ), align 4
+ %a10 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 10), align 4
+ %a11 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 11), align 4
+ %a12 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12), align 4
+ %a13 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 13), align 4
+ %a14 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 14), align 4
+ %a15 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 15), align 4
+ %b0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 0 ), align 4
+ %b1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 1 ), align 4
+ %b2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 2 ), align 4
+ %b3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 3 ), align 4
+ %b4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4 ), align 4
+ %b5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 5 ), align 4
+ %b6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 6 ), align 4
+ %b7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 7 ), align 4
+ %b8 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8 ), align 4
+ %b9 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 9 ), align 4
+ %b10 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 10), align 4
+ %b11 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 11), align 4
+ %b12 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12), align 4
+ %b13 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 13), align 4
+ %b14 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 14), align 4
+ %b15 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 15), align 4
+ %r0 = call i32 @llvm.ssub.sat.i32(i32 %a0 , i32 %b0 )
+ %r1 = call i32 @llvm.ssub.sat.i32(i32 %a1 , i32 %b1 )
+ %r2 = call i32 @llvm.ssub.sat.i32(i32 %a2 , i32 %b2 )
+ %r3 = call i32 @llvm.ssub.sat.i32(i32 %a3 , i32 %b3 )
+ %r4 = call i32 @llvm.ssub.sat.i32(i32 %a4 , i32 %b4 )
+ %r5 = call i32 @llvm.ssub.sat.i32(i32 %a5 , i32 %b5 )
+ %r6 = call i32 @llvm.ssub.sat.i32(i32 %a6 , i32 %b6 )
+ %r7 = call i32 @llvm.ssub.sat.i32(i32 %a7 , i32 %b7 )
+ %r8 = call i32 @llvm.ssub.sat.i32(i32 %a8 , i32 %b8 )
+ %r9 = call i32 @llvm.ssub.sat.i32(i32 %a9 , i32 %b9 )
+ %r10 = call i32 @llvm.ssub.sat.i32(i32 %a10, i32 %b10)
+ %r11 = call i32 @llvm.ssub.sat.i32(i32 %a11, i32 %b11)
+ %r12 = call i32 @llvm.ssub.sat.i32(i32 %a12, i32 %b12)
+ %r13 = call i32 @llvm.ssub.sat.i32(i32 %a13, i32 %b13)
+ %r14 = call i32 @llvm.ssub.sat.i32(i32 %a14, i32 %b14)
+ %r15 = call i32 @llvm.ssub.sat.i32(i32 %a15, i32 %b15)
+ store i32 %r0 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 0 ), align 4
+ store i32 %r1 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 1 ), align 4
+ store i32 %r2 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 2 ), align 4
+ store i32 %r3 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 3 ), align 4
+ store i32 %r4 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4 ), align 4
+ store i32 %r5 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 5 ), align 4
+ store i32 %r6 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 6 ), align 4
+ store i32 %r7 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 7 ), align 4
+ store i32 %r8 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8 ), align 4
+ store i32 %r9 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 9 ), align 4
+ store i32 %r10, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 10), align 4
+ store i32 %r11, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 11), align 4
+ store i32 %r12, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12), align 4
+ store i32 %r13, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 13), align 4
+ store i32 %r14, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 14), align 4
+ store i32 %r15, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 15), align 4
+ ret void
+}
+
+define void @sub_v32i16() {
+; SSE-LABEL: @sub_v32i16(
+; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @a16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @b16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP9:%.*]] = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> [[TMP1]], <8 x i16> [[TMP5]])
+; SSE-NEXT: [[TMP10:%.*]] = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> [[TMP2]], <8 x i16> [[TMP6]])
+; SSE-NEXT: [[TMP11:%.*]] = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> [[TMP3]], <8 x i16> [[TMP7]])
+; SSE-NEXT: [[TMP12:%.*]] = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> [[TMP4]], <8 x i16> [[TMP8]])
+; SSE-NEXT: store <8 x i16> [[TMP9]], <8 x i16>* bitcast ([32 x i16]* @c16 to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP10]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP11]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP12]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24) to <8 x i16>*), align 2
+; SSE-NEXT: ret void
+;
+; SLM-LABEL: @sub_v32i16(
+; SLM-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @a16 to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @b16 to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP9:%.*]] = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> [[TMP1]], <8 x i16> [[TMP5]])
+; SLM-NEXT: [[TMP10:%.*]] = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> [[TMP2]], <8 x i16> [[TMP6]])
+; SLM-NEXT: [[TMP11:%.*]] = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> [[TMP3]], <8 x i16> [[TMP7]])
+; SLM-NEXT: [[TMP12:%.*]] = call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> [[TMP4]], <8 x i16> [[TMP8]])
+; SLM-NEXT: store <8 x i16> [[TMP9]], <8 x i16>* bitcast ([32 x i16]* @c16 to <8 x i16>*), align 2
+; SLM-NEXT: store <8 x i16> [[TMP10]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8) to <8 x i16>*), align 2
+; SLM-NEXT: store <8 x i16> [[TMP11]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <8 x i16>*), align 2
+; SLM-NEXT: store <8 x i16> [[TMP12]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24) to <8 x i16>*), align 2
+; SLM-NEXT: ret void
+;
+; AVX-LABEL: @sub_v32i16(
+; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @a16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP3:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @b16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP4:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP5:%.*]] = call <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16> [[TMP1]], <16 x i16> [[TMP3]])
+; AVX-NEXT: [[TMP6:%.*]] = call <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16> [[TMP2]], <16 x i16> [[TMP4]])
+; AVX-NEXT: store <16 x i16> [[TMP5]], <16 x i16>* bitcast ([32 x i16]* @c16 to <16 x i16>*), align 2
+; AVX-NEXT: store <16 x i16> [[TMP6]], <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX-NEXT: ret void
+;
+; AVX512-LABEL: @sub_v32i16(
+; AVX512-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @a16 to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP2:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP3:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @b16 to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP4:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP5:%.*]] = call <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16> [[TMP1]], <16 x i16> [[TMP3]])
+; AVX512-NEXT: [[TMP6:%.*]] = call <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16> [[TMP2]], <16 x i16> [[TMP4]])
+; AVX512-NEXT: store <16 x i16> [[TMP5]], <16 x i16>* bitcast ([32 x i16]* @c16 to <16 x i16>*), align 2
+; AVX512-NEXT: store <16 x i16> [[TMP6]], <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX512-NEXT: ret void
+;
+ %a0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 0 ), align 2
+ %a1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 1 ), align 2
+ %a2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 2 ), align 2
+ %a3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 3 ), align 2
+ %a4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 4 ), align 2
+ %a5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 5 ), align 2
+ %a6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 6 ), align 2
+ %a7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 7 ), align 2
+ %a8 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8 ), align 2
+ %a9 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 9 ), align 2
+ %a10 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 10), align 2
+ %a11 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 11), align 2
+ %a12 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 12), align 2
+ %a13 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 13), align 2
+ %a14 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 14), align 2
+ %a15 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 15), align 2
+ %a16 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16), align 2
+ %a17 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 17), align 2
+ %a18 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 18), align 2
+ %a19 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 19), align 2
+ %a20 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 20), align 2
+ %a21 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 21), align 2
+ %a22 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 22), align 2
+ %a23 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 23), align 2
+ %a24 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24), align 2
+ %a25 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 25), align 2
+ %a26 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 26), align 2
+ %a27 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 27), align 2
+ %a28 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 28), align 2
+ %a29 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 29), align 2
+ %a30 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 30), align 2
+ %a31 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 31), align 2
+ %b0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 0 ), align 2
+ %b1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 1 ), align 2
+ %b2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 2 ), align 2
+ %b3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 3 ), align 2
+ %b4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 4 ), align 2
+ %b5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 5 ), align 2
+ %b6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 6 ), align 2
+ %b7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 7 ), align 2
+ %b8 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8 ), align 2
+ %b9 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 9 ), align 2
+ %b10 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 10), align 2
+ %b11 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 11), align 2
+ %b12 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 12), align 2
+ %b13 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 13), align 2
+ %b14 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 14), align 2
+ %b15 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 15), align 2
+ %b16 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16), align 2
+ %b17 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 17), align 2
+ %b18 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 18), align 2
+ %b19 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 19), align 2
+ %b20 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 20), align 2
+ %b21 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 21), align 2
+ %b22 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 22), align 2
+ %b23 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 23), align 2
+ %b24 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24), align 2
+ %b25 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 25), align 2
+ %b26 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 26), align 2
+ %b27 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 27), align 2
+ %b28 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 28), align 2
+ %b29 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 29), align 2
+ %b30 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 30), align 2
+ %b31 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 31), align 2
+ %r0 = call i16 @llvm.ssub.sat.i16(i16 %a0 , i16 %b0 )
+ %r1 = call i16 @llvm.ssub.sat.i16(i16 %a1 , i16 %b1 )
+ %r2 = call i16 @llvm.ssub.sat.i16(i16 %a2 , i16 %b2 )
+ %r3 = call i16 @llvm.ssub.sat.i16(i16 %a3 , i16 %b3 )
+ %r4 = call i16 @llvm.ssub.sat.i16(i16 %a4 , i16 %b4 )
+ %r5 = call i16 @llvm.ssub.sat.i16(i16 %a5 , i16 %b5 )
+ %r6 = call i16 @llvm.ssub.sat.i16(i16 %a6 , i16 %b6 )
+ %r7 = call i16 @llvm.ssub.sat.i16(i16 %a7 , i16 %b7 )
+ %r8 = call i16 @llvm.ssub.sat.i16(i16 %a8 , i16 %b8 )
+ %r9 = call i16 @llvm.ssub.sat.i16(i16 %a9 , i16 %b9 )
+ %r10 = call i16 @llvm.ssub.sat.i16(i16 %a10, i16 %b10)
+ %r11 = call i16 @llvm.ssub.sat.i16(i16 %a11, i16 %b11)
+ %r12 = call i16 @llvm.ssub.sat.i16(i16 %a12, i16 %b12)
+ %r13 = call i16 @llvm.ssub.sat.i16(i16 %a13, i16 %b13)
+ %r14 = call i16 @llvm.ssub.sat.i16(i16 %a14, i16 %b14)
+ %r15 = call i16 @llvm.ssub.sat.i16(i16 %a15, i16 %b15)
+ %r16 = call i16 @llvm.ssub.sat.i16(i16 %a16, i16 %b16)
+ %r17 = call i16 @llvm.ssub.sat.i16(i16 %a17, i16 %b17)
+ %r18 = call i16 @llvm.ssub.sat.i16(i16 %a18, i16 %b18)
+ %r19 = call i16 @llvm.ssub.sat.i16(i16 %a19, i16 %b19)
+ %r20 = call i16 @llvm.ssub.sat.i16(i16 %a20, i16 %b20)
+ %r21 = call i16 @llvm.ssub.sat.i16(i16 %a21, i16 %b21)
+ %r22 = call i16 @llvm.ssub.sat.i16(i16 %a22, i16 %b22)
+ %r23 = call i16 @llvm.ssub.sat.i16(i16 %a23, i16 %b23)
+ %r24 = call i16 @llvm.ssub.sat.i16(i16 %a24, i16 %b24)
+ %r25 = call i16 @llvm.ssub.sat.i16(i16 %a25, i16 %b25)
+ %r26 = call i16 @llvm.ssub.sat.i16(i16 %a26, i16 %b26)
+ %r27 = call i16 @llvm.ssub.sat.i16(i16 %a27, i16 %b27)
+ %r28 = call i16 @llvm.ssub.sat.i16(i16 %a28, i16 %b28)
+ %r29 = call i16 @llvm.ssub.sat.i16(i16 %a29, i16 %b29)
+ %r30 = call i16 @llvm.ssub.sat.i16(i16 %a30, i16 %b30)
+ %r31 = call i16 @llvm.ssub.sat.i16(i16 %a31, i16 %b31)
+ store i16 %r0 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 0 ), align 2
+ store i16 %r1 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 1 ), align 2
+ store i16 %r2 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 2 ), align 2
+ store i16 %r3 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 3 ), align 2
+ store i16 %r4 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 4 ), align 2
+ store i16 %r5 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 5 ), align 2
+ store i16 %r6 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 6 ), align 2
+ store i16 %r7 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 7 ), align 2
+ store i16 %r8 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8 ), align 2
+ store i16 %r9 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 9 ), align 2
+ store i16 %r10, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 10), align 2
+ store i16 %r11, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 11), align 2
+ store i16 %r12, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 12), align 2
+ store i16 %r13, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 13), align 2
+ store i16 %r14, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 14), align 2
+ store i16 %r15, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 15), align 2
+ store i16 %r16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16), align 2
+ store i16 %r17, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 17), align 2
+ store i16 %r18, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 18), align 2
+ store i16 %r19, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 19), align 2
+ store i16 %r20, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 20), align 2
+ store i16 %r21, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 21), align 2
+ store i16 %r22, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 22), align 2
+ store i16 %r23, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 23), align 2
+ store i16 %r24, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24), align 2
+ store i16 %r25, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 25), align 2
+ store i16 %r26, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 26), align 2
+ store i16 %r27, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 27), align 2
+ store i16 %r28, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 28), align 2
+ store i16 %r29, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 29), align 2
+ store i16 %r30, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 30), align 2
+ store i16 %r31, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 31), align 2
+ ret void
+}
+
+define void @sub_v64i8() {
+; CHECK-LABEL: @sub_v64i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([64 x i8]* @a8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 32) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 48) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([64 x i8]* @b8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP6:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP7:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 32) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP8:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 48) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP9:%.*]] = call <16 x i8> @llvm.ssub.sat.v16i8(<16 x i8> [[TMP1]], <16 x i8> [[TMP5]])
+; CHECK-NEXT: [[TMP10:%.*]] = call <16 x i8> @llvm.ssub.sat.v16i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP6]])
+; CHECK-NEXT: [[TMP11:%.*]] = call <16 x i8> @llvm.ssub.sat.v16i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP7]])
+; CHECK-NEXT: [[TMP12:%.*]] = call <16 x i8> @llvm.ssub.sat.v16i8(<16 x i8> [[TMP4]], <16 x i8> [[TMP8]])
+; CHECK-NEXT: store <16 x i8> [[TMP9]], <16 x i8>* bitcast ([64 x i8]* @c8 to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP10]], <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP11]], <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 32) to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP12]], <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 48) to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %a0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 0 ), align 1
+ %a1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 1 ), align 1
+ %a2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 2 ), align 1
+ %a3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 3 ), align 1
+ %a4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 4 ), align 1
+ %a5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 5 ), align 1
+ %a6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 6 ), align 1
+ %a7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 7 ), align 1
+ %a8 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 8 ), align 1
+ %a9 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 9 ), align 1
+ %a10 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 10), align 1
+ %a11 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 11), align 1
+ %a12 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 12), align 1
+ %a13 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 13), align 1
+ %a14 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 14), align 1
+ %a15 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 15), align 1
+ %a16 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 16), align 1
+ %a17 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 17), align 1
+ %a18 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 18), align 1
+ %a19 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 19), align 1
+ %a20 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 20), align 1
+ %a21 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 21), align 1
+ %a22 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 22), align 1
+ %a23 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 23), align 1
+ %a24 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 24), align 1
+ %a25 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 25), align 1
+ %a26 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 26), align 1
+ %a27 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 27), align 1
+ %a28 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 28), align 1
+ %a29 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 29), align 1
+ %a30 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 30), align 1
+ %a31 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 31), align 1
+ %a32 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 32), align 1
+ %a33 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 33), align 1
+ %a34 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 34), align 1
+ %a35 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 35), align 1
+ %a36 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 36), align 1
+ %a37 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 37), align 1
+ %a38 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 38), align 1
+ %a39 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 39), align 1
+ %a40 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 40), align 1
+ %a41 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 41), align 1
+ %a42 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 42), align 1
+ %a43 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 43), align 1
+ %a44 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 44), align 1
+ %a45 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 45), align 1
+ %a46 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 46), align 1
+ %a47 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 47), align 1
+ %a48 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 48), align 1
+ %a49 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 49), align 1
+ %a50 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 50), align 1
+ %a51 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 51), align 1
+ %a52 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 52), align 1
+ %a53 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 53), align 1
+ %a54 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 54), align 1
+ %a55 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 55), align 1
+ %a56 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 56), align 1
+ %a57 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 57), align 1
+ %a58 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 58), align 1
+ %a59 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 59), align 1
+ %a60 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 60), align 1
+ %a61 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 61), align 1
+ %a62 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 62), align 1
+ %a63 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 63), align 1
+ %b0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 0 ), align 1
+ %b1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 1 ), align 1
+ %b2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 2 ), align 1
+ %b3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 3 ), align 1
+ %b4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 4 ), align 1
+ %b5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 5 ), align 1
+ %b6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 6 ), align 1
+ %b7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 7 ), align 1
+ %b8 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 8 ), align 1
+ %b9 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 9 ), align 1
+ %b10 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 10), align 1
+ %b11 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 11), align 1
+ %b12 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 12), align 1
+ %b13 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 13), align 1
+ %b14 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 14), align 1
+ %b15 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 15), align 1
+ %b16 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 16), align 1
+ %b17 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 17), align 1
+ %b18 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 18), align 1
+ %b19 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 19), align 1
+ %b20 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 20), align 1
+ %b21 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 21), align 1
+ %b22 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 22), align 1
+ %b23 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 23), align 1
+ %b24 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 24), align 1
+ %b25 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 25), align 1
+ %b26 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 26), align 1
+ %b27 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 27), align 1
+ %b28 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 28), align 1
+ %b29 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 29), align 1
+ %b30 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 30), align 1
+ %b31 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 31), align 1
+ %b32 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 32), align 1
+ %b33 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 33), align 1
+ %b34 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 34), align 1
+ %b35 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 35), align 1
+ %b36 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 36), align 1
+ %b37 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 37), align 1
+ %b38 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 38), align 1
+ %b39 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 39), align 1
+ %b40 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 40), align 1
+ %b41 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 41), align 1
+ %b42 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 42), align 1
+ %b43 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 43), align 1
+ %b44 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 44), align 1
+ %b45 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 45), align 1
+ %b46 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 46), align 1
+ %b47 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 47), align 1
+ %b48 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 48), align 1
+ %b49 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 49), align 1
+ %b50 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 50), align 1
+ %b51 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 51), align 1
+ %b52 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 52), align 1
+ %b53 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 53), align 1
+ %b54 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 54), align 1
+ %b55 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 55), align 1
+ %b56 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 56), align 1
+ %b57 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 57), align 1
+ %b58 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 58), align 1
+ %b59 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 59), align 1
+ %b60 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 60), align 1
+ %b61 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 61), align 1
+ %b62 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 62), align 1
+ %b63 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 63), align 1
+ %r0 = call i8 @llvm.ssub.sat.i8(i8 %a0 , i8 %b0 )
+ %r1 = call i8 @llvm.ssub.sat.i8(i8 %a1 , i8 %b1 )
+ %r2 = call i8 @llvm.ssub.sat.i8(i8 %a2 , i8 %b2 )
+ %r3 = call i8 @llvm.ssub.sat.i8(i8 %a3 , i8 %b3 )
+ %r4 = call i8 @llvm.ssub.sat.i8(i8 %a4 , i8 %b4 )
+ %r5 = call i8 @llvm.ssub.sat.i8(i8 %a5 , i8 %b5 )
+ %r6 = call i8 @llvm.ssub.sat.i8(i8 %a6 , i8 %b6 )
+ %r7 = call i8 @llvm.ssub.sat.i8(i8 %a7 , i8 %b7 )
+ %r8 = call i8 @llvm.ssub.sat.i8(i8 %a8 , i8 %b8 )
+ %r9 = call i8 @llvm.ssub.sat.i8(i8 %a9 , i8 %b9 )
+ %r10 = call i8 @llvm.ssub.sat.i8(i8 %a10, i8 %b10)
+ %r11 = call i8 @llvm.ssub.sat.i8(i8 %a11, i8 %b11)
+ %r12 = call i8 @llvm.ssub.sat.i8(i8 %a12, i8 %b12)
+ %r13 = call i8 @llvm.ssub.sat.i8(i8 %a13, i8 %b13)
+ %r14 = call i8 @llvm.ssub.sat.i8(i8 %a14, i8 %b14)
+ %r15 = call i8 @llvm.ssub.sat.i8(i8 %a15, i8 %b15)
+ %r16 = call i8 @llvm.ssub.sat.i8(i8 %a16, i8 %b16)
+ %r17 = call i8 @llvm.ssub.sat.i8(i8 %a17, i8 %b17)
+ %r18 = call i8 @llvm.ssub.sat.i8(i8 %a18, i8 %b18)
+ %r19 = call i8 @llvm.ssub.sat.i8(i8 %a19, i8 %b19)
+ %r20 = call i8 @llvm.ssub.sat.i8(i8 %a20, i8 %b20)
+ %r21 = call i8 @llvm.ssub.sat.i8(i8 %a21, i8 %b21)
+ %r22 = call i8 @llvm.ssub.sat.i8(i8 %a22, i8 %b22)
+ %r23 = call i8 @llvm.ssub.sat.i8(i8 %a23, i8 %b23)
+ %r24 = call i8 @llvm.ssub.sat.i8(i8 %a24, i8 %b24)
+ %r25 = call i8 @llvm.ssub.sat.i8(i8 %a25, i8 %b25)
+ %r26 = call i8 @llvm.ssub.sat.i8(i8 %a26, i8 %b26)
+ %r27 = call i8 @llvm.ssub.sat.i8(i8 %a27, i8 %b27)
+ %r28 = call i8 @llvm.ssub.sat.i8(i8 %a28, i8 %b28)
+ %r29 = call i8 @llvm.ssub.sat.i8(i8 %a29, i8 %b29)
+ %r30 = call i8 @llvm.ssub.sat.i8(i8 %a30, i8 %b30)
+ %r31 = call i8 @llvm.ssub.sat.i8(i8 %a31, i8 %b31)
+ %r32 = call i8 @llvm.ssub.sat.i8(i8 %a32, i8 %b32)
+ %r33 = call i8 @llvm.ssub.sat.i8(i8 %a33, i8 %b33)
+ %r34 = call i8 @llvm.ssub.sat.i8(i8 %a34, i8 %b34)
+ %r35 = call i8 @llvm.ssub.sat.i8(i8 %a35, i8 %b35)
+ %r36 = call i8 @llvm.ssub.sat.i8(i8 %a36, i8 %b36)
+ %r37 = call i8 @llvm.ssub.sat.i8(i8 %a37, i8 %b37)
+ %r38 = call i8 @llvm.ssub.sat.i8(i8 %a38, i8 %b38)
+ %r39 = call i8 @llvm.ssub.sat.i8(i8 %a39, i8 %b39)
+ %r40 = call i8 @llvm.ssub.sat.i8(i8 %a40, i8 %b40)
+ %r41 = call i8 @llvm.ssub.sat.i8(i8 %a41, i8 %b41)
+ %r42 = call i8 @llvm.ssub.sat.i8(i8 %a42, i8 %b42)
+ %r43 = call i8 @llvm.ssub.sat.i8(i8 %a43, i8 %b43)
+ %r44 = call i8 @llvm.ssub.sat.i8(i8 %a44, i8 %b44)
+ %r45 = call i8 @llvm.ssub.sat.i8(i8 %a45, i8 %b45)
+ %r46 = call i8 @llvm.ssub.sat.i8(i8 %a46, i8 %b46)
+ %r47 = call i8 @llvm.ssub.sat.i8(i8 %a47, i8 %b47)
+ %r48 = call i8 @llvm.ssub.sat.i8(i8 %a48, i8 %b48)
+ %r49 = call i8 @llvm.ssub.sat.i8(i8 %a49, i8 %b49)
+ %r50 = call i8 @llvm.ssub.sat.i8(i8 %a50, i8 %b50)
+ %r51 = call i8 @llvm.ssub.sat.i8(i8 %a51, i8 %b51)
+ %r52 = call i8 @llvm.ssub.sat.i8(i8 %a52, i8 %b52)
+ %r53 = call i8 @llvm.ssub.sat.i8(i8 %a53, i8 %b53)
+ %r54 = call i8 @llvm.ssub.sat.i8(i8 %a54, i8 %b54)
+ %r55 = call i8 @llvm.ssub.sat.i8(i8 %a55, i8 %b55)
+ %r56 = call i8 @llvm.ssub.sat.i8(i8 %a56, i8 %b56)
+ %r57 = call i8 @llvm.ssub.sat.i8(i8 %a57, i8 %b57)
+ %r58 = call i8 @llvm.ssub.sat.i8(i8 %a58, i8 %b58)
+ %r59 = call i8 @llvm.ssub.sat.i8(i8 %a59, i8 %b59)
+ %r60 = call i8 @llvm.ssub.sat.i8(i8 %a60, i8 %b60)
+ %r61 = call i8 @llvm.ssub.sat.i8(i8 %a61, i8 %b61)
+ %r62 = call i8 @llvm.ssub.sat.i8(i8 %a62, i8 %b62)
+ %r63 = call i8 @llvm.ssub.sat.i8(i8 %a63, i8 %b63)
+ store i8 %r0 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 0 ), align 1
+ store i8 %r1 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 1 ), align 1
+ store i8 %r2 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 2 ), align 1
+ store i8 %r3 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 3 ), align 1
+ store i8 %r4 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 4 ), align 1
+ store i8 %r5 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 5 ), align 1
+ store i8 %r6 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 6 ), align 1
+ store i8 %r7 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 7 ), align 1
+ store i8 %r8 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 8 ), align 1
+ store i8 %r9 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 9 ), align 1
+ store i8 %r10, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 10), align 1
+ store i8 %r11, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 11), align 1
+ store i8 %r12, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 12), align 1
+ store i8 %r13, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 13), align 1
+ store i8 %r14, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 14), align 1
+ store i8 %r15, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 15), align 1
+ store i8 %r16, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 16), align 1
+ store i8 %r17, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 17), align 1
+ store i8 %r18, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 18), align 1
+ store i8 %r19, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 19), align 1
+ store i8 %r20, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 20), align 1
+ store i8 %r21, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 21), align 1
+ store i8 %r22, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 22), align 1
+ store i8 %r23, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 23), align 1
+ store i8 %r24, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 24), align 1
+ store i8 %r25, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 25), align 1
+ store i8 %r26, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 26), align 1
+ store i8 %r27, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 27), align 1
+ store i8 %r28, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 28), align 1
+ store i8 %r29, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 29), align 1
+ store i8 %r30, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 30), align 1
+ store i8 %r31, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 31), align 1
+ store i8 %r32, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 32), align 1
+ store i8 %r33, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 33), align 1
+ store i8 %r34, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 34), align 1
+ store i8 %r35, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 35), align 1
+ store i8 %r36, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 36), align 1
+ store i8 %r37, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 37), align 1
+ store i8 %r38, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 38), align 1
+ store i8 %r39, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 39), align 1
+ store i8 %r40, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 40), align 1
+ store i8 %r41, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 41), align 1
+ store i8 %r42, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 42), align 1
+ store i8 %r43, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 43), align 1
+ store i8 %r44, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 44), align 1
+ store i8 %r45, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 45), align 1
+ store i8 %r46, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 46), align 1
+ store i8 %r47, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 47), align 1
+ store i8 %r48, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 48), align 1
+ store i8 %r49, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 49), align 1
+ store i8 %r50, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 50), align 1
+ store i8 %r51, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 51), align 1
+ store i8 %r52, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 52), align 1
+ store i8 %r53, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 53), align 1
+ store i8 %r54, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 54), align 1
+ store i8 %r55, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 55), align 1
+ store i8 %r56, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 56), align 1
+ store i8 %r57, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 57), align 1
+ store i8 %r58, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 58), align 1
+ store i8 %r59, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 59), align 1
+ store i8 %r60, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 60), align 1
+ store i8 %r61, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 61), align 1
+ store i8 %r62, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 62), align 1
+ store i8 %r63, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 63), align 1
+ ret void
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-ssubo.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-ssubo.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-ssubo.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-ssubo.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,1254 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,SSE
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=slm -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,SLM
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=knl -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skx -mattr=-prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skx -mattr=+prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX256BW
+
+ at a64 = common global [8 x i64] zeroinitializer, align 64
+ at b64 = common global [8 x i64] zeroinitializer, align 64
+ at c64 = common global [8 x i64] zeroinitializer, align 64
+ at a32 = common global [16 x i32] zeroinitializer, align 64
+ at b32 = common global [16 x i32] zeroinitializer, align 64
+ at c32 = common global [16 x i32] zeroinitializer, align 64
+ at a16 = common global [32 x i16] zeroinitializer, align 64
+ at b16 = common global [32 x i16] zeroinitializer, align 64
+ at c16 = common global [32 x i16] zeroinitializer, align 64
+ at a8 = common global [64 x i8] zeroinitializer, align 64
+ at b8 = common global [64 x i8] zeroinitializer, align 64
+ at c8 = common global [64 x i8] zeroinitializer, align 64
+
+declare {i64, i1} @llvm.ssub.with.overflow.i64(i64, i64)
+declare {i32, i1} @llvm.ssub.with.overflow.i32(i32, i32)
+declare {i16, i1} @llvm.ssub.with.overflow.i16(i16, i16)
+declare {i8 , i1} @llvm.ssub.with.overflow.i8 (i8 , i8 )
+
+define void @sub_v8i64() {
+; CHECK-LABEL: @sub_v8i64(
+; CHECK-NEXT: [[A0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 0), align 8
+; CHECK-NEXT: [[A1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 1), align 8
+; CHECK-NEXT: [[A2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2), align 8
+; CHECK-NEXT: [[A3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 3), align 8
+; CHECK-NEXT: [[A4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4), align 8
+; CHECK-NEXT: [[A5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 5), align 8
+; CHECK-NEXT: [[A6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6), align 8
+; CHECK-NEXT: [[A7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 7), align 8
+; CHECK-NEXT: [[B0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 0), align 8
+; CHECK-NEXT: [[B1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 1), align 8
+; CHECK-NEXT: [[B2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2), align 8
+; CHECK-NEXT: [[B3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 3), align 8
+; CHECK-NEXT: [[B4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4), align 8
+; CHECK-NEXT: [[B5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 5), align 8
+; CHECK-NEXT: [[B6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6), align 8
+; CHECK-NEXT: [[B7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 7), align 8
+; CHECK-NEXT: [[C0:%.*]] = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 [[A0]], i64 [[B0]])
+; CHECK-NEXT: [[C1:%.*]] = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 [[A1]], i64 [[B1]])
+; CHECK-NEXT: [[C2:%.*]] = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 [[A2]], i64 [[B2]])
+; CHECK-NEXT: [[C3:%.*]] = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 [[A3]], i64 [[B3]])
+; CHECK-NEXT: [[C4:%.*]] = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 [[A4]], i64 [[B4]])
+; CHECK-NEXT: [[C5:%.*]] = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 [[A5]], i64 [[B5]])
+; CHECK-NEXT: [[C6:%.*]] = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 [[A6]], i64 [[B6]])
+; CHECK-NEXT: [[C7:%.*]] = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 [[A7]], i64 [[B7]])
+; CHECK-NEXT: [[R0:%.*]] = extractvalue { i64, i1 } [[C0]], 0
+; CHECK-NEXT: [[R1:%.*]] = extractvalue { i64, i1 } [[C1]], 0
+; CHECK-NEXT: [[R2:%.*]] = extractvalue { i64, i1 } [[C2]], 0
+; CHECK-NEXT: [[R3:%.*]] = extractvalue { i64, i1 } [[C3]], 0
+; CHECK-NEXT: [[R4:%.*]] = extractvalue { i64, i1 } [[C4]], 0
+; CHECK-NEXT: [[R5:%.*]] = extractvalue { i64, i1 } [[C5]], 0
+; CHECK-NEXT: [[R6:%.*]] = extractvalue { i64, i1 } [[C6]], 0
+; CHECK-NEXT: [[R7:%.*]] = extractvalue { i64, i1 } [[C7]], 0
+; CHECK-NEXT: store i64 [[R0]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 0), align 8
+; CHECK-NEXT: store i64 [[R1]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 1), align 8
+; CHECK-NEXT: store i64 [[R2]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2), align 8
+; CHECK-NEXT: store i64 [[R3]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 3), align 8
+; CHECK-NEXT: store i64 [[R4]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4), align 8
+; CHECK-NEXT: store i64 [[R5]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 5), align 8
+; CHECK-NEXT: store i64 [[R6]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6), align 8
+; CHECK-NEXT: store i64 [[R7]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 7), align 8
+; CHECK-NEXT: ret void
+;
+ %a0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 0), align 8
+ %a1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 1), align 8
+ %a2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2), align 8
+ %a3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 3), align 8
+ %a4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4), align 8
+ %a5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 5), align 8
+ %a6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6), align 8
+ %a7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 7), align 8
+ %b0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 0), align 8
+ %b1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 1), align 8
+ %b2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2), align 8
+ %b3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 3), align 8
+ %b4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4), align 8
+ %b5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 5), align 8
+ %b6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6), align 8
+ %b7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 7), align 8
+ %c0 = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a0, i64 %b0)
+ %c1 = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a1, i64 %b1)
+ %c2 = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a2, i64 %b2)
+ %c3 = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a3, i64 %b3)
+ %c4 = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a4, i64 %b4)
+ %c5 = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a5, i64 %b5)
+ %c6 = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a6, i64 %b6)
+ %c7 = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a7, i64 %b7)
+ %r0 = extractvalue {i64, i1} %c0, 0
+ %r1 = extractvalue {i64, i1} %c1, 0
+ %r2 = extractvalue {i64, i1} %c2, 0
+ %r3 = extractvalue {i64, i1} %c3, 0
+ %r4 = extractvalue {i64, i1} %c4, 0
+ %r5 = extractvalue {i64, i1} %c5, 0
+ %r6 = extractvalue {i64, i1} %c6, 0
+ %r7 = extractvalue {i64, i1} %c7, 0
+ store i64 %r0, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 0), align 8
+ store i64 %r1, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 1), align 8
+ store i64 %r2, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2), align 8
+ store i64 %r3, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 3), align 8
+ store i64 %r4, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4), align 8
+ store i64 %r5, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 5), align 8
+ store i64 %r6, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6), align 8
+ store i64 %r7, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 7), align 8
+ ret void
+}
+
+define void @sub_v16i32() {
+; CHECK-LABEL: @sub_v16i32(
+; CHECK-NEXT: [[A0:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 0), align 4
+; CHECK-NEXT: [[A1:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 1), align 4
+; CHECK-NEXT: [[A2:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 2), align 4
+; CHECK-NEXT: [[A3:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 3), align 4
+; CHECK-NEXT: [[A4:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4), align 4
+; CHECK-NEXT: [[A5:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 5), align 4
+; CHECK-NEXT: [[A6:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 6), align 4
+; CHECK-NEXT: [[A7:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 7), align 4
+; CHECK-NEXT: [[A8:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8), align 4
+; CHECK-NEXT: [[A9:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 9), align 4
+; CHECK-NEXT: [[A10:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 10), align 4
+; CHECK-NEXT: [[A11:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 11), align 4
+; CHECK-NEXT: [[A12:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12), align 4
+; CHECK-NEXT: [[A13:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 13), align 4
+; CHECK-NEXT: [[A14:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 14), align 4
+; CHECK-NEXT: [[A15:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 15), align 4
+; CHECK-NEXT: [[B0:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 0), align 4
+; CHECK-NEXT: [[B1:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 1), align 4
+; CHECK-NEXT: [[B2:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 2), align 4
+; CHECK-NEXT: [[B3:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 3), align 4
+; CHECK-NEXT: [[B4:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4), align 4
+; CHECK-NEXT: [[B5:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 5), align 4
+; CHECK-NEXT: [[B6:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 6), align 4
+; CHECK-NEXT: [[B7:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 7), align 4
+; CHECK-NEXT: [[B8:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8), align 4
+; CHECK-NEXT: [[B9:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 9), align 4
+; CHECK-NEXT: [[B10:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 10), align 4
+; CHECK-NEXT: [[B11:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 11), align 4
+; CHECK-NEXT: [[B12:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12), align 4
+; CHECK-NEXT: [[B13:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 13), align 4
+; CHECK-NEXT: [[B14:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 14), align 4
+; CHECK-NEXT: [[B15:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 15), align 4
+; CHECK-NEXT: [[C0:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A0]], i32 [[B0]])
+; CHECK-NEXT: [[C1:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A1]], i32 [[B1]])
+; CHECK-NEXT: [[C2:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A2]], i32 [[B2]])
+; CHECK-NEXT: [[C3:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A3]], i32 [[B3]])
+; CHECK-NEXT: [[C4:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A4]], i32 [[B4]])
+; CHECK-NEXT: [[C5:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A5]], i32 [[B5]])
+; CHECK-NEXT: [[C6:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A6]], i32 [[B6]])
+; CHECK-NEXT: [[C7:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A7]], i32 [[B7]])
+; CHECK-NEXT: [[C8:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A8]], i32 [[B8]])
+; CHECK-NEXT: [[C9:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A9]], i32 [[B9]])
+; CHECK-NEXT: [[C10:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A10]], i32 [[B10]])
+; CHECK-NEXT: [[C11:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A11]], i32 [[B11]])
+; CHECK-NEXT: [[C12:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A12]], i32 [[B12]])
+; CHECK-NEXT: [[C13:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A13]], i32 [[B13]])
+; CHECK-NEXT: [[C14:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A14]], i32 [[B14]])
+; CHECK-NEXT: [[C15:%.*]] = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 [[A15]], i32 [[B15]])
+; CHECK-NEXT: [[R0:%.*]] = extractvalue { i32, i1 } [[C0]], 0
+; CHECK-NEXT: [[R1:%.*]] = extractvalue { i32, i1 } [[C1]], 0
+; CHECK-NEXT: [[R2:%.*]] = extractvalue { i32, i1 } [[C2]], 0
+; CHECK-NEXT: [[R3:%.*]] = extractvalue { i32, i1 } [[C3]], 0
+; CHECK-NEXT: [[R4:%.*]] = extractvalue { i32, i1 } [[C4]], 0
+; CHECK-NEXT: [[R5:%.*]] = extractvalue { i32, i1 } [[C5]], 0
+; CHECK-NEXT: [[R6:%.*]] = extractvalue { i32, i1 } [[C6]], 0
+; CHECK-NEXT: [[R7:%.*]] = extractvalue { i32, i1 } [[C7]], 0
+; CHECK-NEXT: [[R8:%.*]] = extractvalue { i32, i1 } [[C8]], 0
+; CHECK-NEXT: [[R9:%.*]] = extractvalue { i32, i1 } [[C9]], 0
+; CHECK-NEXT: [[R10:%.*]] = extractvalue { i32, i1 } [[C10]], 0
+; CHECK-NEXT: [[R11:%.*]] = extractvalue { i32, i1 } [[C11]], 0
+; CHECK-NEXT: [[R12:%.*]] = extractvalue { i32, i1 } [[C12]], 0
+; CHECK-NEXT: [[R13:%.*]] = extractvalue { i32, i1 } [[C13]], 0
+; CHECK-NEXT: [[R14:%.*]] = extractvalue { i32, i1 } [[C14]], 0
+; CHECK-NEXT: [[R15:%.*]] = extractvalue { i32, i1 } [[C15]], 0
+; CHECK-NEXT: store i32 [[R0]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 0), align 4
+; CHECK-NEXT: store i32 [[R1]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 1), align 4
+; CHECK-NEXT: store i32 [[R2]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 2), align 4
+; CHECK-NEXT: store i32 [[R3]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 3), align 4
+; CHECK-NEXT: store i32 [[R4]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4), align 4
+; CHECK-NEXT: store i32 [[R5]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 5), align 4
+; CHECK-NEXT: store i32 [[R6]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 6), align 4
+; CHECK-NEXT: store i32 [[R7]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 7), align 4
+; CHECK-NEXT: store i32 [[R8]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8), align 4
+; CHECK-NEXT: store i32 [[R9]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 9), align 4
+; CHECK-NEXT: store i32 [[R10]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 10), align 4
+; CHECK-NEXT: store i32 [[R11]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 11), align 4
+; CHECK-NEXT: store i32 [[R12]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12), align 4
+; CHECK-NEXT: store i32 [[R13]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 13), align 4
+; CHECK-NEXT: store i32 [[R14]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 14), align 4
+; CHECK-NEXT: store i32 [[R15]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 15), align 4
+; CHECK-NEXT: ret void
+;
+ %a0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 0 ), align 4
+ %a1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 1 ), align 4
+ %a2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 2 ), align 4
+ %a3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 3 ), align 4
+ %a4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4 ), align 4
+ %a5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 5 ), align 4
+ %a6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 6 ), align 4
+ %a7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 7 ), align 4
+ %a8 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8 ), align 4
+ %a9 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 9 ), align 4
+ %a10 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 10), align 4
+ %a11 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 11), align 4
+ %a12 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12), align 4
+ %a13 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 13), align 4
+ %a14 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 14), align 4
+ %a15 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 15), align 4
+ %b0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 0 ), align 4
+ %b1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 1 ), align 4
+ %b2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 2 ), align 4
+ %b3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 3 ), align 4
+ %b4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4 ), align 4
+ %b5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 5 ), align 4
+ %b6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 6 ), align 4
+ %b7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 7 ), align 4
+ %b8 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8 ), align 4
+ %b9 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 9 ), align 4
+ %b10 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 10), align 4
+ %b11 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 11), align 4
+ %b12 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12), align 4
+ %b13 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 13), align 4
+ %b14 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 14), align 4
+ %b15 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 15), align 4
+ %c0 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a0 , i32 %b0 )
+ %c1 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a1 , i32 %b1 )
+ %c2 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a2 , i32 %b2 )
+ %c3 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a3 , i32 %b3 )
+ %c4 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a4 , i32 %b4 )
+ %c5 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a5 , i32 %b5 )
+ %c6 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a6 , i32 %b6 )
+ %c7 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a7 , i32 %b7 )
+ %c8 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a8 , i32 %b8 )
+ %c9 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a9 , i32 %b9 )
+ %c10 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a10, i32 %b10)
+ %c11 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a11, i32 %b11)
+ %c12 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a12, i32 %b12)
+ %c13 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a13, i32 %b13)
+ %c14 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a14, i32 %b14)
+ %c15 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a15, i32 %b15)
+ %r0 = extractvalue {i32, i1} %c0 , 0
+ %r1 = extractvalue {i32, i1} %c1 , 0
+ %r2 = extractvalue {i32, i1} %c2 , 0
+ %r3 = extractvalue {i32, i1} %c3 , 0
+ %r4 = extractvalue {i32, i1} %c4 , 0
+ %r5 = extractvalue {i32, i1} %c5 , 0
+ %r6 = extractvalue {i32, i1} %c6 , 0
+ %r7 = extractvalue {i32, i1} %c7 , 0
+ %r8 = extractvalue {i32, i1} %c8 , 0
+ %r9 = extractvalue {i32, i1} %c9 , 0
+ %r10 = extractvalue {i32, i1} %c10, 0
+ %r11 = extractvalue {i32, i1} %c11, 0
+ %r12 = extractvalue {i32, i1} %c12, 0
+ %r13 = extractvalue {i32, i1} %c13, 0
+ %r14 = extractvalue {i32, i1} %c14, 0
+ %r15 = extractvalue {i32, i1} %c15, 0
+ store i32 %r0 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 0 ), align 4
+ store i32 %r1 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 1 ), align 4
+ store i32 %r2 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 2 ), align 4
+ store i32 %r3 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 3 ), align 4
+ store i32 %r4 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4 ), align 4
+ store i32 %r5 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 5 ), align 4
+ store i32 %r6 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 6 ), align 4
+ store i32 %r7 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 7 ), align 4
+ store i32 %r8 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8 ), align 4
+ store i32 %r9 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 9 ), align 4
+ store i32 %r10, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 10), align 4
+ store i32 %r11, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 11), align 4
+ store i32 %r12, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12), align 4
+ store i32 %r13, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 13), align 4
+ store i32 %r14, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 14), align 4
+ store i32 %r15, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 15), align 4
+ ret void
+}
+
+define void @sub_v32i16() {
+; CHECK-LABEL: @sub_v32i16(
+; CHECK-NEXT: [[A0:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 0), align 2
+; CHECK-NEXT: [[A1:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 1), align 2
+; CHECK-NEXT: [[A2:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 2), align 2
+; CHECK-NEXT: [[A3:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 3), align 2
+; CHECK-NEXT: [[A4:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 4), align 2
+; CHECK-NEXT: [[A5:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 5), align 2
+; CHECK-NEXT: [[A6:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 6), align 2
+; CHECK-NEXT: [[A7:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 7), align 2
+; CHECK-NEXT: [[A8:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8), align 2
+; CHECK-NEXT: [[A9:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 9), align 2
+; CHECK-NEXT: [[A10:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 10), align 2
+; CHECK-NEXT: [[A11:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 11), align 2
+; CHECK-NEXT: [[A12:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 12), align 2
+; CHECK-NEXT: [[A13:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 13), align 2
+; CHECK-NEXT: [[A14:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 14), align 2
+; CHECK-NEXT: [[A15:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 15), align 2
+; CHECK-NEXT: [[A16:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16), align 2
+; CHECK-NEXT: [[A17:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 17), align 2
+; CHECK-NEXT: [[A18:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 18), align 2
+; CHECK-NEXT: [[A19:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 19), align 2
+; CHECK-NEXT: [[A20:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 20), align 2
+; CHECK-NEXT: [[A21:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 21), align 2
+; CHECK-NEXT: [[A22:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 22), align 2
+; CHECK-NEXT: [[A23:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 23), align 2
+; CHECK-NEXT: [[A24:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24), align 2
+; CHECK-NEXT: [[A25:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 25), align 2
+; CHECK-NEXT: [[A26:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 26), align 2
+; CHECK-NEXT: [[A27:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 27), align 2
+; CHECK-NEXT: [[A28:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 28), align 2
+; CHECK-NEXT: [[A29:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 29), align 2
+; CHECK-NEXT: [[A30:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 30), align 2
+; CHECK-NEXT: [[A31:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 31), align 2
+; CHECK-NEXT: [[B0:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 0), align 2
+; CHECK-NEXT: [[B1:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 1), align 2
+; CHECK-NEXT: [[B2:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 2), align 2
+; CHECK-NEXT: [[B3:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 3), align 2
+; CHECK-NEXT: [[B4:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 4), align 2
+; CHECK-NEXT: [[B5:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 5), align 2
+; CHECK-NEXT: [[B6:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 6), align 2
+; CHECK-NEXT: [[B7:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 7), align 2
+; CHECK-NEXT: [[B8:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8), align 2
+; CHECK-NEXT: [[B9:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 9), align 2
+; CHECK-NEXT: [[B10:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 10), align 2
+; CHECK-NEXT: [[B11:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 11), align 2
+; CHECK-NEXT: [[B12:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 12), align 2
+; CHECK-NEXT: [[B13:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 13), align 2
+; CHECK-NEXT: [[B14:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 14), align 2
+; CHECK-NEXT: [[B15:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 15), align 2
+; CHECK-NEXT: [[B16:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16), align 2
+; CHECK-NEXT: [[B17:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 17), align 2
+; CHECK-NEXT: [[B18:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 18), align 2
+; CHECK-NEXT: [[B19:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 19), align 2
+; CHECK-NEXT: [[B20:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 20), align 2
+; CHECK-NEXT: [[B21:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 21), align 2
+; CHECK-NEXT: [[B22:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 22), align 2
+; CHECK-NEXT: [[B23:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 23), align 2
+; CHECK-NEXT: [[B24:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24), align 2
+; CHECK-NEXT: [[B25:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 25), align 2
+; CHECK-NEXT: [[B26:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 26), align 2
+; CHECK-NEXT: [[B27:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 27), align 2
+; CHECK-NEXT: [[B28:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 28), align 2
+; CHECK-NEXT: [[B29:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 29), align 2
+; CHECK-NEXT: [[B30:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 30), align 2
+; CHECK-NEXT: [[B31:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 31), align 2
+; CHECK-NEXT: [[C0:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A0]], i16 [[B0]])
+; CHECK-NEXT: [[C1:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A1]], i16 [[B1]])
+; CHECK-NEXT: [[C2:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A2]], i16 [[B2]])
+; CHECK-NEXT: [[C3:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A3]], i16 [[B3]])
+; CHECK-NEXT: [[C4:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A4]], i16 [[B4]])
+; CHECK-NEXT: [[C5:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A5]], i16 [[B5]])
+; CHECK-NEXT: [[C6:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A6]], i16 [[B6]])
+; CHECK-NEXT: [[C7:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A7]], i16 [[B7]])
+; CHECK-NEXT: [[C8:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A8]], i16 [[B8]])
+; CHECK-NEXT: [[C9:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A9]], i16 [[B9]])
+; CHECK-NEXT: [[C10:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A10]], i16 [[B10]])
+; CHECK-NEXT: [[C11:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A11]], i16 [[B11]])
+; CHECK-NEXT: [[C12:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A12]], i16 [[B12]])
+; CHECK-NEXT: [[C13:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A13]], i16 [[B13]])
+; CHECK-NEXT: [[C14:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A14]], i16 [[B14]])
+; CHECK-NEXT: [[C15:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A15]], i16 [[B15]])
+; CHECK-NEXT: [[C16:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A16]], i16 [[B16]])
+; CHECK-NEXT: [[C17:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A17]], i16 [[B17]])
+; CHECK-NEXT: [[C18:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A18]], i16 [[B18]])
+; CHECK-NEXT: [[C19:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A19]], i16 [[B19]])
+; CHECK-NEXT: [[C20:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A20]], i16 [[B20]])
+; CHECK-NEXT: [[C21:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A21]], i16 [[B21]])
+; CHECK-NEXT: [[C22:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A22]], i16 [[B22]])
+; CHECK-NEXT: [[C23:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A23]], i16 [[B23]])
+; CHECK-NEXT: [[C24:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A24]], i16 [[B24]])
+; CHECK-NEXT: [[C25:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A25]], i16 [[B25]])
+; CHECK-NEXT: [[C26:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A26]], i16 [[B26]])
+; CHECK-NEXT: [[C27:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A27]], i16 [[B27]])
+; CHECK-NEXT: [[C28:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A28]], i16 [[B28]])
+; CHECK-NEXT: [[C29:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A29]], i16 [[B29]])
+; CHECK-NEXT: [[C30:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A30]], i16 [[B30]])
+; CHECK-NEXT: [[C31:%.*]] = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 [[A31]], i16 [[B31]])
+; CHECK-NEXT: [[R0:%.*]] = extractvalue { i16, i1 } [[C0]], 0
+; CHECK-NEXT: [[R1:%.*]] = extractvalue { i16, i1 } [[C1]], 0
+; CHECK-NEXT: [[R2:%.*]] = extractvalue { i16, i1 } [[C2]], 0
+; CHECK-NEXT: [[R3:%.*]] = extractvalue { i16, i1 } [[C3]], 0
+; CHECK-NEXT: [[R4:%.*]] = extractvalue { i16, i1 } [[C4]], 0
+; CHECK-NEXT: [[R5:%.*]] = extractvalue { i16, i1 } [[C5]], 0
+; CHECK-NEXT: [[R6:%.*]] = extractvalue { i16, i1 } [[C6]], 0
+; CHECK-NEXT: [[R7:%.*]] = extractvalue { i16, i1 } [[C7]], 0
+; CHECK-NEXT: [[R8:%.*]] = extractvalue { i16, i1 } [[C8]], 0
+; CHECK-NEXT: [[R9:%.*]] = extractvalue { i16, i1 } [[C9]], 0
+; CHECK-NEXT: [[R10:%.*]] = extractvalue { i16, i1 } [[C10]], 0
+; CHECK-NEXT: [[R11:%.*]] = extractvalue { i16, i1 } [[C11]], 0
+; CHECK-NEXT: [[R12:%.*]] = extractvalue { i16, i1 } [[C12]], 0
+; CHECK-NEXT: [[R13:%.*]] = extractvalue { i16, i1 } [[C13]], 0
+; CHECK-NEXT: [[R14:%.*]] = extractvalue { i16, i1 } [[C14]], 0
+; CHECK-NEXT: [[R15:%.*]] = extractvalue { i16, i1 } [[C15]], 0
+; CHECK-NEXT: [[R16:%.*]] = extractvalue { i16, i1 } [[C16]], 0
+; CHECK-NEXT: [[R17:%.*]] = extractvalue { i16, i1 } [[C17]], 0
+; CHECK-NEXT: [[R18:%.*]] = extractvalue { i16, i1 } [[C18]], 0
+; CHECK-NEXT: [[R19:%.*]] = extractvalue { i16, i1 } [[C19]], 0
+; CHECK-NEXT: [[R20:%.*]] = extractvalue { i16, i1 } [[C20]], 0
+; CHECK-NEXT: [[R21:%.*]] = extractvalue { i16, i1 } [[C21]], 0
+; CHECK-NEXT: [[R22:%.*]] = extractvalue { i16, i1 } [[C22]], 0
+; CHECK-NEXT: [[R23:%.*]] = extractvalue { i16, i1 } [[C23]], 0
+; CHECK-NEXT: [[R24:%.*]] = extractvalue { i16, i1 } [[C24]], 0
+; CHECK-NEXT: [[R25:%.*]] = extractvalue { i16, i1 } [[C25]], 0
+; CHECK-NEXT: [[R26:%.*]] = extractvalue { i16, i1 } [[C26]], 0
+; CHECK-NEXT: [[R27:%.*]] = extractvalue { i16, i1 } [[C27]], 0
+; CHECK-NEXT: [[R28:%.*]] = extractvalue { i16, i1 } [[C28]], 0
+; CHECK-NEXT: [[R29:%.*]] = extractvalue { i16, i1 } [[C29]], 0
+; CHECK-NEXT: [[R30:%.*]] = extractvalue { i16, i1 } [[C30]], 0
+; CHECK-NEXT: [[R31:%.*]] = extractvalue { i16, i1 } [[C31]], 0
+; CHECK-NEXT: store i16 [[R0]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 0), align 2
+; CHECK-NEXT: store i16 [[R1]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 1), align 2
+; CHECK-NEXT: store i16 [[R2]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 2), align 2
+; CHECK-NEXT: store i16 [[R3]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 3), align 2
+; CHECK-NEXT: store i16 [[R4]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 4), align 2
+; CHECK-NEXT: store i16 [[R5]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 5), align 2
+; CHECK-NEXT: store i16 [[R6]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 6), align 2
+; CHECK-NEXT: store i16 [[R7]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 7), align 2
+; CHECK-NEXT: store i16 [[R8]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8), align 2
+; CHECK-NEXT: store i16 [[R9]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 9), align 2
+; CHECK-NEXT: store i16 [[R10]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 10), align 2
+; CHECK-NEXT: store i16 [[R11]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 11), align 2
+; CHECK-NEXT: store i16 [[R12]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 12), align 2
+; CHECK-NEXT: store i16 [[R13]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 13), align 2
+; CHECK-NEXT: store i16 [[R14]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 14), align 2
+; CHECK-NEXT: store i16 [[R15]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 15), align 2
+; CHECK-NEXT: store i16 [[R16]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16), align 2
+; CHECK-NEXT: store i16 [[R17]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 17), align 2
+; CHECK-NEXT: store i16 [[R18]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 18), align 2
+; CHECK-NEXT: store i16 [[R19]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 19), align 2
+; CHECK-NEXT: store i16 [[R20]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 20), align 2
+; CHECK-NEXT: store i16 [[R21]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 21), align 2
+; CHECK-NEXT: store i16 [[R22]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 22), align 2
+; CHECK-NEXT: store i16 [[R23]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 23), align 2
+; CHECK-NEXT: store i16 [[R24]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24), align 2
+; CHECK-NEXT: store i16 [[R25]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 25), align 2
+; CHECK-NEXT: store i16 [[R26]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 26), align 2
+; CHECK-NEXT: store i16 [[R27]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 27), align 2
+; CHECK-NEXT: store i16 [[R28]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 28), align 2
+; CHECK-NEXT: store i16 [[R29]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 29), align 2
+; CHECK-NEXT: store i16 [[R30]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 30), align 2
+; CHECK-NEXT: store i16 [[R31]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 31), align 2
+; CHECK-NEXT: ret void
+;
+ %a0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 0 ), align 2
+ %a1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 1 ), align 2
+ %a2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 2 ), align 2
+ %a3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 3 ), align 2
+ %a4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 4 ), align 2
+ %a5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 5 ), align 2
+ %a6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 6 ), align 2
+ %a7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 7 ), align 2
+ %a8 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8 ), align 2
+ %a9 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 9 ), align 2
+ %a10 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 10), align 2
+ %a11 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 11), align 2
+ %a12 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 12), align 2
+ %a13 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 13), align 2
+ %a14 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 14), align 2
+ %a15 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 15), align 2
+ %a16 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16), align 2
+ %a17 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 17), align 2
+ %a18 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 18), align 2
+ %a19 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 19), align 2
+ %a20 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 20), align 2
+ %a21 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 21), align 2
+ %a22 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 22), align 2
+ %a23 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 23), align 2
+ %a24 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24), align 2
+ %a25 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 25), align 2
+ %a26 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 26), align 2
+ %a27 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 27), align 2
+ %a28 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 28), align 2
+ %a29 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 29), align 2
+ %a30 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 30), align 2
+ %a31 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 31), align 2
+ %b0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 0 ), align 2
+ %b1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 1 ), align 2
+ %b2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 2 ), align 2
+ %b3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 3 ), align 2
+ %b4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 4 ), align 2
+ %b5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 5 ), align 2
+ %b6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 6 ), align 2
+ %b7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 7 ), align 2
+ %b8 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8 ), align 2
+ %b9 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 9 ), align 2
+ %b10 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 10), align 2
+ %b11 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 11), align 2
+ %b12 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 12), align 2
+ %b13 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 13), align 2
+ %b14 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 14), align 2
+ %b15 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 15), align 2
+ %b16 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16), align 2
+ %b17 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 17), align 2
+ %b18 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 18), align 2
+ %b19 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 19), align 2
+ %b20 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 20), align 2
+ %b21 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 21), align 2
+ %b22 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 22), align 2
+ %b23 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 23), align 2
+ %b24 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24), align 2
+ %b25 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 25), align 2
+ %b26 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 26), align 2
+ %b27 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 27), align 2
+ %b28 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 28), align 2
+ %b29 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 29), align 2
+ %b30 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 30), align 2
+ %b31 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 31), align 2
+ %c0 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a0 , i16 %b0 )
+ %c1 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a1 , i16 %b1 )
+ %c2 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a2 , i16 %b2 )
+ %c3 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a3 , i16 %b3 )
+ %c4 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a4 , i16 %b4 )
+ %c5 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a5 , i16 %b5 )
+ %c6 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a6 , i16 %b6 )
+ %c7 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a7 , i16 %b7 )
+ %c8 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a8 , i16 %b8 )
+ %c9 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a9 , i16 %b9 )
+ %c10 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a10, i16 %b10)
+ %c11 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a11, i16 %b11)
+ %c12 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a12, i16 %b12)
+ %c13 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a13, i16 %b13)
+ %c14 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a14, i16 %b14)
+ %c15 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a15, i16 %b15)
+ %c16 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a16, i16 %b16)
+ %c17 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a17, i16 %b17)
+ %c18 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a18, i16 %b18)
+ %c19 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a19, i16 %b19)
+ %c20 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a20, i16 %b20)
+ %c21 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a21, i16 %b21)
+ %c22 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a22, i16 %b22)
+ %c23 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a23, i16 %b23)
+ %c24 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a24, i16 %b24)
+ %c25 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a25, i16 %b25)
+ %c26 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a26, i16 %b26)
+ %c27 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a27, i16 %b27)
+ %c28 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a28, i16 %b28)
+ %c29 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a29, i16 %b29)
+ %c30 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a30, i16 %b30)
+ %c31 = call {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a31, i16 %b31)
+ %r0 = extractvalue {i16, i1} %c0 , 0
+ %r1 = extractvalue {i16, i1} %c1 , 0
+ %r2 = extractvalue {i16, i1} %c2 , 0
+ %r3 = extractvalue {i16, i1} %c3 , 0
+ %r4 = extractvalue {i16, i1} %c4 , 0
+ %r5 = extractvalue {i16, i1} %c5 , 0
+ %r6 = extractvalue {i16, i1} %c6 , 0
+ %r7 = extractvalue {i16, i1} %c7 , 0
+ %r8 = extractvalue {i16, i1} %c8 , 0
+ %r9 = extractvalue {i16, i1} %c9 , 0
+ %r10 = extractvalue {i16, i1} %c10, 0
+ %r11 = extractvalue {i16, i1} %c11, 0
+ %r12 = extractvalue {i16, i1} %c12, 0
+ %r13 = extractvalue {i16, i1} %c13, 0
+ %r14 = extractvalue {i16, i1} %c14, 0
+ %r15 = extractvalue {i16, i1} %c15, 0
+ %r16 = extractvalue {i16, i1} %c16, 0
+ %r17 = extractvalue {i16, i1} %c17, 0
+ %r18 = extractvalue {i16, i1} %c18, 0
+ %r19 = extractvalue {i16, i1} %c19, 0
+ %r20 = extractvalue {i16, i1} %c20, 0
+ %r21 = extractvalue {i16, i1} %c21, 0
+ %r22 = extractvalue {i16, i1} %c22, 0
+ %r23 = extractvalue {i16, i1} %c23, 0
+ %r24 = extractvalue {i16, i1} %c24, 0
+ %r25 = extractvalue {i16, i1} %c25, 0
+ %r26 = extractvalue {i16, i1} %c26, 0
+ %r27 = extractvalue {i16, i1} %c27, 0
+ %r28 = extractvalue {i16, i1} %c28, 0
+ %r29 = extractvalue {i16, i1} %c29, 0
+ %r30 = extractvalue {i16, i1} %c30, 0
+ %r31 = extractvalue {i16, i1} %c31, 0
+ store i16 %r0 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 0 ), align 2
+ store i16 %r1 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 1 ), align 2
+ store i16 %r2 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 2 ), align 2
+ store i16 %r3 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 3 ), align 2
+ store i16 %r4 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 4 ), align 2
+ store i16 %r5 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 5 ), align 2
+ store i16 %r6 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 6 ), align 2
+ store i16 %r7 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 7 ), align 2
+ store i16 %r8 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8 ), align 2
+ store i16 %r9 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 9 ), align 2
+ store i16 %r10, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 10), align 2
+ store i16 %r11, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 11), align 2
+ store i16 %r12, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 12), align 2
+ store i16 %r13, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 13), align 2
+ store i16 %r14, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 14), align 2
+ store i16 %r15, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 15), align 2
+ store i16 %r16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16), align 2
+ store i16 %r17, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 17), align 2
+ store i16 %r18, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 18), align 2
+ store i16 %r19, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 19), align 2
+ store i16 %r20, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 20), align 2
+ store i16 %r21, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 21), align 2
+ store i16 %r22, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 22), align 2
+ store i16 %r23, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 23), align 2
+ store i16 %r24, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24), align 2
+ store i16 %r25, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 25), align 2
+ store i16 %r26, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 26), align 2
+ store i16 %r27, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 27), align 2
+ store i16 %r28, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 28), align 2
+ store i16 %r29, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 29), align 2
+ store i16 %r30, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 30), align 2
+ store i16 %r31, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 31), align 2
+ ret void
+}
+
+define void @sub_v64i8() {
+; CHECK-LABEL: @sub_v64i8(
+; CHECK-NEXT: [[A0:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 0), align 1
+; CHECK-NEXT: [[A1:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 1), align 1
+; CHECK-NEXT: [[A2:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 2), align 1
+; CHECK-NEXT: [[A3:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 3), align 1
+; CHECK-NEXT: [[A4:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 4), align 1
+; CHECK-NEXT: [[A5:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 5), align 1
+; CHECK-NEXT: [[A6:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 6), align 1
+; CHECK-NEXT: [[A7:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 7), align 1
+; CHECK-NEXT: [[A8:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 8), align 1
+; CHECK-NEXT: [[A9:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 9), align 1
+; CHECK-NEXT: [[A10:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 10), align 1
+; CHECK-NEXT: [[A11:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 11), align 1
+; CHECK-NEXT: [[A12:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 12), align 1
+; CHECK-NEXT: [[A13:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 13), align 1
+; CHECK-NEXT: [[A14:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 14), align 1
+; CHECK-NEXT: [[A15:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 15), align 1
+; CHECK-NEXT: [[A16:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 16), align 1
+; CHECK-NEXT: [[A17:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 17), align 1
+; CHECK-NEXT: [[A18:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 18), align 1
+; CHECK-NEXT: [[A19:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 19), align 1
+; CHECK-NEXT: [[A20:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 20), align 1
+; CHECK-NEXT: [[A21:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 21), align 1
+; CHECK-NEXT: [[A22:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 22), align 1
+; CHECK-NEXT: [[A23:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 23), align 1
+; CHECK-NEXT: [[A24:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 24), align 1
+; CHECK-NEXT: [[A25:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 25), align 1
+; CHECK-NEXT: [[A26:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 26), align 1
+; CHECK-NEXT: [[A27:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 27), align 1
+; CHECK-NEXT: [[A28:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 28), align 1
+; CHECK-NEXT: [[A29:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 29), align 1
+; CHECK-NEXT: [[A30:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 30), align 1
+; CHECK-NEXT: [[A31:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 31), align 1
+; CHECK-NEXT: [[A32:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 32), align 1
+; CHECK-NEXT: [[A33:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 33), align 1
+; CHECK-NEXT: [[A34:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 34), align 1
+; CHECK-NEXT: [[A35:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 35), align 1
+; CHECK-NEXT: [[A36:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 36), align 1
+; CHECK-NEXT: [[A37:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 37), align 1
+; CHECK-NEXT: [[A38:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 38), align 1
+; CHECK-NEXT: [[A39:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 39), align 1
+; CHECK-NEXT: [[A40:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 40), align 1
+; CHECK-NEXT: [[A41:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 41), align 1
+; CHECK-NEXT: [[A42:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 42), align 1
+; CHECK-NEXT: [[A43:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 43), align 1
+; CHECK-NEXT: [[A44:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 44), align 1
+; CHECK-NEXT: [[A45:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 45), align 1
+; CHECK-NEXT: [[A46:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 46), align 1
+; CHECK-NEXT: [[A47:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 47), align 1
+; CHECK-NEXT: [[A48:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 48), align 1
+; CHECK-NEXT: [[A49:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 49), align 1
+; CHECK-NEXT: [[A50:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 50), align 1
+; CHECK-NEXT: [[A51:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 51), align 1
+; CHECK-NEXT: [[A52:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 52), align 1
+; CHECK-NEXT: [[A53:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 53), align 1
+; CHECK-NEXT: [[A54:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 54), align 1
+; CHECK-NEXT: [[A55:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 55), align 1
+; CHECK-NEXT: [[A56:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 56), align 1
+; CHECK-NEXT: [[A57:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 57), align 1
+; CHECK-NEXT: [[A58:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 58), align 1
+; CHECK-NEXT: [[A59:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 59), align 1
+; CHECK-NEXT: [[A60:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 60), align 1
+; CHECK-NEXT: [[A61:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 61), align 1
+; CHECK-NEXT: [[A62:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 62), align 1
+; CHECK-NEXT: [[A63:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 63), align 1
+; CHECK-NEXT: [[B0:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 0), align 1
+; CHECK-NEXT: [[B1:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 1), align 1
+; CHECK-NEXT: [[B2:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 2), align 1
+; CHECK-NEXT: [[B3:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 3), align 1
+; CHECK-NEXT: [[B4:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 4), align 1
+; CHECK-NEXT: [[B5:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 5), align 1
+; CHECK-NEXT: [[B6:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 6), align 1
+; CHECK-NEXT: [[B7:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 7), align 1
+; CHECK-NEXT: [[B8:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 8), align 1
+; CHECK-NEXT: [[B9:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 9), align 1
+; CHECK-NEXT: [[B10:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 10), align 1
+; CHECK-NEXT: [[B11:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 11), align 1
+; CHECK-NEXT: [[B12:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 12), align 1
+; CHECK-NEXT: [[B13:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 13), align 1
+; CHECK-NEXT: [[B14:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 14), align 1
+; CHECK-NEXT: [[B15:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 15), align 1
+; CHECK-NEXT: [[B16:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 16), align 1
+; CHECK-NEXT: [[B17:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 17), align 1
+; CHECK-NEXT: [[B18:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 18), align 1
+; CHECK-NEXT: [[B19:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 19), align 1
+; CHECK-NEXT: [[B20:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 20), align 1
+; CHECK-NEXT: [[B21:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 21), align 1
+; CHECK-NEXT: [[B22:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 22), align 1
+; CHECK-NEXT: [[B23:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 23), align 1
+; CHECK-NEXT: [[B24:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 24), align 1
+; CHECK-NEXT: [[B25:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 25), align 1
+; CHECK-NEXT: [[B26:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 26), align 1
+; CHECK-NEXT: [[B27:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 27), align 1
+; CHECK-NEXT: [[B28:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 28), align 1
+; CHECK-NEXT: [[B29:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 29), align 1
+; CHECK-NEXT: [[B30:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 30), align 1
+; CHECK-NEXT: [[B31:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 31), align 1
+; CHECK-NEXT: [[B32:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 32), align 1
+; CHECK-NEXT: [[B33:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 33), align 1
+; CHECK-NEXT: [[B34:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 34), align 1
+; CHECK-NEXT: [[B35:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 35), align 1
+; CHECK-NEXT: [[B36:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 36), align 1
+; CHECK-NEXT: [[B37:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 37), align 1
+; CHECK-NEXT: [[B38:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 38), align 1
+; CHECK-NEXT: [[B39:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 39), align 1
+; CHECK-NEXT: [[B40:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 40), align 1
+; CHECK-NEXT: [[B41:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 41), align 1
+; CHECK-NEXT: [[B42:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 42), align 1
+; CHECK-NEXT: [[B43:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 43), align 1
+; CHECK-NEXT: [[B44:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 44), align 1
+; CHECK-NEXT: [[B45:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 45), align 1
+; CHECK-NEXT: [[B46:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 46), align 1
+; CHECK-NEXT: [[B47:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 47), align 1
+; CHECK-NEXT: [[B48:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 48), align 1
+; CHECK-NEXT: [[B49:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 49), align 1
+; CHECK-NEXT: [[B50:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 50), align 1
+; CHECK-NEXT: [[B51:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 51), align 1
+; CHECK-NEXT: [[B52:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 52), align 1
+; CHECK-NEXT: [[B53:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 53), align 1
+; CHECK-NEXT: [[B54:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 54), align 1
+; CHECK-NEXT: [[B55:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 55), align 1
+; CHECK-NEXT: [[B56:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 56), align 1
+; CHECK-NEXT: [[B57:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 57), align 1
+; CHECK-NEXT: [[B58:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 58), align 1
+; CHECK-NEXT: [[B59:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 59), align 1
+; CHECK-NEXT: [[B60:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 60), align 1
+; CHECK-NEXT: [[B61:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 61), align 1
+; CHECK-NEXT: [[B62:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 62), align 1
+; CHECK-NEXT: [[B63:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 63), align 1
+; CHECK-NEXT: [[C0:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A0]], i8 [[B0]])
+; CHECK-NEXT: [[C1:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A1]], i8 [[B1]])
+; CHECK-NEXT: [[C2:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A2]], i8 [[B2]])
+; CHECK-NEXT: [[C3:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A3]], i8 [[B3]])
+; CHECK-NEXT: [[C4:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A4]], i8 [[B4]])
+; CHECK-NEXT: [[C5:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A5]], i8 [[B5]])
+; CHECK-NEXT: [[C6:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A6]], i8 [[B6]])
+; CHECK-NEXT: [[C7:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A7]], i8 [[B7]])
+; CHECK-NEXT: [[C8:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A8]], i8 [[B8]])
+; CHECK-NEXT: [[C9:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A9]], i8 [[B9]])
+; CHECK-NEXT: [[C10:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A10]], i8 [[B10]])
+; CHECK-NEXT: [[C11:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A11]], i8 [[B11]])
+; CHECK-NEXT: [[C12:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A12]], i8 [[B12]])
+; CHECK-NEXT: [[C13:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A13]], i8 [[B13]])
+; CHECK-NEXT: [[C14:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A14]], i8 [[B14]])
+; CHECK-NEXT: [[C15:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A15]], i8 [[B15]])
+; CHECK-NEXT: [[C16:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A16]], i8 [[B16]])
+; CHECK-NEXT: [[C17:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A17]], i8 [[B17]])
+; CHECK-NEXT: [[C18:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A18]], i8 [[B18]])
+; CHECK-NEXT: [[C19:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A19]], i8 [[B19]])
+; CHECK-NEXT: [[C20:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A20]], i8 [[B20]])
+; CHECK-NEXT: [[C21:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A21]], i8 [[B21]])
+; CHECK-NEXT: [[C22:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A22]], i8 [[B22]])
+; CHECK-NEXT: [[C23:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A23]], i8 [[B23]])
+; CHECK-NEXT: [[C24:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A24]], i8 [[B24]])
+; CHECK-NEXT: [[C25:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A25]], i8 [[B25]])
+; CHECK-NEXT: [[C26:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A26]], i8 [[B26]])
+; CHECK-NEXT: [[C27:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A27]], i8 [[B27]])
+; CHECK-NEXT: [[C28:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A28]], i8 [[B28]])
+; CHECK-NEXT: [[C29:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A29]], i8 [[B29]])
+; CHECK-NEXT: [[C30:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A30]], i8 [[B30]])
+; CHECK-NEXT: [[C31:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A31]], i8 [[B31]])
+; CHECK-NEXT: [[C32:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A32]], i8 [[B32]])
+; CHECK-NEXT: [[C33:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A33]], i8 [[B33]])
+; CHECK-NEXT: [[C34:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A34]], i8 [[B34]])
+; CHECK-NEXT: [[C35:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A35]], i8 [[B35]])
+; CHECK-NEXT: [[C36:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A36]], i8 [[B36]])
+; CHECK-NEXT: [[C37:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A37]], i8 [[B37]])
+; CHECK-NEXT: [[C38:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A38]], i8 [[B38]])
+; CHECK-NEXT: [[C39:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A39]], i8 [[B39]])
+; CHECK-NEXT: [[C40:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A40]], i8 [[B40]])
+; CHECK-NEXT: [[C41:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A41]], i8 [[B41]])
+; CHECK-NEXT: [[C42:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A42]], i8 [[B42]])
+; CHECK-NEXT: [[C43:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A43]], i8 [[B43]])
+; CHECK-NEXT: [[C44:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A44]], i8 [[B44]])
+; CHECK-NEXT: [[C45:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A45]], i8 [[B45]])
+; CHECK-NEXT: [[C46:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A46]], i8 [[B46]])
+; CHECK-NEXT: [[C47:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A47]], i8 [[B47]])
+; CHECK-NEXT: [[C48:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A48]], i8 [[B48]])
+; CHECK-NEXT: [[C49:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A49]], i8 [[B49]])
+; CHECK-NEXT: [[C50:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A50]], i8 [[B50]])
+; CHECK-NEXT: [[C51:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A51]], i8 [[B51]])
+; CHECK-NEXT: [[C52:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A52]], i8 [[B52]])
+; CHECK-NEXT: [[C53:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A53]], i8 [[B53]])
+; CHECK-NEXT: [[C54:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A54]], i8 [[B54]])
+; CHECK-NEXT: [[C55:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A55]], i8 [[B55]])
+; CHECK-NEXT: [[C56:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A56]], i8 [[B56]])
+; CHECK-NEXT: [[C57:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A57]], i8 [[B57]])
+; CHECK-NEXT: [[C58:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A58]], i8 [[B58]])
+; CHECK-NEXT: [[C59:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A59]], i8 [[B59]])
+; CHECK-NEXT: [[C60:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A60]], i8 [[B60]])
+; CHECK-NEXT: [[C61:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A61]], i8 [[B61]])
+; CHECK-NEXT: [[C62:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A62]], i8 [[B62]])
+; CHECK-NEXT: [[C63:%.*]] = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 [[A63]], i8 [[B63]])
+; CHECK-NEXT: [[R0:%.*]] = extractvalue { i8, i1 } [[C0]], 0
+; CHECK-NEXT: [[R1:%.*]] = extractvalue { i8, i1 } [[C1]], 0
+; CHECK-NEXT: [[R2:%.*]] = extractvalue { i8, i1 } [[C2]], 0
+; CHECK-NEXT: [[R3:%.*]] = extractvalue { i8, i1 } [[C3]], 0
+; CHECK-NEXT: [[R4:%.*]] = extractvalue { i8, i1 } [[C4]], 0
+; CHECK-NEXT: [[R5:%.*]] = extractvalue { i8, i1 } [[C5]], 0
+; CHECK-NEXT: [[R6:%.*]] = extractvalue { i8, i1 } [[C6]], 0
+; CHECK-NEXT: [[R7:%.*]] = extractvalue { i8, i1 } [[C7]], 0
+; CHECK-NEXT: [[R8:%.*]] = extractvalue { i8, i1 } [[C8]], 0
+; CHECK-NEXT: [[R9:%.*]] = extractvalue { i8, i1 } [[C9]], 0
+; CHECK-NEXT: [[R10:%.*]] = extractvalue { i8, i1 } [[C10]], 0
+; CHECK-NEXT: [[R11:%.*]] = extractvalue { i8, i1 } [[C11]], 0
+; CHECK-NEXT: [[R12:%.*]] = extractvalue { i8, i1 } [[C12]], 0
+; CHECK-NEXT: [[R13:%.*]] = extractvalue { i8, i1 } [[C13]], 0
+; CHECK-NEXT: [[R14:%.*]] = extractvalue { i8, i1 } [[C14]], 0
+; CHECK-NEXT: [[R15:%.*]] = extractvalue { i8, i1 } [[C15]], 0
+; CHECK-NEXT: [[R16:%.*]] = extractvalue { i8, i1 } [[C16]], 0
+; CHECK-NEXT: [[R17:%.*]] = extractvalue { i8, i1 } [[C17]], 0
+; CHECK-NEXT: [[R18:%.*]] = extractvalue { i8, i1 } [[C18]], 0
+; CHECK-NEXT: [[R19:%.*]] = extractvalue { i8, i1 } [[C19]], 0
+; CHECK-NEXT: [[R20:%.*]] = extractvalue { i8, i1 } [[C20]], 0
+; CHECK-NEXT: [[R21:%.*]] = extractvalue { i8, i1 } [[C21]], 0
+; CHECK-NEXT: [[R22:%.*]] = extractvalue { i8, i1 } [[C22]], 0
+; CHECK-NEXT: [[R23:%.*]] = extractvalue { i8, i1 } [[C23]], 0
+; CHECK-NEXT: [[R24:%.*]] = extractvalue { i8, i1 } [[C24]], 0
+; CHECK-NEXT: [[R25:%.*]] = extractvalue { i8, i1 } [[C25]], 0
+; CHECK-NEXT: [[R26:%.*]] = extractvalue { i8, i1 } [[C26]], 0
+; CHECK-NEXT: [[R27:%.*]] = extractvalue { i8, i1 } [[C27]], 0
+; CHECK-NEXT: [[R28:%.*]] = extractvalue { i8, i1 } [[C28]], 0
+; CHECK-NEXT: [[R29:%.*]] = extractvalue { i8, i1 } [[C29]], 0
+; CHECK-NEXT: [[R30:%.*]] = extractvalue { i8, i1 } [[C30]], 0
+; CHECK-NEXT: [[R31:%.*]] = extractvalue { i8, i1 } [[C31]], 0
+; CHECK-NEXT: [[R32:%.*]] = extractvalue { i8, i1 } [[C32]], 0
+; CHECK-NEXT: [[R33:%.*]] = extractvalue { i8, i1 } [[C33]], 0
+; CHECK-NEXT: [[R34:%.*]] = extractvalue { i8, i1 } [[C34]], 0
+; CHECK-NEXT: [[R35:%.*]] = extractvalue { i8, i1 } [[C35]], 0
+; CHECK-NEXT: [[R36:%.*]] = extractvalue { i8, i1 } [[C36]], 0
+; CHECK-NEXT: [[R37:%.*]] = extractvalue { i8, i1 } [[C37]], 0
+; CHECK-NEXT: [[R38:%.*]] = extractvalue { i8, i1 } [[C38]], 0
+; CHECK-NEXT: [[R39:%.*]] = extractvalue { i8, i1 } [[C39]], 0
+; CHECK-NEXT: [[R40:%.*]] = extractvalue { i8, i1 } [[C40]], 0
+; CHECK-NEXT: [[R41:%.*]] = extractvalue { i8, i1 } [[C41]], 0
+; CHECK-NEXT: [[R42:%.*]] = extractvalue { i8, i1 } [[C42]], 0
+; CHECK-NEXT: [[R43:%.*]] = extractvalue { i8, i1 } [[C43]], 0
+; CHECK-NEXT: [[R44:%.*]] = extractvalue { i8, i1 } [[C44]], 0
+; CHECK-NEXT: [[R45:%.*]] = extractvalue { i8, i1 } [[C45]], 0
+; CHECK-NEXT: [[R46:%.*]] = extractvalue { i8, i1 } [[C46]], 0
+; CHECK-NEXT: [[R47:%.*]] = extractvalue { i8, i1 } [[C47]], 0
+; CHECK-NEXT: [[R48:%.*]] = extractvalue { i8, i1 } [[C48]], 0
+; CHECK-NEXT: [[R49:%.*]] = extractvalue { i8, i1 } [[C49]], 0
+; CHECK-NEXT: [[R50:%.*]] = extractvalue { i8, i1 } [[C50]], 0
+; CHECK-NEXT: [[R51:%.*]] = extractvalue { i8, i1 } [[C51]], 0
+; CHECK-NEXT: [[R52:%.*]] = extractvalue { i8, i1 } [[C52]], 0
+; CHECK-NEXT: [[R53:%.*]] = extractvalue { i8, i1 } [[C53]], 0
+; CHECK-NEXT: [[R54:%.*]] = extractvalue { i8, i1 } [[C54]], 0
+; CHECK-NEXT: [[R55:%.*]] = extractvalue { i8, i1 } [[C55]], 0
+; CHECK-NEXT: [[R56:%.*]] = extractvalue { i8, i1 } [[C56]], 0
+; CHECK-NEXT: [[R57:%.*]] = extractvalue { i8, i1 } [[C57]], 0
+; CHECK-NEXT: [[R58:%.*]] = extractvalue { i8, i1 } [[C58]], 0
+; CHECK-NEXT: [[R59:%.*]] = extractvalue { i8, i1 } [[C59]], 0
+; CHECK-NEXT: [[R60:%.*]] = extractvalue { i8, i1 } [[C60]], 0
+; CHECK-NEXT: [[R61:%.*]] = extractvalue { i8, i1 } [[C61]], 0
+; CHECK-NEXT: [[R62:%.*]] = extractvalue { i8, i1 } [[C62]], 0
+; CHECK-NEXT: [[R63:%.*]] = extractvalue { i8, i1 } [[C63]], 0
+; CHECK-NEXT: store i8 [[R0]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 0), align 1
+; CHECK-NEXT: store i8 [[R1]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 1), align 1
+; CHECK-NEXT: store i8 [[R2]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 2), align 1
+; CHECK-NEXT: store i8 [[R3]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 3), align 1
+; CHECK-NEXT: store i8 [[R4]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 4), align 1
+; CHECK-NEXT: store i8 [[R5]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 5), align 1
+; CHECK-NEXT: store i8 [[R6]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 6), align 1
+; CHECK-NEXT: store i8 [[R7]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 7), align 1
+; CHECK-NEXT: store i8 [[R8]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 8), align 1
+; CHECK-NEXT: store i8 [[R9]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 9), align 1
+; CHECK-NEXT: store i8 [[R10]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 10), align 1
+; CHECK-NEXT: store i8 [[R11]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 11), align 1
+; CHECK-NEXT: store i8 [[R12]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 12), align 1
+; CHECK-NEXT: store i8 [[R13]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 13), align 1
+; CHECK-NEXT: store i8 [[R14]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 14), align 1
+; CHECK-NEXT: store i8 [[R15]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 15), align 1
+; CHECK-NEXT: store i8 [[R16]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 16), align 1
+; CHECK-NEXT: store i8 [[R17]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 17), align 1
+; CHECK-NEXT: store i8 [[R18]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 18), align 1
+; CHECK-NEXT: store i8 [[R19]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 19), align 1
+; CHECK-NEXT: store i8 [[R20]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 20), align 1
+; CHECK-NEXT: store i8 [[R21]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 21), align 1
+; CHECK-NEXT: store i8 [[R22]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 22), align 1
+; CHECK-NEXT: store i8 [[R23]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 23), align 1
+; CHECK-NEXT: store i8 [[R24]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 24), align 1
+; CHECK-NEXT: store i8 [[R25]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 25), align 1
+; CHECK-NEXT: store i8 [[R26]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 26), align 1
+; CHECK-NEXT: store i8 [[R27]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 27), align 1
+; CHECK-NEXT: store i8 [[R28]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 28), align 1
+; CHECK-NEXT: store i8 [[R29]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 29), align 1
+; CHECK-NEXT: store i8 [[R30]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 30), align 1
+; CHECK-NEXT: store i8 [[R31]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 31), align 1
+; CHECK-NEXT: store i8 [[R32]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 32), align 1
+; CHECK-NEXT: store i8 [[R33]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 33), align 1
+; CHECK-NEXT: store i8 [[R34]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 34), align 1
+; CHECK-NEXT: store i8 [[R35]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 35), align 1
+; CHECK-NEXT: store i8 [[R36]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 36), align 1
+; CHECK-NEXT: store i8 [[R37]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 37), align 1
+; CHECK-NEXT: store i8 [[R38]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 38), align 1
+; CHECK-NEXT: store i8 [[R39]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 39), align 1
+; CHECK-NEXT: store i8 [[R40]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 40), align 1
+; CHECK-NEXT: store i8 [[R41]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 41), align 1
+; CHECK-NEXT: store i8 [[R42]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 42), align 1
+; CHECK-NEXT: store i8 [[R43]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 43), align 1
+; CHECK-NEXT: store i8 [[R44]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 44), align 1
+; CHECK-NEXT: store i8 [[R45]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 45), align 1
+; CHECK-NEXT: store i8 [[R46]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 46), align 1
+; CHECK-NEXT: store i8 [[R47]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 47), align 1
+; CHECK-NEXT: store i8 [[R48]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 48), align 1
+; CHECK-NEXT: store i8 [[R49]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 49), align 1
+; CHECK-NEXT: store i8 [[R50]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 50), align 1
+; CHECK-NEXT: store i8 [[R51]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 51), align 1
+; CHECK-NEXT: store i8 [[R52]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 52), align 1
+; CHECK-NEXT: store i8 [[R53]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 53), align 1
+; CHECK-NEXT: store i8 [[R54]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 54), align 1
+; CHECK-NEXT: store i8 [[R55]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 55), align 1
+; CHECK-NEXT: store i8 [[R56]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 56), align 1
+; CHECK-NEXT: store i8 [[R57]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 57), align 1
+; CHECK-NEXT: store i8 [[R58]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 58), align 1
+; CHECK-NEXT: store i8 [[R59]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 59), align 1
+; CHECK-NEXT: store i8 [[R60]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 60), align 1
+; CHECK-NEXT: store i8 [[R61]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 61), align 1
+; CHECK-NEXT: store i8 [[R62]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 62), align 1
+; CHECK-NEXT: store i8 [[R63]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 63), align 1
+; CHECK-NEXT: ret void
+;
+ %a0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 0 ), align 1
+ %a1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 1 ), align 1
+ %a2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 2 ), align 1
+ %a3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 3 ), align 1
+ %a4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 4 ), align 1
+ %a5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 5 ), align 1
+ %a6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 6 ), align 1
+ %a7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 7 ), align 1
+ %a8 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 8 ), align 1
+ %a9 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 9 ), align 1
+ %a10 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 10), align 1
+ %a11 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 11), align 1
+ %a12 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 12), align 1
+ %a13 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 13), align 1
+ %a14 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 14), align 1
+ %a15 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 15), align 1
+ %a16 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 16), align 1
+ %a17 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 17), align 1
+ %a18 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 18), align 1
+ %a19 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 19), align 1
+ %a20 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 20), align 1
+ %a21 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 21), align 1
+ %a22 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 22), align 1
+ %a23 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 23), align 1
+ %a24 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 24), align 1
+ %a25 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 25), align 1
+ %a26 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 26), align 1
+ %a27 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 27), align 1
+ %a28 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 28), align 1
+ %a29 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 29), align 1
+ %a30 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 30), align 1
+ %a31 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 31), align 1
+ %a32 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 32), align 1
+ %a33 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 33), align 1
+ %a34 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 34), align 1
+ %a35 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 35), align 1
+ %a36 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 36), align 1
+ %a37 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 37), align 1
+ %a38 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 38), align 1
+ %a39 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 39), align 1
+ %a40 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 40), align 1
+ %a41 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 41), align 1
+ %a42 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 42), align 1
+ %a43 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 43), align 1
+ %a44 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 44), align 1
+ %a45 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 45), align 1
+ %a46 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 46), align 1
+ %a47 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 47), align 1
+ %a48 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 48), align 1
+ %a49 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 49), align 1
+ %a50 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 50), align 1
+ %a51 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 51), align 1
+ %a52 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 52), align 1
+ %a53 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 53), align 1
+ %a54 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 54), align 1
+ %a55 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 55), align 1
+ %a56 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 56), align 1
+ %a57 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 57), align 1
+ %a58 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 58), align 1
+ %a59 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 59), align 1
+ %a60 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 60), align 1
+ %a61 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 61), align 1
+ %a62 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 62), align 1
+ %a63 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 63), align 1
+ %b0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 0 ), align 1
+ %b1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 1 ), align 1
+ %b2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 2 ), align 1
+ %b3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 3 ), align 1
+ %b4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 4 ), align 1
+ %b5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 5 ), align 1
+ %b6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 6 ), align 1
+ %b7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 7 ), align 1
+ %b8 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 8 ), align 1
+ %b9 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 9 ), align 1
+ %b10 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 10), align 1
+ %b11 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 11), align 1
+ %b12 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 12), align 1
+ %b13 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 13), align 1
+ %b14 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 14), align 1
+ %b15 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 15), align 1
+ %b16 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 16), align 1
+ %b17 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 17), align 1
+ %b18 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 18), align 1
+ %b19 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 19), align 1
+ %b20 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 20), align 1
+ %b21 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 21), align 1
+ %b22 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 22), align 1
+ %b23 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 23), align 1
+ %b24 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 24), align 1
+ %b25 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 25), align 1
+ %b26 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 26), align 1
+ %b27 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 27), align 1
+ %b28 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 28), align 1
+ %b29 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 29), align 1
+ %b30 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 30), align 1
+ %b31 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 31), align 1
+ %b32 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 32), align 1
+ %b33 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 33), align 1
+ %b34 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 34), align 1
+ %b35 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 35), align 1
+ %b36 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 36), align 1
+ %b37 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 37), align 1
+ %b38 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 38), align 1
+ %b39 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 39), align 1
+ %b40 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 40), align 1
+ %b41 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 41), align 1
+ %b42 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 42), align 1
+ %b43 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 43), align 1
+ %b44 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 44), align 1
+ %b45 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 45), align 1
+ %b46 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 46), align 1
+ %b47 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 47), align 1
+ %b48 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 48), align 1
+ %b49 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 49), align 1
+ %b50 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 50), align 1
+ %b51 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 51), align 1
+ %b52 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 52), align 1
+ %b53 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 53), align 1
+ %b54 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 54), align 1
+ %b55 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 55), align 1
+ %b56 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 56), align 1
+ %b57 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 57), align 1
+ %b58 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 58), align 1
+ %b59 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 59), align 1
+ %b60 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 60), align 1
+ %b61 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 61), align 1
+ %b62 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 62), align 1
+ %b63 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 63), align 1
+ %c0 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a0 , i8 %b0 )
+ %c1 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a1 , i8 %b1 )
+ %c2 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a2 , i8 %b2 )
+ %c3 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a3 , i8 %b3 )
+ %c4 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a4 , i8 %b4 )
+ %c5 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a5 , i8 %b5 )
+ %c6 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a6 , i8 %b6 )
+ %c7 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a7 , i8 %b7 )
+ %c8 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a8 , i8 %b8 )
+ %c9 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a9 , i8 %b9 )
+ %c10 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a10, i8 %b10)
+ %c11 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a11, i8 %b11)
+ %c12 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a12, i8 %b12)
+ %c13 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a13, i8 %b13)
+ %c14 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a14, i8 %b14)
+ %c15 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a15, i8 %b15)
+ %c16 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a16, i8 %b16)
+ %c17 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a17, i8 %b17)
+ %c18 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a18, i8 %b18)
+ %c19 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a19, i8 %b19)
+ %c20 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a20, i8 %b20)
+ %c21 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a21, i8 %b21)
+ %c22 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a22, i8 %b22)
+ %c23 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a23, i8 %b23)
+ %c24 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a24, i8 %b24)
+ %c25 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a25, i8 %b25)
+ %c26 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a26, i8 %b26)
+ %c27 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a27, i8 %b27)
+ %c28 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a28, i8 %b28)
+ %c29 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a29, i8 %b29)
+ %c30 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a30, i8 %b30)
+ %c31 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a31, i8 %b31)
+ %c32 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a32, i8 %b32)
+ %c33 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a33, i8 %b33)
+ %c34 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a34, i8 %b34)
+ %c35 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a35, i8 %b35)
+ %c36 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a36, i8 %b36)
+ %c37 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a37, i8 %b37)
+ %c38 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a38, i8 %b38)
+ %c39 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a39, i8 %b39)
+ %c40 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a40, i8 %b40)
+ %c41 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a41, i8 %b41)
+ %c42 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a42, i8 %b42)
+ %c43 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a43, i8 %b43)
+ %c44 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a44, i8 %b44)
+ %c45 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a45, i8 %b45)
+ %c46 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a46, i8 %b46)
+ %c47 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a47, i8 %b47)
+ %c48 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a48, i8 %b48)
+ %c49 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a49, i8 %b49)
+ %c50 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a50, i8 %b50)
+ %c51 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a51, i8 %b51)
+ %c52 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a52, i8 %b52)
+ %c53 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a53, i8 %b53)
+ %c54 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a54, i8 %b54)
+ %c55 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a55, i8 %b55)
+ %c56 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a56, i8 %b56)
+ %c57 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a57, i8 %b57)
+ %c58 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a58, i8 %b58)
+ %c59 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a59, i8 %b59)
+ %c60 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a60, i8 %b60)
+ %c61 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a61, i8 %b61)
+ %c62 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a62, i8 %b62)
+ %c63 = call {i8, i1} @llvm.ssub.with.overflow.i8(i8 %a63, i8 %b63)
+ %r0 = extractvalue {i8, i1} %c0 , 0
+ %r1 = extractvalue {i8, i1} %c1 , 0
+ %r2 = extractvalue {i8, i1} %c2 , 0
+ %r3 = extractvalue {i8, i1} %c3 , 0
+ %r4 = extractvalue {i8, i1} %c4 , 0
+ %r5 = extractvalue {i8, i1} %c5 , 0
+ %r6 = extractvalue {i8, i1} %c6 , 0
+ %r7 = extractvalue {i8, i1} %c7 , 0
+ %r8 = extractvalue {i8, i1} %c8 , 0
+ %r9 = extractvalue {i8, i1} %c9 , 0
+ %r10 = extractvalue {i8, i1} %c10, 0
+ %r11 = extractvalue {i8, i1} %c11, 0
+ %r12 = extractvalue {i8, i1} %c12, 0
+ %r13 = extractvalue {i8, i1} %c13, 0
+ %r14 = extractvalue {i8, i1} %c14, 0
+ %r15 = extractvalue {i8, i1} %c15, 0
+ %r16 = extractvalue {i8, i1} %c16, 0
+ %r17 = extractvalue {i8, i1} %c17, 0
+ %r18 = extractvalue {i8, i1} %c18, 0
+ %r19 = extractvalue {i8, i1} %c19, 0
+ %r20 = extractvalue {i8, i1} %c20, 0
+ %r21 = extractvalue {i8, i1} %c21, 0
+ %r22 = extractvalue {i8, i1} %c22, 0
+ %r23 = extractvalue {i8, i1} %c23, 0
+ %r24 = extractvalue {i8, i1} %c24, 0
+ %r25 = extractvalue {i8, i1} %c25, 0
+ %r26 = extractvalue {i8, i1} %c26, 0
+ %r27 = extractvalue {i8, i1} %c27, 0
+ %r28 = extractvalue {i8, i1} %c28, 0
+ %r29 = extractvalue {i8, i1} %c29, 0
+ %r30 = extractvalue {i8, i1} %c30, 0
+ %r31 = extractvalue {i8, i1} %c31, 0
+ %r32 = extractvalue {i8, i1} %c32, 0
+ %r33 = extractvalue {i8, i1} %c33, 0
+ %r34 = extractvalue {i8, i1} %c34, 0
+ %r35 = extractvalue {i8, i1} %c35, 0
+ %r36 = extractvalue {i8, i1} %c36, 0
+ %r37 = extractvalue {i8, i1} %c37, 0
+ %r38 = extractvalue {i8, i1} %c38, 0
+ %r39 = extractvalue {i8, i1} %c39, 0
+ %r40 = extractvalue {i8, i1} %c40, 0
+ %r41 = extractvalue {i8, i1} %c41, 0
+ %r42 = extractvalue {i8, i1} %c42, 0
+ %r43 = extractvalue {i8, i1} %c43, 0
+ %r44 = extractvalue {i8, i1} %c44, 0
+ %r45 = extractvalue {i8, i1} %c45, 0
+ %r46 = extractvalue {i8, i1} %c46, 0
+ %r47 = extractvalue {i8, i1} %c47, 0
+ %r48 = extractvalue {i8, i1} %c48, 0
+ %r49 = extractvalue {i8, i1} %c49, 0
+ %r50 = extractvalue {i8, i1} %c50, 0
+ %r51 = extractvalue {i8, i1} %c51, 0
+ %r52 = extractvalue {i8, i1} %c52, 0
+ %r53 = extractvalue {i8, i1} %c53, 0
+ %r54 = extractvalue {i8, i1} %c54, 0
+ %r55 = extractvalue {i8, i1} %c55, 0
+ %r56 = extractvalue {i8, i1} %c56, 0
+ %r57 = extractvalue {i8, i1} %c57, 0
+ %r58 = extractvalue {i8, i1} %c58, 0
+ %r59 = extractvalue {i8, i1} %c59, 0
+ %r60 = extractvalue {i8, i1} %c60, 0
+ %r61 = extractvalue {i8, i1} %c61, 0
+ %r62 = extractvalue {i8, i1} %c62, 0
+ %r63 = extractvalue {i8, i1} %c63, 0
+ store i8 %r0 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 0 ), align 1
+ store i8 %r1 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 1 ), align 1
+ store i8 %r2 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 2 ), align 1
+ store i8 %r3 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 3 ), align 1
+ store i8 %r4 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 4 ), align 1
+ store i8 %r5 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 5 ), align 1
+ store i8 %r6 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 6 ), align 1
+ store i8 %r7 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 7 ), align 1
+ store i8 %r8 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 8 ), align 1
+ store i8 %r9 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 9 ), align 1
+ store i8 %r10, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 10), align 1
+ store i8 %r11, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 11), align 1
+ store i8 %r12, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 12), align 1
+ store i8 %r13, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 13), align 1
+ store i8 %r14, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 14), align 1
+ store i8 %r15, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 15), align 1
+ store i8 %r16, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 16), align 1
+ store i8 %r17, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 17), align 1
+ store i8 %r18, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 18), align 1
+ store i8 %r19, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 19), align 1
+ store i8 %r20, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 20), align 1
+ store i8 %r21, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 21), align 1
+ store i8 %r22, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 22), align 1
+ store i8 %r23, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 23), align 1
+ store i8 %r24, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 24), align 1
+ store i8 %r25, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 25), align 1
+ store i8 %r26, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 26), align 1
+ store i8 %r27, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 27), align 1
+ store i8 %r28, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 28), align 1
+ store i8 %r29, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 29), align 1
+ store i8 %r30, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 30), align 1
+ store i8 %r31, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 31), align 1
+ store i8 %r32, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 32), align 1
+ store i8 %r33, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 33), align 1
+ store i8 %r34, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 34), align 1
+ store i8 %r35, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 35), align 1
+ store i8 %r36, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 36), align 1
+ store i8 %r37, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 37), align 1
+ store i8 %r38, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 38), align 1
+ store i8 %r39, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 39), align 1
+ store i8 %r40, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 40), align 1
+ store i8 %r41, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 41), align 1
+ store i8 %r42, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 42), align 1
+ store i8 %r43, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 43), align 1
+ store i8 %r44, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 44), align 1
+ store i8 %r45, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 45), align 1
+ store i8 %r46, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 46), align 1
+ store i8 %r47, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 47), align 1
+ store i8 %r48, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 48), align 1
+ store i8 %r49, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 49), align 1
+ store i8 %r50, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 50), align 1
+ store i8 %r51, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 51), align 1
+ store i8 %r52, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 52), align 1
+ store i8 %r53, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 53), align 1
+ store i8 %r54, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 54), align 1
+ store i8 %r55, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 55), align 1
+ store i8 %r56, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 56), align 1
+ store i8 %r57, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 57), align 1
+ store i8 %r58, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 58), align 1
+ store i8 %r59, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 59), align 1
+ store i8 %r60, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 60), align 1
+ store i8 %r61, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 61), align 1
+ store i8 %r62, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 62), align 1
+ store i8 %r63, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 63), align 1
+ ret void
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-usat.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-usat.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-usat.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-usat.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,729 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,SSE
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=slm -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,SLM
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=knl -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skx -mattr=-prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skx -mattr=+prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX256BW
+
+ at a64 = common global [8 x i64] zeroinitializer, align 64
+ at b64 = common global [8 x i64] zeroinitializer, align 64
+ at c64 = common global [8 x i64] zeroinitializer, align 64
+ at a32 = common global [16 x i32] zeroinitializer, align 64
+ at b32 = common global [16 x i32] zeroinitializer, align 64
+ at c32 = common global [16 x i32] zeroinitializer, align 64
+ at a16 = common global [32 x i16] zeroinitializer, align 64
+ at b16 = common global [32 x i16] zeroinitializer, align 64
+ at c16 = common global [32 x i16] zeroinitializer, align 64
+ at a8 = common global [64 x i8] zeroinitializer, align 64
+ at b8 = common global [64 x i8] zeroinitializer, align 64
+ at c8 = common global [64 x i8] zeroinitializer, align 64
+
+declare i64 @llvm.usub.sat.i64(i64, i64)
+declare i32 @llvm.usub.sat.i32(i32, i32)
+declare i16 @llvm.usub.sat.i16(i16, i16)
+declare i8 @llvm.usub.sat.i8 (i8 , i8 )
+
+define void @sub_v8i64() {
+; SSE-LABEL: @sub_v8i64(
+; SSE-NEXT: [[A0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 0), align 8
+; SSE-NEXT: [[A1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 1), align 8
+; SSE-NEXT: [[A2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2), align 8
+; SSE-NEXT: [[A3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 3), align 8
+; SSE-NEXT: [[A4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4), align 8
+; SSE-NEXT: [[A5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 5), align 8
+; SSE-NEXT: [[A6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6), align 8
+; SSE-NEXT: [[A7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 7), align 8
+; SSE-NEXT: [[B0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 0), align 8
+; SSE-NEXT: [[B1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 1), align 8
+; SSE-NEXT: [[B2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2), align 8
+; SSE-NEXT: [[B3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 3), align 8
+; SSE-NEXT: [[B4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4), align 8
+; SSE-NEXT: [[B5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 5), align 8
+; SSE-NEXT: [[B6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6), align 8
+; SSE-NEXT: [[B7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 7), align 8
+; SSE-NEXT: [[R0:%.*]] = call i64 @llvm.usub.sat.i64(i64 [[A0]], i64 [[B0]])
+; SSE-NEXT: [[R1:%.*]] = call i64 @llvm.usub.sat.i64(i64 [[A1]], i64 [[B1]])
+; SSE-NEXT: [[R2:%.*]] = call i64 @llvm.usub.sat.i64(i64 [[A2]], i64 [[B2]])
+; SSE-NEXT: [[R3:%.*]] = call i64 @llvm.usub.sat.i64(i64 [[A3]], i64 [[B3]])
+; SSE-NEXT: [[R4:%.*]] = call i64 @llvm.usub.sat.i64(i64 [[A4]], i64 [[B4]])
+; SSE-NEXT: [[R5:%.*]] = call i64 @llvm.usub.sat.i64(i64 [[A5]], i64 [[B5]])
+; SSE-NEXT: [[R6:%.*]] = call i64 @llvm.usub.sat.i64(i64 [[A6]], i64 [[B6]])
+; SSE-NEXT: [[R7:%.*]] = call i64 @llvm.usub.sat.i64(i64 [[A7]], i64 [[B7]])
+; SSE-NEXT: store i64 [[R0]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 0), align 8
+; SSE-NEXT: store i64 [[R1]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 1), align 8
+; SSE-NEXT: store i64 [[R2]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2), align 8
+; SSE-NEXT: store i64 [[R3]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 3), align 8
+; SSE-NEXT: store i64 [[R4]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4), align 8
+; SSE-NEXT: store i64 [[R5]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 5), align 8
+; SSE-NEXT: store i64 [[R6]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6), align 8
+; SSE-NEXT: store i64 [[R7]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 7), align 8
+; SSE-NEXT: ret void
+;
+; SLM-LABEL: @sub_v8i64(
+; SLM-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @a64 to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP2:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @b64 to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP9:%.*]] = call <2 x i64> @llvm.usub.sat.v2i64(<2 x i64> [[TMP1]], <2 x i64> [[TMP5]])
+; SLM-NEXT: [[TMP10:%.*]] = call <2 x i64> @llvm.usub.sat.v2i64(<2 x i64> [[TMP2]], <2 x i64> [[TMP6]])
+; SLM-NEXT: [[TMP11:%.*]] = call <2 x i64> @llvm.usub.sat.v2i64(<2 x i64> [[TMP3]], <2 x i64> [[TMP7]])
+; SLM-NEXT: [[TMP12:%.*]] = call <2 x i64> @llvm.usub.sat.v2i64(<2 x i64> [[TMP4]], <2 x i64> [[TMP8]])
+; SLM-NEXT: store <2 x i64> [[TMP9]], <2 x i64>* bitcast ([8 x i64]* @c64 to <2 x i64>*), align 8
+; SLM-NEXT: store <2 x i64> [[TMP10]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2) to <2 x i64>*), align 8
+; SLM-NEXT: store <2 x i64> [[TMP11]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4) to <2 x i64>*), align 8
+; SLM-NEXT: store <2 x i64> [[TMP12]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6) to <2 x i64>*), align 8
+; SLM-NEXT: ret void
+;
+; AVX-LABEL: @sub_v8i64(
+; AVX-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @a64 to <4 x i64>*), align 8
+; AVX-NEXT: [[TMP2:%.*]] = load <4 x i64>, <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX-NEXT: [[TMP3:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @b64 to <4 x i64>*), align 8
+; AVX-NEXT: [[TMP4:%.*]] = load <4 x i64>, <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX-NEXT: [[TMP5:%.*]] = call <4 x i64> @llvm.usub.sat.v4i64(<4 x i64> [[TMP1]], <4 x i64> [[TMP3]])
+; AVX-NEXT: [[TMP6:%.*]] = call <4 x i64> @llvm.usub.sat.v4i64(<4 x i64> [[TMP2]], <4 x i64> [[TMP4]])
+; AVX-NEXT: store <4 x i64> [[TMP5]], <4 x i64>* bitcast ([8 x i64]* @c64 to <4 x i64>*), align 8
+; AVX-NEXT: store <4 x i64> [[TMP6]], <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX-NEXT: ret void
+;
+; AVX512-LABEL: @sub_v8i64(
+; AVX512-NEXT: [[TMP1:%.*]] = load <8 x i64>, <8 x i64>* bitcast ([8 x i64]* @a64 to <8 x i64>*), align 8
+; AVX512-NEXT: [[TMP2:%.*]] = load <8 x i64>, <8 x i64>* bitcast ([8 x i64]* @b64 to <8 x i64>*), align 8
+; AVX512-NEXT: [[TMP3:%.*]] = call <8 x i64> @llvm.usub.sat.v8i64(<8 x i64> [[TMP1]], <8 x i64> [[TMP2]])
+; AVX512-NEXT: store <8 x i64> [[TMP3]], <8 x i64>* bitcast ([8 x i64]* @c64 to <8 x i64>*), align 8
+; AVX512-NEXT: ret void
+;
+ %a0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 0), align 8
+ %a1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 1), align 8
+ %a2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2), align 8
+ %a3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 3), align 8
+ %a4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4), align 8
+ %a5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 5), align 8
+ %a6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6), align 8
+ %a7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 7), align 8
+ %b0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 0), align 8
+ %b1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 1), align 8
+ %b2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2), align 8
+ %b3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 3), align 8
+ %b4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4), align 8
+ %b5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 5), align 8
+ %b6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6), align 8
+ %b7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 7), align 8
+ %r0 = call i64 @llvm.usub.sat.i64(i64 %a0, i64 %b0)
+ %r1 = call i64 @llvm.usub.sat.i64(i64 %a1, i64 %b1)
+ %r2 = call i64 @llvm.usub.sat.i64(i64 %a2, i64 %b2)
+ %r3 = call i64 @llvm.usub.sat.i64(i64 %a3, i64 %b3)
+ %r4 = call i64 @llvm.usub.sat.i64(i64 %a4, i64 %b4)
+ %r5 = call i64 @llvm.usub.sat.i64(i64 %a5, i64 %b5)
+ %r6 = call i64 @llvm.usub.sat.i64(i64 %a6, i64 %b6)
+ %r7 = call i64 @llvm.usub.sat.i64(i64 %a7, i64 %b7)
+ store i64 %r0, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 0), align 8
+ store i64 %r1, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 1), align 8
+ store i64 %r2, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2), align 8
+ store i64 %r3, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 3), align 8
+ store i64 %r4, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4), align 8
+ store i64 %r5, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 5), align 8
+ store i64 %r6, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6), align 8
+ store i64 %r7, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 7), align 8
+ ret void
+}
+
+define void @sub_v16i32() {
+; SSE-LABEL: @sub_v16i32(
+; SSE-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @a32 to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @b32 to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP9:%.*]] = call <4 x i32> @llvm.usub.sat.v4i32(<4 x i32> [[TMP1]], <4 x i32> [[TMP5]])
+; SSE-NEXT: [[TMP10:%.*]] = call <4 x i32> @llvm.usub.sat.v4i32(<4 x i32> [[TMP2]], <4 x i32> [[TMP6]])
+; SSE-NEXT: [[TMP11:%.*]] = call <4 x i32> @llvm.usub.sat.v4i32(<4 x i32> [[TMP3]], <4 x i32> [[TMP7]])
+; SSE-NEXT: [[TMP12:%.*]] = call <4 x i32> @llvm.usub.sat.v4i32(<4 x i32> [[TMP4]], <4 x i32> [[TMP8]])
+; SSE-NEXT: store <4 x i32> [[TMP9]], <4 x i32>* bitcast ([16 x i32]* @c32 to <4 x i32>*), align 4
+; SSE-NEXT: store <4 x i32> [[TMP10]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4) to <4 x i32>*), align 4
+; SSE-NEXT: store <4 x i32> [[TMP11]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8) to <4 x i32>*), align 4
+; SSE-NEXT: store <4 x i32> [[TMP12]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12) to <4 x i32>*), align 4
+; SSE-NEXT: ret void
+;
+; SLM-LABEL: @sub_v16i32(
+; SLM-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @a32 to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @b32 to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP9:%.*]] = call <4 x i32> @llvm.usub.sat.v4i32(<4 x i32> [[TMP1]], <4 x i32> [[TMP5]])
+; SLM-NEXT: [[TMP10:%.*]] = call <4 x i32> @llvm.usub.sat.v4i32(<4 x i32> [[TMP2]], <4 x i32> [[TMP6]])
+; SLM-NEXT: [[TMP11:%.*]] = call <4 x i32> @llvm.usub.sat.v4i32(<4 x i32> [[TMP3]], <4 x i32> [[TMP7]])
+; SLM-NEXT: [[TMP12:%.*]] = call <4 x i32> @llvm.usub.sat.v4i32(<4 x i32> [[TMP4]], <4 x i32> [[TMP8]])
+; SLM-NEXT: store <4 x i32> [[TMP9]], <4 x i32>* bitcast ([16 x i32]* @c32 to <4 x i32>*), align 4
+; SLM-NEXT: store <4 x i32> [[TMP10]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4) to <4 x i32>*), align 4
+; SLM-NEXT: store <4 x i32> [[TMP11]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8) to <4 x i32>*), align 4
+; SLM-NEXT: store <4 x i32> [[TMP12]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12) to <4 x i32>*), align 4
+; SLM-NEXT: ret void
+;
+; AVX-LABEL: @sub_v16i32(
+; AVX-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([16 x i32]* @a32 to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8) to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP3:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([16 x i32]* @b32 to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP4:%.*]] = load <8 x i32>, <8 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8) to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP5:%.*]] = call <8 x i32> @llvm.usub.sat.v8i32(<8 x i32> [[TMP1]], <8 x i32> [[TMP3]])
+; AVX-NEXT: [[TMP6:%.*]] = call <8 x i32> @llvm.usub.sat.v8i32(<8 x i32> [[TMP2]], <8 x i32> [[TMP4]])
+; AVX-NEXT: store <8 x i32> [[TMP5]], <8 x i32>* bitcast ([16 x i32]* @c32 to <8 x i32>*), align 4
+; AVX-NEXT: store <8 x i32> [[TMP6]], <8 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8) to <8 x i32>*), align 4
+; AVX-NEXT: ret void
+;
+; AVX512-LABEL: @sub_v16i32(
+; AVX512-NEXT: [[TMP1:%.*]] = load <16 x i32>, <16 x i32>* bitcast ([16 x i32]* @a32 to <16 x i32>*), align 4
+; AVX512-NEXT: [[TMP2:%.*]] = load <16 x i32>, <16 x i32>* bitcast ([16 x i32]* @b32 to <16 x i32>*), align 4
+; AVX512-NEXT: [[TMP3:%.*]] = call <16 x i32> @llvm.usub.sat.v16i32(<16 x i32> [[TMP1]], <16 x i32> [[TMP2]])
+; AVX512-NEXT: store <16 x i32> [[TMP3]], <16 x i32>* bitcast ([16 x i32]* @c32 to <16 x i32>*), align 4
+; AVX512-NEXT: ret void
+;
+ %a0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 0 ), align 4
+ %a1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 1 ), align 4
+ %a2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 2 ), align 4
+ %a3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 3 ), align 4
+ %a4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4 ), align 4
+ %a5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 5 ), align 4
+ %a6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 6 ), align 4
+ %a7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 7 ), align 4
+ %a8 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8 ), align 4
+ %a9 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 9 ), align 4
+ %a10 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 10), align 4
+ %a11 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 11), align 4
+ %a12 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12), align 4
+ %a13 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 13), align 4
+ %a14 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 14), align 4
+ %a15 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 15), align 4
+ %b0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 0 ), align 4
+ %b1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 1 ), align 4
+ %b2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 2 ), align 4
+ %b3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 3 ), align 4
+ %b4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4 ), align 4
+ %b5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 5 ), align 4
+ %b6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 6 ), align 4
+ %b7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 7 ), align 4
+ %b8 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8 ), align 4
+ %b9 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 9 ), align 4
+ %b10 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 10), align 4
+ %b11 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 11), align 4
+ %b12 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12), align 4
+ %b13 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 13), align 4
+ %b14 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 14), align 4
+ %b15 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 15), align 4
+ %r0 = call i32 @llvm.usub.sat.i32(i32 %a0 , i32 %b0 )
+ %r1 = call i32 @llvm.usub.sat.i32(i32 %a1 , i32 %b1 )
+ %r2 = call i32 @llvm.usub.sat.i32(i32 %a2 , i32 %b2 )
+ %r3 = call i32 @llvm.usub.sat.i32(i32 %a3 , i32 %b3 )
+ %r4 = call i32 @llvm.usub.sat.i32(i32 %a4 , i32 %b4 )
+ %r5 = call i32 @llvm.usub.sat.i32(i32 %a5 , i32 %b5 )
+ %r6 = call i32 @llvm.usub.sat.i32(i32 %a6 , i32 %b6 )
+ %r7 = call i32 @llvm.usub.sat.i32(i32 %a7 , i32 %b7 )
+ %r8 = call i32 @llvm.usub.sat.i32(i32 %a8 , i32 %b8 )
+ %r9 = call i32 @llvm.usub.sat.i32(i32 %a9 , i32 %b9 )
+ %r10 = call i32 @llvm.usub.sat.i32(i32 %a10, i32 %b10)
+ %r11 = call i32 @llvm.usub.sat.i32(i32 %a11, i32 %b11)
+ %r12 = call i32 @llvm.usub.sat.i32(i32 %a12, i32 %b12)
+ %r13 = call i32 @llvm.usub.sat.i32(i32 %a13, i32 %b13)
+ %r14 = call i32 @llvm.usub.sat.i32(i32 %a14, i32 %b14)
+ %r15 = call i32 @llvm.usub.sat.i32(i32 %a15, i32 %b15)
+ store i32 %r0 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 0 ), align 4
+ store i32 %r1 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 1 ), align 4
+ store i32 %r2 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 2 ), align 4
+ store i32 %r3 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 3 ), align 4
+ store i32 %r4 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4 ), align 4
+ store i32 %r5 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 5 ), align 4
+ store i32 %r6 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 6 ), align 4
+ store i32 %r7 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 7 ), align 4
+ store i32 %r8 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8 ), align 4
+ store i32 %r9 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 9 ), align 4
+ store i32 %r10, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 10), align 4
+ store i32 %r11, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 11), align 4
+ store i32 %r12, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12), align 4
+ store i32 %r13, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 13), align 4
+ store i32 %r14, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 14), align 4
+ store i32 %r15, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 15), align 4
+ ret void
+}
+
+define void @sub_v32i16() {
+; SSE-LABEL: @sub_v32i16(
+; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @a16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @b16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP9:%.*]] = call <8 x i16> @llvm.usub.sat.v8i16(<8 x i16> [[TMP1]], <8 x i16> [[TMP5]])
+; SSE-NEXT: [[TMP10:%.*]] = call <8 x i16> @llvm.usub.sat.v8i16(<8 x i16> [[TMP2]], <8 x i16> [[TMP6]])
+; SSE-NEXT: [[TMP11:%.*]] = call <8 x i16> @llvm.usub.sat.v8i16(<8 x i16> [[TMP3]], <8 x i16> [[TMP7]])
+; SSE-NEXT: [[TMP12:%.*]] = call <8 x i16> @llvm.usub.sat.v8i16(<8 x i16> [[TMP4]], <8 x i16> [[TMP8]])
+; SSE-NEXT: store <8 x i16> [[TMP9]], <8 x i16>* bitcast ([32 x i16]* @c16 to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP10]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP11]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP12]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24) to <8 x i16>*), align 2
+; SSE-NEXT: ret void
+;
+; SLM-LABEL: @sub_v32i16(
+; SLM-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @a16 to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @b16 to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP9:%.*]] = call <8 x i16> @llvm.usub.sat.v8i16(<8 x i16> [[TMP1]], <8 x i16> [[TMP5]])
+; SLM-NEXT: [[TMP10:%.*]] = call <8 x i16> @llvm.usub.sat.v8i16(<8 x i16> [[TMP2]], <8 x i16> [[TMP6]])
+; SLM-NEXT: [[TMP11:%.*]] = call <8 x i16> @llvm.usub.sat.v8i16(<8 x i16> [[TMP3]], <8 x i16> [[TMP7]])
+; SLM-NEXT: [[TMP12:%.*]] = call <8 x i16> @llvm.usub.sat.v8i16(<8 x i16> [[TMP4]], <8 x i16> [[TMP8]])
+; SLM-NEXT: store <8 x i16> [[TMP9]], <8 x i16>* bitcast ([32 x i16]* @c16 to <8 x i16>*), align 2
+; SLM-NEXT: store <8 x i16> [[TMP10]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8) to <8 x i16>*), align 2
+; SLM-NEXT: store <8 x i16> [[TMP11]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <8 x i16>*), align 2
+; SLM-NEXT: store <8 x i16> [[TMP12]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24) to <8 x i16>*), align 2
+; SLM-NEXT: ret void
+;
+; AVX-LABEL: @sub_v32i16(
+; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @a16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP3:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @b16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP4:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP5:%.*]] = call <16 x i16> @llvm.usub.sat.v16i16(<16 x i16> [[TMP1]], <16 x i16> [[TMP3]])
+; AVX-NEXT: [[TMP6:%.*]] = call <16 x i16> @llvm.usub.sat.v16i16(<16 x i16> [[TMP2]], <16 x i16> [[TMP4]])
+; AVX-NEXT: store <16 x i16> [[TMP5]], <16 x i16>* bitcast ([32 x i16]* @c16 to <16 x i16>*), align 2
+; AVX-NEXT: store <16 x i16> [[TMP6]], <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX-NEXT: ret void
+;
+; AVX512-LABEL: @sub_v32i16(
+; AVX512-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @a16 to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP2:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP3:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @b16 to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP4:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP5:%.*]] = call <16 x i16> @llvm.usub.sat.v16i16(<16 x i16> [[TMP1]], <16 x i16> [[TMP3]])
+; AVX512-NEXT: [[TMP6:%.*]] = call <16 x i16> @llvm.usub.sat.v16i16(<16 x i16> [[TMP2]], <16 x i16> [[TMP4]])
+; AVX512-NEXT: store <16 x i16> [[TMP5]], <16 x i16>* bitcast ([32 x i16]* @c16 to <16 x i16>*), align 2
+; AVX512-NEXT: store <16 x i16> [[TMP6]], <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX512-NEXT: ret void
+;
+ %a0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 0 ), align 2
+ %a1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 1 ), align 2
+ %a2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 2 ), align 2
+ %a3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 3 ), align 2
+ %a4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 4 ), align 2
+ %a5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 5 ), align 2
+ %a6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 6 ), align 2
+ %a7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 7 ), align 2
+ %a8 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8 ), align 2
+ %a9 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 9 ), align 2
+ %a10 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 10), align 2
+ %a11 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 11), align 2
+ %a12 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 12), align 2
+ %a13 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 13), align 2
+ %a14 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 14), align 2
+ %a15 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 15), align 2
+ %a16 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16), align 2
+ %a17 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 17), align 2
+ %a18 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 18), align 2
+ %a19 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 19), align 2
+ %a20 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 20), align 2
+ %a21 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 21), align 2
+ %a22 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 22), align 2
+ %a23 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 23), align 2
+ %a24 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24), align 2
+ %a25 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 25), align 2
+ %a26 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 26), align 2
+ %a27 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 27), align 2
+ %a28 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 28), align 2
+ %a29 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 29), align 2
+ %a30 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 30), align 2
+ %a31 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 31), align 2
+ %b0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 0 ), align 2
+ %b1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 1 ), align 2
+ %b2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 2 ), align 2
+ %b3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 3 ), align 2
+ %b4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 4 ), align 2
+ %b5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 5 ), align 2
+ %b6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 6 ), align 2
+ %b7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 7 ), align 2
+ %b8 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8 ), align 2
+ %b9 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 9 ), align 2
+ %b10 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 10), align 2
+ %b11 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 11), align 2
+ %b12 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 12), align 2
+ %b13 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 13), align 2
+ %b14 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 14), align 2
+ %b15 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 15), align 2
+ %b16 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16), align 2
+ %b17 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 17), align 2
+ %b18 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 18), align 2
+ %b19 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 19), align 2
+ %b20 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 20), align 2
+ %b21 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 21), align 2
+ %b22 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 22), align 2
+ %b23 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 23), align 2
+ %b24 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24), align 2
+ %b25 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 25), align 2
+ %b26 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 26), align 2
+ %b27 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 27), align 2
+ %b28 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 28), align 2
+ %b29 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 29), align 2
+ %b30 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 30), align 2
+ %b31 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 31), align 2
+ %r0 = call i16 @llvm.usub.sat.i16(i16 %a0 , i16 %b0 )
+ %r1 = call i16 @llvm.usub.sat.i16(i16 %a1 , i16 %b1 )
+ %r2 = call i16 @llvm.usub.sat.i16(i16 %a2 , i16 %b2 )
+ %r3 = call i16 @llvm.usub.sat.i16(i16 %a3 , i16 %b3 )
+ %r4 = call i16 @llvm.usub.sat.i16(i16 %a4 , i16 %b4 )
+ %r5 = call i16 @llvm.usub.sat.i16(i16 %a5 , i16 %b5 )
+ %r6 = call i16 @llvm.usub.sat.i16(i16 %a6 , i16 %b6 )
+ %r7 = call i16 @llvm.usub.sat.i16(i16 %a7 , i16 %b7 )
+ %r8 = call i16 @llvm.usub.sat.i16(i16 %a8 , i16 %b8 )
+ %r9 = call i16 @llvm.usub.sat.i16(i16 %a9 , i16 %b9 )
+ %r10 = call i16 @llvm.usub.sat.i16(i16 %a10, i16 %b10)
+ %r11 = call i16 @llvm.usub.sat.i16(i16 %a11, i16 %b11)
+ %r12 = call i16 @llvm.usub.sat.i16(i16 %a12, i16 %b12)
+ %r13 = call i16 @llvm.usub.sat.i16(i16 %a13, i16 %b13)
+ %r14 = call i16 @llvm.usub.sat.i16(i16 %a14, i16 %b14)
+ %r15 = call i16 @llvm.usub.sat.i16(i16 %a15, i16 %b15)
+ %r16 = call i16 @llvm.usub.sat.i16(i16 %a16, i16 %b16)
+ %r17 = call i16 @llvm.usub.sat.i16(i16 %a17, i16 %b17)
+ %r18 = call i16 @llvm.usub.sat.i16(i16 %a18, i16 %b18)
+ %r19 = call i16 @llvm.usub.sat.i16(i16 %a19, i16 %b19)
+ %r20 = call i16 @llvm.usub.sat.i16(i16 %a20, i16 %b20)
+ %r21 = call i16 @llvm.usub.sat.i16(i16 %a21, i16 %b21)
+ %r22 = call i16 @llvm.usub.sat.i16(i16 %a22, i16 %b22)
+ %r23 = call i16 @llvm.usub.sat.i16(i16 %a23, i16 %b23)
+ %r24 = call i16 @llvm.usub.sat.i16(i16 %a24, i16 %b24)
+ %r25 = call i16 @llvm.usub.sat.i16(i16 %a25, i16 %b25)
+ %r26 = call i16 @llvm.usub.sat.i16(i16 %a26, i16 %b26)
+ %r27 = call i16 @llvm.usub.sat.i16(i16 %a27, i16 %b27)
+ %r28 = call i16 @llvm.usub.sat.i16(i16 %a28, i16 %b28)
+ %r29 = call i16 @llvm.usub.sat.i16(i16 %a29, i16 %b29)
+ %r30 = call i16 @llvm.usub.sat.i16(i16 %a30, i16 %b30)
+ %r31 = call i16 @llvm.usub.sat.i16(i16 %a31, i16 %b31)
+ store i16 %r0 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 0 ), align 2
+ store i16 %r1 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 1 ), align 2
+ store i16 %r2 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 2 ), align 2
+ store i16 %r3 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 3 ), align 2
+ store i16 %r4 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 4 ), align 2
+ store i16 %r5 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 5 ), align 2
+ store i16 %r6 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 6 ), align 2
+ store i16 %r7 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 7 ), align 2
+ store i16 %r8 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8 ), align 2
+ store i16 %r9 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 9 ), align 2
+ store i16 %r10, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 10), align 2
+ store i16 %r11, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 11), align 2
+ store i16 %r12, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 12), align 2
+ store i16 %r13, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 13), align 2
+ store i16 %r14, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 14), align 2
+ store i16 %r15, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 15), align 2
+ store i16 %r16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16), align 2
+ store i16 %r17, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 17), align 2
+ store i16 %r18, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 18), align 2
+ store i16 %r19, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 19), align 2
+ store i16 %r20, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 20), align 2
+ store i16 %r21, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 21), align 2
+ store i16 %r22, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 22), align 2
+ store i16 %r23, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 23), align 2
+ store i16 %r24, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24), align 2
+ store i16 %r25, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 25), align 2
+ store i16 %r26, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 26), align 2
+ store i16 %r27, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 27), align 2
+ store i16 %r28, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 28), align 2
+ store i16 %r29, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 29), align 2
+ store i16 %r30, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 30), align 2
+ store i16 %r31, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 31), align 2
+ ret void
+}
+
+define void @sub_v64i8() {
+; CHECK-LABEL: @sub_v64i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([64 x i8]* @a8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 32) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 48) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([64 x i8]* @b8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP6:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP7:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 32) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP8:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 48) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP9:%.*]] = call <16 x i8> @llvm.usub.sat.v16i8(<16 x i8> [[TMP1]], <16 x i8> [[TMP5]])
+; CHECK-NEXT: [[TMP10:%.*]] = call <16 x i8> @llvm.usub.sat.v16i8(<16 x i8> [[TMP2]], <16 x i8> [[TMP6]])
+; CHECK-NEXT: [[TMP11:%.*]] = call <16 x i8> @llvm.usub.sat.v16i8(<16 x i8> [[TMP3]], <16 x i8> [[TMP7]])
+; CHECK-NEXT: [[TMP12:%.*]] = call <16 x i8> @llvm.usub.sat.v16i8(<16 x i8> [[TMP4]], <16 x i8> [[TMP8]])
+; CHECK-NEXT: store <16 x i8> [[TMP9]], <16 x i8>* bitcast ([64 x i8]* @c8 to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP10]], <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP11]], <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 32) to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP12]], <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 48) to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %a0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 0 ), align 1
+ %a1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 1 ), align 1
+ %a2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 2 ), align 1
+ %a3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 3 ), align 1
+ %a4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 4 ), align 1
+ %a5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 5 ), align 1
+ %a6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 6 ), align 1
+ %a7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 7 ), align 1
+ %a8 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 8 ), align 1
+ %a9 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 9 ), align 1
+ %a10 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 10), align 1
+ %a11 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 11), align 1
+ %a12 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 12), align 1
+ %a13 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 13), align 1
+ %a14 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 14), align 1
+ %a15 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 15), align 1
+ %a16 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 16), align 1
+ %a17 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 17), align 1
+ %a18 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 18), align 1
+ %a19 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 19), align 1
+ %a20 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 20), align 1
+ %a21 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 21), align 1
+ %a22 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 22), align 1
+ %a23 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 23), align 1
+ %a24 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 24), align 1
+ %a25 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 25), align 1
+ %a26 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 26), align 1
+ %a27 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 27), align 1
+ %a28 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 28), align 1
+ %a29 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 29), align 1
+ %a30 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 30), align 1
+ %a31 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 31), align 1
+ %a32 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 32), align 1
+ %a33 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 33), align 1
+ %a34 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 34), align 1
+ %a35 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 35), align 1
+ %a36 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 36), align 1
+ %a37 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 37), align 1
+ %a38 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 38), align 1
+ %a39 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 39), align 1
+ %a40 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 40), align 1
+ %a41 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 41), align 1
+ %a42 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 42), align 1
+ %a43 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 43), align 1
+ %a44 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 44), align 1
+ %a45 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 45), align 1
+ %a46 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 46), align 1
+ %a47 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 47), align 1
+ %a48 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 48), align 1
+ %a49 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 49), align 1
+ %a50 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 50), align 1
+ %a51 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 51), align 1
+ %a52 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 52), align 1
+ %a53 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 53), align 1
+ %a54 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 54), align 1
+ %a55 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 55), align 1
+ %a56 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 56), align 1
+ %a57 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 57), align 1
+ %a58 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 58), align 1
+ %a59 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 59), align 1
+ %a60 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 60), align 1
+ %a61 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 61), align 1
+ %a62 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 62), align 1
+ %a63 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 63), align 1
+ %b0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 0 ), align 1
+ %b1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 1 ), align 1
+ %b2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 2 ), align 1
+ %b3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 3 ), align 1
+ %b4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 4 ), align 1
+ %b5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 5 ), align 1
+ %b6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 6 ), align 1
+ %b7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 7 ), align 1
+ %b8 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 8 ), align 1
+ %b9 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 9 ), align 1
+ %b10 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 10), align 1
+ %b11 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 11), align 1
+ %b12 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 12), align 1
+ %b13 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 13), align 1
+ %b14 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 14), align 1
+ %b15 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 15), align 1
+ %b16 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 16), align 1
+ %b17 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 17), align 1
+ %b18 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 18), align 1
+ %b19 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 19), align 1
+ %b20 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 20), align 1
+ %b21 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 21), align 1
+ %b22 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 22), align 1
+ %b23 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 23), align 1
+ %b24 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 24), align 1
+ %b25 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 25), align 1
+ %b26 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 26), align 1
+ %b27 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 27), align 1
+ %b28 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 28), align 1
+ %b29 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 29), align 1
+ %b30 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 30), align 1
+ %b31 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 31), align 1
+ %b32 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 32), align 1
+ %b33 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 33), align 1
+ %b34 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 34), align 1
+ %b35 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 35), align 1
+ %b36 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 36), align 1
+ %b37 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 37), align 1
+ %b38 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 38), align 1
+ %b39 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 39), align 1
+ %b40 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 40), align 1
+ %b41 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 41), align 1
+ %b42 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 42), align 1
+ %b43 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 43), align 1
+ %b44 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 44), align 1
+ %b45 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 45), align 1
+ %b46 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 46), align 1
+ %b47 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 47), align 1
+ %b48 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 48), align 1
+ %b49 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 49), align 1
+ %b50 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 50), align 1
+ %b51 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 51), align 1
+ %b52 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 52), align 1
+ %b53 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 53), align 1
+ %b54 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 54), align 1
+ %b55 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 55), align 1
+ %b56 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 56), align 1
+ %b57 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 57), align 1
+ %b58 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 58), align 1
+ %b59 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 59), align 1
+ %b60 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 60), align 1
+ %b61 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 61), align 1
+ %b62 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 62), align 1
+ %b63 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 63), align 1
+ %r0 = call i8 @llvm.usub.sat.i8(i8 %a0 , i8 %b0 )
+ %r1 = call i8 @llvm.usub.sat.i8(i8 %a1 , i8 %b1 )
+ %r2 = call i8 @llvm.usub.sat.i8(i8 %a2 , i8 %b2 )
+ %r3 = call i8 @llvm.usub.sat.i8(i8 %a3 , i8 %b3 )
+ %r4 = call i8 @llvm.usub.sat.i8(i8 %a4 , i8 %b4 )
+ %r5 = call i8 @llvm.usub.sat.i8(i8 %a5 , i8 %b5 )
+ %r6 = call i8 @llvm.usub.sat.i8(i8 %a6 , i8 %b6 )
+ %r7 = call i8 @llvm.usub.sat.i8(i8 %a7 , i8 %b7 )
+ %r8 = call i8 @llvm.usub.sat.i8(i8 %a8 , i8 %b8 )
+ %r9 = call i8 @llvm.usub.sat.i8(i8 %a9 , i8 %b9 )
+ %r10 = call i8 @llvm.usub.sat.i8(i8 %a10, i8 %b10)
+ %r11 = call i8 @llvm.usub.sat.i8(i8 %a11, i8 %b11)
+ %r12 = call i8 @llvm.usub.sat.i8(i8 %a12, i8 %b12)
+ %r13 = call i8 @llvm.usub.sat.i8(i8 %a13, i8 %b13)
+ %r14 = call i8 @llvm.usub.sat.i8(i8 %a14, i8 %b14)
+ %r15 = call i8 @llvm.usub.sat.i8(i8 %a15, i8 %b15)
+ %r16 = call i8 @llvm.usub.sat.i8(i8 %a16, i8 %b16)
+ %r17 = call i8 @llvm.usub.sat.i8(i8 %a17, i8 %b17)
+ %r18 = call i8 @llvm.usub.sat.i8(i8 %a18, i8 %b18)
+ %r19 = call i8 @llvm.usub.sat.i8(i8 %a19, i8 %b19)
+ %r20 = call i8 @llvm.usub.sat.i8(i8 %a20, i8 %b20)
+ %r21 = call i8 @llvm.usub.sat.i8(i8 %a21, i8 %b21)
+ %r22 = call i8 @llvm.usub.sat.i8(i8 %a22, i8 %b22)
+ %r23 = call i8 @llvm.usub.sat.i8(i8 %a23, i8 %b23)
+ %r24 = call i8 @llvm.usub.sat.i8(i8 %a24, i8 %b24)
+ %r25 = call i8 @llvm.usub.sat.i8(i8 %a25, i8 %b25)
+ %r26 = call i8 @llvm.usub.sat.i8(i8 %a26, i8 %b26)
+ %r27 = call i8 @llvm.usub.sat.i8(i8 %a27, i8 %b27)
+ %r28 = call i8 @llvm.usub.sat.i8(i8 %a28, i8 %b28)
+ %r29 = call i8 @llvm.usub.sat.i8(i8 %a29, i8 %b29)
+ %r30 = call i8 @llvm.usub.sat.i8(i8 %a30, i8 %b30)
+ %r31 = call i8 @llvm.usub.sat.i8(i8 %a31, i8 %b31)
+ %r32 = call i8 @llvm.usub.sat.i8(i8 %a32, i8 %b32)
+ %r33 = call i8 @llvm.usub.sat.i8(i8 %a33, i8 %b33)
+ %r34 = call i8 @llvm.usub.sat.i8(i8 %a34, i8 %b34)
+ %r35 = call i8 @llvm.usub.sat.i8(i8 %a35, i8 %b35)
+ %r36 = call i8 @llvm.usub.sat.i8(i8 %a36, i8 %b36)
+ %r37 = call i8 @llvm.usub.sat.i8(i8 %a37, i8 %b37)
+ %r38 = call i8 @llvm.usub.sat.i8(i8 %a38, i8 %b38)
+ %r39 = call i8 @llvm.usub.sat.i8(i8 %a39, i8 %b39)
+ %r40 = call i8 @llvm.usub.sat.i8(i8 %a40, i8 %b40)
+ %r41 = call i8 @llvm.usub.sat.i8(i8 %a41, i8 %b41)
+ %r42 = call i8 @llvm.usub.sat.i8(i8 %a42, i8 %b42)
+ %r43 = call i8 @llvm.usub.sat.i8(i8 %a43, i8 %b43)
+ %r44 = call i8 @llvm.usub.sat.i8(i8 %a44, i8 %b44)
+ %r45 = call i8 @llvm.usub.sat.i8(i8 %a45, i8 %b45)
+ %r46 = call i8 @llvm.usub.sat.i8(i8 %a46, i8 %b46)
+ %r47 = call i8 @llvm.usub.sat.i8(i8 %a47, i8 %b47)
+ %r48 = call i8 @llvm.usub.sat.i8(i8 %a48, i8 %b48)
+ %r49 = call i8 @llvm.usub.sat.i8(i8 %a49, i8 %b49)
+ %r50 = call i8 @llvm.usub.sat.i8(i8 %a50, i8 %b50)
+ %r51 = call i8 @llvm.usub.sat.i8(i8 %a51, i8 %b51)
+ %r52 = call i8 @llvm.usub.sat.i8(i8 %a52, i8 %b52)
+ %r53 = call i8 @llvm.usub.sat.i8(i8 %a53, i8 %b53)
+ %r54 = call i8 @llvm.usub.sat.i8(i8 %a54, i8 %b54)
+ %r55 = call i8 @llvm.usub.sat.i8(i8 %a55, i8 %b55)
+ %r56 = call i8 @llvm.usub.sat.i8(i8 %a56, i8 %b56)
+ %r57 = call i8 @llvm.usub.sat.i8(i8 %a57, i8 %b57)
+ %r58 = call i8 @llvm.usub.sat.i8(i8 %a58, i8 %b58)
+ %r59 = call i8 @llvm.usub.sat.i8(i8 %a59, i8 %b59)
+ %r60 = call i8 @llvm.usub.sat.i8(i8 %a60, i8 %b60)
+ %r61 = call i8 @llvm.usub.sat.i8(i8 %a61, i8 %b61)
+ %r62 = call i8 @llvm.usub.sat.i8(i8 %a62, i8 %b62)
+ %r63 = call i8 @llvm.usub.sat.i8(i8 %a63, i8 %b63)
+ store i8 %r0 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 0 ), align 1
+ store i8 %r1 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 1 ), align 1
+ store i8 %r2 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 2 ), align 1
+ store i8 %r3 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 3 ), align 1
+ store i8 %r4 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 4 ), align 1
+ store i8 %r5 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 5 ), align 1
+ store i8 %r6 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 6 ), align 1
+ store i8 %r7 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 7 ), align 1
+ store i8 %r8 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 8 ), align 1
+ store i8 %r9 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 9 ), align 1
+ store i8 %r10, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 10), align 1
+ store i8 %r11, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 11), align 1
+ store i8 %r12, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 12), align 1
+ store i8 %r13, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 13), align 1
+ store i8 %r14, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 14), align 1
+ store i8 %r15, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 15), align 1
+ store i8 %r16, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 16), align 1
+ store i8 %r17, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 17), align 1
+ store i8 %r18, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 18), align 1
+ store i8 %r19, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 19), align 1
+ store i8 %r20, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 20), align 1
+ store i8 %r21, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 21), align 1
+ store i8 %r22, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 22), align 1
+ store i8 %r23, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 23), align 1
+ store i8 %r24, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 24), align 1
+ store i8 %r25, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 25), align 1
+ store i8 %r26, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 26), align 1
+ store i8 %r27, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 27), align 1
+ store i8 %r28, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 28), align 1
+ store i8 %r29, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 29), align 1
+ store i8 %r30, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 30), align 1
+ store i8 %r31, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 31), align 1
+ store i8 %r32, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 32), align 1
+ store i8 %r33, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 33), align 1
+ store i8 %r34, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 34), align 1
+ store i8 %r35, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 35), align 1
+ store i8 %r36, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 36), align 1
+ store i8 %r37, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 37), align 1
+ store i8 %r38, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 38), align 1
+ store i8 %r39, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 39), align 1
+ store i8 %r40, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 40), align 1
+ store i8 %r41, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 41), align 1
+ store i8 %r42, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 42), align 1
+ store i8 %r43, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 43), align 1
+ store i8 %r44, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 44), align 1
+ store i8 %r45, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 45), align 1
+ store i8 %r46, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 46), align 1
+ store i8 %r47, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 47), align 1
+ store i8 %r48, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 48), align 1
+ store i8 %r49, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 49), align 1
+ store i8 %r50, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 50), align 1
+ store i8 %r51, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 51), align 1
+ store i8 %r52, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 52), align 1
+ store i8 %r53, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 53), align 1
+ store i8 %r54, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 54), align 1
+ store i8 %r55, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 55), align 1
+ store i8 %r56, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 56), align 1
+ store i8 %r57, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 57), align 1
+ store i8 %r58, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 58), align 1
+ store i8 %r59, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 59), align 1
+ store i8 %r60, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 60), align 1
+ store i8 %r61, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 61), align 1
+ store i8 %r62, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 62), align 1
+ store i8 %r63, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 63), align 1
+ ret void
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-usubo.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-usubo.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-usubo.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub-usubo.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,1254 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,SSE
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=slm -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,SLM
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=knl -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skx -mattr=-prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skx -mattr=+prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,AVX,AVX256BW
+
+ at a64 = common global [8 x i64] zeroinitializer, align 64
+ at b64 = common global [8 x i64] zeroinitializer, align 64
+ at c64 = common global [8 x i64] zeroinitializer, align 64
+ at a32 = common global [16 x i32] zeroinitializer, align 64
+ at b32 = common global [16 x i32] zeroinitializer, align 64
+ at c32 = common global [16 x i32] zeroinitializer, align 64
+ at a16 = common global [32 x i16] zeroinitializer, align 64
+ at b16 = common global [32 x i16] zeroinitializer, align 64
+ at c16 = common global [32 x i16] zeroinitializer, align 64
+ at a8 = common global [64 x i8] zeroinitializer, align 64
+ at b8 = common global [64 x i8] zeroinitializer, align 64
+ at c8 = common global [64 x i8] zeroinitializer, align 64
+
+declare {i64, i1} @llvm.usub.with.overflow.i64(i64, i64)
+declare {i32, i1} @llvm.usub.with.overflow.i32(i32, i32)
+declare {i16, i1} @llvm.usub.with.overflow.i16(i16, i16)
+declare {i8 , i1} @llvm.usub.with.overflow.i8 (i8 , i8 )
+
+define void @sub_v8i64() {
+; CHECK-LABEL: @sub_v8i64(
+; CHECK-NEXT: [[A0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 0), align 8
+; CHECK-NEXT: [[A1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 1), align 8
+; CHECK-NEXT: [[A2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2), align 8
+; CHECK-NEXT: [[A3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 3), align 8
+; CHECK-NEXT: [[A4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4), align 8
+; CHECK-NEXT: [[A5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 5), align 8
+; CHECK-NEXT: [[A6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6), align 8
+; CHECK-NEXT: [[A7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 7), align 8
+; CHECK-NEXT: [[B0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 0), align 8
+; CHECK-NEXT: [[B1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 1), align 8
+; CHECK-NEXT: [[B2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2), align 8
+; CHECK-NEXT: [[B3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 3), align 8
+; CHECK-NEXT: [[B4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4), align 8
+; CHECK-NEXT: [[B5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 5), align 8
+; CHECK-NEXT: [[B6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6), align 8
+; CHECK-NEXT: [[B7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 7), align 8
+; CHECK-NEXT: [[C0:%.*]] = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 [[A0]], i64 [[B0]])
+; CHECK-NEXT: [[C1:%.*]] = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 [[A1]], i64 [[B1]])
+; CHECK-NEXT: [[C2:%.*]] = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 [[A2]], i64 [[B2]])
+; CHECK-NEXT: [[C3:%.*]] = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 [[A3]], i64 [[B3]])
+; CHECK-NEXT: [[C4:%.*]] = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 [[A4]], i64 [[B4]])
+; CHECK-NEXT: [[C5:%.*]] = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 [[A5]], i64 [[B5]])
+; CHECK-NEXT: [[C6:%.*]] = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 [[A6]], i64 [[B6]])
+; CHECK-NEXT: [[C7:%.*]] = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 [[A7]], i64 [[B7]])
+; CHECK-NEXT: [[R0:%.*]] = extractvalue { i64, i1 } [[C0]], 0
+; CHECK-NEXT: [[R1:%.*]] = extractvalue { i64, i1 } [[C1]], 0
+; CHECK-NEXT: [[R2:%.*]] = extractvalue { i64, i1 } [[C2]], 0
+; CHECK-NEXT: [[R3:%.*]] = extractvalue { i64, i1 } [[C3]], 0
+; CHECK-NEXT: [[R4:%.*]] = extractvalue { i64, i1 } [[C4]], 0
+; CHECK-NEXT: [[R5:%.*]] = extractvalue { i64, i1 } [[C5]], 0
+; CHECK-NEXT: [[R6:%.*]] = extractvalue { i64, i1 } [[C6]], 0
+; CHECK-NEXT: [[R7:%.*]] = extractvalue { i64, i1 } [[C7]], 0
+; CHECK-NEXT: store i64 [[R0]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 0), align 8
+; CHECK-NEXT: store i64 [[R1]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 1), align 8
+; CHECK-NEXT: store i64 [[R2]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2), align 8
+; CHECK-NEXT: store i64 [[R3]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 3), align 8
+; CHECK-NEXT: store i64 [[R4]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4), align 8
+; CHECK-NEXT: store i64 [[R5]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 5), align 8
+; CHECK-NEXT: store i64 [[R6]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6), align 8
+; CHECK-NEXT: store i64 [[R7]], i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 7), align 8
+; CHECK-NEXT: ret void
+;
+ %a0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 0), align 8
+ %a1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 1), align 8
+ %a2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2), align 8
+ %a3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 3), align 8
+ %a4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4), align 8
+ %a5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 5), align 8
+ %a6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6), align 8
+ %a7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 7), align 8
+ %b0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 0), align 8
+ %b1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 1), align 8
+ %b2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2), align 8
+ %b3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 3), align 8
+ %b4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4), align 8
+ %b5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 5), align 8
+ %b6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6), align 8
+ %b7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 7), align 8
+ %c0 = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a0, i64 %b0)
+ %c1 = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a1, i64 %b1)
+ %c2 = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a2, i64 %b2)
+ %c3 = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a3, i64 %b3)
+ %c4 = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a4, i64 %b4)
+ %c5 = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a5, i64 %b5)
+ %c6 = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a6, i64 %b6)
+ %c7 = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a7, i64 %b7)
+ %r0 = extractvalue {i64, i1} %c0, 0
+ %r1 = extractvalue {i64, i1} %c1, 0
+ %r2 = extractvalue {i64, i1} %c2, 0
+ %r3 = extractvalue {i64, i1} %c3, 0
+ %r4 = extractvalue {i64, i1} %c4, 0
+ %r5 = extractvalue {i64, i1} %c5, 0
+ %r6 = extractvalue {i64, i1} %c6, 0
+ %r7 = extractvalue {i64, i1} %c7, 0
+ store i64 %r0, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 0), align 8
+ store i64 %r1, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 1), align 8
+ store i64 %r2, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2), align 8
+ store i64 %r3, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 3), align 8
+ store i64 %r4, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4), align 8
+ store i64 %r5, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 5), align 8
+ store i64 %r6, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6), align 8
+ store i64 %r7, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 7), align 8
+ ret void
+}
+
+define void @sub_v16i32() {
+; CHECK-LABEL: @sub_v16i32(
+; CHECK-NEXT: [[A0:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 0), align 4
+; CHECK-NEXT: [[A1:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 1), align 4
+; CHECK-NEXT: [[A2:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 2), align 4
+; CHECK-NEXT: [[A3:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 3), align 4
+; CHECK-NEXT: [[A4:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4), align 4
+; CHECK-NEXT: [[A5:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 5), align 4
+; CHECK-NEXT: [[A6:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 6), align 4
+; CHECK-NEXT: [[A7:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 7), align 4
+; CHECK-NEXT: [[A8:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8), align 4
+; CHECK-NEXT: [[A9:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 9), align 4
+; CHECK-NEXT: [[A10:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 10), align 4
+; CHECK-NEXT: [[A11:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 11), align 4
+; CHECK-NEXT: [[A12:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12), align 4
+; CHECK-NEXT: [[A13:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 13), align 4
+; CHECK-NEXT: [[A14:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 14), align 4
+; CHECK-NEXT: [[A15:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 15), align 4
+; CHECK-NEXT: [[B0:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 0), align 4
+; CHECK-NEXT: [[B1:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 1), align 4
+; CHECK-NEXT: [[B2:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 2), align 4
+; CHECK-NEXT: [[B3:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 3), align 4
+; CHECK-NEXT: [[B4:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4), align 4
+; CHECK-NEXT: [[B5:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 5), align 4
+; CHECK-NEXT: [[B6:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 6), align 4
+; CHECK-NEXT: [[B7:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 7), align 4
+; CHECK-NEXT: [[B8:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8), align 4
+; CHECK-NEXT: [[B9:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 9), align 4
+; CHECK-NEXT: [[B10:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 10), align 4
+; CHECK-NEXT: [[B11:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 11), align 4
+; CHECK-NEXT: [[B12:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12), align 4
+; CHECK-NEXT: [[B13:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 13), align 4
+; CHECK-NEXT: [[B14:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 14), align 4
+; CHECK-NEXT: [[B15:%.*]] = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 15), align 4
+; CHECK-NEXT: [[C0:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A0]], i32 [[B0]])
+; CHECK-NEXT: [[C1:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A1]], i32 [[B1]])
+; CHECK-NEXT: [[C2:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A2]], i32 [[B2]])
+; CHECK-NEXT: [[C3:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A3]], i32 [[B3]])
+; CHECK-NEXT: [[C4:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A4]], i32 [[B4]])
+; CHECK-NEXT: [[C5:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A5]], i32 [[B5]])
+; CHECK-NEXT: [[C6:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A6]], i32 [[B6]])
+; CHECK-NEXT: [[C7:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A7]], i32 [[B7]])
+; CHECK-NEXT: [[C8:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A8]], i32 [[B8]])
+; CHECK-NEXT: [[C9:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A9]], i32 [[B9]])
+; CHECK-NEXT: [[C10:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A10]], i32 [[B10]])
+; CHECK-NEXT: [[C11:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A11]], i32 [[B11]])
+; CHECK-NEXT: [[C12:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A12]], i32 [[B12]])
+; CHECK-NEXT: [[C13:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A13]], i32 [[B13]])
+; CHECK-NEXT: [[C14:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A14]], i32 [[B14]])
+; CHECK-NEXT: [[C15:%.*]] = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 [[A15]], i32 [[B15]])
+; CHECK-NEXT: [[R0:%.*]] = extractvalue { i32, i1 } [[C0]], 0
+; CHECK-NEXT: [[R1:%.*]] = extractvalue { i32, i1 } [[C1]], 0
+; CHECK-NEXT: [[R2:%.*]] = extractvalue { i32, i1 } [[C2]], 0
+; CHECK-NEXT: [[R3:%.*]] = extractvalue { i32, i1 } [[C3]], 0
+; CHECK-NEXT: [[R4:%.*]] = extractvalue { i32, i1 } [[C4]], 0
+; CHECK-NEXT: [[R5:%.*]] = extractvalue { i32, i1 } [[C5]], 0
+; CHECK-NEXT: [[R6:%.*]] = extractvalue { i32, i1 } [[C6]], 0
+; CHECK-NEXT: [[R7:%.*]] = extractvalue { i32, i1 } [[C7]], 0
+; CHECK-NEXT: [[R8:%.*]] = extractvalue { i32, i1 } [[C8]], 0
+; CHECK-NEXT: [[R9:%.*]] = extractvalue { i32, i1 } [[C9]], 0
+; CHECK-NEXT: [[R10:%.*]] = extractvalue { i32, i1 } [[C10]], 0
+; CHECK-NEXT: [[R11:%.*]] = extractvalue { i32, i1 } [[C11]], 0
+; CHECK-NEXT: [[R12:%.*]] = extractvalue { i32, i1 } [[C12]], 0
+; CHECK-NEXT: [[R13:%.*]] = extractvalue { i32, i1 } [[C13]], 0
+; CHECK-NEXT: [[R14:%.*]] = extractvalue { i32, i1 } [[C14]], 0
+; CHECK-NEXT: [[R15:%.*]] = extractvalue { i32, i1 } [[C15]], 0
+; CHECK-NEXT: store i32 [[R0]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 0), align 4
+; CHECK-NEXT: store i32 [[R1]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 1), align 4
+; CHECK-NEXT: store i32 [[R2]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 2), align 4
+; CHECK-NEXT: store i32 [[R3]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 3), align 4
+; CHECK-NEXT: store i32 [[R4]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4), align 4
+; CHECK-NEXT: store i32 [[R5]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 5), align 4
+; CHECK-NEXT: store i32 [[R6]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 6), align 4
+; CHECK-NEXT: store i32 [[R7]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 7), align 4
+; CHECK-NEXT: store i32 [[R8]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8), align 4
+; CHECK-NEXT: store i32 [[R9]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 9), align 4
+; CHECK-NEXT: store i32 [[R10]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 10), align 4
+; CHECK-NEXT: store i32 [[R11]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 11), align 4
+; CHECK-NEXT: store i32 [[R12]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12), align 4
+; CHECK-NEXT: store i32 [[R13]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 13), align 4
+; CHECK-NEXT: store i32 [[R14]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 14), align 4
+; CHECK-NEXT: store i32 [[R15]], i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 15), align 4
+; CHECK-NEXT: ret void
+;
+ %a0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 0 ), align 4
+ %a1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 1 ), align 4
+ %a2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 2 ), align 4
+ %a3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 3 ), align 4
+ %a4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4 ), align 4
+ %a5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 5 ), align 4
+ %a6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 6 ), align 4
+ %a7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 7 ), align 4
+ %a8 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8 ), align 4
+ %a9 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 9 ), align 4
+ %a10 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 10), align 4
+ %a11 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 11), align 4
+ %a12 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12), align 4
+ %a13 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 13), align 4
+ %a14 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 14), align 4
+ %a15 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 15), align 4
+ %b0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 0 ), align 4
+ %b1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 1 ), align 4
+ %b2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 2 ), align 4
+ %b3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 3 ), align 4
+ %b4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4 ), align 4
+ %b5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 5 ), align 4
+ %b6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 6 ), align 4
+ %b7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 7 ), align 4
+ %b8 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8 ), align 4
+ %b9 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 9 ), align 4
+ %b10 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 10), align 4
+ %b11 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 11), align 4
+ %b12 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12), align 4
+ %b13 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 13), align 4
+ %b14 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 14), align 4
+ %b15 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 15), align 4
+ %c0 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a0 , i32 %b0 )
+ %c1 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a1 , i32 %b1 )
+ %c2 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a2 , i32 %b2 )
+ %c3 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a3 , i32 %b3 )
+ %c4 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a4 , i32 %b4 )
+ %c5 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a5 , i32 %b5 )
+ %c6 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a6 , i32 %b6 )
+ %c7 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a7 , i32 %b7 )
+ %c8 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a8 , i32 %b8 )
+ %c9 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a9 , i32 %b9 )
+ %c10 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a10, i32 %b10)
+ %c11 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a11, i32 %b11)
+ %c12 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a12, i32 %b12)
+ %c13 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a13, i32 %b13)
+ %c14 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a14, i32 %b14)
+ %c15 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a15, i32 %b15)
+ %r0 = extractvalue {i32, i1} %c0 , 0
+ %r1 = extractvalue {i32, i1} %c1 , 0
+ %r2 = extractvalue {i32, i1} %c2 , 0
+ %r3 = extractvalue {i32, i1} %c3 , 0
+ %r4 = extractvalue {i32, i1} %c4 , 0
+ %r5 = extractvalue {i32, i1} %c5 , 0
+ %r6 = extractvalue {i32, i1} %c6 , 0
+ %r7 = extractvalue {i32, i1} %c7 , 0
+ %r8 = extractvalue {i32, i1} %c8 , 0
+ %r9 = extractvalue {i32, i1} %c9 , 0
+ %r10 = extractvalue {i32, i1} %c10, 0
+ %r11 = extractvalue {i32, i1} %c11, 0
+ %r12 = extractvalue {i32, i1} %c12, 0
+ %r13 = extractvalue {i32, i1} %c13, 0
+ %r14 = extractvalue {i32, i1} %c14, 0
+ %r15 = extractvalue {i32, i1} %c15, 0
+ store i32 %r0 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 0 ), align 4
+ store i32 %r1 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 1 ), align 4
+ store i32 %r2 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 2 ), align 4
+ store i32 %r3 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 3 ), align 4
+ store i32 %r4 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4 ), align 4
+ store i32 %r5 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 5 ), align 4
+ store i32 %r6 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 6 ), align 4
+ store i32 %r7 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 7 ), align 4
+ store i32 %r8 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8 ), align 4
+ store i32 %r9 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 9 ), align 4
+ store i32 %r10, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 10), align 4
+ store i32 %r11, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 11), align 4
+ store i32 %r12, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12), align 4
+ store i32 %r13, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 13), align 4
+ store i32 %r14, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 14), align 4
+ store i32 %r15, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 15), align 4
+ ret void
+}
+
+define void @sub_v32i16() {
+; CHECK-LABEL: @sub_v32i16(
+; CHECK-NEXT: [[A0:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 0), align 2
+; CHECK-NEXT: [[A1:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 1), align 2
+; CHECK-NEXT: [[A2:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 2), align 2
+; CHECK-NEXT: [[A3:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 3), align 2
+; CHECK-NEXT: [[A4:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 4), align 2
+; CHECK-NEXT: [[A5:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 5), align 2
+; CHECK-NEXT: [[A6:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 6), align 2
+; CHECK-NEXT: [[A7:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 7), align 2
+; CHECK-NEXT: [[A8:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8), align 2
+; CHECK-NEXT: [[A9:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 9), align 2
+; CHECK-NEXT: [[A10:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 10), align 2
+; CHECK-NEXT: [[A11:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 11), align 2
+; CHECK-NEXT: [[A12:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 12), align 2
+; CHECK-NEXT: [[A13:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 13), align 2
+; CHECK-NEXT: [[A14:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 14), align 2
+; CHECK-NEXT: [[A15:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 15), align 2
+; CHECK-NEXT: [[A16:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16), align 2
+; CHECK-NEXT: [[A17:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 17), align 2
+; CHECK-NEXT: [[A18:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 18), align 2
+; CHECK-NEXT: [[A19:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 19), align 2
+; CHECK-NEXT: [[A20:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 20), align 2
+; CHECK-NEXT: [[A21:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 21), align 2
+; CHECK-NEXT: [[A22:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 22), align 2
+; CHECK-NEXT: [[A23:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 23), align 2
+; CHECK-NEXT: [[A24:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24), align 2
+; CHECK-NEXT: [[A25:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 25), align 2
+; CHECK-NEXT: [[A26:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 26), align 2
+; CHECK-NEXT: [[A27:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 27), align 2
+; CHECK-NEXT: [[A28:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 28), align 2
+; CHECK-NEXT: [[A29:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 29), align 2
+; CHECK-NEXT: [[A30:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 30), align 2
+; CHECK-NEXT: [[A31:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 31), align 2
+; CHECK-NEXT: [[B0:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 0), align 2
+; CHECK-NEXT: [[B1:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 1), align 2
+; CHECK-NEXT: [[B2:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 2), align 2
+; CHECK-NEXT: [[B3:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 3), align 2
+; CHECK-NEXT: [[B4:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 4), align 2
+; CHECK-NEXT: [[B5:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 5), align 2
+; CHECK-NEXT: [[B6:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 6), align 2
+; CHECK-NEXT: [[B7:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 7), align 2
+; CHECK-NEXT: [[B8:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8), align 2
+; CHECK-NEXT: [[B9:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 9), align 2
+; CHECK-NEXT: [[B10:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 10), align 2
+; CHECK-NEXT: [[B11:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 11), align 2
+; CHECK-NEXT: [[B12:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 12), align 2
+; CHECK-NEXT: [[B13:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 13), align 2
+; CHECK-NEXT: [[B14:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 14), align 2
+; CHECK-NEXT: [[B15:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 15), align 2
+; CHECK-NEXT: [[B16:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16), align 2
+; CHECK-NEXT: [[B17:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 17), align 2
+; CHECK-NEXT: [[B18:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 18), align 2
+; CHECK-NEXT: [[B19:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 19), align 2
+; CHECK-NEXT: [[B20:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 20), align 2
+; CHECK-NEXT: [[B21:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 21), align 2
+; CHECK-NEXT: [[B22:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 22), align 2
+; CHECK-NEXT: [[B23:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 23), align 2
+; CHECK-NEXT: [[B24:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24), align 2
+; CHECK-NEXT: [[B25:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 25), align 2
+; CHECK-NEXT: [[B26:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 26), align 2
+; CHECK-NEXT: [[B27:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 27), align 2
+; CHECK-NEXT: [[B28:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 28), align 2
+; CHECK-NEXT: [[B29:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 29), align 2
+; CHECK-NEXT: [[B30:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 30), align 2
+; CHECK-NEXT: [[B31:%.*]] = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 31), align 2
+; CHECK-NEXT: [[C0:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A0]], i16 [[B0]])
+; CHECK-NEXT: [[C1:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A1]], i16 [[B1]])
+; CHECK-NEXT: [[C2:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A2]], i16 [[B2]])
+; CHECK-NEXT: [[C3:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A3]], i16 [[B3]])
+; CHECK-NEXT: [[C4:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A4]], i16 [[B4]])
+; CHECK-NEXT: [[C5:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A5]], i16 [[B5]])
+; CHECK-NEXT: [[C6:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A6]], i16 [[B6]])
+; CHECK-NEXT: [[C7:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A7]], i16 [[B7]])
+; CHECK-NEXT: [[C8:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A8]], i16 [[B8]])
+; CHECK-NEXT: [[C9:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A9]], i16 [[B9]])
+; CHECK-NEXT: [[C10:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A10]], i16 [[B10]])
+; CHECK-NEXT: [[C11:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A11]], i16 [[B11]])
+; CHECK-NEXT: [[C12:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A12]], i16 [[B12]])
+; CHECK-NEXT: [[C13:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A13]], i16 [[B13]])
+; CHECK-NEXT: [[C14:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A14]], i16 [[B14]])
+; CHECK-NEXT: [[C15:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A15]], i16 [[B15]])
+; CHECK-NEXT: [[C16:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A16]], i16 [[B16]])
+; CHECK-NEXT: [[C17:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A17]], i16 [[B17]])
+; CHECK-NEXT: [[C18:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A18]], i16 [[B18]])
+; CHECK-NEXT: [[C19:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A19]], i16 [[B19]])
+; CHECK-NEXT: [[C20:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A20]], i16 [[B20]])
+; CHECK-NEXT: [[C21:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A21]], i16 [[B21]])
+; CHECK-NEXT: [[C22:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A22]], i16 [[B22]])
+; CHECK-NEXT: [[C23:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A23]], i16 [[B23]])
+; CHECK-NEXT: [[C24:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A24]], i16 [[B24]])
+; CHECK-NEXT: [[C25:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A25]], i16 [[B25]])
+; CHECK-NEXT: [[C26:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A26]], i16 [[B26]])
+; CHECK-NEXT: [[C27:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A27]], i16 [[B27]])
+; CHECK-NEXT: [[C28:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A28]], i16 [[B28]])
+; CHECK-NEXT: [[C29:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A29]], i16 [[B29]])
+; CHECK-NEXT: [[C30:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A30]], i16 [[B30]])
+; CHECK-NEXT: [[C31:%.*]] = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 [[A31]], i16 [[B31]])
+; CHECK-NEXT: [[R0:%.*]] = extractvalue { i16, i1 } [[C0]], 0
+; CHECK-NEXT: [[R1:%.*]] = extractvalue { i16, i1 } [[C1]], 0
+; CHECK-NEXT: [[R2:%.*]] = extractvalue { i16, i1 } [[C2]], 0
+; CHECK-NEXT: [[R3:%.*]] = extractvalue { i16, i1 } [[C3]], 0
+; CHECK-NEXT: [[R4:%.*]] = extractvalue { i16, i1 } [[C4]], 0
+; CHECK-NEXT: [[R5:%.*]] = extractvalue { i16, i1 } [[C5]], 0
+; CHECK-NEXT: [[R6:%.*]] = extractvalue { i16, i1 } [[C6]], 0
+; CHECK-NEXT: [[R7:%.*]] = extractvalue { i16, i1 } [[C7]], 0
+; CHECK-NEXT: [[R8:%.*]] = extractvalue { i16, i1 } [[C8]], 0
+; CHECK-NEXT: [[R9:%.*]] = extractvalue { i16, i1 } [[C9]], 0
+; CHECK-NEXT: [[R10:%.*]] = extractvalue { i16, i1 } [[C10]], 0
+; CHECK-NEXT: [[R11:%.*]] = extractvalue { i16, i1 } [[C11]], 0
+; CHECK-NEXT: [[R12:%.*]] = extractvalue { i16, i1 } [[C12]], 0
+; CHECK-NEXT: [[R13:%.*]] = extractvalue { i16, i1 } [[C13]], 0
+; CHECK-NEXT: [[R14:%.*]] = extractvalue { i16, i1 } [[C14]], 0
+; CHECK-NEXT: [[R15:%.*]] = extractvalue { i16, i1 } [[C15]], 0
+; CHECK-NEXT: [[R16:%.*]] = extractvalue { i16, i1 } [[C16]], 0
+; CHECK-NEXT: [[R17:%.*]] = extractvalue { i16, i1 } [[C17]], 0
+; CHECK-NEXT: [[R18:%.*]] = extractvalue { i16, i1 } [[C18]], 0
+; CHECK-NEXT: [[R19:%.*]] = extractvalue { i16, i1 } [[C19]], 0
+; CHECK-NEXT: [[R20:%.*]] = extractvalue { i16, i1 } [[C20]], 0
+; CHECK-NEXT: [[R21:%.*]] = extractvalue { i16, i1 } [[C21]], 0
+; CHECK-NEXT: [[R22:%.*]] = extractvalue { i16, i1 } [[C22]], 0
+; CHECK-NEXT: [[R23:%.*]] = extractvalue { i16, i1 } [[C23]], 0
+; CHECK-NEXT: [[R24:%.*]] = extractvalue { i16, i1 } [[C24]], 0
+; CHECK-NEXT: [[R25:%.*]] = extractvalue { i16, i1 } [[C25]], 0
+; CHECK-NEXT: [[R26:%.*]] = extractvalue { i16, i1 } [[C26]], 0
+; CHECK-NEXT: [[R27:%.*]] = extractvalue { i16, i1 } [[C27]], 0
+; CHECK-NEXT: [[R28:%.*]] = extractvalue { i16, i1 } [[C28]], 0
+; CHECK-NEXT: [[R29:%.*]] = extractvalue { i16, i1 } [[C29]], 0
+; CHECK-NEXT: [[R30:%.*]] = extractvalue { i16, i1 } [[C30]], 0
+; CHECK-NEXT: [[R31:%.*]] = extractvalue { i16, i1 } [[C31]], 0
+; CHECK-NEXT: store i16 [[R0]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 0), align 2
+; CHECK-NEXT: store i16 [[R1]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 1), align 2
+; CHECK-NEXT: store i16 [[R2]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 2), align 2
+; CHECK-NEXT: store i16 [[R3]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 3), align 2
+; CHECK-NEXT: store i16 [[R4]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 4), align 2
+; CHECK-NEXT: store i16 [[R5]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 5), align 2
+; CHECK-NEXT: store i16 [[R6]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 6), align 2
+; CHECK-NEXT: store i16 [[R7]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 7), align 2
+; CHECK-NEXT: store i16 [[R8]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8), align 2
+; CHECK-NEXT: store i16 [[R9]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 9), align 2
+; CHECK-NEXT: store i16 [[R10]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 10), align 2
+; CHECK-NEXT: store i16 [[R11]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 11), align 2
+; CHECK-NEXT: store i16 [[R12]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 12), align 2
+; CHECK-NEXT: store i16 [[R13]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 13), align 2
+; CHECK-NEXT: store i16 [[R14]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 14), align 2
+; CHECK-NEXT: store i16 [[R15]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 15), align 2
+; CHECK-NEXT: store i16 [[R16]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16), align 2
+; CHECK-NEXT: store i16 [[R17]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 17), align 2
+; CHECK-NEXT: store i16 [[R18]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 18), align 2
+; CHECK-NEXT: store i16 [[R19]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 19), align 2
+; CHECK-NEXT: store i16 [[R20]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 20), align 2
+; CHECK-NEXT: store i16 [[R21]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 21), align 2
+; CHECK-NEXT: store i16 [[R22]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 22), align 2
+; CHECK-NEXT: store i16 [[R23]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 23), align 2
+; CHECK-NEXT: store i16 [[R24]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24), align 2
+; CHECK-NEXT: store i16 [[R25]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 25), align 2
+; CHECK-NEXT: store i16 [[R26]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 26), align 2
+; CHECK-NEXT: store i16 [[R27]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 27), align 2
+; CHECK-NEXT: store i16 [[R28]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 28), align 2
+; CHECK-NEXT: store i16 [[R29]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 29), align 2
+; CHECK-NEXT: store i16 [[R30]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 30), align 2
+; CHECK-NEXT: store i16 [[R31]], i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 31), align 2
+; CHECK-NEXT: ret void
+;
+ %a0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 0 ), align 2
+ %a1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 1 ), align 2
+ %a2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 2 ), align 2
+ %a3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 3 ), align 2
+ %a4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 4 ), align 2
+ %a5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 5 ), align 2
+ %a6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 6 ), align 2
+ %a7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 7 ), align 2
+ %a8 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8 ), align 2
+ %a9 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 9 ), align 2
+ %a10 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 10), align 2
+ %a11 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 11), align 2
+ %a12 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 12), align 2
+ %a13 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 13), align 2
+ %a14 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 14), align 2
+ %a15 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 15), align 2
+ %a16 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16), align 2
+ %a17 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 17), align 2
+ %a18 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 18), align 2
+ %a19 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 19), align 2
+ %a20 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 20), align 2
+ %a21 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 21), align 2
+ %a22 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 22), align 2
+ %a23 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 23), align 2
+ %a24 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24), align 2
+ %a25 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 25), align 2
+ %a26 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 26), align 2
+ %a27 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 27), align 2
+ %a28 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 28), align 2
+ %a29 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 29), align 2
+ %a30 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 30), align 2
+ %a31 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 31), align 2
+ %b0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 0 ), align 2
+ %b1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 1 ), align 2
+ %b2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 2 ), align 2
+ %b3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 3 ), align 2
+ %b4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 4 ), align 2
+ %b5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 5 ), align 2
+ %b6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 6 ), align 2
+ %b7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 7 ), align 2
+ %b8 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8 ), align 2
+ %b9 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 9 ), align 2
+ %b10 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 10), align 2
+ %b11 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 11), align 2
+ %b12 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 12), align 2
+ %b13 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 13), align 2
+ %b14 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 14), align 2
+ %b15 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 15), align 2
+ %b16 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16), align 2
+ %b17 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 17), align 2
+ %b18 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 18), align 2
+ %b19 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 19), align 2
+ %b20 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 20), align 2
+ %b21 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 21), align 2
+ %b22 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 22), align 2
+ %b23 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 23), align 2
+ %b24 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24), align 2
+ %b25 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 25), align 2
+ %b26 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 26), align 2
+ %b27 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 27), align 2
+ %b28 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 28), align 2
+ %b29 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 29), align 2
+ %b30 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 30), align 2
+ %b31 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 31), align 2
+ %c0 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a0 , i16 %b0 )
+ %c1 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a1 , i16 %b1 )
+ %c2 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a2 , i16 %b2 )
+ %c3 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a3 , i16 %b3 )
+ %c4 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a4 , i16 %b4 )
+ %c5 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a5 , i16 %b5 )
+ %c6 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a6 , i16 %b6 )
+ %c7 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a7 , i16 %b7 )
+ %c8 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a8 , i16 %b8 )
+ %c9 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a9 , i16 %b9 )
+ %c10 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a10, i16 %b10)
+ %c11 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a11, i16 %b11)
+ %c12 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a12, i16 %b12)
+ %c13 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a13, i16 %b13)
+ %c14 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a14, i16 %b14)
+ %c15 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a15, i16 %b15)
+ %c16 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a16, i16 %b16)
+ %c17 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a17, i16 %b17)
+ %c18 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a18, i16 %b18)
+ %c19 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a19, i16 %b19)
+ %c20 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a20, i16 %b20)
+ %c21 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a21, i16 %b21)
+ %c22 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a22, i16 %b22)
+ %c23 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a23, i16 %b23)
+ %c24 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a24, i16 %b24)
+ %c25 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a25, i16 %b25)
+ %c26 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a26, i16 %b26)
+ %c27 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a27, i16 %b27)
+ %c28 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a28, i16 %b28)
+ %c29 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a29, i16 %b29)
+ %c30 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a30, i16 %b30)
+ %c31 = call {i16, i1} @llvm.usub.with.overflow.i16(i16 %a31, i16 %b31)
+ %r0 = extractvalue {i16, i1} %c0 , 0
+ %r1 = extractvalue {i16, i1} %c1 , 0
+ %r2 = extractvalue {i16, i1} %c2 , 0
+ %r3 = extractvalue {i16, i1} %c3 , 0
+ %r4 = extractvalue {i16, i1} %c4 , 0
+ %r5 = extractvalue {i16, i1} %c5 , 0
+ %r6 = extractvalue {i16, i1} %c6 , 0
+ %r7 = extractvalue {i16, i1} %c7 , 0
+ %r8 = extractvalue {i16, i1} %c8 , 0
+ %r9 = extractvalue {i16, i1} %c9 , 0
+ %r10 = extractvalue {i16, i1} %c10, 0
+ %r11 = extractvalue {i16, i1} %c11, 0
+ %r12 = extractvalue {i16, i1} %c12, 0
+ %r13 = extractvalue {i16, i1} %c13, 0
+ %r14 = extractvalue {i16, i1} %c14, 0
+ %r15 = extractvalue {i16, i1} %c15, 0
+ %r16 = extractvalue {i16, i1} %c16, 0
+ %r17 = extractvalue {i16, i1} %c17, 0
+ %r18 = extractvalue {i16, i1} %c18, 0
+ %r19 = extractvalue {i16, i1} %c19, 0
+ %r20 = extractvalue {i16, i1} %c20, 0
+ %r21 = extractvalue {i16, i1} %c21, 0
+ %r22 = extractvalue {i16, i1} %c22, 0
+ %r23 = extractvalue {i16, i1} %c23, 0
+ %r24 = extractvalue {i16, i1} %c24, 0
+ %r25 = extractvalue {i16, i1} %c25, 0
+ %r26 = extractvalue {i16, i1} %c26, 0
+ %r27 = extractvalue {i16, i1} %c27, 0
+ %r28 = extractvalue {i16, i1} %c28, 0
+ %r29 = extractvalue {i16, i1} %c29, 0
+ %r30 = extractvalue {i16, i1} %c30, 0
+ %r31 = extractvalue {i16, i1} %c31, 0
+ store i16 %r0 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 0 ), align 2
+ store i16 %r1 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 1 ), align 2
+ store i16 %r2 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 2 ), align 2
+ store i16 %r3 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 3 ), align 2
+ store i16 %r4 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 4 ), align 2
+ store i16 %r5 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 5 ), align 2
+ store i16 %r6 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 6 ), align 2
+ store i16 %r7 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 7 ), align 2
+ store i16 %r8 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8 ), align 2
+ store i16 %r9 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 9 ), align 2
+ store i16 %r10, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 10), align 2
+ store i16 %r11, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 11), align 2
+ store i16 %r12, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 12), align 2
+ store i16 %r13, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 13), align 2
+ store i16 %r14, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 14), align 2
+ store i16 %r15, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 15), align 2
+ store i16 %r16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16), align 2
+ store i16 %r17, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 17), align 2
+ store i16 %r18, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 18), align 2
+ store i16 %r19, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 19), align 2
+ store i16 %r20, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 20), align 2
+ store i16 %r21, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 21), align 2
+ store i16 %r22, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 22), align 2
+ store i16 %r23, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 23), align 2
+ store i16 %r24, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24), align 2
+ store i16 %r25, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 25), align 2
+ store i16 %r26, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 26), align 2
+ store i16 %r27, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 27), align 2
+ store i16 %r28, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 28), align 2
+ store i16 %r29, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 29), align 2
+ store i16 %r30, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 30), align 2
+ store i16 %r31, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 31), align 2
+ ret void
+}
+
+define void @sub_v64i8() {
+; CHECK-LABEL: @sub_v64i8(
+; CHECK-NEXT: [[A0:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 0), align 1
+; CHECK-NEXT: [[A1:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 1), align 1
+; CHECK-NEXT: [[A2:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 2), align 1
+; CHECK-NEXT: [[A3:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 3), align 1
+; CHECK-NEXT: [[A4:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 4), align 1
+; CHECK-NEXT: [[A5:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 5), align 1
+; CHECK-NEXT: [[A6:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 6), align 1
+; CHECK-NEXT: [[A7:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 7), align 1
+; CHECK-NEXT: [[A8:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 8), align 1
+; CHECK-NEXT: [[A9:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 9), align 1
+; CHECK-NEXT: [[A10:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 10), align 1
+; CHECK-NEXT: [[A11:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 11), align 1
+; CHECK-NEXT: [[A12:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 12), align 1
+; CHECK-NEXT: [[A13:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 13), align 1
+; CHECK-NEXT: [[A14:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 14), align 1
+; CHECK-NEXT: [[A15:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 15), align 1
+; CHECK-NEXT: [[A16:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 16), align 1
+; CHECK-NEXT: [[A17:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 17), align 1
+; CHECK-NEXT: [[A18:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 18), align 1
+; CHECK-NEXT: [[A19:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 19), align 1
+; CHECK-NEXT: [[A20:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 20), align 1
+; CHECK-NEXT: [[A21:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 21), align 1
+; CHECK-NEXT: [[A22:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 22), align 1
+; CHECK-NEXT: [[A23:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 23), align 1
+; CHECK-NEXT: [[A24:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 24), align 1
+; CHECK-NEXT: [[A25:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 25), align 1
+; CHECK-NEXT: [[A26:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 26), align 1
+; CHECK-NEXT: [[A27:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 27), align 1
+; CHECK-NEXT: [[A28:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 28), align 1
+; CHECK-NEXT: [[A29:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 29), align 1
+; CHECK-NEXT: [[A30:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 30), align 1
+; CHECK-NEXT: [[A31:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 31), align 1
+; CHECK-NEXT: [[A32:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 32), align 1
+; CHECK-NEXT: [[A33:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 33), align 1
+; CHECK-NEXT: [[A34:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 34), align 1
+; CHECK-NEXT: [[A35:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 35), align 1
+; CHECK-NEXT: [[A36:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 36), align 1
+; CHECK-NEXT: [[A37:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 37), align 1
+; CHECK-NEXT: [[A38:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 38), align 1
+; CHECK-NEXT: [[A39:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 39), align 1
+; CHECK-NEXT: [[A40:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 40), align 1
+; CHECK-NEXT: [[A41:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 41), align 1
+; CHECK-NEXT: [[A42:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 42), align 1
+; CHECK-NEXT: [[A43:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 43), align 1
+; CHECK-NEXT: [[A44:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 44), align 1
+; CHECK-NEXT: [[A45:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 45), align 1
+; CHECK-NEXT: [[A46:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 46), align 1
+; CHECK-NEXT: [[A47:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 47), align 1
+; CHECK-NEXT: [[A48:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 48), align 1
+; CHECK-NEXT: [[A49:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 49), align 1
+; CHECK-NEXT: [[A50:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 50), align 1
+; CHECK-NEXT: [[A51:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 51), align 1
+; CHECK-NEXT: [[A52:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 52), align 1
+; CHECK-NEXT: [[A53:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 53), align 1
+; CHECK-NEXT: [[A54:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 54), align 1
+; CHECK-NEXT: [[A55:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 55), align 1
+; CHECK-NEXT: [[A56:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 56), align 1
+; CHECK-NEXT: [[A57:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 57), align 1
+; CHECK-NEXT: [[A58:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 58), align 1
+; CHECK-NEXT: [[A59:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 59), align 1
+; CHECK-NEXT: [[A60:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 60), align 1
+; CHECK-NEXT: [[A61:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 61), align 1
+; CHECK-NEXT: [[A62:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 62), align 1
+; CHECK-NEXT: [[A63:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 63), align 1
+; CHECK-NEXT: [[B0:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 0), align 1
+; CHECK-NEXT: [[B1:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 1), align 1
+; CHECK-NEXT: [[B2:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 2), align 1
+; CHECK-NEXT: [[B3:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 3), align 1
+; CHECK-NEXT: [[B4:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 4), align 1
+; CHECK-NEXT: [[B5:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 5), align 1
+; CHECK-NEXT: [[B6:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 6), align 1
+; CHECK-NEXT: [[B7:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 7), align 1
+; CHECK-NEXT: [[B8:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 8), align 1
+; CHECK-NEXT: [[B9:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 9), align 1
+; CHECK-NEXT: [[B10:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 10), align 1
+; CHECK-NEXT: [[B11:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 11), align 1
+; CHECK-NEXT: [[B12:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 12), align 1
+; CHECK-NEXT: [[B13:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 13), align 1
+; CHECK-NEXT: [[B14:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 14), align 1
+; CHECK-NEXT: [[B15:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 15), align 1
+; CHECK-NEXT: [[B16:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 16), align 1
+; CHECK-NEXT: [[B17:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 17), align 1
+; CHECK-NEXT: [[B18:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 18), align 1
+; CHECK-NEXT: [[B19:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 19), align 1
+; CHECK-NEXT: [[B20:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 20), align 1
+; CHECK-NEXT: [[B21:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 21), align 1
+; CHECK-NEXT: [[B22:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 22), align 1
+; CHECK-NEXT: [[B23:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 23), align 1
+; CHECK-NEXT: [[B24:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 24), align 1
+; CHECK-NEXT: [[B25:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 25), align 1
+; CHECK-NEXT: [[B26:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 26), align 1
+; CHECK-NEXT: [[B27:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 27), align 1
+; CHECK-NEXT: [[B28:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 28), align 1
+; CHECK-NEXT: [[B29:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 29), align 1
+; CHECK-NEXT: [[B30:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 30), align 1
+; CHECK-NEXT: [[B31:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 31), align 1
+; CHECK-NEXT: [[B32:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 32), align 1
+; CHECK-NEXT: [[B33:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 33), align 1
+; CHECK-NEXT: [[B34:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 34), align 1
+; CHECK-NEXT: [[B35:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 35), align 1
+; CHECK-NEXT: [[B36:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 36), align 1
+; CHECK-NEXT: [[B37:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 37), align 1
+; CHECK-NEXT: [[B38:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 38), align 1
+; CHECK-NEXT: [[B39:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 39), align 1
+; CHECK-NEXT: [[B40:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 40), align 1
+; CHECK-NEXT: [[B41:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 41), align 1
+; CHECK-NEXT: [[B42:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 42), align 1
+; CHECK-NEXT: [[B43:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 43), align 1
+; CHECK-NEXT: [[B44:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 44), align 1
+; CHECK-NEXT: [[B45:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 45), align 1
+; CHECK-NEXT: [[B46:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 46), align 1
+; CHECK-NEXT: [[B47:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 47), align 1
+; CHECK-NEXT: [[B48:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 48), align 1
+; CHECK-NEXT: [[B49:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 49), align 1
+; CHECK-NEXT: [[B50:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 50), align 1
+; CHECK-NEXT: [[B51:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 51), align 1
+; CHECK-NEXT: [[B52:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 52), align 1
+; CHECK-NEXT: [[B53:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 53), align 1
+; CHECK-NEXT: [[B54:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 54), align 1
+; CHECK-NEXT: [[B55:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 55), align 1
+; CHECK-NEXT: [[B56:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 56), align 1
+; CHECK-NEXT: [[B57:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 57), align 1
+; CHECK-NEXT: [[B58:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 58), align 1
+; CHECK-NEXT: [[B59:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 59), align 1
+; CHECK-NEXT: [[B60:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 60), align 1
+; CHECK-NEXT: [[B61:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 61), align 1
+; CHECK-NEXT: [[B62:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 62), align 1
+; CHECK-NEXT: [[B63:%.*]] = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 63), align 1
+; CHECK-NEXT: [[C0:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A0]], i8 [[B0]])
+; CHECK-NEXT: [[C1:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A1]], i8 [[B1]])
+; CHECK-NEXT: [[C2:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A2]], i8 [[B2]])
+; CHECK-NEXT: [[C3:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A3]], i8 [[B3]])
+; CHECK-NEXT: [[C4:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A4]], i8 [[B4]])
+; CHECK-NEXT: [[C5:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A5]], i8 [[B5]])
+; CHECK-NEXT: [[C6:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A6]], i8 [[B6]])
+; CHECK-NEXT: [[C7:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A7]], i8 [[B7]])
+; CHECK-NEXT: [[C8:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A8]], i8 [[B8]])
+; CHECK-NEXT: [[C9:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A9]], i8 [[B9]])
+; CHECK-NEXT: [[C10:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A10]], i8 [[B10]])
+; CHECK-NEXT: [[C11:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A11]], i8 [[B11]])
+; CHECK-NEXT: [[C12:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A12]], i8 [[B12]])
+; CHECK-NEXT: [[C13:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A13]], i8 [[B13]])
+; CHECK-NEXT: [[C14:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A14]], i8 [[B14]])
+; CHECK-NEXT: [[C15:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A15]], i8 [[B15]])
+; CHECK-NEXT: [[C16:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A16]], i8 [[B16]])
+; CHECK-NEXT: [[C17:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A17]], i8 [[B17]])
+; CHECK-NEXT: [[C18:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A18]], i8 [[B18]])
+; CHECK-NEXT: [[C19:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A19]], i8 [[B19]])
+; CHECK-NEXT: [[C20:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A20]], i8 [[B20]])
+; CHECK-NEXT: [[C21:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A21]], i8 [[B21]])
+; CHECK-NEXT: [[C22:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A22]], i8 [[B22]])
+; CHECK-NEXT: [[C23:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A23]], i8 [[B23]])
+; CHECK-NEXT: [[C24:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A24]], i8 [[B24]])
+; CHECK-NEXT: [[C25:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A25]], i8 [[B25]])
+; CHECK-NEXT: [[C26:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A26]], i8 [[B26]])
+; CHECK-NEXT: [[C27:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A27]], i8 [[B27]])
+; CHECK-NEXT: [[C28:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A28]], i8 [[B28]])
+; CHECK-NEXT: [[C29:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A29]], i8 [[B29]])
+; CHECK-NEXT: [[C30:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A30]], i8 [[B30]])
+; CHECK-NEXT: [[C31:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A31]], i8 [[B31]])
+; CHECK-NEXT: [[C32:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A32]], i8 [[B32]])
+; CHECK-NEXT: [[C33:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A33]], i8 [[B33]])
+; CHECK-NEXT: [[C34:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A34]], i8 [[B34]])
+; CHECK-NEXT: [[C35:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A35]], i8 [[B35]])
+; CHECK-NEXT: [[C36:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A36]], i8 [[B36]])
+; CHECK-NEXT: [[C37:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A37]], i8 [[B37]])
+; CHECK-NEXT: [[C38:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A38]], i8 [[B38]])
+; CHECK-NEXT: [[C39:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A39]], i8 [[B39]])
+; CHECK-NEXT: [[C40:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A40]], i8 [[B40]])
+; CHECK-NEXT: [[C41:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A41]], i8 [[B41]])
+; CHECK-NEXT: [[C42:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A42]], i8 [[B42]])
+; CHECK-NEXT: [[C43:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A43]], i8 [[B43]])
+; CHECK-NEXT: [[C44:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A44]], i8 [[B44]])
+; CHECK-NEXT: [[C45:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A45]], i8 [[B45]])
+; CHECK-NEXT: [[C46:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A46]], i8 [[B46]])
+; CHECK-NEXT: [[C47:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A47]], i8 [[B47]])
+; CHECK-NEXT: [[C48:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A48]], i8 [[B48]])
+; CHECK-NEXT: [[C49:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A49]], i8 [[B49]])
+; CHECK-NEXT: [[C50:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A50]], i8 [[B50]])
+; CHECK-NEXT: [[C51:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A51]], i8 [[B51]])
+; CHECK-NEXT: [[C52:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A52]], i8 [[B52]])
+; CHECK-NEXT: [[C53:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A53]], i8 [[B53]])
+; CHECK-NEXT: [[C54:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A54]], i8 [[B54]])
+; CHECK-NEXT: [[C55:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A55]], i8 [[B55]])
+; CHECK-NEXT: [[C56:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A56]], i8 [[B56]])
+; CHECK-NEXT: [[C57:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A57]], i8 [[B57]])
+; CHECK-NEXT: [[C58:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A58]], i8 [[B58]])
+; CHECK-NEXT: [[C59:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A59]], i8 [[B59]])
+; CHECK-NEXT: [[C60:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A60]], i8 [[B60]])
+; CHECK-NEXT: [[C61:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A61]], i8 [[B61]])
+; CHECK-NEXT: [[C62:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A62]], i8 [[B62]])
+; CHECK-NEXT: [[C63:%.*]] = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 [[A63]], i8 [[B63]])
+; CHECK-NEXT: [[R0:%.*]] = extractvalue { i8, i1 } [[C0]], 0
+; CHECK-NEXT: [[R1:%.*]] = extractvalue { i8, i1 } [[C1]], 0
+; CHECK-NEXT: [[R2:%.*]] = extractvalue { i8, i1 } [[C2]], 0
+; CHECK-NEXT: [[R3:%.*]] = extractvalue { i8, i1 } [[C3]], 0
+; CHECK-NEXT: [[R4:%.*]] = extractvalue { i8, i1 } [[C4]], 0
+; CHECK-NEXT: [[R5:%.*]] = extractvalue { i8, i1 } [[C5]], 0
+; CHECK-NEXT: [[R6:%.*]] = extractvalue { i8, i1 } [[C6]], 0
+; CHECK-NEXT: [[R7:%.*]] = extractvalue { i8, i1 } [[C7]], 0
+; CHECK-NEXT: [[R8:%.*]] = extractvalue { i8, i1 } [[C8]], 0
+; CHECK-NEXT: [[R9:%.*]] = extractvalue { i8, i1 } [[C9]], 0
+; CHECK-NEXT: [[R10:%.*]] = extractvalue { i8, i1 } [[C10]], 0
+; CHECK-NEXT: [[R11:%.*]] = extractvalue { i8, i1 } [[C11]], 0
+; CHECK-NEXT: [[R12:%.*]] = extractvalue { i8, i1 } [[C12]], 0
+; CHECK-NEXT: [[R13:%.*]] = extractvalue { i8, i1 } [[C13]], 0
+; CHECK-NEXT: [[R14:%.*]] = extractvalue { i8, i1 } [[C14]], 0
+; CHECK-NEXT: [[R15:%.*]] = extractvalue { i8, i1 } [[C15]], 0
+; CHECK-NEXT: [[R16:%.*]] = extractvalue { i8, i1 } [[C16]], 0
+; CHECK-NEXT: [[R17:%.*]] = extractvalue { i8, i1 } [[C17]], 0
+; CHECK-NEXT: [[R18:%.*]] = extractvalue { i8, i1 } [[C18]], 0
+; CHECK-NEXT: [[R19:%.*]] = extractvalue { i8, i1 } [[C19]], 0
+; CHECK-NEXT: [[R20:%.*]] = extractvalue { i8, i1 } [[C20]], 0
+; CHECK-NEXT: [[R21:%.*]] = extractvalue { i8, i1 } [[C21]], 0
+; CHECK-NEXT: [[R22:%.*]] = extractvalue { i8, i1 } [[C22]], 0
+; CHECK-NEXT: [[R23:%.*]] = extractvalue { i8, i1 } [[C23]], 0
+; CHECK-NEXT: [[R24:%.*]] = extractvalue { i8, i1 } [[C24]], 0
+; CHECK-NEXT: [[R25:%.*]] = extractvalue { i8, i1 } [[C25]], 0
+; CHECK-NEXT: [[R26:%.*]] = extractvalue { i8, i1 } [[C26]], 0
+; CHECK-NEXT: [[R27:%.*]] = extractvalue { i8, i1 } [[C27]], 0
+; CHECK-NEXT: [[R28:%.*]] = extractvalue { i8, i1 } [[C28]], 0
+; CHECK-NEXT: [[R29:%.*]] = extractvalue { i8, i1 } [[C29]], 0
+; CHECK-NEXT: [[R30:%.*]] = extractvalue { i8, i1 } [[C30]], 0
+; CHECK-NEXT: [[R31:%.*]] = extractvalue { i8, i1 } [[C31]], 0
+; CHECK-NEXT: [[R32:%.*]] = extractvalue { i8, i1 } [[C32]], 0
+; CHECK-NEXT: [[R33:%.*]] = extractvalue { i8, i1 } [[C33]], 0
+; CHECK-NEXT: [[R34:%.*]] = extractvalue { i8, i1 } [[C34]], 0
+; CHECK-NEXT: [[R35:%.*]] = extractvalue { i8, i1 } [[C35]], 0
+; CHECK-NEXT: [[R36:%.*]] = extractvalue { i8, i1 } [[C36]], 0
+; CHECK-NEXT: [[R37:%.*]] = extractvalue { i8, i1 } [[C37]], 0
+; CHECK-NEXT: [[R38:%.*]] = extractvalue { i8, i1 } [[C38]], 0
+; CHECK-NEXT: [[R39:%.*]] = extractvalue { i8, i1 } [[C39]], 0
+; CHECK-NEXT: [[R40:%.*]] = extractvalue { i8, i1 } [[C40]], 0
+; CHECK-NEXT: [[R41:%.*]] = extractvalue { i8, i1 } [[C41]], 0
+; CHECK-NEXT: [[R42:%.*]] = extractvalue { i8, i1 } [[C42]], 0
+; CHECK-NEXT: [[R43:%.*]] = extractvalue { i8, i1 } [[C43]], 0
+; CHECK-NEXT: [[R44:%.*]] = extractvalue { i8, i1 } [[C44]], 0
+; CHECK-NEXT: [[R45:%.*]] = extractvalue { i8, i1 } [[C45]], 0
+; CHECK-NEXT: [[R46:%.*]] = extractvalue { i8, i1 } [[C46]], 0
+; CHECK-NEXT: [[R47:%.*]] = extractvalue { i8, i1 } [[C47]], 0
+; CHECK-NEXT: [[R48:%.*]] = extractvalue { i8, i1 } [[C48]], 0
+; CHECK-NEXT: [[R49:%.*]] = extractvalue { i8, i1 } [[C49]], 0
+; CHECK-NEXT: [[R50:%.*]] = extractvalue { i8, i1 } [[C50]], 0
+; CHECK-NEXT: [[R51:%.*]] = extractvalue { i8, i1 } [[C51]], 0
+; CHECK-NEXT: [[R52:%.*]] = extractvalue { i8, i1 } [[C52]], 0
+; CHECK-NEXT: [[R53:%.*]] = extractvalue { i8, i1 } [[C53]], 0
+; CHECK-NEXT: [[R54:%.*]] = extractvalue { i8, i1 } [[C54]], 0
+; CHECK-NEXT: [[R55:%.*]] = extractvalue { i8, i1 } [[C55]], 0
+; CHECK-NEXT: [[R56:%.*]] = extractvalue { i8, i1 } [[C56]], 0
+; CHECK-NEXT: [[R57:%.*]] = extractvalue { i8, i1 } [[C57]], 0
+; CHECK-NEXT: [[R58:%.*]] = extractvalue { i8, i1 } [[C58]], 0
+; CHECK-NEXT: [[R59:%.*]] = extractvalue { i8, i1 } [[C59]], 0
+; CHECK-NEXT: [[R60:%.*]] = extractvalue { i8, i1 } [[C60]], 0
+; CHECK-NEXT: [[R61:%.*]] = extractvalue { i8, i1 } [[C61]], 0
+; CHECK-NEXT: [[R62:%.*]] = extractvalue { i8, i1 } [[C62]], 0
+; CHECK-NEXT: [[R63:%.*]] = extractvalue { i8, i1 } [[C63]], 0
+; CHECK-NEXT: store i8 [[R0]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 0), align 1
+; CHECK-NEXT: store i8 [[R1]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 1), align 1
+; CHECK-NEXT: store i8 [[R2]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 2), align 1
+; CHECK-NEXT: store i8 [[R3]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 3), align 1
+; CHECK-NEXT: store i8 [[R4]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 4), align 1
+; CHECK-NEXT: store i8 [[R5]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 5), align 1
+; CHECK-NEXT: store i8 [[R6]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 6), align 1
+; CHECK-NEXT: store i8 [[R7]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 7), align 1
+; CHECK-NEXT: store i8 [[R8]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 8), align 1
+; CHECK-NEXT: store i8 [[R9]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 9), align 1
+; CHECK-NEXT: store i8 [[R10]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 10), align 1
+; CHECK-NEXT: store i8 [[R11]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 11), align 1
+; CHECK-NEXT: store i8 [[R12]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 12), align 1
+; CHECK-NEXT: store i8 [[R13]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 13), align 1
+; CHECK-NEXT: store i8 [[R14]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 14), align 1
+; CHECK-NEXT: store i8 [[R15]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 15), align 1
+; CHECK-NEXT: store i8 [[R16]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 16), align 1
+; CHECK-NEXT: store i8 [[R17]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 17), align 1
+; CHECK-NEXT: store i8 [[R18]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 18), align 1
+; CHECK-NEXT: store i8 [[R19]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 19), align 1
+; CHECK-NEXT: store i8 [[R20]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 20), align 1
+; CHECK-NEXT: store i8 [[R21]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 21), align 1
+; CHECK-NEXT: store i8 [[R22]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 22), align 1
+; CHECK-NEXT: store i8 [[R23]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 23), align 1
+; CHECK-NEXT: store i8 [[R24]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 24), align 1
+; CHECK-NEXT: store i8 [[R25]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 25), align 1
+; CHECK-NEXT: store i8 [[R26]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 26), align 1
+; CHECK-NEXT: store i8 [[R27]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 27), align 1
+; CHECK-NEXT: store i8 [[R28]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 28), align 1
+; CHECK-NEXT: store i8 [[R29]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 29), align 1
+; CHECK-NEXT: store i8 [[R30]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 30), align 1
+; CHECK-NEXT: store i8 [[R31]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 31), align 1
+; CHECK-NEXT: store i8 [[R32]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 32), align 1
+; CHECK-NEXT: store i8 [[R33]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 33), align 1
+; CHECK-NEXT: store i8 [[R34]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 34), align 1
+; CHECK-NEXT: store i8 [[R35]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 35), align 1
+; CHECK-NEXT: store i8 [[R36]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 36), align 1
+; CHECK-NEXT: store i8 [[R37]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 37), align 1
+; CHECK-NEXT: store i8 [[R38]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 38), align 1
+; CHECK-NEXT: store i8 [[R39]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 39), align 1
+; CHECK-NEXT: store i8 [[R40]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 40), align 1
+; CHECK-NEXT: store i8 [[R41]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 41), align 1
+; CHECK-NEXT: store i8 [[R42]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 42), align 1
+; CHECK-NEXT: store i8 [[R43]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 43), align 1
+; CHECK-NEXT: store i8 [[R44]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 44), align 1
+; CHECK-NEXT: store i8 [[R45]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 45), align 1
+; CHECK-NEXT: store i8 [[R46]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 46), align 1
+; CHECK-NEXT: store i8 [[R47]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 47), align 1
+; CHECK-NEXT: store i8 [[R48]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 48), align 1
+; CHECK-NEXT: store i8 [[R49]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 49), align 1
+; CHECK-NEXT: store i8 [[R50]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 50), align 1
+; CHECK-NEXT: store i8 [[R51]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 51), align 1
+; CHECK-NEXT: store i8 [[R52]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 52), align 1
+; CHECK-NEXT: store i8 [[R53]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 53), align 1
+; CHECK-NEXT: store i8 [[R54]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 54), align 1
+; CHECK-NEXT: store i8 [[R55]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 55), align 1
+; CHECK-NEXT: store i8 [[R56]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 56), align 1
+; CHECK-NEXT: store i8 [[R57]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 57), align 1
+; CHECK-NEXT: store i8 [[R58]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 58), align 1
+; CHECK-NEXT: store i8 [[R59]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 59), align 1
+; CHECK-NEXT: store i8 [[R60]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 60), align 1
+; CHECK-NEXT: store i8 [[R61]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 61), align 1
+; CHECK-NEXT: store i8 [[R62]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 62), align 1
+; CHECK-NEXT: store i8 [[R63]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 63), align 1
+; CHECK-NEXT: ret void
+;
+ %a0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 0 ), align 1
+ %a1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 1 ), align 1
+ %a2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 2 ), align 1
+ %a3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 3 ), align 1
+ %a4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 4 ), align 1
+ %a5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 5 ), align 1
+ %a6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 6 ), align 1
+ %a7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 7 ), align 1
+ %a8 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 8 ), align 1
+ %a9 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 9 ), align 1
+ %a10 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 10), align 1
+ %a11 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 11), align 1
+ %a12 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 12), align 1
+ %a13 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 13), align 1
+ %a14 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 14), align 1
+ %a15 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 15), align 1
+ %a16 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 16), align 1
+ %a17 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 17), align 1
+ %a18 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 18), align 1
+ %a19 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 19), align 1
+ %a20 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 20), align 1
+ %a21 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 21), align 1
+ %a22 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 22), align 1
+ %a23 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 23), align 1
+ %a24 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 24), align 1
+ %a25 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 25), align 1
+ %a26 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 26), align 1
+ %a27 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 27), align 1
+ %a28 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 28), align 1
+ %a29 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 29), align 1
+ %a30 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 30), align 1
+ %a31 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 31), align 1
+ %a32 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 32), align 1
+ %a33 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 33), align 1
+ %a34 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 34), align 1
+ %a35 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 35), align 1
+ %a36 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 36), align 1
+ %a37 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 37), align 1
+ %a38 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 38), align 1
+ %a39 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 39), align 1
+ %a40 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 40), align 1
+ %a41 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 41), align 1
+ %a42 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 42), align 1
+ %a43 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 43), align 1
+ %a44 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 44), align 1
+ %a45 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 45), align 1
+ %a46 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 46), align 1
+ %a47 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 47), align 1
+ %a48 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 48), align 1
+ %a49 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 49), align 1
+ %a50 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 50), align 1
+ %a51 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 51), align 1
+ %a52 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 52), align 1
+ %a53 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 53), align 1
+ %a54 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 54), align 1
+ %a55 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 55), align 1
+ %a56 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 56), align 1
+ %a57 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 57), align 1
+ %a58 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 58), align 1
+ %a59 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 59), align 1
+ %a60 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 60), align 1
+ %a61 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 61), align 1
+ %a62 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 62), align 1
+ %a63 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 63), align 1
+ %b0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 0 ), align 1
+ %b1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 1 ), align 1
+ %b2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 2 ), align 1
+ %b3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 3 ), align 1
+ %b4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 4 ), align 1
+ %b5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 5 ), align 1
+ %b6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 6 ), align 1
+ %b7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 7 ), align 1
+ %b8 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 8 ), align 1
+ %b9 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 9 ), align 1
+ %b10 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 10), align 1
+ %b11 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 11), align 1
+ %b12 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 12), align 1
+ %b13 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 13), align 1
+ %b14 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 14), align 1
+ %b15 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 15), align 1
+ %b16 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 16), align 1
+ %b17 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 17), align 1
+ %b18 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 18), align 1
+ %b19 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 19), align 1
+ %b20 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 20), align 1
+ %b21 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 21), align 1
+ %b22 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 22), align 1
+ %b23 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 23), align 1
+ %b24 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 24), align 1
+ %b25 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 25), align 1
+ %b26 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 26), align 1
+ %b27 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 27), align 1
+ %b28 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 28), align 1
+ %b29 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 29), align 1
+ %b30 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 30), align 1
+ %b31 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 31), align 1
+ %b32 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 32), align 1
+ %b33 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 33), align 1
+ %b34 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 34), align 1
+ %b35 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 35), align 1
+ %b36 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 36), align 1
+ %b37 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 37), align 1
+ %b38 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 38), align 1
+ %b39 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 39), align 1
+ %b40 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 40), align 1
+ %b41 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 41), align 1
+ %b42 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 42), align 1
+ %b43 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 43), align 1
+ %b44 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 44), align 1
+ %b45 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 45), align 1
+ %b46 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 46), align 1
+ %b47 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 47), align 1
+ %b48 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 48), align 1
+ %b49 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 49), align 1
+ %b50 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 50), align 1
+ %b51 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 51), align 1
+ %b52 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 52), align 1
+ %b53 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 53), align 1
+ %b54 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 54), align 1
+ %b55 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 55), align 1
+ %b56 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 56), align 1
+ %b57 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 57), align 1
+ %b58 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 58), align 1
+ %b59 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 59), align 1
+ %b60 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 60), align 1
+ %b61 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 61), align 1
+ %b62 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 62), align 1
+ %b63 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 63), align 1
+ %c0 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a0 , i8 %b0 )
+ %c1 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a1 , i8 %b1 )
+ %c2 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a2 , i8 %b2 )
+ %c3 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a3 , i8 %b3 )
+ %c4 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a4 , i8 %b4 )
+ %c5 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a5 , i8 %b5 )
+ %c6 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a6 , i8 %b6 )
+ %c7 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a7 , i8 %b7 )
+ %c8 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a8 , i8 %b8 )
+ %c9 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a9 , i8 %b9 )
+ %c10 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a10, i8 %b10)
+ %c11 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a11, i8 %b11)
+ %c12 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a12, i8 %b12)
+ %c13 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a13, i8 %b13)
+ %c14 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a14, i8 %b14)
+ %c15 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a15, i8 %b15)
+ %c16 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a16, i8 %b16)
+ %c17 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a17, i8 %b17)
+ %c18 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a18, i8 %b18)
+ %c19 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a19, i8 %b19)
+ %c20 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a20, i8 %b20)
+ %c21 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a21, i8 %b21)
+ %c22 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a22, i8 %b22)
+ %c23 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a23, i8 %b23)
+ %c24 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a24, i8 %b24)
+ %c25 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a25, i8 %b25)
+ %c26 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a26, i8 %b26)
+ %c27 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a27, i8 %b27)
+ %c28 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a28, i8 %b28)
+ %c29 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a29, i8 %b29)
+ %c30 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a30, i8 %b30)
+ %c31 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a31, i8 %b31)
+ %c32 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a32, i8 %b32)
+ %c33 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a33, i8 %b33)
+ %c34 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a34, i8 %b34)
+ %c35 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a35, i8 %b35)
+ %c36 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a36, i8 %b36)
+ %c37 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a37, i8 %b37)
+ %c38 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a38, i8 %b38)
+ %c39 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a39, i8 %b39)
+ %c40 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a40, i8 %b40)
+ %c41 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a41, i8 %b41)
+ %c42 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a42, i8 %b42)
+ %c43 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a43, i8 %b43)
+ %c44 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a44, i8 %b44)
+ %c45 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a45, i8 %b45)
+ %c46 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a46, i8 %b46)
+ %c47 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a47, i8 %b47)
+ %c48 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a48, i8 %b48)
+ %c49 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a49, i8 %b49)
+ %c50 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a50, i8 %b50)
+ %c51 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a51, i8 %b51)
+ %c52 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a52, i8 %b52)
+ %c53 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a53, i8 %b53)
+ %c54 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a54, i8 %b54)
+ %c55 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a55, i8 %b55)
+ %c56 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a56, i8 %b56)
+ %c57 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a57, i8 %b57)
+ %c58 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a58, i8 %b58)
+ %c59 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a59, i8 %b59)
+ %c60 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a60, i8 %b60)
+ %c61 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a61, i8 %b61)
+ %c62 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a62, i8 %b62)
+ %c63 = call {i8, i1} @llvm.usub.with.overflow.i8(i8 %a63, i8 %b63)
+ %r0 = extractvalue {i8, i1} %c0 , 0
+ %r1 = extractvalue {i8, i1} %c1 , 0
+ %r2 = extractvalue {i8, i1} %c2 , 0
+ %r3 = extractvalue {i8, i1} %c3 , 0
+ %r4 = extractvalue {i8, i1} %c4 , 0
+ %r5 = extractvalue {i8, i1} %c5 , 0
+ %r6 = extractvalue {i8, i1} %c6 , 0
+ %r7 = extractvalue {i8, i1} %c7 , 0
+ %r8 = extractvalue {i8, i1} %c8 , 0
+ %r9 = extractvalue {i8, i1} %c9 , 0
+ %r10 = extractvalue {i8, i1} %c10, 0
+ %r11 = extractvalue {i8, i1} %c11, 0
+ %r12 = extractvalue {i8, i1} %c12, 0
+ %r13 = extractvalue {i8, i1} %c13, 0
+ %r14 = extractvalue {i8, i1} %c14, 0
+ %r15 = extractvalue {i8, i1} %c15, 0
+ %r16 = extractvalue {i8, i1} %c16, 0
+ %r17 = extractvalue {i8, i1} %c17, 0
+ %r18 = extractvalue {i8, i1} %c18, 0
+ %r19 = extractvalue {i8, i1} %c19, 0
+ %r20 = extractvalue {i8, i1} %c20, 0
+ %r21 = extractvalue {i8, i1} %c21, 0
+ %r22 = extractvalue {i8, i1} %c22, 0
+ %r23 = extractvalue {i8, i1} %c23, 0
+ %r24 = extractvalue {i8, i1} %c24, 0
+ %r25 = extractvalue {i8, i1} %c25, 0
+ %r26 = extractvalue {i8, i1} %c26, 0
+ %r27 = extractvalue {i8, i1} %c27, 0
+ %r28 = extractvalue {i8, i1} %c28, 0
+ %r29 = extractvalue {i8, i1} %c29, 0
+ %r30 = extractvalue {i8, i1} %c30, 0
+ %r31 = extractvalue {i8, i1} %c31, 0
+ %r32 = extractvalue {i8, i1} %c32, 0
+ %r33 = extractvalue {i8, i1} %c33, 0
+ %r34 = extractvalue {i8, i1} %c34, 0
+ %r35 = extractvalue {i8, i1} %c35, 0
+ %r36 = extractvalue {i8, i1} %c36, 0
+ %r37 = extractvalue {i8, i1} %c37, 0
+ %r38 = extractvalue {i8, i1} %c38, 0
+ %r39 = extractvalue {i8, i1} %c39, 0
+ %r40 = extractvalue {i8, i1} %c40, 0
+ %r41 = extractvalue {i8, i1} %c41, 0
+ %r42 = extractvalue {i8, i1} %c42, 0
+ %r43 = extractvalue {i8, i1} %c43, 0
+ %r44 = extractvalue {i8, i1} %c44, 0
+ %r45 = extractvalue {i8, i1} %c45, 0
+ %r46 = extractvalue {i8, i1} %c46, 0
+ %r47 = extractvalue {i8, i1} %c47, 0
+ %r48 = extractvalue {i8, i1} %c48, 0
+ %r49 = extractvalue {i8, i1} %c49, 0
+ %r50 = extractvalue {i8, i1} %c50, 0
+ %r51 = extractvalue {i8, i1} %c51, 0
+ %r52 = extractvalue {i8, i1} %c52, 0
+ %r53 = extractvalue {i8, i1} %c53, 0
+ %r54 = extractvalue {i8, i1} %c54, 0
+ %r55 = extractvalue {i8, i1} %c55, 0
+ %r56 = extractvalue {i8, i1} %c56, 0
+ %r57 = extractvalue {i8, i1} %c57, 0
+ %r58 = extractvalue {i8, i1} %c58, 0
+ %r59 = extractvalue {i8, i1} %c59, 0
+ %r60 = extractvalue {i8, i1} %c60, 0
+ %r61 = extractvalue {i8, i1} %c61, 0
+ %r62 = extractvalue {i8, i1} %c62, 0
+ %r63 = extractvalue {i8, i1} %c63, 0
+ store i8 %r0 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 0 ), align 1
+ store i8 %r1 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 1 ), align 1
+ store i8 %r2 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 2 ), align 1
+ store i8 %r3 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 3 ), align 1
+ store i8 %r4 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 4 ), align 1
+ store i8 %r5 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 5 ), align 1
+ store i8 %r6 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 6 ), align 1
+ store i8 %r7 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 7 ), align 1
+ store i8 %r8 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 8 ), align 1
+ store i8 %r9 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 9 ), align 1
+ store i8 %r10, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 10), align 1
+ store i8 %r11, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 11), align 1
+ store i8 %r12, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 12), align 1
+ store i8 %r13, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 13), align 1
+ store i8 %r14, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 14), align 1
+ store i8 %r15, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 15), align 1
+ store i8 %r16, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 16), align 1
+ store i8 %r17, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 17), align 1
+ store i8 %r18, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 18), align 1
+ store i8 %r19, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 19), align 1
+ store i8 %r20, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 20), align 1
+ store i8 %r21, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 21), align 1
+ store i8 %r22, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 22), align 1
+ store i8 %r23, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 23), align 1
+ store i8 %r24, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 24), align 1
+ store i8 %r25, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 25), align 1
+ store i8 %r26, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 26), align 1
+ store i8 %r27, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 27), align 1
+ store i8 %r28, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 28), align 1
+ store i8 %r29, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 29), align 1
+ store i8 %r30, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 30), align 1
+ store i8 %r31, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 31), align 1
+ store i8 %r32, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 32), align 1
+ store i8 %r33, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 33), align 1
+ store i8 %r34, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 34), align 1
+ store i8 %r35, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 35), align 1
+ store i8 %r36, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 36), align 1
+ store i8 %r37, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 37), align 1
+ store i8 %r38, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 38), align 1
+ store i8 %r39, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 39), align 1
+ store i8 %r40, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 40), align 1
+ store i8 %r41, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 41), align 1
+ store i8 %r42, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 42), align 1
+ store i8 %r43, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 43), align 1
+ store i8 %r44, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 44), align 1
+ store i8 %r45, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 45), align 1
+ store i8 %r46, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 46), align 1
+ store i8 %r47, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 47), align 1
+ store i8 %r48, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 48), align 1
+ store i8 %r49, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 49), align 1
+ store i8 %r50, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 50), align 1
+ store i8 %r51, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 51), align 1
+ store i8 %r52, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 52), align 1
+ store i8 %r53, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 53), align 1
+ store i8 %r54, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 54), align 1
+ store i8 %r55, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 55), align 1
+ store i8 %r56, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 56), align 1
+ store i8 %r57, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 57), align 1
+ store i8 %r58, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 58), align 1
+ store i8 %r59, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 59), align 1
+ store i8 %r60, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 60), align 1
+ store i8 %r61, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 61), align 1
+ store i8 %r62, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 62), align 1
+ store i8 %r63, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 63), align 1
+ ret void
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/arith-sub.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,708 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=slm -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SLM
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=knl -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skx -mattr=-prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512 --check-prefix=AVX512BW
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skx -mattr=+prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2
+
+ at a64 = common global [8 x i64] zeroinitializer, align 64
+ at b64 = common global [8 x i64] zeroinitializer, align 64
+ at c64 = common global [8 x i64] zeroinitializer, align 64
+ at a32 = common global [16 x i32] zeroinitializer, align 64
+ at b32 = common global [16 x i32] zeroinitializer, align 64
+ at c32 = common global [16 x i32] zeroinitializer, align 64
+ at a16 = common global [32 x i16] zeroinitializer, align 64
+ at b16 = common global [32 x i16] zeroinitializer, align 64
+ at c16 = common global [32 x i16] zeroinitializer, align 64
+ at a8 = common global [64 x i8] zeroinitializer, align 64
+ at b8 = common global [64 x i8] zeroinitializer, align 64
+ at c8 = common global [64 x i8] zeroinitializer, align 64
+
+define void @sub_v8i64() {
+; SSE-LABEL: @sub_v8i64(
+; SSE-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @a64 to <2 x i64>*), align 8
+; SSE-NEXT: [[TMP2:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2) to <2 x i64>*), align 8
+; SSE-NEXT: [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4) to <2 x i64>*), align 8
+; SSE-NEXT: [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6) to <2 x i64>*), align 8
+; SSE-NEXT: [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @b64 to <2 x i64>*), align 8
+; SSE-NEXT: [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2) to <2 x i64>*), align 8
+; SSE-NEXT: [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4) to <2 x i64>*), align 8
+; SSE-NEXT: [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6) to <2 x i64>*), align 8
+; SSE-NEXT: [[TMP9:%.*]] = sub <2 x i64> [[TMP1]], [[TMP5]]
+; SSE-NEXT: [[TMP10:%.*]] = sub <2 x i64> [[TMP2]], [[TMP6]]
+; SSE-NEXT: [[TMP11:%.*]] = sub <2 x i64> [[TMP3]], [[TMP7]]
+; SSE-NEXT: [[TMP12:%.*]] = sub <2 x i64> [[TMP4]], [[TMP8]]
+; SSE-NEXT: store <2 x i64> [[TMP9]], <2 x i64>* bitcast ([8 x i64]* @c64 to <2 x i64>*), align 8
+; SSE-NEXT: store <2 x i64> [[TMP10]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2) to <2 x i64>*), align 8
+; SSE-NEXT: store <2 x i64> [[TMP11]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4) to <2 x i64>*), align 8
+; SSE-NEXT: store <2 x i64> [[TMP12]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6) to <2 x i64>*), align 8
+; SSE-NEXT: ret void
+;
+; SLM-LABEL: @sub_v8i64(
+; SLM-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @a64 to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP2:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP3:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP4:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP5:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @b64 to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP6:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP7:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP8:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6) to <2 x i64>*), align 8
+; SLM-NEXT: [[TMP9:%.*]] = sub <2 x i64> [[TMP1]], [[TMP5]]
+; SLM-NEXT: [[TMP10:%.*]] = sub <2 x i64> [[TMP2]], [[TMP6]]
+; SLM-NEXT: [[TMP11:%.*]] = sub <2 x i64> [[TMP3]], [[TMP7]]
+; SLM-NEXT: [[TMP12:%.*]] = sub <2 x i64> [[TMP4]], [[TMP8]]
+; SLM-NEXT: store <2 x i64> [[TMP9]], <2 x i64>* bitcast ([8 x i64]* @c64 to <2 x i64>*), align 8
+; SLM-NEXT: store <2 x i64> [[TMP10]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2) to <2 x i64>*), align 8
+; SLM-NEXT: store <2 x i64> [[TMP11]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4) to <2 x i64>*), align 8
+; SLM-NEXT: store <2 x i64> [[TMP12]], <2 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6) to <2 x i64>*), align 8
+; SLM-NEXT: ret void
+;
+; AVX-LABEL: @sub_v8i64(
+; AVX-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @a64 to <4 x i64>*), align 8
+; AVX-NEXT: [[TMP2:%.*]] = load <4 x i64>, <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX-NEXT: [[TMP3:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @b64 to <4 x i64>*), align 8
+; AVX-NEXT: [[TMP4:%.*]] = load <4 x i64>, <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX-NEXT: [[TMP5:%.*]] = sub <4 x i64> [[TMP1]], [[TMP3]]
+; AVX-NEXT: [[TMP6:%.*]] = sub <4 x i64> [[TMP2]], [[TMP4]]
+; AVX-NEXT: store <4 x i64> [[TMP5]], <4 x i64>* bitcast ([8 x i64]* @c64 to <4 x i64>*), align 8
+; AVX-NEXT: store <4 x i64> [[TMP6]], <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4) to <4 x i64>*), align 8
+; AVX-NEXT: ret void
+;
+; AVX512-LABEL: @sub_v8i64(
+; AVX512-NEXT: [[TMP1:%.*]] = load <8 x i64>, <8 x i64>* bitcast ([8 x i64]* @a64 to <8 x i64>*), align 8
+; AVX512-NEXT: [[TMP2:%.*]] = load <8 x i64>, <8 x i64>* bitcast ([8 x i64]* @b64 to <8 x i64>*), align 8
+; AVX512-NEXT: [[TMP3:%.*]] = sub <8 x i64> [[TMP1]], [[TMP2]]
+; AVX512-NEXT: store <8 x i64> [[TMP3]], <8 x i64>* bitcast ([8 x i64]* @c64 to <8 x i64>*), align 8
+; AVX512-NEXT: ret void
+;
+ %a0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 0), align 8
+ %a1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 1), align 8
+ %a2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 2), align 8
+ %a3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 3), align 8
+ %a4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 4), align 8
+ %a5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 5), align 8
+ %a6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 6), align 8
+ %a7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @a64, i32 0, i64 7), align 8
+ %b0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 0), align 8
+ %b1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 1), align 8
+ %b2 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 2), align 8
+ %b3 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 3), align 8
+ %b4 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 4), align 8
+ %b5 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 5), align 8
+ %b6 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 6), align 8
+ %b7 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @b64, i32 0, i64 7), align 8
+ %r0 = sub i64 %a0, %b0
+ %r1 = sub i64 %a1, %b1
+ %r2 = sub i64 %a2, %b2
+ %r3 = sub i64 %a3, %b3
+ %r4 = sub i64 %a4, %b4
+ %r5 = sub i64 %a5, %b5
+ %r6 = sub i64 %a6, %b6
+ %r7 = sub i64 %a7, %b7
+ store i64 %r0, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 0), align 8
+ store i64 %r1, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 1), align 8
+ store i64 %r2, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 2), align 8
+ store i64 %r3, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 3), align 8
+ store i64 %r4, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 4), align 8
+ store i64 %r5, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 5), align 8
+ store i64 %r6, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 6), align 8
+ store i64 %r7, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @c64, i32 0, i64 7), align 8
+ ret void
+}
+
+define void @sub_v16i32() {
+; SSE-LABEL: @sub_v16i32(
+; SSE-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @a32 to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @b32 to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12) to <4 x i32>*), align 4
+; SSE-NEXT: [[TMP9:%.*]] = sub <4 x i32> [[TMP1]], [[TMP5]]
+; SSE-NEXT: [[TMP10:%.*]] = sub <4 x i32> [[TMP2]], [[TMP6]]
+; SSE-NEXT: [[TMP11:%.*]] = sub <4 x i32> [[TMP3]], [[TMP7]]
+; SSE-NEXT: [[TMP12:%.*]] = sub <4 x i32> [[TMP4]], [[TMP8]]
+; SSE-NEXT: store <4 x i32> [[TMP9]], <4 x i32>* bitcast ([16 x i32]* @c32 to <4 x i32>*), align 4
+; SSE-NEXT: store <4 x i32> [[TMP10]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4) to <4 x i32>*), align 4
+; SSE-NEXT: store <4 x i32> [[TMP11]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8) to <4 x i32>*), align 4
+; SSE-NEXT: store <4 x i32> [[TMP12]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12) to <4 x i32>*), align 4
+; SSE-NEXT: ret void
+;
+; SLM-LABEL: @sub_v16i32(
+; SLM-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @a32 to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([16 x i32]* @b32 to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP8:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12) to <4 x i32>*), align 4
+; SLM-NEXT: [[TMP9:%.*]] = sub <4 x i32> [[TMP1]], [[TMP5]]
+; SLM-NEXT: [[TMP10:%.*]] = sub <4 x i32> [[TMP2]], [[TMP6]]
+; SLM-NEXT: [[TMP11:%.*]] = sub <4 x i32> [[TMP3]], [[TMP7]]
+; SLM-NEXT: [[TMP12:%.*]] = sub <4 x i32> [[TMP4]], [[TMP8]]
+; SLM-NEXT: store <4 x i32> [[TMP9]], <4 x i32>* bitcast ([16 x i32]* @c32 to <4 x i32>*), align 4
+; SLM-NEXT: store <4 x i32> [[TMP10]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4) to <4 x i32>*), align 4
+; SLM-NEXT: store <4 x i32> [[TMP11]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8) to <4 x i32>*), align 4
+; SLM-NEXT: store <4 x i32> [[TMP12]], <4 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12) to <4 x i32>*), align 4
+; SLM-NEXT: ret void
+;
+; AVX-LABEL: @sub_v16i32(
+; AVX-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([16 x i32]* @a32 to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8) to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP3:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([16 x i32]* @b32 to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP4:%.*]] = load <8 x i32>, <8 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8) to <8 x i32>*), align 4
+; AVX-NEXT: [[TMP5:%.*]] = sub <8 x i32> [[TMP1]], [[TMP3]]
+; AVX-NEXT: [[TMP6:%.*]] = sub <8 x i32> [[TMP2]], [[TMP4]]
+; AVX-NEXT: store <8 x i32> [[TMP5]], <8 x i32>* bitcast ([16 x i32]* @c32 to <8 x i32>*), align 4
+; AVX-NEXT: store <8 x i32> [[TMP6]], <8 x i32>* bitcast (i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8) to <8 x i32>*), align 4
+; AVX-NEXT: ret void
+;
+; AVX512-LABEL: @sub_v16i32(
+; AVX512-NEXT: [[TMP1:%.*]] = load <16 x i32>, <16 x i32>* bitcast ([16 x i32]* @a32 to <16 x i32>*), align 4
+; AVX512-NEXT: [[TMP2:%.*]] = load <16 x i32>, <16 x i32>* bitcast ([16 x i32]* @b32 to <16 x i32>*), align 4
+; AVX512-NEXT: [[TMP3:%.*]] = sub <16 x i32> [[TMP1]], [[TMP2]]
+; AVX512-NEXT: store <16 x i32> [[TMP3]], <16 x i32>* bitcast ([16 x i32]* @c32 to <16 x i32>*), align 4
+; AVX512-NEXT: ret void
+;
+ %a0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 0 ), align 4
+ %a1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 1 ), align 4
+ %a2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 2 ), align 4
+ %a3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 3 ), align 4
+ %a4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 4 ), align 4
+ %a5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 5 ), align 4
+ %a6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 6 ), align 4
+ %a7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 7 ), align 4
+ %a8 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 8 ), align 4
+ %a9 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 9 ), align 4
+ %a10 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 10), align 4
+ %a11 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 11), align 4
+ %a12 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 12), align 4
+ %a13 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 13), align 4
+ %a14 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 14), align 4
+ %a15 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @a32, i32 0, i64 15), align 4
+ %b0 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 0 ), align 4
+ %b1 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 1 ), align 4
+ %b2 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 2 ), align 4
+ %b3 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 3 ), align 4
+ %b4 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 4 ), align 4
+ %b5 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 5 ), align 4
+ %b6 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 6 ), align 4
+ %b7 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 7 ), align 4
+ %b8 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 8 ), align 4
+ %b9 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 9 ), align 4
+ %b10 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 10), align 4
+ %b11 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 11), align 4
+ %b12 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 12), align 4
+ %b13 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 13), align 4
+ %b14 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 14), align 4
+ %b15 = load i32, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @b32, i32 0, i64 15), align 4
+ %r0 = sub i32 %a0 , %b0
+ %r1 = sub i32 %a1 , %b1
+ %r2 = sub i32 %a2 , %b2
+ %r3 = sub i32 %a3 , %b3
+ %r4 = sub i32 %a4 , %b4
+ %r5 = sub i32 %a5 , %b5
+ %r6 = sub i32 %a6 , %b6
+ %r7 = sub i32 %a7 , %b7
+ %r8 = sub i32 %a8 , %b8
+ %r9 = sub i32 %a9 , %b9
+ %r10 = sub i32 %a10, %b10
+ %r11 = sub i32 %a11, %b11
+ %r12 = sub i32 %a12, %b12
+ %r13 = sub i32 %a13, %b13
+ %r14 = sub i32 %a14, %b14
+ %r15 = sub i32 %a15, %b15
+ store i32 %r0 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 0 ), align 4
+ store i32 %r1 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 1 ), align 4
+ store i32 %r2 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 2 ), align 4
+ store i32 %r3 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 3 ), align 4
+ store i32 %r4 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 4 ), align 4
+ store i32 %r5 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 5 ), align 4
+ store i32 %r6 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 6 ), align 4
+ store i32 %r7 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 7 ), align 4
+ store i32 %r8 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 8 ), align 4
+ store i32 %r9 , i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 9 ), align 4
+ store i32 %r10, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 10), align 4
+ store i32 %r11, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 11), align 4
+ store i32 %r12, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 12), align 4
+ store i32 %r13, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 13), align 4
+ store i32 %r14, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 14), align 4
+ store i32 %r15, i32* getelementptr inbounds ([16 x i32], [16 x i32]* @c32, i32 0, i64 15), align 4
+ ret void
+}
+
+define void @sub_v32i16() {
+; SSE-LABEL: @sub_v32i16(
+; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @a16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @b16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP9:%.*]] = sub <8 x i16> [[TMP1]], [[TMP5]]
+; SSE-NEXT: [[TMP10:%.*]] = sub <8 x i16> [[TMP2]], [[TMP6]]
+; SSE-NEXT: [[TMP11:%.*]] = sub <8 x i16> [[TMP3]], [[TMP7]]
+; SSE-NEXT: [[TMP12:%.*]] = sub <8 x i16> [[TMP4]], [[TMP8]]
+; SSE-NEXT: store <8 x i16> [[TMP9]], <8 x i16>* bitcast ([32 x i16]* @c16 to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP10]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP11]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP12]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24) to <8 x i16>*), align 2
+; SSE-NEXT: ret void
+;
+; SLM-LABEL: @sub_v32i16(
+; SLM-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @a16 to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP3:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP4:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP5:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([32 x i16]* @b16 to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP6:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP7:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP8:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24) to <8 x i16>*), align 2
+; SLM-NEXT: [[TMP9:%.*]] = sub <8 x i16> [[TMP1]], [[TMP5]]
+; SLM-NEXT: [[TMP10:%.*]] = sub <8 x i16> [[TMP2]], [[TMP6]]
+; SLM-NEXT: [[TMP11:%.*]] = sub <8 x i16> [[TMP3]], [[TMP7]]
+; SLM-NEXT: [[TMP12:%.*]] = sub <8 x i16> [[TMP4]], [[TMP8]]
+; SLM-NEXT: store <8 x i16> [[TMP9]], <8 x i16>* bitcast ([32 x i16]* @c16 to <8 x i16>*), align 2
+; SLM-NEXT: store <8 x i16> [[TMP10]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8) to <8 x i16>*), align 2
+; SLM-NEXT: store <8 x i16> [[TMP11]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <8 x i16>*), align 2
+; SLM-NEXT: store <8 x i16> [[TMP12]], <8 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24) to <8 x i16>*), align 2
+; SLM-NEXT: ret void
+;
+; AVX-LABEL: @sub_v32i16(
+; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @a16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP3:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @b16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP4:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP5:%.*]] = sub <16 x i16> [[TMP1]], [[TMP3]]
+; AVX-NEXT: [[TMP6:%.*]] = sub <16 x i16> [[TMP2]], [[TMP4]]
+; AVX-NEXT: store <16 x i16> [[TMP5]], <16 x i16>* bitcast ([32 x i16]* @c16 to <16 x i16>*), align 2
+; AVX-NEXT: store <16 x i16> [[TMP6]], <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX-NEXT: ret void
+;
+; AVX512-LABEL: @sub_v32i16(
+; AVX512-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @a16 to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP2:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP3:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([32 x i16]* @b16 to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP4:%.*]] = load <16 x i16>, <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX512-NEXT: [[TMP5:%.*]] = sub <16 x i16> [[TMP1]], [[TMP3]]
+; AVX512-NEXT: [[TMP6:%.*]] = sub <16 x i16> [[TMP2]], [[TMP4]]
+; AVX512-NEXT: store <16 x i16> [[TMP5]], <16 x i16>* bitcast ([32 x i16]* @c16 to <16 x i16>*), align 2
+; AVX512-NEXT: store <16 x i16> [[TMP6]], <16 x i16>* bitcast (i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16) to <16 x i16>*), align 2
+; AVX512-NEXT: ret void
+;
+ %a0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 0 ), align 2
+ %a1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 1 ), align 2
+ %a2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 2 ), align 2
+ %a3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 3 ), align 2
+ %a4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 4 ), align 2
+ %a5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 5 ), align 2
+ %a6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 6 ), align 2
+ %a7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 7 ), align 2
+ %a8 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 8 ), align 2
+ %a9 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 9 ), align 2
+ %a10 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 10), align 2
+ %a11 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 11), align 2
+ %a12 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 12), align 2
+ %a13 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 13), align 2
+ %a14 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 14), align 2
+ %a15 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 15), align 2
+ %a16 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 16), align 2
+ %a17 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 17), align 2
+ %a18 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 18), align 2
+ %a19 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 19), align 2
+ %a20 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 20), align 2
+ %a21 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 21), align 2
+ %a22 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 22), align 2
+ %a23 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 23), align 2
+ %a24 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 24), align 2
+ %a25 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 25), align 2
+ %a26 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 26), align 2
+ %a27 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 27), align 2
+ %a28 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 28), align 2
+ %a29 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 29), align 2
+ %a30 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 30), align 2
+ %a31 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @a16, i32 0, i64 31), align 2
+ %b0 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 0 ), align 2
+ %b1 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 1 ), align 2
+ %b2 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 2 ), align 2
+ %b3 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 3 ), align 2
+ %b4 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 4 ), align 2
+ %b5 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 5 ), align 2
+ %b6 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 6 ), align 2
+ %b7 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 7 ), align 2
+ %b8 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 8 ), align 2
+ %b9 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 9 ), align 2
+ %b10 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 10), align 2
+ %b11 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 11), align 2
+ %b12 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 12), align 2
+ %b13 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 13), align 2
+ %b14 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 14), align 2
+ %b15 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 15), align 2
+ %b16 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 16), align 2
+ %b17 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 17), align 2
+ %b18 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 18), align 2
+ %b19 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 19), align 2
+ %b20 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 20), align 2
+ %b21 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 21), align 2
+ %b22 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 22), align 2
+ %b23 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 23), align 2
+ %b24 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 24), align 2
+ %b25 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 25), align 2
+ %b26 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 26), align 2
+ %b27 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 27), align 2
+ %b28 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 28), align 2
+ %b29 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 29), align 2
+ %b30 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 30), align 2
+ %b31 = load i16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @b16, i32 0, i64 31), align 2
+ %r0 = sub i16 %a0 , %b0
+ %r1 = sub i16 %a1 , %b1
+ %r2 = sub i16 %a2 , %b2
+ %r3 = sub i16 %a3 , %b3
+ %r4 = sub i16 %a4 , %b4
+ %r5 = sub i16 %a5 , %b5
+ %r6 = sub i16 %a6 , %b6
+ %r7 = sub i16 %a7 , %b7
+ %r8 = sub i16 %a8 , %b8
+ %r9 = sub i16 %a9 , %b9
+ %r10 = sub i16 %a10, %b10
+ %r11 = sub i16 %a11, %b11
+ %r12 = sub i16 %a12, %b12
+ %r13 = sub i16 %a13, %b13
+ %r14 = sub i16 %a14, %b14
+ %r15 = sub i16 %a15, %b15
+ %r16 = sub i16 %a16, %b16
+ %r17 = sub i16 %a17, %b17
+ %r18 = sub i16 %a18, %b18
+ %r19 = sub i16 %a19, %b19
+ %r20 = sub i16 %a20, %b20
+ %r21 = sub i16 %a21, %b21
+ %r22 = sub i16 %a22, %b22
+ %r23 = sub i16 %a23, %b23
+ %r24 = sub i16 %a24, %b24
+ %r25 = sub i16 %a25, %b25
+ %r26 = sub i16 %a26, %b26
+ %r27 = sub i16 %a27, %b27
+ %r28 = sub i16 %a28, %b28
+ %r29 = sub i16 %a29, %b29
+ %r30 = sub i16 %a30, %b30
+ %r31 = sub i16 %a31, %b31
+ store i16 %r0 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 0 ), align 2
+ store i16 %r1 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 1 ), align 2
+ store i16 %r2 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 2 ), align 2
+ store i16 %r3 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 3 ), align 2
+ store i16 %r4 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 4 ), align 2
+ store i16 %r5 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 5 ), align 2
+ store i16 %r6 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 6 ), align 2
+ store i16 %r7 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 7 ), align 2
+ store i16 %r8 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 8 ), align 2
+ store i16 %r9 , i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 9 ), align 2
+ store i16 %r10, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 10), align 2
+ store i16 %r11, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 11), align 2
+ store i16 %r12, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 12), align 2
+ store i16 %r13, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 13), align 2
+ store i16 %r14, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 14), align 2
+ store i16 %r15, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 15), align 2
+ store i16 %r16, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 16), align 2
+ store i16 %r17, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 17), align 2
+ store i16 %r18, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 18), align 2
+ store i16 %r19, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 19), align 2
+ store i16 %r20, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 20), align 2
+ store i16 %r21, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 21), align 2
+ store i16 %r22, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 22), align 2
+ store i16 %r23, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 23), align 2
+ store i16 %r24, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 24), align 2
+ store i16 %r25, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 25), align 2
+ store i16 %r26, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 26), align 2
+ store i16 %r27, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 27), align 2
+ store i16 %r28, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 28), align 2
+ store i16 %r29, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 29), align 2
+ store i16 %r30, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 30), align 2
+ store i16 %r31, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @c16, i32 0, i64 31), align 2
+ ret void
+}
+
+define void @sub_v64i8() {
+; CHECK-LABEL: @sub_v64i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([64 x i8]* @a8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP3:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 32) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 48) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP5:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([64 x i8]* @b8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP6:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP7:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 32) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP8:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 48) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP9:%.*]] = sub <16 x i8> [[TMP1]], [[TMP5]]
+; CHECK-NEXT: [[TMP10:%.*]] = sub <16 x i8> [[TMP2]], [[TMP6]]
+; CHECK-NEXT: [[TMP11:%.*]] = sub <16 x i8> [[TMP3]], [[TMP7]]
+; CHECK-NEXT: [[TMP12:%.*]] = sub <16 x i8> [[TMP4]], [[TMP8]]
+; CHECK-NEXT: store <16 x i8> [[TMP9]], <16 x i8>* bitcast ([64 x i8]* @c8 to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP10]], <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP11]], <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 32) to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP12]], <16 x i8>* bitcast (i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 48) to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %a0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 0 ), align 1
+ %a1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 1 ), align 1
+ %a2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 2 ), align 1
+ %a3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 3 ), align 1
+ %a4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 4 ), align 1
+ %a5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 5 ), align 1
+ %a6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 6 ), align 1
+ %a7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 7 ), align 1
+ %a8 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 8 ), align 1
+ %a9 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 9 ), align 1
+ %a10 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 10), align 1
+ %a11 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 11), align 1
+ %a12 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 12), align 1
+ %a13 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 13), align 1
+ %a14 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 14), align 1
+ %a15 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 15), align 1
+ %a16 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 16), align 1
+ %a17 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 17), align 1
+ %a18 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 18), align 1
+ %a19 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 19), align 1
+ %a20 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 20), align 1
+ %a21 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 21), align 1
+ %a22 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 22), align 1
+ %a23 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 23), align 1
+ %a24 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 24), align 1
+ %a25 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 25), align 1
+ %a26 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 26), align 1
+ %a27 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 27), align 1
+ %a28 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 28), align 1
+ %a29 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 29), align 1
+ %a30 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 30), align 1
+ %a31 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 31), align 1
+ %a32 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 32), align 1
+ %a33 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 33), align 1
+ %a34 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 34), align 1
+ %a35 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 35), align 1
+ %a36 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 36), align 1
+ %a37 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 37), align 1
+ %a38 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 38), align 1
+ %a39 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 39), align 1
+ %a40 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 40), align 1
+ %a41 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 41), align 1
+ %a42 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 42), align 1
+ %a43 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 43), align 1
+ %a44 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 44), align 1
+ %a45 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 45), align 1
+ %a46 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 46), align 1
+ %a47 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 47), align 1
+ %a48 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 48), align 1
+ %a49 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 49), align 1
+ %a50 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 50), align 1
+ %a51 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 51), align 1
+ %a52 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 52), align 1
+ %a53 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 53), align 1
+ %a54 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 54), align 1
+ %a55 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 55), align 1
+ %a56 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 56), align 1
+ %a57 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 57), align 1
+ %a58 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 58), align 1
+ %a59 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 59), align 1
+ %a60 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 60), align 1
+ %a61 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 61), align 1
+ %a62 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 62), align 1
+ %a63 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @a8, i32 0, i64 63), align 1
+ %b0 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 0 ), align 1
+ %b1 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 1 ), align 1
+ %b2 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 2 ), align 1
+ %b3 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 3 ), align 1
+ %b4 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 4 ), align 1
+ %b5 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 5 ), align 1
+ %b6 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 6 ), align 1
+ %b7 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 7 ), align 1
+ %b8 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 8 ), align 1
+ %b9 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 9 ), align 1
+ %b10 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 10), align 1
+ %b11 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 11), align 1
+ %b12 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 12), align 1
+ %b13 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 13), align 1
+ %b14 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 14), align 1
+ %b15 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 15), align 1
+ %b16 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 16), align 1
+ %b17 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 17), align 1
+ %b18 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 18), align 1
+ %b19 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 19), align 1
+ %b20 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 20), align 1
+ %b21 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 21), align 1
+ %b22 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 22), align 1
+ %b23 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 23), align 1
+ %b24 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 24), align 1
+ %b25 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 25), align 1
+ %b26 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 26), align 1
+ %b27 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 27), align 1
+ %b28 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 28), align 1
+ %b29 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 29), align 1
+ %b30 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 30), align 1
+ %b31 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 31), align 1
+ %b32 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 32), align 1
+ %b33 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 33), align 1
+ %b34 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 34), align 1
+ %b35 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 35), align 1
+ %b36 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 36), align 1
+ %b37 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 37), align 1
+ %b38 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 38), align 1
+ %b39 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 39), align 1
+ %b40 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 40), align 1
+ %b41 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 41), align 1
+ %b42 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 42), align 1
+ %b43 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 43), align 1
+ %b44 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 44), align 1
+ %b45 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 45), align 1
+ %b46 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 46), align 1
+ %b47 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 47), align 1
+ %b48 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 48), align 1
+ %b49 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 49), align 1
+ %b50 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 50), align 1
+ %b51 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 51), align 1
+ %b52 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 52), align 1
+ %b53 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 53), align 1
+ %b54 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 54), align 1
+ %b55 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 55), align 1
+ %b56 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 56), align 1
+ %b57 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 57), align 1
+ %b58 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 58), align 1
+ %b59 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 59), align 1
+ %b60 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 60), align 1
+ %b61 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 61), align 1
+ %b62 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 62), align 1
+ %b63 = load i8, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @b8, i32 0, i64 63), align 1
+ %r0 = sub i8 %a0 , %b0
+ %r1 = sub i8 %a1 , %b1
+ %r2 = sub i8 %a2 , %b2
+ %r3 = sub i8 %a3 , %b3
+ %r4 = sub i8 %a4 , %b4
+ %r5 = sub i8 %a5 , %b5
+ %r6 = sub i8 %a6 , %b6
+ %r7 = sub i8 %a7 , %b7
+ %r8 = sub i8 %a8 , %b8
+ %r9 = sub i8 %a9 , %b9
+ %r10 = sub i8 %a10, %b10
+ %r11 = sub i8 %a11, %b11
+ %r12 = sub i8 %a12, %b12
+ %r13 = sub i8 %a13, %b13
+ %r14 = sub i8 %a14, %b14
+ %r15 = sub i8 %a15, %b15
+ %r16 = sub i8 %a16, %b16
+ %r17 = sub i8 %a17, %b17
+ %r18 = sub i8 %a18, %b18
+ %r19 = sub i8 %a19, %b19
+ %r20 = sub i8 %a20, %b20
+ %r21 = sub i8 %a21, %b21
+ %r22 = sub i8 %a22, %b22
+ %r23 = sub i8 %a23, %b23
+ %r24 = sub i8 %a24, %b24
+ %r25 = sub i8 %a25, %b25
+ %r26 = sub i8 %a26, %b26
+ %r27 = sub i8 %a27, %b27
+ %r28 = sub i8 %a28, %b28
+ %r29 = sub i8 %a29, %b29
+ %r30 = sub i8 %a30, %b30
+ %r31 = sub i8 %a31, %b31
+ %r32 = sub i8 %a32, %b32
+ %r33 = sub i8 %a33, %b33
+ %r34 = sub i8 %a34, %b34
+ %r35 = sub i8 %a35, %b35
+ %r36 = sub i8 %a36, %b36
+ %r37 = sub i8 %a37, %b37
+ %r38 = sub i8 %a38, %b38
+ %r39 = sub i8 %a39, %b39
+ %r40 = sub i8 %a40, %b40
+ %r41 = sub i8 %a41, %b41
+ %r42 = sub i8 %a42, %b42
+ %r43 = sub i8 %a43, %b43
+ %r44 = sub i8 %a44, %b44
+ %r45 = sub i8 %a45, %b45
+ %r46 = sub i8 %a46, %b46
+ %r47 = sub i8 %a47, %b47
+ %r48 = sub i8 %a48, %b48
+ %r49 = sub i8 %a49, %b49
+ %r50 = sub i8 %a50, %b50
+ %r51 = sub i8 %a51, %b51
+ %r52 = sub i8 %a52, %b52
+ %r53 = sub i8 %a53, %b53
+ %r54 = sub i8 %a54, %b54
+ %r55 = sub i8 %a55, %b55
+ %r56 = sub i8 %a56, %b56
+ %r57 = sub i8 %a57, %b57
+ %r58 = sub i8 %a58, %b58
+ %r59 = sub i8 %a59, %b59
+ %r60 = sub i8 %a60, %b60
+ %r61 = sub i8 %a61, %b61
+ %r62 = sub i8 %a62, %b62
+ %r63 = sub i8 %a63, %b63
+ store i8 %r0 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 0 ), align 1
+ store i8 %r1 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 1 ), align 1
+ store i8 %r2 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 2 ), align 1
+ store i8 %r3 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 3 ), align 1
+ store i8 %r4 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 4 ), align 1
+ store i8 %r5 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 5 ), align 1
+ store i8 %r6 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 6 ), align 1
+ store i8 %r7 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 7 ), align 1
+ store i8 %r8 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 8 ), align 1
+ store i8 %r9 , i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 9 ), align 1
+ store i8 %r10, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 10), align 1
+ store i8 %r11, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 11), align 1
+ store i8 %r12, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 12), align 1
+ store i8 %r13, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 13), align 1
+ store i8 %r14, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 14), align 1
+ store i8 %r15, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 15), align 1
+ store i8 %r16, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 16), align 1
+ store i8 %r17, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 17), align 1
+ store i8 %r18, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 18), align 1
+ store i8 %r19, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 19), align 1
+ store i8 %r20, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 20), align 1
+ store i8 %r21, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 21), align 1
+ store i8 %r22, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 22), align 1
+ store i8 %r23, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 23), align 1
+ store i8 %r24, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 24), align 1
+ store i8 %r25, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 25), align 1
+ store i8 %r26, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 26), align 1
+ store i8 %r27, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 27), align 1
+ store i8 %r28, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 28), align 1
+ store i8 %r29, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 29), align 1
+ store i8 %r30, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 30), align 1
+ store i8 %r31, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 31), align 1
+ store i8 %r32, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 32), align 1
+ store i8 %r33, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 33), align 1
+ store i8 %r34, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 34), align 1
+ store i8 %r35, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 35), align 1
+ store i8 %r36, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 36), align 1
+ store i8 %r37, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 37), align 1
+ store i8 %r38, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 38), align 1
+ store i8 %r39, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 39), align 1
+ store i8 %r40, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 40), align 1
+ store i8 %r41, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 41), align 1
+ store i8 %r42, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 42), align 1
+ store i8 %r43, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 43), align 1
+ store i8 %r44, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 44), align 1
+ store i8 %r45, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 45), align 1
+ store i8 %r46, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 46), align 1
+ store i8 %r47, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 47), align 1
+ store i8 %r48, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 48), align 1
+ store i8 %r49, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 49), align 1
+ store i8 %r50, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 50), align 1
+ store i8 %r51, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 51), align 1
+ store i8 %r52, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 52), align 1
+ store i8 %r53, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 53), align 1
+ store i8 %r54, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 54), align 1
+ store i8 %r55, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 55), align 1
+ store i8 %r56, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 56), align 1
+ store i8 %r57, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 57), align 1
+ store i8 %r58, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 58), align 1
+ store i8 %r59, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 59), align 1
+ store i8 %r60, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 60), align 1
+ store i8 %r61, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 61), align 1
+ store i8 %r62, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 62), align 1
+ store i8 %r63, i8* getelementptr inbounds ([64 x i8], [64 x i8]* @c8, i32 0, i64 63), align 1
+ ret void
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/atomics.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/atomics.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/atomics.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/atomics.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,35 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -S |FileCheck %s
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+
+ at x = global [4 x i32] zeroinitializer, align 16
+ at a = global [4 x i32] zeroinitializer, align 16
+
+; The SLPVectorizer should not vectorize atomic stores and it should not
+; schedule regular stores around atomic stores.
+
+define void @test() {
+; CHECK-LABEL: @test(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: store i32 0, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 0), align 16
+; CHECK-NEXT: store atomic i32 0, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @x, i64 0, i64 0) release, align 16
+; CHECK-NEXT: store i32 0, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 1), align 4
+; CHECK-NEXT: store atomic i32 1, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @x, i64 0, i64 1) release, align 4
+; CHECK-NEXT: store i32 0, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 2), align 8
+; CHECK-NEXT: store atomic i32 2, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @x, i64 0, i64 2) release, align 8
+; CHECK-NEXT: store i32 0, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 3), align 4
+; CHECK-NEXT: store atomic i32 3, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @x, i64 0, i64 3) release, align 4
+; CHECK-NEXT: ret void
+;
+entry:
+ store i32 0, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 0), align 16
+ store atomic i32 0, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @x, i64 0, i64 0) release, align 16
+ store i32 0, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 1), align 4
+ store atomic i32 1, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @x, i64 0, i64 1) release, align 4
+ store i32 0, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 2), align 8
+ store atomic i32 2, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @x, i64 0, i64 2) release, align 8
+ store i32 0, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 3), align 4
+ store atomic i32 3, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @x, i64 0, i64 3) release, align 4
+ ret void
+}
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/bad_types.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/bad_types.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/bad_types.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/bad_types.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,112 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -S -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @test1(x86_mmx %a, x86_mmx %b, i64* %ptr) {
+; Ensure we can handle x86_mmx values which are primitive and can be bitcast
+; with integer types but can't be put into a vector.
+;
+; CHECK-LABEL: @test1(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[A_CAST:%.*]] = bitcast x86_mmx [[A:%.*]] to i64
+; CHECK-NEXT: [[B_CAST:%.*]] = bitcast x86_mmx [[B:%.*]] to i64
+; CHECK-NEXT: [[A_AND:%.*]] = and i64 [[A_CAST]], 42
+; CHECK-NEXT: [[B_AND:%.*]] = and i64 [[B_CAST]], 42
+; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, i64* [[PTR:%.*]], i32 1
+; CHECK-NEXT: store i64 [[A_AND]], i64* [[PTR]]
+; CHECK-NEXT: store i64 [[B_AND]], i64* [[GEP]]
+; CHECK-NEXT: ret void
+;
+entry:
+ %a.cast = bitcast x86_mmx %a to i64
+ %b.cast = bitcast x86_mmx %b to i64
+ %a.and = and i64 %a.cast, 42
+ %b.and = and i64 %b.cast, 42
+ %gep = getelementptr i64, i64* %ptr, i32 1
+ store i64 %a.and, i64* %ptr
+ store i64 %b.and, i64* %gep
+ ret void
+}
+
+define void @test2(x86_mmx %a, x86_mmx %b) {
+; Same as @test1 but using phi-input vectorization instead of store
+; vectorization.
+;
+; CHECK-LABEL: @test2(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN:%.*]], label [[EXIT:%.*]]
+; CHECK: if.then:
+; CHECK-NEXT: [[A_CAST:%.*]] = bitcast x86_mmx [[A:%.*]] to i64
+; CHECK-NEXT: [[B_CAST:%.*]] = bitcast x86_mmx [[B:%.*]] to i64
+; CHECK-NEXT: [[A_AND:%.*]] = and i64 [[A_CAST]], 42
+; CHECK-NEXT: [[B_AND:%.*]] = and i64 [[B_CAST]], 42
+; CHECK-NEXT: br label [[EXIT]]
+; CHECK: exit:
+; CHECK-NEXT: [[A_PHI:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[A_AND]], [[IF_THEN]] ]
+; CHECK-NEXT: [[B_PHI:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[B_AND]], [[IF_THEN]] ]
+; CHECK-NEXT: tail call void @f(i64 [[A_PHI]], i64 [[B_PHI]])
+; CHECK-NEXT: ret void
+;
+entry:
+ br i1 undef, label %if.then, label %exit
+
+if.then:
+ %a.cast = bitcast x86_mmx %a to i64
+ %b.cast = bitcast x86_mmx %b to i64
+ %a.and = and i64 %a.cast, 42
+ %b.and = and i64 %b.cast, 42
+ br label %exit
+
+exit:
+ %a.phi = phi i64 [ 0, %entry ], [ %a.and, %if.then ]
+ %b.phi = phi i64 [ 0, %entry ], [ %b.and, %if.then ]
+ tail call void @f(i64 %a.phi, i64 %b.phi)
+ ret void
+}
+
+define i8 @test3(i8 *%addr) {
+; Check that we do not vectorize types that are padded to a bigger ones.
+;
+; CHECK-LABEL: @test3(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[A:%.*]] = bitcast i8* [[ADDR:%.*]] to i2*
+; CHECK-NEXT: [[A0:%.*]] = getelementptr inbounds i2, i2* [[A]], i64 0
+; CHECK-NEXT: [[A1:%.*]] = getelementptr inbounds i2, i2* [[A]], i64 1
+; CHECK-NEXT: [[A2:%.*]] = getelementptr inbounds i2, i2* [[A]], i64 2
+; CHECK-NEXT: [[A3:%.*]] = getelementptr inbounds i2, i2* [[A]], i64 3
+; CHECK-NEXT: [[L0:%.*]] = load i2, i2* [[A0]], align 1
+; CHECK-NEXT: [[L1:%.*]] = load i2, i2* [[A1]], align 1
+; CHECK-NEXT: [[L2:%.*]] = load i2, i2* [[A2]], align 1
+; CHECK-NEXT: [[L3:%.*]] = load i2, i2* [[A3]], align 1
+; CHECK-NEXT: br label [[BB1:%.*]]
+; CHECK: bb1:
+; CHECK-NEXT: [[P0:%.*]] = phi i2 [ [[L0]], [[ENTRY:%.*]] ]
+; CHECK-NEXT: [[P1:%.*]] = phi i2 [ [[L1]], [[ENTRY]] ]
+; CHECK-NEXT: [[P2:%.*]] = phi i2 [ [[L2]], [[ENTRY]] ]
+; CHECK-NEXT: [[P3:%.*]] = phi i2 [ [[L3]], [[ENTRY]] ]
+; CHECK-NEXT: [[R:%.*]] = zext i2 [[P2]] to i8
+; CHECK-NEXT: ret i8 [[R]]
+;
+entry:
+ %a = bitcast i8* %addr to i2*
+ %a0 = getelementptr inbounds i2, i2* %a, i64 0
+ %a1 = getelementptr inbounds i2, i2* %a, i64 1
+ %a2 = getelementptr inbounds i2, i2* %a, i64 2
+ %a3 = getelementptr inbounds i2, i2* %a, i64 3
+ %l0 = load i2, i2* %a0, align 1
+ %l1 = load i2, i2* %a1, align 1
+ %l2 = load i2, i2* %a2, align 1
+ %l3 = load i2, i2* %a3, align 1
+ br label %bb1
+bb1: ; preds = %entry
+ %p0 = phi i2 [ %l0, %entry ]
+ %p1 = phi i2 [ %l1, %entry ]
+ %p2 = phi i2 [ %l2, %entry ]
+ %p3 = phi i2 [ %l3, %entry ]
+ %r = zext i2 %p2 to i8
+ ret i8 %r
+}
+
+declare void @f(i64, i64)
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/barriercall.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/barriercall.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/barriercall.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/barriercall.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,45 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+define i32 @foo(i32* nocapture %A, i32 %n) {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[CALL:%.*]] = tail call i32 (...) @bar()
+; CHECK-NEXT: [[TMP0:%.*]] = insertelement <4 x i32> undef, i32 [[N:%.*]], i32 0
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x i32> [[TMP0]], i32 [[N]], i32 1
+; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> [[TMP1]], i32 [[N]], i32 2
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[N]], i32 3
+; CHECK-NEXT: [[TMP4:%.*]] = mul nsw <4 x i32> [[TMP3]], <i32 5, i32 9, i32 3, i32 10>
+; CHECK-NEXT: [[TMP5:%.*]] = shl <4 x i32> [[TMP3]], <i32 5, i32 9, i32 3, i32 10>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <4 x i32> [[TMP4]], <4 x i32> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 6, i32 3>
+; CHECK-NEXT: [[TMP7:%.*]] = add nsw <4 x i32> [[TMP6]], <i32 9, i32 9, i32 9, i32 9>
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast i32* [[A:%.*]] to <4 x i32>*
+; CHECK-NEXT: store <4 x i32> [[TMP7]], <4 x i32>* [[TMP8]], align 4
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %call = tail call i32 (...) @bar() #2
+ %mul = mul nsw i32 %n, 5
+ %add = add nsw i32 %mul, 9
+ store i32 %add, i32* %A, align 4
+ %mul1 = mul nsw i32 %n, 9
+ %add2 = add nsw i32 %mul1, 9
+ %arrayidx3 = getelementptr inbounds i32, i32* %A, i64 1
+ store i32 %add2, i32* %arrayidx3, align 4
+ %mul4 = shl i32 %n, 3
+ %add5 = add nsw i32 %mul4, 9
+ %arrayidx6 = getelementptr inbounds i32, i32* %A, i64 2
+ store i32 %add5, i32* %arrayidx6, align 4
+ %mul7 = mul nsw i32 %n, 10
+ %add8 = add nsw i32 %mul7, 9
+ %arrayidx9 = getelementptr inbounds i32, i32* %A, i64 3
+ store i32 %add8, i32* %arrayidx9, align 4
+ ret i32 undef
+}
+
+ ; We can still vectorize the stores below.
+
+declare i32 @bar(...)
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/bitreverse.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/bitreverse.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/bitreverse.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/bitreverse.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,423 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=bdver2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=XOP
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=bdver4 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=XOP
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+ at src64 = common global [4 x i64] zeroinitializer, align 32
+ at dst64 = common global [4 x i64] zeroinitializer, align 32
+ at src32 = common global [8 x i32] zeroinitializer, align 32
+ at dst32 = common global [8 x i32] zeroinitializer, align 32
+ at src16 = common global [16 x i16] zeroinitializer, align 32
+ at dst16 = common global [16 x i16] zeroinitializer, align 32
+ at src8 = common global [32 x i8] zeroinitializer, align 32
+ at dst8 = common global [32 x i8] zeroinitializer, align 32
+
+declare i64 @llvm.bitreverse.i64(i64)
+declare i32 @llvm.bitreverse.i32(i32)
+declare i16 @llvm.bitreverse.i16(i16)
+declare i8 @llvm.bitreverse.i8(i8)
+
+define void @bitreverse_2i64() #0 {
+; CHECK-LABEL: @bitreverse_2i64(
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([4 x i64]* @src64 to <2 x i64>*), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> [[TMP1]])
+; CHECK-NEXT: store <2 x i64> [[TMP2]], <2 x i64>* bitcast ([4 x i64]* @dst64 to <2 x i64>*), align 8
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+ %bitreverse0 = call i64 @llvm.bitreverse.i64(i64 %ld0)
+ %bitreverse1 = call i64 @llvm.bitreverse.i64(i64 %ld1)
+ store i64 %bitreverse0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+ store i64 %bitreverse1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+ ret void
+}
+
+define void @bitreverse_4i64() #0 {
+; SSE-LABEL: @bitreverse_4i64(
+; SSE-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([4 x i64]* @src64 to <2 x i64>*), align 4
+; SSE-NEXT: [[TMP2:%.*]] = load <2 x i64>, <2 x i64>* bitcast (i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2) to <2 x i64>*), align 4
+; SSE-NEXT: [[TMP3:%.*]] = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> [[TMP1]])
+; SSE-NEXT: [[TMP4:%.*]] = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> [[TMP2]])
+; SSE-NEXT: store <2 x i64> [[TMP3]], <2 x i64>* bitcast ([4 x i64]* @dst64 to <2 x i64>*), align 4
+; SSE-NEXT: store <2 x i64> [[TMP4]], <2 x i64>* bitcast (i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2) to <2 x i64>*), align 4
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @bitreverse_4i64(
+; AVX-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([4 x i64]* @src64 to <4 x i64>*), align 4
+; AVX-NEXT: [[TMP2:%.*]] = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> [[TMP1]])
+; AVX-NEXT: store <4 x i64> [[TMP2]], <4 x i64>* bitcast ([4 x i64]* @dst64 to <4 x i64>*), align 4
+; AVX-NEXT: ret void
+;
+; XOP-LABEL: @bitreverse_4i64(
+; XOP-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([4 x i64]* @src64 to <4 x i64>*), align 4
+; XOP-NEXT: [[TMP2:%.*]] = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> [[TMP1]])
+; XOP-NEXT: store <4 x i64> [[TMP2]], <4 x i64>* bitcast ([4 x i64]* @dst64 to <4 x i64>*), align 4
+; XOP-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+ %ld2 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+ %ld3 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+ %bitreverse0 = call i64 @llvm.bitreverse.i64(i64 %ld0)
+ %bitreverse1 = call i64 @llvm.bitreverse.i64(i64 %ld1)
+ %bitreverse2 = call i64 @llvm.bitreverse.i64(i64 %ld2)
+ %bitreverse3 = call i64 @llvm.bitreverse.i64(i64 %ld3)
+ store i64 %bitreverse0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+ store i64 %bitreverse1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+ store i64 %bitreverse2, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+ store i64 %bitreverse3, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+ ret void
+}
+
+define void @bitreverse_4i32() #0 {
+; CHECK-LABEL: @bitreverse_4i32(
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([8 x i32]* @src32 to <4 x i32>*), align 4
+; CHECK-NEXT: [[TMP2:%.*]] = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> [[TMP1]])
+; CHECK-NEXT: store <4 x i32> [[TMP2]], <4 x i32>* bitcast ([8 x i32]* @dst32 to <4 x i32>*), align 4
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+ %bitreverse0 = call i32 @llvm.bitreverse.i32(i32 %ld0)
+ %bitreverse1 = call i32 @llvm.bitreverse.i32(i32 %ld1)
+ %bitreverse2 = call i32 @llvm.bitreverse.i32(i32 %ld2)
+ %bitreverse3 = call i32 @llvm.bitreverse.i32(i32 %ld3)
+ store i32 %bitreverse0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+ store i32 %bitreverse1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+ store i32 %bitreverse2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+ store i32 %bitreverse3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+ ret void
+}
+
+define void @bitreverse_8i32() #0 {
+; SSE-LABEL: @bitreverse_8i32(
+; SSE-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([8 x i32]* @src32 to <4 x i32>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4) to <4 x i32>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> [[TMP1]])
+; SSE-NEXT: [[TMP4:%.*]] = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> [[TMP2]])
+; SSE-NEXT: store <4 x i32> [[TMP3]], <4 x i32>* bitcast ([8 x i32]* @dst32 to <4 x i32>*), align 2
+; SSE-NEXT: store <4 x i32> [[TMP4]], <4 x i32>* bitcast (i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4) to <4 x i32>*), align 2
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @bitreverse_8i32(
+; AVX-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> [[TMP1]])
+; AVX-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2
+; AVX-NEXT: ret void
+;
+; XOP-LABEL: @bitreverse_8i32(
+; XOP-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2
+; XOP-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> [[TMP1]])
+; XOP-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2
+; XOP-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+ %ld4 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+ %ld5 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+ %ld6 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+ %ld7 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+ %bitreverse0 = call i32 @llvm.bitreverse.i32(i32 %ld0)
+ %bitreverse1 = call i32 @llvm.bitreverse.i32(i32 %ld1)
+ %bitreverse2 = call i32 @llvm.bitreverse.i32(i32 %ld2)
+ %bitreverse3 = call i32 @llvm.bitreverse.i32(i32 %ld3)
+ %bitreverse4 = call i32 @llvm.bitreverse.i32(i32 %ld4)
+ %bitreverse5 = call i32 @llvm.bitreverse.i32(i32 %ld5)
+ %bitreverse6 = call i32 @llvm.bitreverse.i32(i32 %ld6)
+ %bitreverse7 = call i32 @llvm.bitreverse.i32(i32 %ld7)
+ store i32 %bitreverse0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+ store i32 %bitreverse1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+ store i32 %bitreverse2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+ store i32 %bitreverse3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+ store i32 %bitreverse4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+ store i32 %bitreverse5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+ store i32 %bitreverse6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+ store i32 %bitreverse7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+ ret void
+}
+
+define void @bitreverse_8i16() #0 {
+; CHECK-LABEL: @bitreverse_8i16(
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> [[TMP1]])
+; CHECK-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %bitreverse0 = call i16 @llvm.bitreverse.i16(i16 %ld0)
+ %bitreverse1 = call i16 @llvm.bitreverse.i16(i16 %ld1)
+ %bitreverse2 = call i16 @llvm.bitreverse.i16(i16 %ld2)
+ %bitreverse3 = call i16 @llvm.bitreverse.i16(i16 %ld3)
+ %bitreverse4 = call i16 @llvm.bitreverse.i16(i16 %ld4)
+ %bitreverse5 = call i16 @llvm.bitreverse.i16(i16 %ld5)
+ %bitreverse6 = call i16 @llvm.bitreverse.i16(i16 %ld6)
+ %bitreverse7 = call i16 @llvm.bitreverse.i16(i16 %ld7)
+ store i16 %bitreverse0, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %bitreverse1, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %bitreverse2, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %bitreverse3, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %bitreverse4, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %bitreverse5, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %bitreverse6, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %bitreverse7, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ ret void
+}
+
+define void @bitreverse_16i16() #0 {
+; SSE-LABEL: @bitreverse_16i16(
+; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> [[TMP1]])
+; SSE-NEXT: [[TMP4:%.*]] = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> [[TMP2]])
+; SSE-NEXT: store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @bitreverse_16i16(
+; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> [[TMP1]])
+; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2
+; AVX-NEXT: ret void
+;
+; XOP-LABEL: @bitreverse_16i16(
+; XOP-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2
+; XOP-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> [[TMP1]])
+; XOP-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2
+; XOP-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %ld8 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2
+ %ld9 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2
+ %ld10 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2
+ %ld11 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2
+ %ld12 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2
+ %ld13 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2
+ %ld14 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2
+ %ld15 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2
+ %bitreverse0 = call i16 @llvm.bitreverse.i16(i16 %ld0)
+ %bitreverse1 = call i16 @llvm.bitreverse.i16(i16 %ld1)
+ %bitreverse2 = call i16 @llvm.bitreverse.i16(i16 %ld2)
+ %bitreverse3 = call i16 @llvm.bitreverse.i16(i16 %ld3)
+ %bitreverse4 = call i16 @llvm.bitreverse.i16(i16 %ld4)
+ %bitreverse5 = call i16 @llvm.bitreverse.i16(i16 %ld5)
+ %bitreverse6 = call i16 @llvm.bitreverse.i16(i16 %ld6)
+ %bitreverse7 = call i16 @llvm.bitreverse.i16(i16 %ld7)
+ %bitreverse8 = call i16 @llvm.bitreverse.i16(i16 %ld8)
+ %bitreverse9 = call i16 @llvm.bitreverse.i16(i16 %ld9)
+ %bitreverse10 = call i16 @llvm.bitreverse.i16(i16 %ld10)
+ %bitreverse11 = call i16 @llvm.bitreverse.i16(i16 %ld11)
+ %bitreverse12 = call i16 @llvm.bitreverse.i16(i16 %ld12)
+ %bitreverse13 = call i16 @llvm.bitreverse.i16(i16 %ld13)
+ %bitreverse14 = call i16 @llvm.bitreverse.i16(i16 %ld14)
+ %bitreverse15 = call i16 @llvm.bitreverse.i16(i16 %ld15)
+ store i16 %bitreverse0 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %bitreverse1 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %bitreverse2 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %bitreverse3 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %bitreverse4 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %bitreverse5 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %bitreverse6 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %bitreverse7 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ store i16 %bitreverse8 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2
+ store i16 %bitreverse9 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2
+ store i16 %bitreverse10, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2
+ store i16 %bitreverse11, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2
+ store i16 %bitreverse12, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2
+ store i16 %bitreverse13, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2
+ store i16 %bitreverse14, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2
+ store i16 %bitreverse15, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2
+ ret void
+}
+
+define void @bitreverse_16i8() #0 {
+; CHECK-LABEL: @bitreverse_16i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP1]])
+; CHECK-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %bitreverse0 = call i8 @llvm.bitreverse.i8(i8 %ld0)
+ %bitreverse1 = call i8 @llvm.bitreverse.i8(i8 %ld1)
+ %bitreverse2 = call i8 @llvm.bitreverse.i8(i8 %ld2)
+ %bitreverse3 = call i8 @llvm.bitreverse.i8(i8 %ld3)
+ %bitreverse4 = call i8 @llvm.bitreverse.i8(i8 %ld4)
+ %bitreverse5 = call i8 @llvm.bitreverse.i8(i8 %ld5)
+ %bitreverse6 = call i8 @llvm.bitreverse.i8(i8 %ld6)
+ %bitreverse7 = call i8 @llvm.bitreverse.i8(i8 %ld7)
+ %bitreverse8 = call i8 @llvm.bitreverse.i8(i8 %ld8)
+ %bitreverse9 = call i8 @llvm.bitreverse.i8(i8 %ld9)
+ %bitreverse10 = call i8 @llvm.bitreverse.i8(i8 %ld10)
+ %bitreverse11 = call i8 @llvm.bitreverse.i8(i8 %ld11)
+ %bitreverse12 = call i8 @llvm.bitreverse.i8(i8 %ld12)
+ %bitreverse13 = call i8 @llvm.bitreverse.i8(i8 %ld13)
+ %bitreverse14 = call i8 @llvm.bitreverse.i8(i8 %ld14)
+ %bitreverse15 = call i8 @llvm.bitreverse.i8(i8 %ld15)
+ store i8 %bitreverse0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %bitreverse1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %bitreverse2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %bitreverse3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %bitreverse4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %bitreverse5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %bitreverse6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %bitreverse7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %bitreverse8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %bitreverse9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %bitreverse10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %bitreverse11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %bitreverse12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %bitreverse13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %bitreverse14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %bitreverse15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ ret void
+}
+
+define void @bitreverse_32i8() #0 {
+; CHECK-LABEL: @bitreverse_32i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP1]])
+; CHECK-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP2]])
+; CHECK-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %ld16 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1
+ %ld17 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1
+ %ld18 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1
+ %ld19 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1
+ %ld20 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1
+ %ld21 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1
+ %ld22 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1
+ %ld23 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1
+ %ld24 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1
+ %ld25 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1
+ %ld26 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1
+ %ld27 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1
+ %ld28 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1
+ %ld29 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1
+ %ld30 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1
+ %ld31 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1
+ %bitreverse0 = call i8 @llvm.bitreverse.i8(i8 %ld0)
+ %bitreverse1 = call i8 @llvm.bitreverse.i8(i8 %ld1)
+ %bitreverse2 = call i8 @llvm.bitreverse.i8(i8 %ld2)
+ %bitreverse3 = call i8 @llvm.bitreverse.i8(i8 %ld3)
+ %bitreverse4 = call i8 @llvm.bitreverse.i8(i8 %ld4)
+ %bitreverse5 = call i8 @llvm.bitreverse.i8(i8 %ld5)
+ %bitreverse6 = call i8 @llvm.bitreverse.i8(i8 %ld6)
+ %bitreverse7 = call i8 @llvm.bitreverse.i8(i8 %ld7)
+ %bitreverse8 = call i8 @llvm.bitreverse.i8(i8 %ld8)
+ %bitreverse9 = call i8 @llvm.bitreverse.i8(i8 %ld9)
+ %bitreverse10 = call i8 @llvm.bitreverse.i8(i8 %ld10)
+ %bitreverse11 = call i8 @llvm.bitreverse.i8(i8 %ld11)
+ %bitreverse12 = call i8 @llvm.bitreverse.i8(i8 %ld12)
+ %bitreverse13 = call i8 @llvm.bitreverse.i8(i8 %ld13)
+ %bitreverse14 = call i8 @llvm.bitreverse.i8(i8 %ld14)
+ %bitreverse15 = call i8 @llvm.bitreverse.i8(i8 %ld15)
+ %bitreverse16 = call i8 @llvm.bitreverse.i8(i8 %ld16)
+ %bitreverse17 = call i8 @llvm.bitreverse.i8(i8 %ld17)
+ %bitreverse18 = call i8 @llvm.bitreverse.i8(i8 %ld18)
+ %bitreverse19 = call i8 @llvm.bitreverse.i8(i8 %ld19)
+ %bitreverse20 = call i8 @llvm.bitreverse.i8(i8 %ld20)
+ %bitreverse21 = call i8 @llvm.bitreverse.i8(i8 %ld21)
+ %bitreverse22 = call i8 @llvm.bitreverse.i8(i8 %ld22)
+ %bitreverse23 = call i8 @llvm.bitreverse.i8(i8 %ld23)
+ %bitreverse24 = call i8 @llvm.bitreverse.i8(i8 %ld24)
+ %bitreverse25 = call i8 @llvm.bitreverse.i8(i8 %ld25)
+ %bitreverse26 = call i8 @llvm.bitreverse.i8(i8 %ld26)
+ %bitreverse27 = call i8 @llvm.bitreverse.i8(i8 %ld27)
+ %bitreverse28 = call i8 @llvm.bitreverse.i8(i8 %ld28)
+ %bitreverse29 = call i8 @llvm.bitreverse.i8(i8 %ld29)
+ %bitreverse30 = call i8 @llvm.bitreverse.i8(i8 %ld30)
+ %bitreverse31 = call i8 @llvm.bitreverse.i8(i8 %ld31)
+ store i8 %bitreverse0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %bitreverse1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %bitreverse2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %bitreverse3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %bitreverse4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %bitreverse5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %bitreverse6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %bitreverse7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %bitreverse8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %bitreverse9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %bitreverse10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %bitreverse11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %bitreverse12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %bitreverse13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %bitreverse14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %bitreverse15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ store i8 %bitreverse16, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1
+ store i8 %bitreverse17, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1
+ store i8 %bitreverse18, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1
+ store i8 %bitreverse19, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1
+ store i8 %bitreverse20, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1
+ store i8 %bitreverse21, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1
+ store i8 %bitreverse22, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1
+ store i8 %bitreverse23, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1
+ store i8 %bitreverse24, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1
+ store i8 %bitreverse25, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1
+ store i8 %bitreverse26, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1
+ store i8 %bitreverse27, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1
+ store i8 %bitreverse28, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1
+ store i8 %bitreverse29, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1
+ store i8 %bitreverse30, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1
+ store i8 %bitreverse31, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1
+ ret void
+}
+
+attributes #0 = { nounwind }
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/blending-shuffle.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/blending-shuffle.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/blending-shuffle.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/blending-shuffle.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,179 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -slp-vectorizer -S -o - -mtriple=x86_64-unknown-linux -mcpu=bdver2 -instcombine | FileCheck %s
+
+define <2 x i8> @g(<2 x i8> %x, <2 x i8> %y) {
+; CHECK-LABEL: @g(
+; CHECK-NEXT: [[X0:%.*]] = extractelement <2 x i8> [[X:%.*]], i32 0
+; CHECK-NEXT: [[Y1:%.*]] = extractelement <2 x i8> [[Y:%.*]], i32 1
+; CHECK-NEXT: [[X0X0:%.*]] = mul i8 [[X0]], [[X0]]
+; CHECK-NEXT: [[Y1Y1:%.*]] = mul i8 [[Y1]], [[Y1]]
+; CHECK-NEXT: [[INS1:%.*]] = insertelement <2 x i8> undef, i8 [[X0X0]], i32 0
+; CHECK-NEXT: [[INS2:%.*]] = insertelement <2 x i8> [[INS1]], i8 [[Y1Y1]], i32 1
+; CHECK-NEXT: ret <2 x i8> [[INS2]]
+;
+ %x0 = extractelement <2 x i8> %x, i32 0
+ %y1 = extractelement <2 x i8> %y, i32 1
+ %x0x0 = mul i8 %x0, %x0
+ %y1y1 = mul i8 %y1, %y1
+ %ins1 = insertelement <2 x i8> undef, i8 %x0x0, i32 0
+ %ins2 = insertelement <2 x i8> %ins1, i8 %y1y1, i32 1
+ ret <2 x i8> %ins2
+}
+
+define <4 x i8> @h(<4 x i8> %x, <4 x i8> %y) {
+; CHECK-LABEL: @h(
+; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x i8> [[X:%.*]], <4 x i8> [[Y:%.*]], <4 x i32> <i32 0, i32 3, i32 5, i32 6>
+; CHECK-NEXT: [[TMP2:%.*]] = mul <4 x i8> [[TMP1]], [[TMP1]]
+; CHECK-NEXT: ret <4 x i8> [[TMP2]]
+;
+ %x0 = extractelement <4 x i8> %x, i32 0
+ %x3 = extractelement <4 x i8> %x, i32 3
+ %y1 = extractelement <4 x i8> %y, i32 1
+ %y2 = extractelement <4 x i8> %y, i32 2
+ %x0x0 = mul i8 %x0, %x0
+ %x3x3 = mul i8 %x3, %x3
+ %y1y1 = mul i8 %y1, %y1
+ %y2y2 = mul i8 %y2, %y2
+ %ins1 = insertelement <4 x i8> undef, i8 %x0x0, i32 0
+ %ins2 = insertelement <4 x i8> %ins1, i8 %x3x3, i32 1
+ %ins3 = insertelement <4 x i8> %ins2, i8 %y1y1, i32 2
+ %ins4 = insertelement <4 x i8> %ins3, i8 %y2y2, i32 3
+ ret <4 x i8> %ins4
+}
+
+define <4 x i8> @h_undef(<4 x i8> %x, <4 x i8> %y) {
+; CHECK-LABEL: @h_undef(
+; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x i8> [[X:%.*]], <4 x i8> [[Y:%.*]], <4 x i32> <i32 undef, i32 3, i32 5, i32 6>
+; CHECK-NEXT: [[TMP2:%.*]] = mul <4 x i8> [[TMP1]], [[TMP1]]
+; CHECK-NEXT: ret <4 x i8> [[TMP2]]
+;
+ %x0 = extractelement <4 x i8> undef, i32 0
+ %x3 = extractelement <4 x i8> %x, i32 3
+ %y1 = extractelement <4 x i8> %y, i32 1
+ %y2 = extractelement <4 x i8> %y, i32 2
+ %x0x0 = mul i8 %x0, %x0
+ %x3x3 = mul i8 %x3, %x3
+ %y1y1 = mul i8 %y1, %y1
+ %y2y2 = mul i8 %y2, %y2
+ %ins1 = insertelement <4 x i8> undef, i8 %x0x0, i32 0
+ %ins2 = insertelement <4 x i8> %ins1, i8 %x3x3, i32 1
+ %ins3 = insertelement <4 x i8> %ins2, i8 %y1y1, i32 2
+ %ins4 = insertelement <4 x i8> %ins3, i8 %y2y2, i32 3
+ ret <4 x i8> %ins4
+}
+
+define i8 @i(<4 x i8> %x, <4 x i8> %y) {
+; CHECK-LABEL: @i(
+; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <4 x i8> [[X:%.*]], <4 x i8> [[Y:%.*]], <4 x i32> <i32 0, i32 3, i32 5, i32 6>
+; CHECK-NEXT: [[TMP2:%.*]] = mul <4 x i8> [[TMP1]], [[TMP1]]
+; CHECK-NEXT: [[RDX_SHUF:%.*]] = shufflevector <4 x i8> [[TMP2]], <4 x i8> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
+; CHECK-NEXT: [[BIN_RDX:%.*]] = add <4 x i8> [[TMP2]], [[RDX_SHUF]]
+; CHECK-NEXT: [[RDX_SHUF1:%.*]] = shufflevector <4 x i8> [[BIN_RDX]], <4 x i8> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT: [[BIN_RDX2:%.*]] = add <4 x i8> [[BIN_RDX]], [[RDX_SHUF1]]
+; CHECK-NEXT: [[TMP3:%.*]] = extractelement <4 x i8> [[BIN_RDX2]], i32 0
+; CHECK-NEXT: ret i8 [[TMP3]]
+;
+ %x0 = extractelement <4 x i8> %x, i32 0
+ %x3 = extractelement <4 x i8> %x, i32 3
+ %y1 = extractelement <4 x i8> %y, i32 1
+ %y2 = extractelement <4 x i8> %y, i32 2
+ %x0x0 = mul i8 %x0, %x0
+ %x3x3 = mul i8 %x3, %x3
+ %y1y1 = mul i8 %y1, %y1
+ %y2y2 = mul i8 %y2, %y2
+ %1 = add i8 %x0x0, %x3x3
+ %2 = add i8 %y1y1, %y2y2
+ %3 = add i8 %1, %2
+ ret i8 %3
+}
+
+define i8 @j(<4 x i8> %x, <4 x i8> %y) {
+; CHECK-LABEL: @j(
+; CHECK-NEXT: [[X0:%.*]] = extractelement <4 x i8> [[X:%.*]], i32 0
+; CHECK-NEXT: [[X3:%.*]] = extractelement <4 x i8> [[X]], i32 3
+; CHECK-NEXT: [[Y1:%.*]] = extractelement <4 x i8> [[Y:%.*]], i32 1
+; CHECK-NEXT: [[Y2:%.*]] = extractelement <4 x i8> [[Y]], i32 2
+; CHECK-NEXT: [[X0X0:%.*]] = mul i8 [[X0]], [[X0]]
+; CHECK-NEXT: [[X3X3:%.*]] = mul i8 [[X3]], [[X3]]
+; CHECK-NEXT: [[Y1Y1:%.*]] = mul i8 [[Y1]], [[Y1]]
+; CHECK-NEXT: [[Y2Y2:%.*]] = mul i8 [[Y2]], [[Y2]]
+; CHECK-NEXT: [[TMP1:%.*]] = add i8 [[X0X0]], [[X3X3]]
+; CHECK-NEXT: [[TMP2:%.*]] = add i8 [[Y1Y1]], [[Y2Y2]]
+; CHECK-NEXT: [[TMP3:%.*]] = sdiv i8 [[TMP1]], [[TMP2]]
+; CHECK-NEXT: ret i8 [[TMP3]]
+;
+ %x0 = extractelement <4 x i8> %x, i32 0
+ %x3 = extractelement <4 x i8> %x, i32 3
+ %y1 = extractelement <4 x i8> %y, i32 1
+ %y2 = extractelement <4 x i8> %y, i32 2
+ %x0x0 = mul i8 %x0, %x0
+ %x3x3 = mul i8 %x3, %x3
+ %y1y1 = mul i8 %y1, %y1
+ %y2y2 = mul i8 %y2, %y2
+ %1 = add i8 %x0x0, %x3x3
+ %2 = add i8 %y1y1, %y2y2
+ %3 = sdiv i8 %1, %2
+ ret i8 %3
+}
+
+define i8 @k(<4 x i8> %x) {
+; CHECK-LABEL: @k(
+; CHECK-NEXT: [[X0:%.*]] = extractelement <4 x i8> [[X:%.*]], i32 0
+; CHECK-NEXT: [[X3:%.*]] = extractelement <4 x i8> [[X]], i32 3
+; CHECK-NEXT: [[X1:%.*]] = extractelement <4 x i8> [[X]], i32 1
+; CHECK-NEXT: [[X2:%.*]] = extractelement <4 x i8> [[X]], i32 2
+; CHECK-NEXT: [[X0X0:%.*]] = mul i8 [[X0]], [[X0]]
+; CHECK-NEXT: [[X3X3:%.*]] = mul i8 [[X3]], [[X3]]
+; CHECK-NEXT: [[X1X1:%.*]] = mul i8 [[X1]], [[X1]]
+; CHECK-NEXT: [[X2X2:%.*]] = mul i8 [[X2]], [[X2]]
+; CHECK-NEXT: [[TMP1:%.*]] = add i8 [[X0X0]], [[X3X3]]
+; CHECK-NEXT: [[TMP2:%.*]] = add i8 [[X1X1]], [[X2X2]]
+; CHECK-NEXT: [[TMP3:%.*]] = sdiv i8 [[TMP1]], [[TMP2]]
+; CHECK-NEXT: ret i8 [[TMP3]]
+;
+ %x0 = extractelement <4 x i8> %x, i32 0
+ %x3 = extractelement <4 x i8> %x, i32 3
+ %x1 = extractelement <4 x i8> %x, i32 1
+ %x2 = extractelement <4 x i8> %x, i32 2
+ %x0x0 = mul i8 %x0, %x0
+ %x3x3 = mul i8 %x3, %x3
+ %x1x1 = mul i8 %x1, %x1
+ %x2x2 = mul i8 %x2, %x2
+ %1 = add i8 %x0x0, %x3x3
+ %2 = add i8 %x1x1, %x2x2
+ %3 = sdiv i8 %1, %2
+ ret i8 %3
+}
+
+define i8 @k_bb(<4 x i8> %x) {
+; CHECK-LABEL: @k_bb(
+; CHECK-NEXT: [[X0:%.*]] = extractelement <4 x i8> [[X:%.*]], i32 0
+; CHECK-NEXT: br label [[BB1:%.*]]
+; CHECK: bb1:
+; CHECK-NEXT: [[X3:%.*]] = extractelement <4 x i8> [[X]], i32 3
+; CHECK-NEXT: [[X1:%.*]] = extractelement <4 x i8> [[X]], i32 1
+; CHECK-NEXT: [[X2:%.*]] = extractelement <4 x i8> [[X]], i32 2
+; CHECK-NEXT: [[X0X0:%.*]] = mul i8 [[X0]], [[X0]]
+; CHECK-NEXT: [[X3X3:%.*]] = mul i8 [[X3]], [[X3]]
+; CHECK-NEXT: [[X1X1:%.*]] = mul i8 [[X1]], [[X1]]
+; CHECK-NEXT: [[X2X2:%.*]] = mul i8 [[X2]], [[X2]]
+; CHECK-NEXT: [[TMP1:%.*]] = add i8 [[X0X0]], [[X3X3]]
+; CHECK-NEXT: [[TMP2:%.*]] = add i8 [[X1X1]], [[X2X2]]
+; CHECK-NEXT: [[TMP3:%.*]] = sdiv i8 [[TMP1]], [[TMP2]]
+; CHECK-NEXT: ret i8 [[TMP3]]
+;
+ %x0 = extractelement <4 x i8> %x, i32 0
+ br label %bb1
+bb1:
+ %x3 = extractelement <4 x i8> %x, i32 3
+ %x1 = extractelement <4 x i8> %x, i32 1
+ %x2 = extractelement <4 x i8> %x, i32 2
+ %x0x0 = mul i8 %x0, %x0
+ %x3x3 = mul i8 %x3, %x3
+ %x1x1 = mul i8 %x1, %x1
+ %x2x2 = mul i8 %x2, %x2
+ %1 = add i8 %x0x0, %x3x3
+ %2 = add i8 %x1x1, %x2x2
+ %3 = sdiv i8 %1, %2
+ ret i8 %3
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/bswap.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/bswap.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/bswap.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/bswap.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,247 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+ at src64 = common global [4 x i64] zeroinitializer, align 32
+ at dst64 = common global [4 x i64] zeroinitializer, align 32
+ at src32 = common global [8 x i32] zeroinitializer, align 32
+ at dst32 = common global [8 x i32] zeroinitializer, align 32
+ at src16 = common global [16 x i16] zeroinitializer, align 32
+ at dst16 = common global [16 x i16] zeroinitializer, align 32
+
+declare i64 @llvm.bswap.i64(i64)
+declare i32 @llvm.bswap.i32(i32)
+declare i16 @llvm.bswap.i16(i16)
+
+define void @bswap_2i64() #0 {
+; SSE-LABEL: @bswap_2i64(
+; SSE-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+; SSE-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+; SSE-NEXT: [[BSWAP0:%.*]] = call i64 @llvm.bswap.i64(i64 [[LD0]])
+; SSE-NEXT: [[BSWAP1:%.*]] = call i64 @llvm.bswap.i64(i64 [[LD1]])
+; SSE-NEXT: store i64 [[BSWAP0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+; SSE-NEXT: store i64 [[BSWAP1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @bswap_2i64(
+; AVX-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([4 x i64]* @src64 to <2 x i64>*), align 8
+; AVX-NEXT: [[TMP2:%.*]] = call <2 x i64> @llvm.bswap.v2i64(<2 x i64> [[TMP1]])
+; AVX-NEXT: store <2 x i64> [[TMP2]], <2 x i64>* bitcast ([4 x i64]* @dst64 to <2 x i64>*), align 8
+; AVX-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+ %bswap0 = call i64 @llvm.bswap.i64(i64 %ld0)
+ %bswap1 = call i64 @llvm.bswap.i64(i64 %ld1)
+ store i64 %bswap0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+ store i64 %bswap1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+ ret void
+}
+
+define void @bswap_4i64() #0 {
+; SSE-LABEL: @bswap_4i64(
+; SSE-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+; SSE-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+; SSE-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+; SSE-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+; SSE-NEXT: [[BSWAP0:%.*]] = call i64 @llvm.bswap.i64(i64 [[LD0]])
+; SSE-NEXT: [[BSWAP1:%.*]] = call i64 @llvm.bswap.i64(i64 [[LD1]])
+; SSE-NEXT: [[BSWAP2:%.*]] = call i64 @llvm.bswap.i64(i64 [[LD2]])
+; SSE-NEXT: [[BSWAP3:%.*]] = call i64 @llvm.bswap.i64(i64 [[LD3]])
+; SSE-NEXT: store i64 [[BSWAP0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+; SSE-NEXT: store i64 [[BSWAP1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+; SSE-NEXT: store i64 [[BSWAP2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+; SSE-NEXT: store i64 [[BSWAP3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @bswap_4i64(
+; AVX-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([4 x i64]* @src64 to <4 x i64>*), align 4
+; AVX-NEXT: [[TMP2:%.*]] = call <4 x i64> @llvm.bswap.v4i64(<4 x i64> [[TMP1]])
+; AVX-NEXT: store <4 x i64> [[TMP2]], <4 x i64>* bitcast ([4 x i64]* @dst64 to <4 x i64>*), align 4
+; AVX-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+ %ld2 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+ %ld3 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+ %bswap0 = call i64 @llvm.bswap.i64(i64 %ld0)
+ %bswap1 = call i64 @llvm.bswap.i64(i64 %ld1)
+ %bswap2 = call i64 @llvm.bswap.i64(i64 %ld2)
+ %bswap3 = call i64 @llvm.bswap.i64(i64 %ld3)
+ store i64 %bswap0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+ store i64 %bswap1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+ store i64 %bswap2, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+ store i64 %bswap3, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+ ret void
+}
+
+define void @bswap_4i32() #0 {
+; CHECK-LABEL: @bswap_4i32(
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([8 x i32]* @src32 to <4 x i32>*), align 4
+; CHECK-NEXT: [[TMP2:%.*]] = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> [[TMP1]])
+; CHECK-NEXT: store <4 x i32> [[TMP2]], <4 x i32>* bitcast ([8 x i32]* @dst32 to <4 x i32>*), align 4
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+ %bswap0 = call i32 @llvm.bswap.i32(i32 %ld0)
+ %bswap1 = call i32 @llvm.bswap.i32(i32 %ld1)
+ %bswap2 = call i32 @llvm.bswap.i32(i32 %ld2)
+ %bswap3 = call i32 @llvm.bswap.i32(i32 %ld3)
+ store i32 %bswap0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+ store i32 %bswap1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+ store i32 %bswap2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+ store i32 %bswap3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+ ret void
+}
+
+define void @bswap_8i32() #0 {
+; SSE-LABEL: @bswap_8i32(
+; SSE-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([8 x i32]* @src32 to <4 x i32>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4) to <4 x i32>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> [[TMP1]])
+; SSE-NEXT: [[TMP4:%.*]] = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> [[TMP2]])
+; SSE-NEXT: store <4 x i32> [[TMP3]], <4 x i32>* bitcast ([8 x i32]* @dst32 to <4 x i32>*), align 2
+; SSE-NEXT: store <4 x i32> [[TMP4]], <4 x i32>* bitcast (i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4) to <4 x i32>*), align 2
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @bswap_8i32(
+; AVX-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.bswap.v8i32(<8 x i32> [[TMP1]])
+; AVX-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2
+; AVX-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+ %ld4 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+ %ld5 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+ %ld6 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+ %ld7 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+ %bswap0 = call i32 @llvm.bswap.i32(i32 %ld0)
+ %bswap1 = call i32 @llvm.bswap.i32(i32 %ld1)
+ %bswap2 = call i32 @llvm.bswap.i32(i32 %ld2)
+ %bswap3 = call i32 @llvm.bswap.i32(i32 %ld3)
+ %bswap4 = call i32 @llvm.bswap.i32(i32 %ld4)
+ %bswap5 = call i32 @llvm.bswap.i32(i32 %ld5)
+ %bswap6 = call i32 @llvm.bswap.i32(i32 %ld6)
+ %bswap7 = call i32 @llvm.bswap.i32(i32 %ld7)
+ store i32 %bswap0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+ store i32 %bswap1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+ store i32 %bswap2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+ store i32 %bswap3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+ store i32 %bswap4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+ store i32 %bswap5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+ store i32 %bswap6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+ store i32 %bswap7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+ ret void
+}
+
+define void @bswap_8i16() #0 {
+; CHECK-LABEL: @bswap_8i16(
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> [[TMP1]])
+; CHECK-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %bswap0 = call i16 @llvm.bswap.i16(i16 %ld0)
+ %bswap1 = call i16 @llvm.bswap.i16(i16 %ld1)
+ %bswap2 = call i16 @llvm.bswap.i16(i16 %ld2)
+ %bswap3 = call i16 @llvm.bswap.i16(i16 %ld3)
+ %bswap4 = call i16 @llvm.bswap.i16(i16 %ld4)
+ %bswap5 = call i16 @llvm.bswap.i16(i16 %ld5)
+ %bswap6 = call i16 @llvm.bswap.i16(i16 %ld6)
+ %bswap7 = call i16 @llvm.bswap.i16(i16 %ld7)
+ store i16 %bswap0, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %bswap1, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %bswap2, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %bswap3, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %bswap4, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %bswap5, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %bswap6, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %bswap7, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ ret void
+}
+
+define void @bswap_16i16() #0 {
+; SSE-LABEL: @bswap_16i16(
+; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> [[TMP1]])
+; SSE-NEXT: [[TMP4:%.*]] = call <8 x i16> @llvm.bswap.v8i16(<8 x i16> [[TMP2]])
+; SSE-NEXT: store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @bswap_16i16(
+; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.bswap.v16i16(<16 x i16> [[TMP1]])
+; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2
+; AVX-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %ld8 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2
+ %ld9 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2
+ %ld10 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2
+ %ld11 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2
+ %ld12 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2
+ %ld13 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2
+ %ld14 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2
+ %ld15 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2
+ %bswap0 = call i16 @llvm.bswap.i16(i16 %ld0)
+ %bswap1 = call i16 @llvm.bswap.i16(i16 %ld1)
+ %bswap2 = call i16 @llvm.bswap.i16(i16 %ld2)
+ %bswap3 = call i16 @llvm.bswap.i16(i16 %ld3)
+ %bswap4 = call i16 @llvm.bswap.i16(i16 %ld4)
+ %bswap5 = call i16 @llvm.bswap.i16(i16 %ld5)
+ %bswap6 = call i16 @llvm.bswap.i16(i16 %ld6)
+ %bswap7 = call i16 @llvm.bswap.i16(i16 %ld7)
+ %bswap8 = call i16 @llvm.bswap.i16(i16 %ld8)
+ %bswap9 = call i16 @llvm.bswap.i16(i16 %ld9)
+ %bswap10 = call i16 @llvm.bswap.i16(i16 %ld10)
+ %bswap11 = call i16 @llvm.bswap.i16(i16 %ld11)
+ %bswap12 = call i16 @llvm.bswap.i16(i16 %ld12)
+ %bswap13 = call i16 @llvm.bswap.i16(i16 %ld13)
+ %bswap14 = call i16 @llvm.bswap.i16(i16 %ld14)
+ %bswap15 = call i16 @llvm.bswap.i16(i16 %ld15)
+ store i16 %bswap0 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %bswap1 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %bswap2 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %bswap3 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %bswap4 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %bswap5 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %bswap6 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %bswap7 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ store i16 %bswap8 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2
+ store i16 %bswap9 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2
+ store i16 %bswap10, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2
+ store i16 %bswap11, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2
+ store i16 %bswap12, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2
+ store i16 %bswap13, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2
+ store i16 %bswap14, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2
+ store i16 %bswap15, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2
+ ret void
+}
+
+attributes #0 = { nounwind }
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/call.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/call.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/call.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/call.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,177 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -slp-threshold=-999 -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+declare double @sin(double)
+declare double @cos(double)
+declare double @pow(double, double)
+declare double @exp2(double)
+declare double @sqrt(double)
+declare i64 @round(i64)
+
+
+define void @sin_libm(double* %a, double* %b) {
+; CHECK-LABEL: @sin_libm(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast double* [[A:%.*]] to <2 x double>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* [[TMP1]], align 8
+; CHECK-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.sin.v2f64(<2 x double> [[TMP2]])
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast double* [[B:%.*]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP3]], <2 x double>* [[TMP4]], align 8
+; CHECK-NEXT: ret void
+;
+ %a0 = load double, double* %a, align 8
+ %idx1 = getelementptr inbounds double, double* %a, i64 1
+ %a1 = load double, double* %idx1, align 8
+ %sin1 = tail call double @sin(double %a0) nounwind readnone
+ %sin2 = tail call double @sin(double %a1) nounwind readnone
+ store double %sin1, double* %b, align 8
+ %idx2 = getelementptr inbounds double, double* %b, i64 1
+ store double %sin2, double* %idx2, align 8
+ ret void
+}
+
+define void @cos_libm(double* %a, double* %b) {
+; CHECK-LABEL: @cos_libm(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast double* [[A:%.*]] to <2 x double>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* [[TMP1]], align 8
+; CHECK-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.cos.v2f64(<2 x double> [[TMP2]])
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast double* [[B:%.*]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP3]], <2 x double>* [[TMP4]], align 8
+; CHECK-NEXT: ret void
+;
+ %a0 = load double, double* %a, align 8
+ %idx1 = getelementptr inbounds double, double* %a, i64 1
+ %a1 = load double, double* %idx1, align 8
+ %cos1 = tail call double @cos(double %a0) nounwind readnone
+ %cos2 = tail call double @cos(double %a1) nounwind readnone
+ store double %cos1, double* %b, align 8
+ %idx2 = getelementptr inbounds double, double* %b, i64 1
+ store double %cos2, double* %idx2, align 8
+ ret void
+}
+
+define void @pow_libm(double* %a, double* %b) {
+; CHECK-LABEL: @pow_libm(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast double* [[A:%.*]] to <2 x double>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* [[TMP1]], align 8
+; CHECK-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.pow.v2f64(<2 x double> [[TMP2]], <2 x double> [[TMP2]])
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast double* [[B:%.*]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP3]], <2 x double>* [[TMP4]], align 8
+; CHECK-NEXT: ret void
+;
+ %a0 = load double, double* %a, align 8
+ %idx1 = getelementptr inbounds double, double* %a, i64 1
+ %a1 = load double, double* %idx1, align 8
+ %pow1 = tail call double @pow(double %a0, double %a0) nounwind readnone
+ %pow2 = tail call double @pow(double %a1, double %a1) nounwind readnone
+ store double %pow1, double* %b, align 8
+ %idx2 = getelementptr inbounds double, double* %b, i64 1
+ store double %pow2, double* %idx2, align 8
+ ret void
+}
+
+define void @exp_libm(double* %a, double* %b) {
+; CHECK-LABEL: @exp_libm(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast double* [[A:%.*]] to <2 x double>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* [[TMP1]], align 8
+; CHECK-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.exp2.v2f64(<2 x double> [[TMP2]])
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast double* [[B:%.*]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP3]], <2 x double>* [[TMP4]], align 8
+; CHECK-NEXT: ret void
+;
+ %a0 = load double, double* %a, align 8
+ %idx1 = getelementptr inbounds double, double* %a, i64 1
+ %a1 = load double, double* %idx1, align 8
+ %exp1 = tail call double @exp2(double %a0) nounwind readnone
+ %exp2 = tail call double @exp2(double %a1) nounwind readnone
+ store double %exp1, double* %b, align 8
+ %idx2 = getelementptr inbounds double, double* %b, i64 1
+ store double %exp2, double* %idx2, align 8
+ ret void
+}
+
+; No fast-math-flags are required to convert sqrt library calls to an intrinsic.
+; We just need to know that errno is not set (readnone).
+
+define void @sqrt_libm_no_errno(double* %a, double* %b) {
+; CHECK-LABEL: @sqrt_libm_no_errno(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast double* [[A:%.*]] to <2 x double>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* [[TMP1]], align 8
+; CHECK-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.sqrt.v2f64(<2 x double> [[TMP2]])
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast double* [[B:%.*]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP3]], <2 x double>* [[TMP4]], align 8
+; CHECK-NEXT: ret void
+;
+ %a0 = load double, double* %a, align 8
+ %idx1 = getelementptr inbounds double, double* %a, i64 1
+ %a1 = load double, double* %idx1, align 8
+ %sqrt1 = tail call double @sqrt(double %a0) nounwind readnone
+ %sqrt2 = tail call double @sqrt(double %a1) nounwind readnone
+ store double %sqrt1, double* %b, align 8
+ %idx2 = getelementptr inbounds double, double* %b, i64 1
+ store double %sqrt2, double* %idx2, align 8
+ ret void
+}
+
+; The sqrt intrinsic does not set errno, but a non-constant sqrt call might, so this can't vectorize.
+; The nnan on the call does not matter because there's no guarantee in the C standard that a negative
+; input would result in a nan output ("On a domain error, the function returns an
+; implementation-defined value.")
+
+define void @sqrt_libm_errno(double* %a, double* %b) {
+; CHECK-LABEL: @sqrt_libm_errno(
+; CHECK-NEXT: [[A0:%.*]] = load double, double* [[A:%.*]], align 8
+; CHECK-NEXT: [[IDX1:%.*]] = getelementptr inbounds double, double* [[A]], i64 1
+; CHECK-NEXT: [[A1:%.*]] = load double, double* [[IDX1]], align 8
+; CHECK-NEXT: [[SQRT1:%.*]] = tail call nnan double @sqrt(double [[A0]]) #2
+; CHECK-NEXT: [[SQRT2:%.*]] = tail call nnan double @sqrt(double [[A1]]) #2
+; CHECK-NEXT: store double [[SQRT1]], double* [[B:%.*]], align 8
+; CHECK-NEXT: [[IDX2:%.*]] = getelementptr inbounds double, double* [[B]], i64 1
+; CHECK-NEXT: store double [[SQRT2]], double* [[IDX2]], align 8
+; CHECK-NEXT: ret void
+;
+ %a0 = load double, double* %a, align 8
+ %idx1 = getelementptr inbounds double, double* %a, i64 1
+ %a1 = load double, double* %idx1, align 8
+ %sqrt1 = tail call nnan double @sqrt(double %a0) nounwind
+ %sqrt2 = tail call nnan double @sqrt(double %a1) nounwind
+ store double %sqrt1, double* %b, align 8
+ %idx2 = getelementptr inbounds double, double* %b, i64 1
+ store double %sqrt2, double* %idx2, align 8
+ ret void
+}
+
+; Negative test case
+define void @round_custom(i64* %a, i64* %b) {
+; CHECK-LABEL: @round_custom(
+; CHECK-NEXT: [[A0:%.*]] = load i64, i64* [[A:%.*]], align 8
+; CHECK-NEXT: [[IDX1:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 1
+; CHECK-NEXT: [[A1:%.*]] = load i64, i64* [[IDX1]], align 8
+; CHECK-NEXT: [[ROUND1:%.*]] = tail call i64 @round(i64 [[A0]]) #3
+; CHECK-NEXT: [[ROUND2:%.*]] = tail call i64 @round(i64 [[A1]]) #3
+; CHECK-NEXT: store i64 [[ROUND1]], i64* [[B:%.*]], align 8
+; CHECK-NEXT: [[IDX2:%.*]] = getelementptr inbounds i64, i64* [[B]], i64 1
+; CHECK-NEXT: store i64 [[ROUND2]], i64* [[IDX2]], align 8
+; CHECK-NEXT: ret void
+;
+ %a0 = load i64, i64* %a, align 8
+ %idx1 = getelementptr inbounds i64, i64* %a, i64 1
+ %a1 = load i64, i64* %idx1, align 8
+ %round1 = tail call i64 @round(i64 %a0) nounwind readnone
+ %round2 = tail call i64 @round(i64 %a1) nounwind readnone
+ store i64 %round1, i64* %b, align 8
+ %idx2 = getelementptr inbounds i64, i64* %b, i64 1
+ store i64 %round2, i64* %idx2, align 8
+ ret void
+}
+
+
+; CHECK: declare <2 x double> @llvm.sin.v2f64(<2 x double>) [[ATTR0:#[0-9]+]]
+; CHECK: declare <2 x double> @llvm.cos.v2f64(<2 x double>) [[ATTR0]]
+; CHECK: declare <2 x double> @llvm.pow.v2f64(<2 x double>, <2 x double>) [[ATTR0]]
+; CHECK: declare <2 x double> @llvm.exp2.v2f64(<2 x double>) [[ATTR0]]
+
+; CHECK: attributes [[ATTR0]] = { nounwind readnone speculatable }
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/cast.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/cast.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/cast.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/cast.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,115 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 -basicaa -slp-vectorizer -dce -S | FileCheck %s
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx -basicaa -slp-vectorizer -dce -S | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+
+; int test_sext_4i8_to_4i32(int * restrict A, char * restrict B) {
+; A[0] = B[0];
+; A[1] = B[1];
+; A[2] = B[2];
+; A[3] = B[3];
+; }
+
+define i32 @test_sext_4i8_to_4i32(i32* noalias nocapture %A, i8* noalias nocapture %B) {
+; CHECK-LABEL: @test_sext_4i8_to_4i32(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast i8* [[B:%.*]] to <4 x i8>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i8>, <4 x i8>* [[TMP0]], align 1
+; CHECK-NEXT: [[TMP2:%.*]] = sext <4 x i8> [[TMP1]] to <4 x i32>
+; CHECK-NEXT: [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to <4 x i32>*
+; CHECK-NEXT: store <4 x i32> [[TMP2]], <4 x i32>* [[TMP3]], align 4
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %0 = load i8, i8* %B, align 1
+ %conv = sext i8 %0 to i32
+ store i32 %conv, i32* %A, align 4
+ %arrayidx2 = getelementptr inbounds i8, i8* %B, i64 1
+ %1 = load i8, i8* %arrayidx2, align 1
+ %conv3 = sext i8 %1 to i32
+ %arrayidx4 = getelementptr inbounds i32, i32* %A, i64 1
+ store i32 %conv3, i32* %arrayidx4, align 4
+ %arrayidx5 = getelementptr inbounds i8, i8* %B, i64 2
+ %2 = load i8, i8* %arrayidx5, align 1
+ %conv6 = sext i8 %2 to i32
+ %arrayidx7 = getelementptr inbounds i32, i32* %A, i64 2
+ store i32 %conv6, i32* %arrayidx7, align 4
+ %arrayidx8 = getelementptr inbounds i8, i8* %B, i64 3
+ %3 = load i8, i8* %arrayidx8, align 1
+ %conv9 = sext i8 %3 to i32
+ %arrayidx10 = getelementptr inbounds i32, i32* %A, i64 3
+ store i32 %conv9, i32* %arrayidx10, align 4
+ ret i32 undef
+}
+
+define i32 @test_zext_4i16_to_4i32(i32* noalias nocapture %A, i16* noalias nocapture %B) {
+; CHECK-LABEL: @test_zext_4i16_to_4i32(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast i16* [[B:%.*]] to <4 x i16>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 1
+; CHECK-NEXT: [[TMP2:%.*]] = zext <4 x i16> [[TMP1]] to <4 x i32>
+; CHECK-NEXT: [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to <4 x i32>*
+; CHECK-NEXT: store <4 x i32> [[TMP2]], <4 x i32>* [[TMP3]], align 4
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %0 = load i16, i16* %B, align 1
+ %conv = zext i16 %0 to i32
+ store i32 %conv, i32* %A, align 4
+ %arrayidx2 = getelementptr inbounds i16, i16* %B, i64 1
+ %1 = load i16, i16* %arrayidx2, align 1
+ %conv3 = zext i16 %1 to i32
+ %arrayidx4 = getelementptr inbounds i32, i32* %A, i64 1
+ store i32 %conv3, i32* %arrayidx4, align 4
+ %arrayidx5 = getelementptr inbounds i16, i16* %B, i64 2
+ %2 = load i16, i16* %arrayidx5, align 1
+ %conv6 = zext i16 %2 to i32
+ %arrayidx7 = getelementptr inbounds i32, i32* %A, i64 2
+ store i32 %conv6, i32* %arrayidx7, align 4
+ %arrayidx8 = getelementptr inbounds i16, i16* %B, i64 3
+ %3 = load i16, i16* %arrayidx8, align 1
+ %conv9 = zext i16 %3 to i32
+ %arrayidx10 = getelementptr inbounds i32, i32* %A, i64 3
+ store i32 %conv9, i32* %arrayidx10, align 4
+ ret i32 undef
+}
+
+define i64 @test_sext_4i16_to_4i64(i64* noalias nocapture %A, i16* noalias nocapture %B) {
+; CHECK-LABEL: @test_sext_4i16_to_4i64(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast i16* [[B:%.*]] to <2 x i16>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i16>, <2 x i16>* [[TMP0]], align 1
+; CHECK-NEXT: [[TMP2:%.*]] = sext <2 x i16> [[TMP1]] to <2 x i64>
+; CHECK-NEXT: [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to <2 x i64>*
+; CHECK-NEXT: store <2 x i64> [[TMP2]], <2 x i64>* [[TMP3]], align 4
+; CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[B]], i64 2
+; CHECK-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 2
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast i16* [[ARRAYIDX5]] to <2 x i16>*
+; CHECK-NEXT: [[TMP5:%.*]] = load <2 x i16>, <2 x i16>* [[TMP4]], align 1
+; CHECK-NEXT: [[TMP6:%.*]] = sext <2 x i16> [[TMP5]] to <2 x i64>
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast i64* [[ARRAYIDX7]] to <2 x i64>*
+; CHECK-NEXT: store <2 x i64> [[TMP6]], <2 x i64>* [[TMP7]], align 4
+; CHECK-NEXT: ret i64 undef
+;
+entry:
+ %0 = load i16, i16* %B, align 1
+ %conv = sext i16 %0 to i64
+ store i64 %conv, i64* %A, align 4
+ %arrayidx2 = getelementptr inbounds i16, i16* %B, i64 1
+ %1 = load i16, i16* %arrayidx2, align 1
+ %conv3 = sext i16 %1 to i64
+ %arrayidx4 = getelementptr inbounds i64, i64* %A, i64 1
+ store i64 %conv3, i64* %arrayidx4, align 4
+ %arrayidx5 = getelementptr inbounds i16, i16* %B, i64 2
+ %2 = load i16, i16* %arrayidx5, align 1
+ %conv6 = sext i16 %2 to i64
+ %arrayidx7 = getelementptr inbounds i64, i64* %A, i64 2
+ store i64 %conv6, i64* %arrayidx7, align 4
+ %arrayidx8 = getelementptr inbounds i16, i16* %B, i64 3
+ %3 = load i16, i16* %arrayidx8, align 1
+ %conv9 = sext i16 %3 to i64
+ %arrayidx10 = getelementptr inbounds i64, i64* %A, i64 3
+ store i64 %conv9, i64* %arrayidx10, align 4
+ ret i64 undef
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/cmp_commute.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/cmp_commute.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/cmp_commute.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/cmp_commute.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,283 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64--- -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE
+; RUN: opt < %s -slp-vectorizer -instcombine -S -mtriple=x86_64--- -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX
+
+;
+; Check that we can commute operands based on the predicate.
+;
+
+define <4 x i32> @icmp_eq_v4i32(<4 x i32> %a, i32* %b) {
+; CHECK-LABEL: @icmp_eq_v4i32(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast i32* [[B:%.*]] to <4 x i32>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* [[TMP1]], align 4
+; CHECK-NEXT: [[TMP3:%.*]] = icmp eq <4 x i32> [[TMP2]], [[A:%.*]]
+; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[TMP3]] to <4 x i32>
+; CHECK-NEXT: ret <4 x i32> [[R]]
+;
+ %a0 = extractelement <4 x i32> %a, i32 0
+ %a1 = extractelement <4 x i32> %a, i32 1
+ %a2 = extractelement <4 x i32> %a, i32 2
+ %a3 = extractelement <4 x i32> %a, i32 3
+ %p0 = getelementptr inbounds i32, i32* %b, i32 0
+ %p1 = getelementptr inbounds i32, i32* %b, i32 1
+ %p2 = getelementptr inbounds i32, i32* %b, i32 2
+ %p3 = getelementptr inbounds i32, i32* %b, i32 3
+ %b0 = load i32, i32* %p0, align 4
+ %b1 = load i32, i32* %p1, align 4
+ %b2 = load i32, i32* %p2, align 4
+ %b3 = load i32, i32* %p3, align 4
+ %c0 = icmp eq i32 %a0, %b0
+ %c1 = icmp eq i32 %b1, %a1
+ %c2 = icmp eq i32 %b2, %a2
+ %c3 = icmp eq i32 %a3, %b3
+ %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0
+ %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1
+ %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2
+ %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3
+ %r = sext <4 x i1> %d3 to <4 x i32>
+ ret <4 x i32> %r
+}
+
+define <4 x i32> @icmp_ne_v4i32(<4 x i32> %a, i32* %b) {
+; CHECK-LABEL: @icmp_ne_v4i32(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast i32* [[B:%.*]] to <4 x i32>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* [[TMP1]], align 4
+; CHECK-NEXT: [[TMP3:%.*]] = icmp ne <4 x i32> [[TMP2]], [[A:%.*]]
+; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[TMP3]] to <4 x i32>
+; CHECK-NEXT: ret <4 x i32> [[R]]
+;
+ %a0 = extractelement <4 x i32> %a, i32 0
+ %a1 = extractelement <4 x i32> %a, i32 1
+ %a2 = extractelement <4 x i32> %a, i32 2
+ %a3 = extractelement <4 x i32> %a, i32 3
+ %p0 = getelementptr inbounds i32, i32* %b, i32 0
+ %p1 = getelementptr inbounds i32, i32* %b, i32 1
+ %p2 = getelementptr inbounds i32, i32* %b, i32 2
+ %p3 = getelementptr inbounds i32, i32* %b, i32 3
+ %b0 = load i32, i32* %p0, align 4
+ %b1 = load i32, i32* %p1, align 4
+ %b2 = load i32, i32* %p2, align 4
+ %b3 = load i32, i32* %p3, align 4
+ %c0 = icmp ne i32 %a0, %b0
+ %c1 = icmp ne i32 %b1, %a1
+ %c2 = icmp ne i32 %b2, %a2
+ %c3 = icmp ne i32 %a3, %b3
+ %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0
+ %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1
+ %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2
+ %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3
+ %r = sext <4 x i1> %d3 to <4 x i32>
+ ret <4 x i32> %r
+}
+
+define <4 x i32> @fcmp_oeq_v4i32(<4 x float> %a, float* %b) {
+; CHECK-LABEL: @fcmp_oeq_v4i32(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast float* [[B:%.*]] to <4 x float>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* [[TMP1]], align 4
+; CHECK-NEXT: [[TMP3:%.*]] = fcmp oeq <4 x float> [[TMP2]], [[A:%.*]]
+; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[TMP3]] to <4 x i32>
+; CHECK-NEXT: ret <4 x i32> [[R]]
+;
+ %a0 = extractelement <4 x float> %a, i32 0
+ %a1 = extractelement <4 x float> %a, i32 1
+ %a2 = extractelement <4 x float> %a, i32 2
+ %a3 = extractelement <4 x float> %a, i32 3
+ %p0 = getelementptr inbounds float, float* %b, i32 0
+ %p1 = getelementptr inbounds float, float* %b, i32 1
+ %p2 = getelementptr inbounds float, float* %b, i32 2
+ %p3 = getelementptr inbounds float, float* %b, i32 3
+ %b0 = load float, float* %p0, align 4
+ %b1 = load float, float* %p1, align 4
+ %b2 = load float, float* %p2, align 4
+ %b3 = load float, float* %p3, align 4
+ %c0 = fcmp oeq float %a0, %b0
+ %c1 = fcmp oeq float %b1, %a1
+ %c2 = fcmp oeq float %b2, %a2
+ %c3 = fcmp oeq float %a3, %b3
+ %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0
+ %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1
+ %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2
+ %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3
+ %r = sext <4 x i1> %d3 to <4 x i32>
+ ret <4 x i32> %r
+}
+
+define <4 x i32> @fcmp_uno_v4i32(<4 x float> %a, float* %b) {
+; CHECK-LABEL: @fcmp_uno_v4i32(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast float* [[B:%.*]] to <4 x float>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* [[TMP1]], align 4
+; CHECK-NEXT: [[TMP3:%.*]] = fcmp uno <4 x float> [[TMP2]], [[A:%.*]]
+; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[TMP3]] to <4 x i32>
+; CHECK-NEXT: ret <4 x i32> [[R]]
+;
+ %a0 = extractelement <4 x float> %a, i32 0
+ %a1 = extractelement <4 x float> %a, i32 1
+ %a2 = extractelement <4 x float> %a, i32 2
+ %a3 = extractelement <4 x float> %a, i32 3
+ %p0 = getelementptr inbounds float, float* %b, i32 0
+ %p1 = getelementptr inbounds float, float* %b, i32 1
+ %p2 = getelementptr inbounds float, float* %b, i32 2
+ %p3 = getelementptr inbounds float, float* %b, i32 3
+ %b0 = load float, float* %p0, align 4
+ %b1 = load float, float* %p1, align 4
+ %b2 = load float, float* %p2, align 4
+ %b3 = load float, float* %p3, align 4
+ %c0 = fcmp uno float %a0, %b0
+ %c1 = fcmp uno float %b1, %a1
+ %c2 = fcmp uno float %b2, %a2
+ %c3 = fcmp uno float %a3, %b3
+ %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0
+ %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1
+ %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2
+ %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3
+ %r = sext <4 x i1> %d3 to <4 x i32>
+ ret <4 x i32> %r
+}
+
+;
+; Check that we can commute operands by swapping the predicate.
+;
+
+define <4 x i32> @icmp_sgt_slt_v4i32(<4 x i32> %a, i32* %b) {
+; CHECK-LABEL: @icmp_sgt_slt_v4i32(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast i32* [[B:%.*]] to <4 x i32>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* [[TMP1]], align 4
+; CHECK-NEXT: [[TMP3:%.*]] = icmp slt <4 x i32> [[TMP2]], [[A:%.*]]
+; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[TMP3]] to <4 x i32>
+; CHECK-NEXT: ret <4 x i32> [[R]]
+;
+ %a0 = extractelement <4 x i32> %a, i32 0
+ %a1 = extractelement <4 x i32> %a, i32 1
+ %a2 = extractelement <4 x i32> %a, i32 2
+ %a3 = extractelement <4 x i32> %a, i32 3
+ %p0 = getelementptr inbounds i32, i32* %b, i32 0
+ %p1 = getelementptr inbounds i32, i32* %b, i32 1
+ %p2 = getelementptr inbounds i32, i32* %b, i32 2
+ %p3 = getelementptr inbounds i32, i32* %b, i32 3
+ %b0 = load i32, i32* %p0, align 4
+ %b1 = load i32, i32* %p1, align 4
+ %b2 = load i32, i32* %p2, align 4
+ %b3 = load i32, i32* %p3, align 4
+ %c0 = icmp sgt i32 %a0, %b0
+ %c1 = icmp slt i32 %b1, %a1
+ %c2 = icmp slt i32 %b2, %a2
+ %c3 = icmp sgt i32 %a3, %b3
+ %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0
+ %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1
+ %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2
+ %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3
+ %r = sext <4 x i1> %d3 to <4 x i32>
+ ret <4 x i32> %r
+}
+
+define <4 x i32> @icmp_uge_ule_v4i32(<4 x i32> %a, i32* %b) {
+; CHECK-LABEL: @icmp_uge_ule_v4i32(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast i32* [[B:%.*]] to <4 x i32>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* [[TMP1]], align 4
+; CHECK-NEXT: [[TMP3:%.*]] = icmp ule <4 x i32> [[TMP2]], [[A:%.*]]
+; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[TMP3]] to <4 x i32>
+; CHECK-NEXT: ret <4 x i32> [[R]]
+;
+ %a0 = extractelement <4 x i32> %a, i32 0
+ %a1 = extractelement <4 x i32> %a, i32 1
+ %a2 = extractelement <4 x i32> %a, i32 2
+ %a3 = extractelement <4 x i32> %a, i32 3
+ %p0 = getelementptr inbounds i32, i32* %b, i32 0
+ %p1 = getelementptr inbounds i32, i32* %b, i32 1
+ %p2 = getelementptr inbounds i32, i32* %b, i32 2
+ %p3 = getelementptr inbounds i32, i32* %b, i32 3
+ %b0 = load i32, i32* %p0, align 4
+ %b1 = load i32, i32* %p1, align 4
+ %b2 = load i32, i32* %p2, align 4
+ %b3 = load i32, i32* %p3, align 4
+ %c0 = icmp uge i32 %a0, %b0
+ %c1 = icmp ule i32 %b1, %a1
+ %c2 = icmp ule i32 %b2, %a2
+ %c3 = icmp uge i32 %a3, %b3
+ %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0
+ %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1
+ %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2
+ %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3
+ %r = sext <4 x i1> %d3 to <4 x i32>
+ ret <4 x i32> %r
+}
+
+define <4 x i32> @fcmp_ogt_olt_v4i32(<4 x float> %a, float* %b) {
+; CHECK-LABEL: @fcmp_ogt_olt_v4i32(
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast float* [[B:%.*]] to <4 x float>*
+; CHECK-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* [[TMP1]], align 4
+; CHECK-NEXT: [[TMP3:%.*]] = fcmp olt <4 x float> [[TMP2]], [[A:%.*]]
+; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[TMP3]] to <4 x i32>
+; CHECK-NEXT: ret <4 x i32> [[R]]
+;
+ %a0 = extractelement <4 x float> %a, i32 0
+ %a1 = extractelement <4 x float> %a, i32 1
+ %a2 = extractelement <4 x float> %a, i32 2
+ %a3 = extractelement <4 x float> %a, i32 3
+ %p0 = getelementptr inbounds float, float* %b, i32 0
+ %p1 = getelementptr inbounds float, float* %b, i32 1
+ %p2 = getelementptr inbounds float, float* %b, i32 2
+ %p3 = getelementptr inbounds float, float* %b, i32 3
+ %b0 = load float, float* %p0, align 4
+ %b1 = load float, float* %p1, align 4
+ %b2 = load float, float* %p2, align 4
+ %b3 = load float, float* %p3, align 4
+ %c0 = fcmp ogt float %a0, %b0
+ %c1 = fcmp olt float %b1, %a1
+ %c2 = fcmp olt float %b2, %a2
+ %c3 = fcmp ogt float %a3, %b3
+ %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0
+ %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1
+ %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2
+ %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3
+ %r = sext <4 x i1> %d3 to <4 x i32>
+ ret <4 x i32> %r
+}
+
+define <4 x i32> @fcmp_ord_uno_v4i32(<4 x float> %a, float* %b) {
+; CHECK-LABEL: @fcmp_ord_uno_v4i32(
+; CHECK-NEXT: [[A0:%.*]] = extractelement <4 x float> [[A:%.*]], i32 0
+; CHECK-NEXT: [[A1:%.*]] = extractelement <4 x float> [[A]], i32 1
+; CHECK-NEXT: [[A2:%.*]] = extractelement <4 x float> [[A]], i32 2
+; CHECK-NEXT: [[A3:%.*]] = extractelement <4 x float> [[A]], i32 3
+; CHECK-NEXT: [[P1:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 1
+; CHECK-NEXT: [[P2:%.*]] = getelementptr inbounds float, float* [[B]], i64 2
+; CHECK-NEXT: [[P3:%.*]] = getelementptr inbounds float, float* [[B]], i64 3
+; CHECK-NEXT: [[B0:%.*]] = load float, float* [[B]], align 4
+; CHECK-NEXT: [[B1:%.*]] = load float, float* [[P1]], align 4
+; CHECK-NEXT: [[B2:%.*]] = load float, float* [[P2]], align 4
+; CHECK-NEXT: [[B3:%.*]] = load float, float* [[P3]], align 4
+; CHECK-NEXT: [[C0:%.*]] = fcmp ord float [[A0]], [[B0]]
+; CHECK-NEXT: [[C1:%.*]] = fcmp uno float [[B1]], [[A1]]
+; CHECK-NEXT: [[C2:%.*]] = fcmp uno float [[B2]], [[A2]]
+; CHECK-NEXT: [[C3:%.*]] = fcmp ord float [[A3]], [[B3]]
+; CHECK-NEXT: [[D0:%.*]] = insertelement <4 x i1> undef, i1 [[C0]], i32 0
+; CHECK-NEXT: [[D1:%.*]] = insertelement <4 x i1> [[D0]], i1 [[C1]], i32 1
+; CHECK-NEXT: [[D2:%.*]] = insertelement <4 x i1> [[D1]], i1 [[C2]], i32 2
+; CHECK-NEXT: [[D3:%.*]] = insertelement <4 x i1> [[D2]], i1 [[C3]], i32 3
+; CHECK-NEXT: [[R:%.*]] = sext <4 x i1> [[D3]] to <4 x i32>
+; CHECK-NEXT: ret <4 x i32> [[R]]
+;
+ %a0 = extractelement <4 x float> %a, i32 0
+ %a1 = extractelement <4 x float> %a, i32 1
+ %a2 = extractelement <4 x float> %a, i32 2
+ %a3 = extractelement <4 x float> %a, i32 3
+ %p0 = getelementptr inbounds float, float* %b, i32 0
+ %p1 = getelementptr inbounds float, float* %b, i32 1
+ %p2 = getelementptr inbounds float, float* %b, i32 2
+ %p3 = getelementptr inbounds float, float* %b, i32 3
+ %b0 = load float, float* %p0, align 4
+ %b1 = load float, float* %p1, align 4
+ %b2 = load float, float* %p2, align 4
+ %b3 = load float, float* %p3, align 4
+ %c0 = fcmp ord float %a0, %b0
+ %c1 = fcmp uno float %b1, %a1
+ %c2 = fcmp uno float %b2, %a2
+ %c3 = fcmp ord float %a3, %b3
+ %d0 = insertelement <4 x i1> undef, i1 %c0, i32 0
+ %d1 = insertelement <4 x i1> %d0, i1 %c1, i32 1
+ %d2 = insertelement <4 x i1> %d1, i1 %c2, i32 2
+ %d3 = insertelement <4 x i1> %d2, i1 %c3, i32 3
+ %r = sext <4 x i1> %d3 to <4 x i32>
+ ret <4 x i32> %r
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/cmp_sel.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/cmp_sel.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/cmp_sel.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/cmp_sel.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,37 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+; int foo(double * restrict A, double * restrict B, double G) {
+; A[0] = (B[10] ? G : 1);
+; A[1] = (B[11] ? G : 1);
+; }
+
+define i32 @foo(double* noalias nocapture %A, double* noalias nocapture %B, double %G) {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[B:%.*]], i64 10
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
+; CHECK-NEXT: [[TMP2:%.*]] = fcmp une <2 x double> [[TMP1]], zeroinitializer
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <2 x double> undef, double [[G:%.*]], i32 0
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x double> [[TMP3]], double [[G]], i32 1
+; CHECK-NEXT: [[TMP5:%.*]] = select <2 x i1> [[TMP2]], <2 x double> [[TMP4]], <2 x double> <double 1.000000e+00, double 1.000000e+00>
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast double* [[A:%.*]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP5]], <2 x double>* [[TMP6]], align 8
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %arrayidx = getelementptr inbounds double, double* %B, i64 10
+ %0 = load double, double* %arrayidx, align 8
+ %tobool = fcmp une double %0, 0.000000e+00
+ %cond = select i1 %tobool, double %G, double 1.000000e+00
+ store double %cond, double* %A, align 8
+ %arrayidx2 = getelementptr inbounds double, double* %B, i64 11
+ %1 = load double, double* %arrayidx2, align 8
+ %tobool3 = fcmp une double %1, 0.000000e+00
+ %cond7 = select i1 %tobool3, double %G, double 1.000000e+00
+ %arrayidx8 = getelementptr inbounds double, double* %A, i64 1
+ store double %cond7, double* %arrayidx8, align 8
+ ret i32 undef
+}
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/commutativity.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/commutativity.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/commutativity.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/commutativity.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,116 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -slp-vectorizer < %s -S | FileCheck %s
+
+; Verify that the SLP vectorizer is able to figure out that commutativity
+; offers the possibility to splat/broadcast %c and thus make it profitable
+; to vectorize this case
+
+
+; ModuleID = 'bugpoint-reduced-simplified.bc'
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.11.0"
+
+ at cle = external unnamed_addr global [32 x i8], align 16
+ at cle32 = external unnamed_addr global [32 x i32], align 16
+
+
+; Check that we correctly detect a splat/broadcast by leveraging the
+; commutativity property of `xor`.
+
+define void @splat(i8 %a, i8 %b, i8 %c) {
+; CHECK-LABEL: @splat(
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <16 x i8> undef, i8 [[C:%.*]], i32 0
+; CHECK-NEXT: [[TMP2:%.*]] = insertelement <16 x i8> [[TMP1]], i8 [[C]], i32 1
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <16 x i8> [[TMP2]], i8 [[C]], i32 2
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <16 x i8> [[TMP3]], i8 [[C]], i32 3
+; CHECK-NEXT: [[TMP5:%.*]] = insertelement <16 x i8> [[TMP4]], i8 [[C]], i32 4
+; CHECK-NEXT: [[TMP6:%.*]] = insertelement <16 x i8> [[TMP5]], i8 [[C]], i32 5
+; CHECK-NEXT: [[TMP7:%.*]] = insertelement <16 x i8> [[TMP6]], i8 [[C]], i32 6
+; CHECK-NEXT: [[TMP8:%.*]] = insertelement <16 x i8> [[TMP7]], i8 [[C]], i32 7
+; CHECK-NEXT: [[TMP9:%.*]] = insertelement <16 x i8> [[TMP8]], i8 [[C]], i32 8
+; CHECK-NEXT: [[TMP10:%.*]] = insertelement <16 x i8> [[TMP9]], i8 [[C]], i32 9
+; CHECK-NEXT: [[TMP11:%.*]] = insertelement <16 x i8> [[TMP10]], i8 [[C]], i32 10
+; CHECK-NEXT: [[TMP12:%.*]] = insertelement <16 x i8> [[TMP11]], i8 [[C]], i32 11
+; CHECK-NEXT: [[TMP13:%.*]] = insertelement <16 x i8> [[TMP12]], i8 [[C]], i32 12
+; CHECK-NEXT: [[TMP14:%.*]] = insertelement <16 x i8> [[TMP13]], i8 [[C]], i32 13
+; CHECK-NEXT: [[TMP15:%.*]] = insertelement <16 x i8> [[TMP14]], i8 [[C]], i32 14
+; CHECK-NEXT: [[TMP16:%.*]] = insertelement <16 x i8> [[TMP15]], i8 [[C]], i32 15
+; CHECK-NEXT: [[TMP17:%.*]] = insertelement <2 x i8> undef, i8 [[A:%.*]], i32 0
+; CHECK-NEXT: [[TMP18:%.*]] = insertelement <2 x i8> [[TMP17]], i8 [[B:%.*]], i32 1
+; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <2 x i8> [[TMP18]], <2 x i8> undef, <16 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0, i32 1, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+; CHECK-NEXT: [[TMP19:%.*]] = xor <16 x i8> [[TMP16]], [[SHUFFLE]]
+; CHECK-NEXT: store <16 x i8> [[TMP19]], <16 x i8>* bitcast ([32 x i8]* @cle to <16 x i8>*), align 16
+; CHECK-NEXT: ret void
+;
+ %1 = xor i8 %c, %a
+ store i8 %1, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 0), align 16
+ %2 = xor i8 %a, %c
+ store i8 %2, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 1)
+ %3 = xor i8 %a, %c
+ store i8 %3, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 2)
+ %4 = xor i8 %a, %c
+ store i8 %4, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 3)
+ %5 = xor i8 %c, %a
+ store i8 %5, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 4)
+ %6 = xor i8 %c, %b
+ store i8 %6, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 5)
+ %7 = xor i8 %c, %a
+ store i8 %7, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 6)
+ %8 = xor i8 %c, %b
+ store i8 %8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 7)
+ %9 = xor i8 %a, %c
+ store i8 %9, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 8)
+ %10 = xor i8 %a, %c
+ store i8 %10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 9)
+ %11 = xor i8 %a, %c
+ store i8 %11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 10)
+ %12 = xor i8 %a, %c
+ store i8 %12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 11)
+ %13 = xor i8 %a, %c
+ store i8 %13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 12)
+ %14 = xor i8 %a, %c
+ store i8 %14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 13)
+ %15 = xor i8 %a, %c
+ store i8 %15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 14)
+ %16 = xor i8 %a, %c
+ store i8 %16, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @cle, i64 0, i64 15)
+ ret void
+}
+
+
+
+; Check that we correctly detect that we can have the same opcode on one side by
+; leveraging the commutativity property of `xor`.
+
+define void @same_opcode_on_one_side(i32 %a, i32 %b, i32 %c) {
+; CHECK-LABEL: @same_opcode_on_one_side(
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x i32> undef, i32 [[C:%.*]], i32 0
+; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> [[TMP1]], i32 [[C]], i32 1
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[C]], i32 2
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <4 x i32> [[TMP3]], i32 [[C]], i32 3
+; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x i32> undef, i32 [[A:%.*]], i32 0
+; CHECK-NEXT: [[TMP6:%.*]] = insertelement <4 x i32> [[TMP5]], i32 [[A]], i32 1
+; CHECK-NEXT: [[TMP7:%.*]] = insertelement <4 x i32> [[TMP6]], i32 [[A]], i32 2
+; CHECK-NEXT: [[TMP8:%.*]] = insertelement <4 x i32> [[TMP7]], i32 [[A]], i32 3
+; CHECK-NEXT: [[TMP9:%.*]] = add <4 x i32> [[TMP4]], [[TMP8]]
+; CHECK-NEXT: [[TMP10:%.*]] = insertelement <4 x i32> [[TMP5]], i32 [[B:%.*]], i32 1
+; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x i32> [[TMP10]], i32 [[C]], i32 2
+; CHECK-NEXT: [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[A]], i32 3
+; CHECK-NEXT: [[TMP13:%.*]] = xor <4 x i32> [[TMP9]], [[TMP12]]
+; CHECK-NEXT: store <4 x i32> [[TMP13]], <4 x i32>* bitcast ([32 x i32]* @cle32 to <4 x i32>*), align 16
+; CHECK-NEXT: ret void
+;
+ %add1 = add i32 %c, %a
+ %add2 = add i32 %c, %a
+ %add3 = add i32 %a, %c
+ %add4 = add i32 %c, %a
+ %1 = xor i32 %add1, %a
+ store i32 %1, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @cle32, i64 0, i64 0), align 16
+ %2 = xor i32 %b, %add2
+ store i32 %2, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @cle32, i64 0, i64 1)
+ %3 = xor i32 %c, %add3
+ store i32 %3, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @cle32, i64 0, i64 2)
+ %4 = xor i32 %a, %add4
+ store i32 %4, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @cle32, i64 0, i64 3)
+ ret void
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/compare-reduce.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/compare-reduce.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/compare-reduce.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/compare-reduce.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,76 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.7.0"
+
+ at .str = private unnamed_addr constant [6 x i8] c"bingo\00", align 1
+
+define void @reduce_compare(double* nocapture %A, i32 %n) {
+; CHECK-LABEL: @reduce_compare(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[CONV:%.*]] = sitofp i32 [[N:%.*]] to double
+; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x double> undef, double [[CONV]], i32 0
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x double> [[TMP0]], double [[CONV]], i32 1
+; CHECK-NEXT: br label [[FOR_BODY:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
+; CHECK-NEXT: [[TMP2:%.*]] = shl nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[A:%.*]], i64 [[TMP2]]
+; CHECK-NEXT: [[TMP3:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: [[TMP4:%.*]] = load <2 x double>, <2 x double>* [[TMP3]], align 8
+; CHECK-NEXT: [[TMP5:%.*]] = fmul <2 x double> [[TMP1]], [[TMP4]]
+; CHECK-NEXT: [[TMP6:%.*]] = fmul <2 x double> [[TMP5]], <double 7.000000e+00, double 4.000000e+00>
+; CHECK-NEXT: [[TMP7:%.*]] = fadd <2 x double> [[TMP6]], <double 5.000000e+00, double 9.000000e+00>
+; CHECK-NEXT: [[TMP8:%.*]] = extractelement <2 x double> [[TMP7]], i32 0
+; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x double> [[TMP7]], i32 1
+; CHECK-NEXT: [[CMP11:%.*]] = fcmp ogt double [[TMP8]], [[TMP9]]
+; CHECK-NEXT: br i1 [[CMP11]], label [[IF_THEN:%.*]], label [[FOR_INC]]
+; CHECK: if.then:
+; CHECK-NEXT: [[CALL:%.*]] = tail call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([6 x i8], [6 x i8]* @.str, i64 0, i64 0))
+; CHECK-NEXT: br label [[FOR_INC]]
+; CHECK: for.inc:
+; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
+; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
+; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], 100
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; CHECK: for.end:
+; CHECK-NEXT: ret void
+;
+entry:
+ %conv = sitofp i32 %n to double
+ br label %for.body
+
+for.body: ; preds = %for.inc, %entry
+ %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.inc ]
+ %0 = shl nsw i64 %indvars.iv, 1
+ %arrayidx = getelementptr inbounds double, double* %A, i64 %0
+ %1 = load double, double* %arrayidx, align 8
+ %mul1 = fmul double %conv, %1
+ %mul2 = fmul double %mul1, 7.000000e+00
+ %add = fadd double %mul2, 5.000000e+00
+ %2 = or i64 %0, 1
+ %arrayidx6 = getelementptr inbounds double, double* %A, i64 %2
+ %3 = load double, double* %arrayidx6, align 8
+ %mul8 = fmul double %conv, %3
+ %mul9 = fmul double %mul8, 4.000000e+00
+ %add10 = fadd double %mul9, 9.000000e+00
+ %cmp11 = fcmp ogt double %add, %add10
+ br i1 %cmp11, label %if.then, label %for.inc
+
+if.then: ; preds = %for.body
+ %call = tail call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([6 x i8], [6 x i8]* @.str, i64 0, i64 0))
+ br label %for.inc
+
+for.inc: ; preds = %for.body, %if.then
+ %indvars.iv.next = add i64 %indvars.iv, 1
+ %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+ %exitcond = icmp eq i32 %lftr.wideiv, 100
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.inc
+ ret void
+}
+
+declare i32 @printf(i8* nocapture, ...)
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/consecutive-access.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/consecutive-access.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/consecutive-access.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/consecutive-access.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,556 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -S | FileCheck %s
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.9.0"
+
+ at A = common global [2000 x double] zeroinitializer, align 16
+ at B = common global [2000 x double] zeroinitializer, align 16
+ at C = common global [2000 x float] zeroinitializer, align 16
+ at D = common global [2000 x float] zeroinitializer, align 16
+
+; Currently SCEV isn't smart enough to figure out that accesses
+; A[3*i], A[3*i+1] and A[3*i+2] are consecutive, but in future
+; that would hopefully be fixed. For now, check that this isn't
+; vectorized.
+; Function Attrs: nounwind ssp uwtable
+define void @foo_3double(i32 %u) #0 {
+; CHECK-LABEL: @foo_3double(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[U_ADDR:%.*]] = alloca i32, align 4
+; CHECK-NEXT: store i32 [[U:%.*]], i32* [[U_ADDR]], align 4
+; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[U]], 3
+; CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[MUL]] to i64
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 [[IDXPROM]]
+; CHECK-NEXT: [[TMP0:%.*]] = load double, double* [[ARRAYIDX]], align 8
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 [[IDXPROM]]
+; CHECK-NEXT: [[TMP1:%.*]] = load double, double* [[ARRAYIDX4]], align 8
+; CHECK-NEXT: [[ADD5:%.*]] = fadd double [[TMP0]], [[TMP1]]
+; CHECK-NEXT: store double [[ADD5]], double* [[ARRAYIDX]], align 8
+; CHECK-NEXT: [[ADD11:%.*]] = add nsw i32 [[MUL]], 1
+; CHECK-NEXT: [[IDXPROM12:%.*]] = sext i32 [[ADD11]] to i64
+; CHECK-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 [[IDXPROM12]]
+; CHECK-NEXT: [[TMP2:%.*]] = load double, double* [[ARRAYIDX13]], align 8
+; CHECK-NEXT: [[ARRAYIDX17:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 [[IDXPROM12]]
+; CHECK-NEXT: [[TMP3:%.*]] = load double, double* [[ARRAYIDX17]], align 8
+; CHECK-NEXT: [[ADD18:%.*]] = fadd double [[TMP2]], [[TMP3]]
+; CHECK-NEXT: store double [[ADD18]], double* [[ARRAYIDX13]], align 8
+; CHECK-NEXT: [[ADD24:%.*]] = add nsw i32 [[MUL]], 2
+; CHECK-NEXT: [[IDXPROM25:%.*]] = sext i32 [[ADD24]] to i64
+; CHECK-NEXT: [[ARRAYIDX26:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 [[IDXPROM25]]
+; CHECK-NEXT: [[TMP4:%.*]] = load double, double* [[ARRAYIDX26]], align 8
+; CHECK-NEXT: [[ARRAYIDX30:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 [[IDXPROM25]]
+; CHECK-NEXT: [[TMP5:%.*]] = load double, double* [[ARRAYIDX30]], align 8
+; CHECK-NEXT: [[ADD31:%.*]] = fadd double [[TMP4]], [[TMP5]]
+; CHECK-NEXT: store double [[ADD31]], double* [[ARRAYIDX26]], align 8
+; CHECK-NEXT: ret void
+;
+entry:
+ %u.addr = alloca i32, align 4
+ store i32 %u, i32* %u.addr, align 4
+ %mul = mul nsw i32 %u, 3
+ %idxprom = sext i32 %mul to i64
+ %arrayidx = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 %idxprom
+ %0 = load double, double* %arrayidx, align 8
+ %arrayidx4 = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 %idxprom
+ %1 = load double, double* %arrayidx4, align 8
+ %add5 = fadd double %0, %1
+ store double %add5, double* %arrayidx, align 8
+ %add11 = add nsw i32 %mul, 1
+ %idxprom12 = sext i32 %add11 to i64
+ %arrayidx13 = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 %idxprom12
+ %2 = load double, double* %arrayidx13, align 8
+ %arrayidx17 = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 %idxprom12
+ %3 = load double, double* %arrayidx17, align 8
+ %add18 = fadd double %2, %3
+ store double %add18, double* %arrayidx13, align 8
+ %add24 = add nsw i32 %mul, 2
+ %idxprom25 = sext i32 %add24 to i64
+ %arrayidx26 = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 %idxprom25
+ %4 = load double, double* %arrayidx26, align 8
+ %arrayidx30 = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 %idxprom25
+ %5 = load double, double* %arrayidx30, align 8
+ %add31 = fadd double %4, %5
+ store double %add31, double* %arrayidx26, align 8
+ ret void
+}
+
+; SCEV should be able to tell that accesses A[C1 + C2*i], A[C1 + C2*i], ...
+; A[C1 + C2*i] are consecutive, if C2 is a power of 2, and C2 > C1 > 0.
+; Thus, the following code should be vectorized.
+; Function Attrs: nounwind ssp uwtable
+define void @foo_2double(i32 %u) #0 {
+; CHECK-LABEL: @foo_2double(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[U_ADDR:%.*]] = alloca i32, align 4
+; CHECK-NEXT: store i32 [[U:%.*]], i32* [[U_ADDR]], align 4
+; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[U]], 2
+; CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[MUL]] to i64
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 [[IDXPROM]]
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 [[IDXPROM]]
+; CHECK-NEXT: [[ADD11:%.*]] = add nsw i32 [[MUL]], 1
+; CHECK-NEXT: [[IDXPROM12:%.*]] = sext i32 [[ADD11]] to i64
+; CHECK-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 [[IDXPROM12]]
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
+; CHECK-NEXT: [[ARRAYIDX17:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 [[IDXPROM12]]
+; CHECK-NEXT: [[TMP2:%.*]] = bitcast double* [[ARRAYIDX4]] to <2 x double>*
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[TMP2]], align 8
+; CHECK-NEXT: [[TMP4:%.*]] = fadd <2 x double> [[TMP1]], [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP4]], <2 x double>* [[TMP5]], align 8
+; CHECK-NEXT: ret void
+;
+entry:
+ %u.addr = alloca i32, align 4
+ store i32 %u, i32* %u.addr, align 4
+ %mul = mul nsw i32 %u, 2
+ %idxprom = sext i32 %mul to i64
+ %arrayidx = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 %idxprom
+ %0 = load double, double* %arrayidx, align 8
+ %arrayidx4 = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 %idxprom
+ %1 = load double, double* %arrayidx4, align 8
+ %add5 = fadd double %0, %1
+ store double %add5, double* %arrayidx, align 8
+ %add11 = add nsw i32 %mul, 1
+ %idxprom12 = sext i32 %add11 to i64
+ %arrayidx13 = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 %idxprom12
+ %2 = load double, double* %arrayidx13, align 8
+ %arrayidx17 = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 %idxprom12
+ %3 = load double, double* %arrayidx17, align 8
+ %add18 = fadd double %2, %3
+ store double %add18, double* %arrayidx13, align 8
+ ret void
+}
+
+; Similar to the previous test, but with different datatype.
+; Function Attrs: nounwind ssp uwtable
+define void @foo_4float(i32 %u) #0 {
+; CHECK-LABEL: @foo_4float(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[U_ADDR:%.*]] = alloca i32, align 4
+; CHECK-NEXT: store i32 [[U:%.*]], i32* [[U_ADDR]], align 4
+; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[U]], 4
+; CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[MUL]] to i64
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2000 x float], [2000 x float]* @C, i32 0, i64 [[IDXPROM]]
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [2000 x float], [2000 x float]* @D, i32 0, i64 [[IDXPROM]]
+; CHECK-NEXT: [[ADD11:%.*]] = add nsw i32 [[MUL]], 1
+; CHECK-NEXT: [[IDXPROM12:%.*]] = sext i32 [[ADD11]] to i64
+; CHECK-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds [2000 x float], [2000 x float]* @C, i32 0, i64 [[IDXPROM12]]
+; CHECK-NEXT: [[ARRAYIDX17:%.*]] = getelementptr inbounds [2000 x float], [2000 x float]* @D, i32 0, i64 [[IDXPROM12]]
+; CHECK-NEXT: [[ADD24:%.*]] = add nsw i32 [[MUL]], 2
+; CHECK-NEXT: [[IDXPROM25:%.*]] = sext i32 [[ADD24]] to i64
+; CHECK-NEXT: [[ARRAYIDX26:%.*]] = getelementptr inbounds [2000 x float], [2000 x float]* @C, i32 0, i64 [[IDXPROM25]]
+; CHECK-NEXT: [[ARRAYIDX30:%.*]] = getelementptr inbounds [2000 x float], [2000 x float]* @D, i32 0, i64 [[IDXPROM25]]
+; CHECK-NEXT: [[ADD37:%.*]] = add nsw i32 [[MUL]], 3
+; CHECK-NEXT: [[IDXPROM38:%.*]] = sext i32 [[ADD37]] to i64
+; CHECK-NEXT: [[ARRAYIDX39:%.*]] = getelementptr inbounds [2000 x float], [2000 x float]* @C, i32 0, i64 [[IDXPROM38]]
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast float* [[ARRAYIDX]] to <4 x float>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* [[TMP0]], align 4
+; CHECK-NEXT: [[ARRAYIDX43:%.*]] = getelementptr inbounds [2000 x float], [2000 x float]* @D, i32 0, i64 [[IDXPROM38]]
+; CHECK-NEXT: [[TMP2:%.*]] = bitcast float* [[ARRAYIDX4]] to <4 x float>*
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x float>, <4 x float>* [[TMP2]], align 4
+; CHECK-NEXT: [[TMP4:%.*]] = fadd <4 x float> [[TMP1]], [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast float* [[ARRAYIDX]] to <4 x float>*
+; CHECK-NEXT: store <4 x float> [[TMP4]], <4 x float>* [[TMP5]], align 4
+; CHECK-NEXT: ret void
+;
+entry:
+ %u.addr = alloca i32, align 4
+ store i32 %u, i32* %u.addr, align 4
+ %mul = mul nsw i32 %u, 4
+ %idxprom = sext i32 %mul to i64
+ %arrayidx = getelementptr inbounds [2000 x float], [2000 x float]* @C, i32 0, i64 %idxprom
+ %0 = load float, float* %arrayidx, align 4
+ %arrayidx4 = getelementptr inbounds [2000 x float], [2000 x float]* @D, i32 0, i64 %idxprom
+ %1 = load float, float* %arrayidx4, align 4
+ %add5 = fadd float %0, %1
+ store float %add5, float* %arrayidx, align 4
+ %add11 = add nsw i32 %mul, 1
+ %idxprom12 = sext i32 %add11 to i64
+ %arrayidx13 = getelementptr inbounds [2000 x float], [2000 x float]* @C, i32 0, i64 %idxprom12
+ %2 = load float, float* %arrayidx13, align 4
+ %arrayidx17 = getelementptr inbounds [2000 x float], [2000 x float]* @D, i32 0, i64 %idxprom12
+ %3 = load float, float* %arrayidx17, align 4
+ %add18 = fadd float %2, %3
+ store float %add18, float* %arrayidx13, align 4
+ %add24 = add nsw i32 %mul, 2
+ %idxprom25 = sext i32 %add24 to i64
+ %arrayidx26 = getelementptr inbounds [2000 x float], [2000 x float]* @C, i32 0, i64 %idxprom25
+ %4 = load float, float* %arrayidx26, align 4
+ %arrayidx30 = getelementptr inbounds [2000 x float], [2000 x float]* @D, i32 0, i64 %idxprom25
+ %5 = load float, float* %arrayidx30, align 4
+ %add31 = fadd float %4, %5
+ store float %add31, float* %arrayidx26, align 4
+ %add37 = add nsw i32 %mul, 3
+ %idxprom38 = sext i32 %add37 to i64
+ %arrayidx39 = getelementptr inbounds [2000 x float], [2000 x float]* @C, i32 0, i64 %idxprom38
+ %6 = load float, float* %arrayidx39, align 4
+ %arrayidx43 = getelementptr inbounds [2000 x float], [2000 x float]* @D, i32 0, i64 %idxprom38
+ %7 = load float, float* %arrayidx43, align 4
+ %add44 = fadd float %6, %7
+ store float %add44, float* %arrayidx39, align 4
+ ret void
+}
+
+; Similar to the previous tests, but now we are dealing with AddRec SCEV.
+; Function Attrs: nounwind ssp uwtable
+define i32 @foo_loop(double* %A, i32 %n) #0 {
+; CHECK-LABEL: @foo_loop(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[A_ADDR:%.*]] = alloca double*, align 8
+; CHECK-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
+; CHECK-NEXT: [[SUM:%.*]] = alloca double, align 8
+; CHECK-NEXT: [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT: store double* [[A:%.*]], double** [[A_ADDR]], align 8
+; CHECK-NEXT: store i32 [[N:%.*]], i32* [[N_ADDR]], align 4
+; CHECK-NEXT: store double 0.000000e+00, double* [[SUM]], align 8
+; CHECK-NEXT: store i32 0, i32* [[I]], align 4
+; CHECK-NEXT: [[CMP1:%.*]] = icmp slt i32 0, [[N]]
+; CHECK-NEXT: br i1 [[CMP1]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END:%.*]]
+; CHECK: for.body.lr.ph:
+; CHECK-NEXT: br label [[FOR_BODY:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: [[TMP0:%.*]] = phi i32 [ 0, [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[TMP1:%.*]] = phi double [ 0.000000e+00, [[FOR_BODY_LR_PH]] ], [ [[ADD7:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP0]], 2
+; CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[MUL]] to i64
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[A]], i64 [[IDXPROM]]
+; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[MUL]], 1
+; CHECK-NEXT: [[IDXPROM3:%.*]] = sext i32 [[ADD]] to i64
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds double, double* [[A]], i64 [[IDXPROM3]]
+; CHECK-NEXT: [[TMP2:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[TMP2]], align 8
+; CHECK-NEXT: [[TMP4:%.*]] = fmul <2 x double> <double 7.000000e+00, double 7.000000e+00>, [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x double> [[TMP4]], i32 0
+; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x double> [[TMP4]], i32 1
+; CHECK-NEXT: [[ADD6:%.*]] = fadd double [[TMP5]], [[TMP6]]
+; CHECK-NEXT: [[ADD7]] = fadd double [[TMP1]], [[ADD6]]
+; CHECK-NEXT: store double [[ADD7]], double* [[SUM]], align 8
+; CHECK-NEXT: [[INC]] = add nsw i32 [[TMP0]], 1
+; CHECK-NEXT: store i32 [[INC]], i32* [[I]], align 4
+; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[INC]], [[N]]
+; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]]
+; CHECK: for.cond.for.end_crit_edge:
+; CHECK-NEXT: [[SPLIT:%.*]] = phi double [ [[ADD7]], [[FOR_BODY]] ]
+; CHECK-NEXT: br label [[FOR_END]]
+; CHECK: for.end:
+; CHECK-NEXT: [[DOTLCSSA:%.*]] = phi double [ [[SPLIT]], [[FOR_COND_FOR_END_CRIT_EDGE]] ], [ 0.000000e+00, [[ENTRY:%.*]] ]
+; CHECK-NEXT: [[CONV:%.*]] = fptosi double [[DOTLCSSA]] to i32
+; CHECK-NEXT: ret i32 [[CONV]]
+;
+entry:
+ %A.addr = alloca double*, align 8
+ %n.addr = alloca i32, align 4
+ %sum = alloca double, align 8
+ %i = alloca i32, align 4
+ store double* %A, double** %A.addr, align 8
+ store i32 %n, i32* %n.addr, align 4
+ store double 0.000000e+00, double* %sum, align 8
+ store i32 0, i32* %i, align 4
+ %cmp1 = icmp slt i32 0, %n
+ br i1 %cmp1, label %for.body.lr.ph, label %for.end
+
+for.body.lr.ph: ; preds = %entry
+ br label %for.body
+
+for.body: ; preds = %for.body.lr.ph, %for.body
+ %0 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.body ]
+ %1 = phi double [ 0.000000e+00, %for.body.lr.ph ], [ %add7, %for.body ]
+ %mul = mul nsw i32 %0, 2
+ %idxprom = sext i32 %mul to i64
+ %arrayidx = getelementptr inbounds double, double* %A, i64 %idxprom
+ %2 = load double, double* %arrayidx, align 8
+ %mul1 = fmul double 7.000000e+00, %2
+ %add = add nsw i32 %mul, 1
+ %idxprom3 = sext i32 %add to i64
+ %arrayidx4 = getelementptr inbounds double, double* %A, i64 %idxprom3
+ %3 = load double, double* %arrayidx4, align 8
+ %mul5 = fmul double 7.000000e+00, %3
+ %add6 = fadd double %mul1, %mul5
+ %add7 = fadd double %1, %add6
+ store double %add7, double* %sum, align 8
+ %inc = add nsw i32 %0, 1
+ store i32 %inc, i32* %i, align 4
+ %cmp = icmp slt i32 %inc, %n
+ br i1 %cmp, label %for.body, label %for.cond.for.end_crit_edge
+
+for.cond.for.end_crit_edge: ; preds = %for.body
+ %split = phi double [ %add7, %for.body ]
+ br label %for.end
+
+for.end: ; preds = %for.cond.for.end_crit_edge, %entry
+ %.lcssa = phi double [ %split, %for.cond.for.end_crit_edge ], [ 0.000000e+00, %entry ]
+ %conv = fptosi double %.lcssa to i32
+ ret i32 %conv
+}
+
+; Similar to foo_2double but with a non-power-of-2 factor and potential
+; wrapping (both indices wrap or both don't in the same time)
+; Function Attrs: nounwind ssp uwtable
+define void @foo_2double_non_power_of_2(i32 %u) #0 {
+; CHECK-LABEL: @foo_2double_non_power_of_2(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[U_ADDR:%.*]] = alloca i32, align 4
+; CHECK-NEXT: store i32 [[U:%.*]], i32* [[U_ADDR]], align 4
+; CHECK-NEXT: [[MUL:%.*]] = mul i32 [[U]], 6
+; CHECK-NEXT: [[ADD6:%.*]] = add i32 [[MUL]], 6
+; CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[ADD6]] to i64
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 [[IDXPROM]]
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 [[IDXPROM]]
+; CHECK-NEXT: [[ADD7:%.*]] = add i32 [[MUL]], 7
+; CHECK-NEXT: [[IDXPROM12:%.*]] = sext i32 [[ADD7]] to i64
+; CHECK-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 [[IDXPROM12]]
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
+; CHECK-NEXT: [[ARRAYIDX17:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 [[IDXPROM12]]
+; CHECK-NEXT: [[TMP2:%.*]] = bitcast double* [[ARRAYIDX4]] to <2 x double>*
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[TMP2]], align 8
+; CHECK-NEXT: [[TMP4:%.*]] = fadd <2 x double> [[TMP1]], [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP4]], <2 x double>* [[TMP5]], align 8
+; CHECK-NEXT: ret void
+;
+entry:
+ %u.addr = alloca i32, align 4
+ store i32 %u, i32* %u.addr, align 4
+ %mul = mul i32 %u, 6
+ %add6 = add i32 %mul, 6
+ %idxprom = sext i32 %add6 to i64
+ %arrayidx = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 %idxprom
+ %0 = load double, double* %arrayidx, align 8
+ %arrayidx4 = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 %idxprom
+ %1 = load double, double* %arrayidx4, align 8
+ %add5 = fadd double %0, %1
+ store double %add5, double* %arrayidx, align 8
+ %add7 = add i32 %mul, 7
+ %idxprom12 = sext i32 %add7 to i64
+ %arrayidx13 = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 %idxprom12
+ %2 = load double, double* %arrayidx13, align 8
+ %arrayidx17 = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 %idxprom12
+ %3 = load double, double* %arrayidx17, align 8
+ %add18 = fadd double %2, %3
+ store double %add18, double* %arrayidx13, align 8
+ ret void
+}
+
+; Similar to foo_2double_non_power_of_2 but with zext's instead of sext's
+; Function Attrs: nounwind ssp uwtable
+define void @foo_2double_non_power_of_2_zext(i32 %u) #0 {
+; CHECK-LABEL: @foo_2double_non_power_of_2_zext(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[U_ADDR:%.*]] = alloca i32, align 4
+; CHECK-NEXT: store i32 [[U:%.*]], i32* [[U_ADDR]], align 4
+; CHECK-NEXT: [[MUL:%.*]] = mul i32 [[U]], 6
+; CHECK-NEXT: [[ADD6:%.*]] = add i32 [[MUL]], 6
+; CHECK-NEXT: [[IDXPROM:%.*]] = zext i32 [[ADD6]] to i64
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 [[IDXPROM]]
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 [[IDXPROM]]
+; CHECK-NEXT: [[ADD7:%.*]] = add i32 [[MUL]], 7
+; CHECK-NEXT: [[IDXPROM12:%.*]] = zext i32 [[ADD7]] to i64
+; CHECK-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 [[IDXPROM12]]
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
+; CHECK-NEXT: [[ARRAYIDX17:%.*]] = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 [[IDXPROM12]]
+; CHECK-NEXT: [[TMP2:%.*]] = bitcast double* [[ARRAYIDX4]] to <2 x double>*
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[TMP2]], align 8
+; CHECK-NEXT: [[TMP4:%.*]] = fadd <2 x double> [[TMP1]], [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP4]], <2 x double>* [[TMP5]], align 8
+; CHECK-NEXT: ret void
+;
+entry:
+ %u.addr = alloca i32, align 4
+ store i32 %u, i32* %u.addr, align 4
+ %mul = mul i32 %u, 6
+ %add6 = add i32 %mul, 6
+ %idxprom = zext i32 %add6 to i64
+ %arrayidx = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 %idxprom
+ %0 = load double, double* %arrayidx, align 8
+ %arrayidx4 = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 %idxprom
+ %1 = load double, double* %arrayidx4, align 8
+ %add5 = fadd double %0, %1
+ store double %add5, double* %arrayidx, align 8
+ %add7 = add i32 %mul, 7
+ %idxprom12 = zext i32 %add7 to i64
+ %arrayidx13 = getelementptr inbounds [2000 x double], [2000 x double]* @A, i32 0, i64 %idxprom12
+ %2 = load double, double* %arrayidx13, align 8
+ %arrayidx17 = getelementptr inbounds [2000 x double], [2000 x double]* @B, i32 0, i64 %idxprom12
+ %3 = load double, double* %arrayidx17, align 8
+ %add18 = fadd double %2, %3
+ store double %add18, double* %arrayidx13, align 8
+ ret void
+}
+
+; Similar to foo_2double_non_power_of_2, but now we are dealing with AddRec SCEV.
+; Alternatively, this is like foo_loop, but with a non-power-of-2 factor and
+; potential wrapping (both indices wrap or both don't in the same time)
+; Function Attrs: nounwind ssp uwtable
+define i32 @foo_loop_non_power_of_2(double* %A, i32 %n) #0 {
+; CHECK-LABEL: @foo_loop_non_power_of_2(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[A_ADDR:%.*]] = alloca double*, align 8
+; CHECK-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
+; CHECK-NEXT: [[SUM:%.*]] = alloca double, align 8
+; CHECK-NEXT: [[I:%.*]] = alloca i32, align 4
+; CHECK-NEXT: store double* [[A:%.*]], double** [[A_ADDR]], align 8
+; CHECK-NEXT: store i32 [[N:%.*]], i32* [[N_ADDR]], align 4
+; CHECK-NEXT: store double 0.000000e+00, double* [[SUM]], align 8
+; CHECK-NEXT: store i32 0, i32* [[I]], align 4
+; CHECK-NEXT: [[CMP1:%.*]] = icmp slt i32 0, [[N]]
+; CHECK-NEXT: br i1 [[CMP1]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END:%.*]]
+; CHECK: for.body.lr.ph:
+; CHECK-NEXT: br label [[FOR_BODY:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: [[TMP0:%.*]] = phi i32 [ 0, [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[TMP1:%.*]] = phi double [ 0.000000e+00, [[FOR_BODY_LR_PH]] ], [ [[ADD7:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[MUL:%.*]] = mul i32 [[TMP0]], 12
+; CHECK-NEXT: [[ADD_5:%.*]] = add i32 [[MUL]], 5
+; CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[ADD_5]] to i64
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[A]], i64 [[IDXPROM]]
+; CHECK-NEXT: [[ADD_6:%.*]] = add i32 [[MUL]], 6
+; CHECK-NEXT: [[IDXPROM3:%.*]] = sext i32 [[ADD_6]] to i64
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds double, double* [[A]], i64 [[IDXPROM3]]
+; CHECK-NEXT: [[TMP2:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[TMP2]], align 8
+; CHECK-NEXT: [[TMP4:%.*]] = fmul <2 x double> <double 7.000000e+00, double 7.000000e+00>, [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x double> [[TMP4]], i32 0
+; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x double> [[TMP4]], i32 1
+; CHECK-NEXT: [[ADD6:%.*]] = fadd double [[TMP5]], [[TMP6]]
+; CHECK-NEXT: [[ADD7]] = fadd double [[TMP1]], [[ADD6]]
+; CHECK-NEXT: store double [[ADD7]], double* [[SUM]], align 8
+; CHECK-NEXT: [[INC]] = add i32 [[TMP0]], 1
+; CHECK-NEXT: store i32 [[INC]], i32* [[I]], align 4
+; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[INC]], [[N]]
+; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]]
+; CHECK: for.cond.for.end_crit_edge:
+; CHECK-NEXT: [[SPLIT:%.*]] = phi double [ [[ADD7]], [[FOR_BODY]] ]
+; CHECK-NEXT: br label [[FOR_END]]
+; CHECK: for.end:
+; CHECK-NEXT: [[DOTLCSSA:%.*]] = phi double [ [[SPLIT]], [[FOR_COND_FOR_END_CRIT_EDGE]] ], [ 0.000000e+00, [[ENTRY:%.*]] ]
+; CHECK-NEXT: [[CONV:%.*]] = fptosi double [[DOTLCSSA]] to i32
+; CHECK-NEXT: ret i32 [[CONV]]
+;
+entry:
+ %A.addr = alloca double*, align 8
+ %n.addr = alloca i32, align 4
+ %sum = alloca double, align 8
+ %i = alloca i32, align 4
+ store double* %A, double** %A.addr, align 8
+ store i32 %n, i32* %n.addr, align 4
+ store double 0.000000e+00, double* %sum, align 8
+ store i32 0, i32* %i, align 4
+ %cmp1 = icmp slt i32 0, %n
+ br i1 %cmp1, label %for.body.lr.ph, label %for.end
+
+for.body.lr.ph: ; preds = %entry
+ br label %for.body
+
+for.body: ; preds = %for.body.lr.ph, %for.body
+ %0 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.body ]
+ %1 = phi double [ 0.000000e+00, %for.body.lr.ph ], [ %add7, %for.body ]
+ %mul = mul i32 %0, 12
+ %add.5 = add i32 %mul, 5
+ %idxprom = sext i32 %add.5 to i64
+ %arrayidx = getelementptr inbounds double, double* %A, i64 %idxprom
+ %2 = load double, double* %arrayidx, align 8
+ %mul1 = fmul double 7.000000e+00, %2
+ %add.6 = add i32 %mul, 6
+ %idxprom3 = sext i32 %add.6 to i64
+ %arrayidx4 = getelementptr inbounds double, double* %A, i64 %idxprom3
+ %3 = load double, double* %arrayidx4, align 8
+ %mul5 = fmul double 7.000000e+00, %3
+ %add6 = fadd double %mul1, %mul5
+ %add7 = fadd double %1, %add6
+ store double %add7, double* %sum, align 8
+ %inc = add i32 %0, 1
+ store i32 %inc, i32* %i, align 4
+ %cmp = icmp slt i32 %inc, %n
+ br i1 %cmp, label %for.body, label %for.cond.for.end_crit_edge
+
+for.cond.for.end_crit_edge: ; preds = %for.body
+ %split = phi double [ %add7, %for.body ]
+ br label %for.end
+
+for.end: ; preds = %for.cond.for.end_crit_edge, %entry
+ %.lcssa = phi double [ %split, %for.cond.for.end_crit_edge ], [ 0.000000e+00, %entry ]
+ %conv = fptosi double %.lcssa to i32
+ ret i32 %conv
+}
+
+; This is generated by `clang -std=c11 -Wpedantic -Wall -O3 main.c -S -o - -emit-llvm`
+; with !{!"clang version 7.0.0 (trunk 337339) (llvm/trunk 337344)"} and stripping off
+; the !tbaa metadata nodes to fit the rest of the test file, where `cat main.c` is:
+;
+; double bar(double *a, unsigned n) {
+; double x = 0.0;
+; double y = 0.0;
+; for (unsigned i = 0; i < n; i += 2) {
+; x += a[i];
+; y += a[i + 1];
+; }
+; return x * y;
+; }
+;
+; The resulting IR is similar to @foo_loop, but with zext's instead of sext's.
+;
+; Make sure we are able to vectorize this from now on:
+;
+define double @bar(double* nocapture readonly %a, i32 %n) local_unnamed_addr #0 {
+; CHECK-LABEL: @bar(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[CMP15:%.*]] = icmp eq i32 [[N:%.*]], 0
+; CHECK-NEXT: br i1 [[CMP15]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY:%.*]]
+; CHECK: for.cond.cleanup:
+; CHECK-NEXT: [[TMP0:%.*]] = phi <2 x double> [ zeroinitializer, [[ENTRY:%.*]] ], [ [[TMP6:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[TMP1:%.*]] = extractelement <2 x double> [[TMP0]], i32 0
+; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x double> [[TMP0]], i32 1
+; CHECK-NEXT: [[MUL:%.*]] = fmul double [[TMP1]], [[TMP2]]
+; CHECK-NEXT: ret double [[MUL]]
+; CHECK: for.body:
+; CHECK-NEXT: [[I_018:%.*]] = phi i32 [ [[ADD5:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT: [[TMP3:%.*]] = phi <2 x double> [ [[TMP6]], [[FOR_BODY]] ], [ zeroinitializer, [[ENTRY]] ]
+; CHECK-NEXT: [[IDXPROM:%.*]] = zext i32 [[I_018]] to i64
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[A:%.*]], i64 [[IDXPROM]]
+; CHECK-NEXT: [[ADD1:%.*]] = or i32 [[I_018]], 1
+; CHECK-NEXT: [[IDXPROM2:%.*]] = zext i32 [[ADD1]] to i64
+; CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds double, double* [[A]], i64 [[IDXPROM2]]
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: [[TMP5:%.*]] = load <2 x double>, <2 x double>* [[TMP4]], align 8
+; CHECK-NEXT: [[TMP6]] = fadd <2 x double> [[TMP3]], [[TMP5]]
+; CHECK-NEXT: [[ADD5]] = add i32 [[I_018]], 2
+; CHECK-NEXT: [[CMP:%.*]] = icmp ult i32 [[ADD5]], [[N]]
+; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP]]
+;
+entry:
+ %cmp15 = icmp eq i32 %n, 0
+ br i1 %cmp15, label %for.cond.cleanup, label %for.body
+
+for.cond.cleanup: ; preds = %for.body, %entry
+ %x.0.lcssa = phi double [ 0.000000e+00, %entry ], [ %add, %for.body ]
+ %y.0.lcssa = phi double [ 0.000000e+00, %entry ], [ %add4, %for.body ]
+ %mul = fmul double %x.0.lcssa, %y.0.lcssa
+ ret double %mul
+
+for.body: ; preds = %entry, %for.body
+ %i.018 = phi i32 [ %add5, %for.body ], [ 0, %entry ]
+ %y.017 = phi double [ %add4, %for.body ], [ 0.000000e+00, %entry ]
+ %x.016 = phi double [ %add, %for.body ], [ 0.000000e+00, %entry ]
+ %idxprom = zext i32 %i.018 to i64
+ %arrayidx = getelementptr inbounds double, double* %a, i64 %idxprom
+ %0 = load double, double* %arrayidx, align 8
+ %add = fadd double %x.016, %0
+ %add1 = or i32 %i.018, 1
+ %idxprom2 = zext i32 %add1 to i64
+ %arrayidx3 = getelementptr inbounds double, double* %a, i64 %idxprom2
+ %1 = load double, double* %arrayidx3, align 8
+ %add4 = fadd double %y.017, %1
+ %add5 = add i32 %i.018, 2
+ %cmp = icmp ult i32 %add5, %n
+ br i1 %cmp, label %for.body, label %for.cond.cleanup
+}
+
+attributes #0 = { nounwind ssp uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
+
+!llvm.ident = !{!0}
+
+!0 = !{!"clang version 3.5.0 "}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/continue_vectorizing.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/continue_vectorizing.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/continue_vectorizing.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/continue_vectorizing.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,50 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+; We will keep trying to vectorize the basic block even we already find vectorized store.
+define void @test1(double* %a, double* %b, double* %c, double* %d) {
+; CHECK-LABEL: @test1(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds double, double* [[A:%.*]], i64 1
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[A]] to <2 x double>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds double, double* [[B:%.*]], i64 1
+; CHECK-NEXT: [[TMP2:%.*]] = bitcast double* [[B]] to <2 x double>*
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[TMP2]], align 8
+; CHECK-NEXT: [[TMP4:%.*]] = fmul <2 x double> [[TMP1]], [[TMP3]]
+; CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds double, double* [[C:%.*]], i64 1
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast double* [[C]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP4]], <2 x double>* [[TMP5]], align 8
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast double* [[A]] to <4 x i32>*
+; CHECK-NEXT: [[TMP7:%.*]] = load <4 x i32>, <4 x i32>* [[TMP6]], align 8
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast double* [[B]] to <4 x i32>*
+; CHECK-NEXT: [[TMP9:%.*]] = load <4 x i32>, <4 x i32>* [[TMP8]], align 8
+; CHECK-NEXT: [[TMP10:%.*]] = mul <4 x i32> [[TMP7]], [[TMP9]]
+; CHECK-NEXT: [[TMP11:%.*]] = bitcast double* [[D:%.*]] to <4 x i32>*
+; CHECK-NEXT: store <4 x i32> [[TMP10]], <4 x i32>* [[TMP11]], align 8
+; CHECK-NEXT: ret void
+;
+entry:
+ %i0 = load double, double* %a, align 8
+ %i1 = load double, double* %b, align 8
+ %mul = fmul double %i0, %i1
+ %arrayidx3 = getelementptr inbounds double, double* %a, i64 1
+ %i3 = load double, double* %arrayidx3, align 8
+ %arrayidx4 = getelementptr inbounds double, double* %b, i64 1
+ %i4 = load double, double* %arrayidx4, align 8
+ %mul5 = fmul double %i3, %i4
+ store double %mul, double* %c, align 8
+ %arrayidx5 = getelementptr inbounds double, double* %c, i64 1
+ store double %mul5, double* %arrayidx5, align 8
+ %0 = bitcast double* %a to <4 x i32>*
+ %1 = load <4 x i32>, <4 x i32>* %0, align 8
+ %2 = bitcast double* %b to <4 x i32>*
+ %3 = load <4 x i32>, <4 x i32>* %2, align 8
+ %4 = mul <4 x i32> %1, %3
+ %5 = bitcast double* %d to <4 x i32>*
+ store <4 x i32> %4, <4 x i32>* %5, align 8
+ ret void
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_7zip.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_7zip.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_7zip.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_7zip.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,65 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+%struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334 = type { %struct._CLzmaProps.0.27.54.81.102.123.144.165.180.195.228.258.333, i16*, i8*, i8*, i32, i32, i64, i64, i32, i32, i32, [4 x i32], i32, i32, i32, i32, i32, [20 x i8] }
+%struct._CLzmaProps.0.27.54.81.102.123.144.165.180.195.228.258.333 = type { i32, i32, i32, i32 }
+
+define fastcc void @LzmaDec_DecodeReal2(%struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* %p) {
+; CHECK-LABEL: @LzmaDec_DecodeReal2(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[RANGE20_I:%.*]] = getelementptr inbounds [[STRUCT_CLZMADEC_1_28_55_82_103_124_145_166_181_196_229_259_334:%.*]], %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* [[P:%.*]], i64 0, i32 4
+; CHECK-NEXT: [[CODE21_I:%.*]] = getelementptr inbounds [[STRUCT_CLZMADEC_1_28_55_82_103_124_145_166_181_196_229_259_334]], %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* [[P]], i64 0, i32 5
+; CHECK-NEXT: br label [[DO_BODY66_I:%.*]]
+; CHECK: do.body66.i:
+; CHECK-NEXT: [[RANGE_2_I:%.*]] = phi i32 [ [[RANGE_4_I:%.*]], [[DO_COND_I:%.*]] ], [ undef, [[ENTRY:%.*]] ]
+; CHECK-NEXT: [[CODE_2_I:%.*]] = phi i32 [ [[CODE_4_I:%.*]], [[DO_COND_I]] ], [ undef, [[ENTRY]] ]
+; CHECK-NEXT: [[DOTRANGE_2_I:%.*]] = select i1 undef, i32 undef, i32 [[RANGE_2_I]]
+; CHECK-NEXT: [[DOTCODE_2_I:%.*]] = select i1 undef, i32 undef, i32 [[CODE_2_I]]
+; CHECK-NEXT: br i1 undef, label [[DO_COND_I]], label [[IF_ELSE_I:%.*]]
+; CHECK: if.else.i:
+; CHECK-NEXT: [[SUB91_I:%.*]] = sub i32 [[DOTRANGE_2_I]], undef
+; CHECK-NEXT: [[SUB92_I:%.*]] = sub i32 [[DOTCODE_2_I]], undef
+; CHECK-NEXT: br label [[DO_COND_I]]
+; CHECK: do.cond.i:
+; CHECK-NEXT: [[RANGE_4_I]] = phi i32 [ [[SUB91_I]], [[IF_ELSE_I]] ], [ undef, [[DO_BODY66_I]] ]
+; CHECK-NEXT: [[CODE_4_I]] = phi i32 [ [[SUB92_I]], [[IF_ELSE_I]] ], [ [[DOTCODE_2_I]], [[DO_BODY66_I]] ]
+; CHECK-NEXT: br i1 undef, label [[DO_BODY66_I]], label [[DO_END1006_I:%.*]]
+; CHECK: do.end1006.i:
+; CHECK-NEXT: [[DOTRANGE_4_I:%.*]] = select i1 undef, i32 undef, i32 [[RANGE_4_I]]
+; CHECK-NEXT: [[DOTCODE_4_I:%.*]] = select i1 undef, i32 undef, i32 [[CODE_4_I]]
+; CHECK-NEXT: store i32 [[DOTRANGE_4_I]], i32* [[RANGE20_I]], align 4
+; CHECK-NEXT: store i32 [[DOTCODE_4_I]], i32* [[CODE21_I]], align 4
+; CHECK-NEXT: ret void
+;
+entry:
+ %range20.i = getelementptr inbounds %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334, %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* %p, i64 0, i32 4
+ %code21.i = getelementptr inbounds %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334, %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* %p, i64 0, i32 5
+ br label %do.body66.i
+
+do.body66.i: ; preds = %do.cond.i, %entry
+ %range.2.i = phi i32 [ %range.4.i, %do.cond.i ], [ undef, %entry ]
+ %code.2.i = phi i32 [ %code.4.i, %do.cond.i ], [ undef, %entry ]
+ %.range.2.i = select i1 undef, i32 undef, i32 %range.2.i
+ %.code.2.i = select i1 undef, i32 undef, i32 %code.2.i
+ br i1 undef, label %do.cond.i, label %if.else.i
+
+if.else.i: ; preds = %do.body66.i
+ %sub91.i = sub i32 %.range.2.i, undef
+ %sub92.i = sub i32 %.code.2.i, undef
+ br label %do.cond.i
+
+do.cond.i: ; preds = %if.else.i, %do.body66.i
+ %range.4.i = phi i32 [ %sub91.i, %if.else.i ], [ undef, %do.body66.i ]
+ %code.4.i = phi i32 [ %sub92.i, %if.else.i ], [ %.code.2.i, %do.body66.i ]
+ br i1 undef, label %do.body66.i, label %do.end1006.i
+
+do.end1006.i: ; preds = %do.cond.i
+ %.range.4.i = select i1 undef, i32 undef, i32 %range.4.i
+ %.code.4.i = select i1 undef, i32 undef, i32 %code.4.i
+ store i32 %.range.4.i, i32* %range20.i, align 4
+ store i32 %.code.4.i, i32* %code21.i, align 4
+ ret void
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_binaryop.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_binaryop.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_binaryop.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_binaryop.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,64 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-darwin13.3.0"
+
+ at a = common global double 0.000000e+00, align 8
+
+define i32 @fn1() {
+; CHECK-LABEL: @fn1(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[INIT:%.*]] = load double, double* @a, align 8
+; CHECK-NEXT: br label [[LOOP:%.*]]
+; CHECK: loop:
+; CHECK-NEXT: [[PHI:%.*]] = phi double [ [[ADD2:%.*]], [[LOOP]] ], [ [[INIT]], [[ENTRY:%.*]] ]
+; CHECK-NEXT: [[POSTADD1_PHI:%.*]] = phi double [ [[POSTADD1:%.*]], [[LOOP]] ], [ [[INIT]], [[ENTRY]] ]
+; CHECK-NEXT: [[POSTADD2_PHI:%.*]] = phi double [ [[POSTADD2:%.*]], [[LOOP]] ], [ [[INIT]], [[ENTRY]] ]
+; CHECK-NEXT: [[ADD1:%.*]] = fadd double [[POSTADD1_PHI]], undef
+; CHECK-NEXT: [[ADD2]] = fadd double [[POSTADD2_PHI]], [[PHI]]
+; CHECK-NEXT: [[MUL2:%.*]] = fmul double [[ADD2]], 0.000000e+00
+; CHECK-NEXT: [[BINARYOP_B:%.*]] = fadd double [[POSTADD1_PHI]], [[MUL2]]
+; CHECK-NEXT: [[MUL1:%.*]] = fmul double [[ADD1]], 0.000000e+00
+; CHECK-NEXT: [[TMP:%.*]] = fadd double [[POSTADD2_PHI]], 0.000000e+00
+; CHECK-NEXT: [[BINARY_V:%.*]] = fadd double [[MUL1]], [[BINARYOP_B]]
+; CHECK-NEXT: [[POSTADD1]] = fadd double [[BINARY_V]], 0.000000e+00
+; CHECK-NEXT: [[POSTADD2]] = fadd double [[TMP]], 1.000000e+00
+; CHECK-NEXT: [[TOBOOL:%.*]] = fcmp une double [[POSTADD1]], 0.000000e+00
+; CHECK-NEXT: br i1 [[TOBOOL]], label [[EXIT:%.*]], label [[LOOP]]
+; CHECK: exit:
+; CHECK-NEXT: ret i32 1
+;
+entry:
+ %init = load double, double* @a, align 8
+ br label %loop
+
+loop:
+ %phi = phi double [ %add2, %loop ], [ %init, %entry ]
+ %postadd1_phi = phi double [ %postadd1, %loop ], [ %init, %entry ]
+ %postadd2_phi = phi double [ %postadd2, %loop ], [ %init, %entry ]
+ %add1 = fadd double %postadd1_phi, undef
+ %add2 = fadd double %postadd2_phi, %phi
+ %mul2 = fmul double %add2, 0.000000e+00
+ %binaryop_B = fadd double %postadd1_phi, %mul2
+ %mul1 = fmul double %add1, 0.000000e+00
+ %tmp = fadd double %postadd2_phi, 0.000000e+00
+
+ ; tryToVectorize() starts with this binary instruction.
+ ; At the same time vectorization wraps around the loop, vectorizes
+ ; postadd1/2 and eventually binary_V and tmp. So binary_V itself is replaced
+ ; with a vector instruction.
+ ; The SLPVectorizer crashed because it tried to use binary_V
+ ; after vectorization to re-arrange instructions.
+ %binary_V = fadd double %mul1, %binaryop_B
+
+ %postadd1 = fadd double %binary_V, 0.000000e+00
+ %postadd2 = fadd double %tmp, 1.000000e+00
+ %tobool = fcmp une double %postadd1, 0.000000e+00
+ br i1 %tobool, label %exit, label %loop
+
+exit:
+ ret i32 1
+}
+
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_bullet.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_bullet.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_bullet.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_bullet.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,213 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+%"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960" = type { i32, i32 }
+
+define void @_ZN23btGeneric6DofConstraint8getInfo1EPN17btTypedConstraint17btConstraintInfo1E(%"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960"* nocapture %info) {
+; CHECK-LABEL: @_ZN23btGeneric6DofConstraint8getInfo1EPN17btTypedConstraint17btConstraintInfo1E(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 undef, label [[IF_ELSE:%.*]], label [[IF_THEN:%.*]]
+; CHECK: if.then:
+; CHECK-NEXT: ret void
+; CHECK: if.else:
+; CHECK-NEXT: [[M_NUMCONSTRAINTROWS4:%.*]] = getelementptr inbounds %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960", %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960"* [[INFO:%.*]], i64 0, i32 0
+; CHECK-NEXT: [[NUB5:%.*]] = getelementptr inbounds %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960", %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960"* [[INFO]], i64 0, i32 1
+; CHECK-NEXT: br i1 undef, label [[LAND_LHS_TRUE_I_1:%.*]], label [[IF_THEN7_1:%.*]]
+; CHECK: land.lhs.true.i.1:
+; CHECK-NEXT: br i1 undef, label [[FOR_INC_1:%.*]], label [[IF_THEN7_1]]
+; CHECK: if.then7.1:
+; CHECK-NEXT: [[INC_1:%.*]] = add nsw i32 0, 1
+; CHECK-NEXT: store i32 [[INC_1]], i32* [[M_NUMCONSTRAINTROWS4]], align 4
+; CHECK-NEXT: [[DEC_1:%.*]] = add nsw i32 6, -1
+; CHECK-NEXT: store i32 [[DEC_1]], i32* [[NUB5]], align 4
+; CHECK-NEXT: br label [[FOR_INC_1]]
+; CHECK: for.inc.1:
+; CHECK-NEXT: [[TMP0:%.*]] = phi i32 [ [[DEC_1]], [[IF_THEN7_1]] ], [ 6, [[LAND_LHS_TRUE_I_1]] ]
+; CHECK-NEXT: [[TMP1:%.*]] = phi i32 [ [[INC_1]], [[IF_THEN7_1]] ], [ 0, [[LAND_LHS_TRUE_I_1]] ]
+; CHECK-NEXT: [[INC_2:%.*]] = add nsw i32 [[TMP1]], 1
+; CHECK-NEXT: store i32 [[INC_2]], i32* [[M_NUMCONSTRAINTROWS4]], align 4
+; CHECK-NEXT: [[DEC_2:%.*]] = add nsw i32 [[TMP0]], -1
+; CHECK-NEXT: store i32 [[DEC_2]], i32* [[NUB5]], align 4
+; CHECK-NEXT: unreachable
+;
+entry:
+ br i1 undef, label %if.else, label %if.then
+
+if.then: ; preds = %entry
+ ret void
+
+if.else: ; preds = %entry
+ %m_numConstraintRows4 = getelementptr inbounds %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960", %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960"* %info, i64 0, i32 0
+ %nub5 = getelementptr inbounds %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960", %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960"* %info, i64 0, i32 1
+ br i1 undef, label %land.lhs.true.i.1, label %if.then7.1
+
+land.lhs.true.i.1: ; preds = %if.else
+ br i1 undef, label %for.inc.1, label %if.then7.1
+
+if.then7.1: ; preds = %land.lhs.true.i.1, %if.else
+ %inc.1 = add nsw i32 0, 1
+ store i32 %inc.1, i32* %m_numConstraintRows4, align 4
+ %dec.1 = add nsw i32 6, -1
+ store i32 %dec.1, i32* %nub5, align 4
+ br label %for.inc.1
+
+for.inc.1: ; preds = %if.then7.1, %land.lhs.true.i.1
+ %0 = phi i32 [ %dec.1, %if.then7.1 ], [ 6, %land.lhs.true.i.1 ]
+ %1 = phi i32 [ %inc.1, %if.then7.1 ], [ 0, %land.lhs.true.i.1 ]
+ %inc.2 = add nsw i32 %1, 1
+ store i32 %inc.2, i32* %m_numConstraintRows4, align 4
+ %dec.2 = add nsw i32 %0, -1
+ store i32 %dec.2, i32* %nub5, align 4
+ unreachable
+}
+
+%class.GIM_TRIANGLE_CALCULATION_CACHE.9.34.69.94.119.144.179.189.264.284.332 = type { float, [3 x %class.btVector3.5.30.65.90.115.140.175.185.260.280.330], [3 x %class.btVector3.5.30.65.90.115.140.175.185.260.280.330], %class.btVector4.7.32.67.92.117.142.177.187.262.282.331, %class.btVector4.7.32.67.92.117.142.177.187.262.282.331, %class.btVector3.5.30.65.90.115.140.175.185.260.280.330, %class.btVector3.5.30.65.90.115.140.175.185.260.280.330, %class.btVector3.5.30.65.90.115.140.175.185.260.280.330, %class.btVector3.5.30.65.90.115.140.175.185.260.280.330, [4 x float], float, float, [4 x float], float, float, [16 x %class.btVector3.5.30.65.90.115.140.175.185.260.280.330], [16 x %class.btVector3.5.30.65.90.115.140.175.185.260.280.330], [16 x %class.btVector3.5.30.65.90.115.140.175.185.260.280.330] }
+%class.btVector3.5.30.65.90.115.140.175.185.260.280.330 = type { [4 x float] }
+%class.btVector4.7.32.67.92.117.142.177.187.262.282.331 = type { %class.btVector3.5.30.65.90.115.140.175.185.260.280.330 }
+
+define void @_ZN30GIM_TRIANGLE_CALCULATION_CACHE18triangle_collisionERK9btVector3S2_S2_fS2_S2_S2_fR25GIM_TRIANGLE_CONTACT_DATA(%class.GIM_TRIANGLE_CALCULATION_CACHE.9.34.69.94.119.144.179.189.264.284.332* %this) {
+; CHECK-LABEL: @_ZN30GIM_TRIANGLE_CALCULATION_CACHE18triangle_collisionERK9btVector3S2_S2_fS2_S2_S2_fR25GIM_TRIANGLE_CONTACT_DATA(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[ARRAYIDX26:%.*]] = getelementptr inbounds [[CLASS_GIM_TRIANGLE_CALCULATION_CACHE_9_34_69_94_119_144_179_189_264_284_332:%.*]], %class.GIM_TRIANGLE_CALCULATION_CACHE.9.34.69.94.119.144.179.189.264.284.332* [[THIS:%.*]], i64 0, i32 2, i64 0, i32 0, i64 1
+; CHECK-NEXT: [[ARRAYIDX36:%.*]] = getelementptr inbounds [[CLASS_GIM_TRIANGLE_CALCULATION_CACHE_9_34_69_94_119_144_179_189_264_284_332]], %class.GIM_TRIANGLE_CALCULATION_CACHE.9.34.69.94.119.144.179.189.264.284.332* [[THIS]], i64 0, i32 2, i64 0, i32 0, i64 2
+; CHECK-NEXT: [[TMP0:%.*]] = load float, float* [[ARRAYIDX36]], align 4
+; CHECK-NEXT: [[ADD587:%.*]] = fadd float undef, undef
+; CHECK-NEXT: [[SUB600:%.*]] = fsub float [[ADD587]], undef
+; CHECK-NEXT: store float [[SUB600]], float* undef, align 4
+; CHECK-NEXT: [[SUB613:%.*]] = fsub float [[ADD587]], [[SUB600]]
+; CHECK-NEXT: store float [[SUB613]], float* [[ARRAYIDX26]], align 4
+; CHECK-NEXT: [[ADD626:%.*]] = fadd float [[TMP0]], undef
+; CHECK-NEXT: [[SUB639:%.*]] = fsub float [[ADD626]], undef
+; CHECK-NEXT: [[SUB652:%.*]] = fsub float [[ADD626]], [[SUB639]]
+; CHECK-NEXT: store float [[SUB652]], float* [[ARRAYIDX36]], align 4
+; CHECK-NEXT: br i1 undef, label [[IF_ELSE1609:%.*]], label [[IF_THEN1595:%.*]]
+; CHECK: if.then1595:
+; CHECK-NEXT: br i1 undef, label [[RETURN:%.*]], label [[FOR_BODY_LR_PH_I_I1702:%.*]]
+; CHECK: for.body.lr.ph.i.i1702:
+; CHECK-NEXT: unreachable
+; CHECK: if.else1609:
+; CHECK-NEXT: unreachable
+; CHECK: return:
+; CHECK-NEXT: ret void
+;
+entry:
+ %arrayidx26 = getelementptr inbounds %class.GIM_TRIANGLE_CALCULATION_CACHE.9.34.69.94.119.144.179.189.264.284.332, %class.GIM_TRIANGLE_CALCULATION_CACHE.9.34.69.94.119.144.179.189.264.284.332* %this, i64 0, i32 2, i64 0, i32 0, i64 1
+ %arrayidx36 = getelementptr inbounds %class.GIM_TRIANGLE_CALCULATION_CACHE.9.34.69.94.119.144.179.189.264.284.332, %class.GIM_TRIANGLE_CALCULATION_CACHE.9.34.69.94.119.144.179.189.264.284.332* %this, i64 0, i32 2, i64 0, i32 0, i64 2
+ %0 = load float, float* %arrayidx36, align 4
+ %add587 = fadd float undef, undef
+ %sub600 = fsub float %add587, undef
+ store float %sub600, float* undef, align 4
+ %sub613 = fsub float %add587, %sub600
+ store float %sub613, float* %arrayidx26, align 4
+ %add626 = fadd float %0, undef
+ %sub639 = fsub float %add626, undef
+ %sub652 = fsub float %add626, %sub639
+ store float %sub652, float* %arrayidx36, align 4
+ br i1 undef, label %if.else1609, label %if.then1595
+
+if.then1595: ; preds = %entry
+ br i1 undef, label %return, label %for.body.lr.ph.i.i1702
+
+for.body.lr.ph.i.i1702: ; preds = %if.then1595
+ unreachable
+
+if.else1609: ; preds = %entry
+ unreachable
+
+return: ; preds = %if.then1595
+ ret void
+}
+
+define void @_Z8dBoxBox2RK9btVector3PKfS1_S1_S3_S1_RS_PfPiiP12dContactGeomiRN36btDiscreteCollisionDetectorInterface6ResultE() {
+; CHECK-LABEL: @_Z8dBoxBox2RK9btVector3PKfS1_S1_S3_S1_RS_PfPiiP12dContactGeomiRN36btDiscreteCollisionDetectorInterface6ResultE(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 undef, label [[RETURN:%.*]], label [[IF_END:%.*]]
+; CHECK: if.end:
+; CHECK-NEXT: br i1 undef, label [[RETURN]], label [[IF_END111:%.*]]
+; CHECK: if.end111:
+; CHECK-NEXT: br i1 undef, label [[RETURN]], label [[IF_END136:%.*]]
+; CHECK: if.end136:
+; CHECK-NEXT: br i1 undef, label [[RETURN]], label [[IF_END162:%.*]]
+; CHECK: if.end162:
+; CHECK-NEXT: br i1 undef, label [[RETURN]], label [[IF_END189:%.*]]
+; CHECK: if.end189:
+; CHECK-NEXT: br i1 undef, label [[RETURN]], label [[IF_END216:%.*]]
+; CHECK: if.end216:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN218:%.*]], label [[IF_END225:%.*]]
+; CHECK: if.then218:
+; CHECK-NEXT: br label [[IF_END225]]
+; CHECK: if.end225:
+; CHECK-NEXT: br i1 undef, label [[RETURN]], label [[IF_END248:%.*]]
+; CHECK: if.end248:
+; CHECK-NEXT: br i1 undef, label [[RETURN]], label [[IF_END304:%.*]]
+; CHECK: if.end304:
+; CHECK-NEXT: br i1 undef, label [[RETURN]], label [[IF_END361:%.*]]
+; CHECK: if.end361:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN370:%.*]], label [[IF_END395:%.*]]
+; CHECK: if.then370:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN374:%.*]], label [[IF_END395]]
+; CHECK: if.then374:
+; CHECK-NEXT: br label [[IF_END395]]
+; CHECK: if.end395:
+; CHECK-NEXT: unreachable
+; CHECK: return:
+; CHECK-NEXT: ret void
+;
+entry:
+ %add8.i2343 = fadd float undef, undef
+ %add8.i2381 = fadd float undef, undef
+ br i1 undef, label %return, label %if.end
+
+if.end: ; preds = %entry
+ br i1 undef, label %return, label %if.end111
+
+if.end111: ; preds = %if.end
+ br i1 undef, label %return, label %if.end136
+
+if.end136: ; preds = %if.end111
+ br i1 undef, label %return, label %if.end162
+
+if.end162: ; preds = %if.end136
+ br i1 undef, label %return, label %if.end189
+
+if.end189: ; preds = %if.end162
+ br i1 undef, label %return, label %if.end216
+
+if.end216: ; preds = %if.end189
+ br i1 undef, label %if.then218, label %if.end225
+
+if.then218: ; preds = %if.end216
+ br label %if.end225
+
+if.end225: ; preds = %if.then218, %if.end216
+ br i1 undef, label %return, label %if.end248
+
+if.end248: ; preds = %if.end225
+ br i1 undef, label %return, label %if.end304
+
+if.end304: ; preds = %if.end248
+ %mul341 = fmul float undef, %add8.i2343
+ %mul344 = fmul float undef, %add8.i2381
+ %sub345 = fsub float %mul341, %mul344
+ br i1 undef, label %return, label %if.end361
+
+if.end361: ; preds = %if.end304
+ %mul364 = fmul float %add8.i2381, %add8.i2381
+ br i1 undef, label %if.then370, label %if.end395
+
+if.then370: ; preds = %if.end361
+ br i1 undef, label %if.then374, label %if.end395
+
+if.then374: ; preds = %if.then370
+ %cmp392 = fcmp olt float %sub345, 0.000000e+00
+ br label %if.end395
+
+if.end395: ; preds = %if.then374, %if.then370, %if.end361
+ unreachable
+
+return: ; preds = %if.end304, %if.end248, %if.end225, %if.end189, %if.end162, %if.end136, %if.end111, %if.end, %entry
+ ret void
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_bullet3.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_bullet3.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_bullet3.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_bullet3.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,141 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+%class.btVector3.23.221.463.485.507.573.595.683.727.749.815.837.991.1585.1607.1629.1651.1849.2047.2069.2091.2113 = type { [4 x float] }
+
+; Function Attrs: ssp uwtable
+define void @_ZN11HullLibrary15CleanupVerticesEjPK9btVector3jRjPS0_fRS0_(%class.btVector3.23.221.463.485.507.573.595.683.727.749.815.837.991.1585.1607.1629.1651.1849.2047.2069.2091.2113* %vertices) #0 align 2 {
+; CHECK-LABEL: @_ZN11HullLibrary15CleanupVerticesEjPK9btVector3jRjPS0_fRS0_(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 undef, label [[RETURN:%.*]], label [[IF_END:%.*]]
+; CHECK: if.end:
+; CHECK-NEXT: br label [[FOR_BODY:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN17_1:%.*]], label [[IF_END22_1:%.*]]
+; CHECK: for.end36:
+; CHECK-NEXT: br label [[FOR_BODY144:%.*]]
+; CHECK: for.body144:
+; CHECK-NEXT: br i1 undef, label [[FOR_END227:%.*]], label [[FOR_BODY144]]
+; CHECK: for.end227:
+; CHECK-NEXT: br i1 undef, label [[FOR_END271:%.*]], label [[FOR_BODY233:%.*]]
+; CHECK: for.body233:
+; CHECK-NEXT: br i1 undef, label [[FOR_BODY233]], label [[FOR_END271]]
+; CHECK: for.end271:
+; CHECK-NEXT: [[TMP0:%.*]] = phi float [ 0x47EFFFFFE0000000, [[FOR_END227]] ], [ undef, [[FOR_BODY233]] ]
+; CHECK-NEXT: [[TMP1:%.*]] = phi float [ 0x47EFFFFFE0000000, [[FOR_END227]] ], [ undef, [[FOR_BODY233]] ]
+; CHECK-NEXT: [[SUB275:%.*]] = fsub float undef, [[TMP1]]
+; CHECK-NEXT: [[SUB279:%.*]] = fsub float undef, [[TMP0]]
+; CHECK-NEXT: br i1 undef, label [[IF_THEN291:%.*]], label [[RETURN]]
+; CHECK: if.then291:
+; CHECK-NEXT: [[MUL292:%.*]] = fmul float [[SUB275]], 5.000000e-01
+; CHECK-NEXT: [[ADD294:%.*]] = fadd float [[TMP1]], [[MUL292]]
+; CHECK-NEXT: [[MUL295:%.*]] = fmul float [[SUB279]], 5.000000e-01
+; CHECK-NEXT: [[ADD297:%.*]] = fadd float [[TMP0]], [[MUL295]]
+; CHECK-NEXT: br i1 undef, label [[IF_END332:%.*]], label [[IF_ELSE319:%.*]]
+; CHECK: if.else319:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN325:%.*]], label [[IF_END327:%.*]]
+; CHECK: if.then325:
+; CHECK-NEXT: br label [[IF_END327]]
+; CHECK: if.end327:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN329:%.*]], label [[IF_END332]]
+; CHECK: if.then329:
+; CHECK-NEXT: br label [[IF_END332]]
+; CHECK: if.end332:
+; CHECK-NEXT: [[DX272_1:%.*]] = phi float [ [[SUB275]], [[IF_THEN329]] ], [ [[SUB275]], [[IF_END327]] ], [ 0x3F847AE140000000, [[IF_THEN291]] ]
+; CHECK-NEXT: [[DY276_1:%.*]] = phi float [ undef, [[IF_THEN329]] ], [ undef, [[IF_END327]] ], [ 0x3F847AE140000000, [[IF_THEN291]] ]
+; CHECK-NEXT: [[SUB334:%.*]] = fsub float [[ADD294]], [[DX272_1]]
+; CHECK-NEXT: [[SUB338:%.*]] = fsub float [[ADD297]], [[DY276_1]]
+; CHECK-NEXT: [[ARRAYIDX_I_I606:%.*]] = getelementptr inbounds [[CLASS_BTVECTOR3_23_221_463_485_507_573_595_683_727_749_815_837_991_1585_1607_1629_1651_1849_2047_2069_2091_2113:%.*]], %class.btVector3.23.221.463.485.507.573.595.683.727.749.815.837.991.1585.1607.1629.1651.1849.2047.2069.2091.2113* [[VERTICES:%.*]], i64 0, i32 0, i64 0
+; CHECK-NEXT: store float [[SUB334]], float* [[ARRAYIDX_I_I606]], align 4
+; CHECK-NEXT: [[ARRAYIDX3_I607:%.*]] = getelementptr inbounds [[CLASS_BTVECTOR3_23_221_463_485_507_573_595_683_727_749_815_837_991_1585_1607_1629_1651_1849_2047_2069_2091_2113]], %class.btVector3.23.221.463.485.507.573.595.683.727.749.815.837.991.1585.1607.1629.1651.1849.2047.2069.2091.2113* [[VERTICES]], i64 0, i32 0, i64 1
+; CHECK-NEXT: store float [[SUB338]], float* [[ARRAYIDX3_I607]], align 4
+; CHECK-NEXT: br label [[RETURN]]
+; CHECK: return:
+; CHECK-NEXT: ret void
+; CHECK: if.then17.1:
+; CHECK-NEXT: br label [[IF_END22_1]]
+; CHECK: if.end22.1:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN17_2:%.*]], label [[IF_END22_2:%.*]]
+; CHECK: if.then17.2:
+; CHECK-NEXT: br label [[IF_END22_2]]
+; CHECK: if.end22.2:
+; CHECK-NEXT: br i1 undef, label [[FOR_END36:%.*]], label [[FOR_BODY]]
+;
+entry:
+ br i1 undef, label %return, label %if.end
+
+if.end: ; preds = %entry
+ br label %for.body
+
+for.body: ; preds = %if.end22.2, %if.end
+ br i1 undef, label %if.then17.1, label %if.end22.1
+
+for.end36: ; preds = %if.end22.2
+ br label %for.body144
+
+for.body144: ; preds = %for.body144, %for.end36
+ br i1 undef, label %for.end227, label %for.body144
+
+for.end227: ; preds = %for.body144
+ br i1 undef, label %for.end271, label %for.body233
+
+for.body233: ; preds = %for.body233, %for.end227
+ br i1 undef, label %for.body233, label %for.end271
+
+for.end271: ; preds = %for.body233, %for.end227
+ %0 = phi float [ 0x47EFFFFFE0000000, %for.end227 ], [ undef, %for.body233 ]
+ %1 = phi float [ 0x47EFFFFFE0000000, %for.end227 ], [ undef, %for.body233 ]
+ %sub275 = fsub float undef, %1
+ %sub279 = fsub float undef, %0
+ br i1 undef, label %if.then291, label %return
+
+if.then291: ; preds = %for.end271
+ %mul292 = fmul float %sub275, 5.000000e-01
+ %add294 = fadd float %1, %mul292
+ %mul295 = fmul float %sub279, 5.000000e-01
+ %add297 = fadd float %0, %mul295
+ br i1 undef, label %if.end332, label %if.else319
+
+if.else319: ; preds = %if.then291
+ br i1 undef, label %if.then325, label %if.end327
+
+if.then325: ; preds = %if.else319
+ br label %if.end327
+
+if.end327: ; preds = %if.then325, %if.else319
+ br i1 undef, label %if.then329, label %if.end332
+
+if.then329: ; preds = %if.end327
+ br label %if.end332
+
+if.end332: ; preds = %if.then329, %if.end327, %if.then291
+ %dx272.1 = phi float [ %sub275, %if.then329 ], [ %sub275, %if.end327 ], [ 0x3F847AE140000000, %if.then291 ]
+ %dy276.1 = phi float [ undef, %if.then329 ], [ undef, %if.end327 ], [ 0x3F847AE140000000, %if.then291 ]
+ %sub334 = fsub float %add294, %dx272.1
+ %sub338 = fsub float %add297, %dy276.1
+ %arrayidx.i.i606 = getelementptr inbounds %class.btVector3.23.221.463.485.507.573.595.683.727.749.815.837.991.1585.1607.1629.1651.1849.2047.2069.2091.2113, %class.btVector3.23.221.463.485.507.573.595.683.727.749.815.837.991.1585.1607.1629.1651.1849.2047.2069.2091.2113* %vertices, i64 0, i32 0, i64 0
+ store float %sub334, float* %arrayidx.i.i606, align 4
+ %arrayidx3.i607 = getelementptr inbounds %class.btVector3.23.221.463.485.507.573.595.683.727.749.815.837.991.1585.1607.1629.1651.1849.2047.2069.2091.2113, %class.btVector3.23.221.463.485.507.573.595.683.727.749.815.837.991.1585.1607.1629.1651.1849.2047.2069.2091.2113* %vertices, i64 0, i32 0, i64 1
+ store float %sub338, float* %arrayidx3.i607, align 4
+ br label %return
+
+return: ; preds = %if.end332, %for.end271, %entry
+ ret void
+
+if.then17.1: ; preds = %for.body
+ br label %if.end22.1
+
+if.end22.1: ; preds = %if.then17.1, %for.body
+ br i1 undef, label %if.then17.2, label %if.end22.2
+
+if.then17.2: ; preds = %if.end22.1
+ br label %if.end22.2
+
+if.end22.2: ; preds = %if.then17.2, %if.end22.1
+ br i1 undef, label %for.end36, label %for.body
+}
+
+attributes #0 = { ssp uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-frame-pointer-elim-non-leaf"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "unsafe-fp-math"="false" "use-soft-float"="false" }
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_cmpop.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_cmpop.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_cmpop.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_cmpop.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,141 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -S | FileCheck %s -check-prefixes=CHECK,SSE
+; RUN: opt < %s -basicaa -slp-vectorizer -S -mattr=+avx | FileCheck %s -check-prefixes=CHECK,AVX
+
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.10.0"
+
+define void @testfunc(float* nocapture %dest, float* nocapture readonly %src) {
+; SSE-LABEL: @testfunc(
+; SSE-NEXT: entry:
+; SSE-NEXT: br label [[FOR_BODY:%.*]]
+; SSE: for.body:
+; SSE-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; SSE-NEXT: [[ACC1_056:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[ADD13:%.*]], [[FOR_BODY]] ]
+; SSE-NEXT: [[S1_055:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[COND_I40:%.*]], [[FOR_BODY]] ]
+; SSE-NEXT: [[S0_054:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[COND_I44:%.*]], [[FOR_BODY]] ]
+; SSE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 [[INDVARS_IV]]
+; SSE-NEXT: [[TMP0:%.*]] = load float, float* [[ARRAYIDX]], align 4
+; SSE-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; SSE-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[DEST:%.*]], i64 [[INDVARS_IV]]
+; SSE-NEXT: store float [[ACC1_056]], float* [[ARRAYIDX2]], align 4
+; SSE-NEXT: [[ADD:%.*]] = fadd float [[S0_054]], [[TMP0]]
+; SSE-NEXT: [[ADD3:%.*]] = fadd float [[S1_055]], [[TMP0]]
+; SSE-NEXT: [[MUL:%.*]] = fmul float [[S0_054]], 0.000000e+00
+; SSE-NEXT: [[ADD4:%.*]] = fadd float [[MUL]], [[ADD3]]
+; SSE-NEXT: [[MUL5:%.*]] = fmul float [[S1_055]], 0.000000e+00
+; SSE-NEXT: [[ADD6:%.*]] = fadd float [[MUL5]], [[ADD]]
+; SSE-NEXT: [[CMP_I:%.*]] = fcmp olt float [[ADD6]], 1.000000e+00
+; SSE-NEXT: [[COND_I:%.*]] = select i1 [[CMP_I]], float [[ADD6]], float 1.000000e+00
+; SSE-NEXT: [[CMP_I51:%.*]] = fcmp olt float [[COND_I]], -1.000000e+00
+; SSE-NEXT: [[CMP_I49:%.*]] = fcmp olt float [[ADD4]], 1.000000e+00
+; SSE-NEXT: [[COND_I50:%.*]] = select i1 [[CMP_I49]], float [[ADD4]], float 1.000000e+00
+; SSE-NEXT: [[CMP_I47:%.*]] = fcmp olt float [[COND_I50]], -1.000000e+00
+; SSE-NEXT: [[COND_I_OP:%.*]] = fmul float [[COND_I]], 0.000000e+00
+; SSE-NEXT: [[MUL10:%.*]] = select i1 [[CMP_I51]], float -0.000000e+00, float [[COND_I_OP]]
+; SSE-NEXT: [[COND_I50_OP:%.*]] = fmul float [[COND_I50]], 0.000000e+00
+; SSE-NEXT: [[MUL11:%.*]] = select i1 [[CMP_I47]], float -0.000000e+00, float [[COND_I50_OP]]
+; SSE-NEXT: [[ADD13]] = fadd float [[MUL10]], [[MUL11]]
+; SSE-NEXT: [[CMP_I45:%.*]] = fcmp olt float [[ADD13]], 1.000000e+00
+; SSE-NEXT: [[COND_I46:%.*]] = select i1 [[CMP_I45]], float [[ADD13]], float 1.000000e+00
+; SSE-NEXT: [[CMP_I43:%.*]] = fcmp olt float [[COND_I46]], -1.000000e+00
+; SSE-NEXT: [[COND_I44]] = select i1 [[CMP_I43]], float -1.000000e+00, float [[COND_I46]]
+; SSE-NEXT: [[CMP_I41:%.*]] = fcmp olt float [[MUL11]], 1.000000e+00
+; SSE-NEXT: [[COND_I42:%.*]] = select i1 [[CMP_I41]], float [[MUL11]], float 1.000000e+00
+; SSE-NEXT: [[CMP_I39:%.*]] = fcmp olt float [[COND_I42]], -1.000000e+00
+; SSE-NEXT: [[COND_I40]] = select i1 [[CMP_I39]], float -1.000000e+00, float [[COND_I42]]
+; SSE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 32
+; SSE-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; SSE: for.end:
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @testfunc(
+; AVX-NEXT: entry:
+; AVX-NEXT: br label [[FOR_BODY:%.*]]
+; AVX: for.body:
+; AVX-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; AVX-NEXT: [[ACC1_056:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[ADD13:%.*]], [[FOR_BODY]] ]
+; AVX-NEXT: [[TMP0:%.*]] = phi <2 x float> [ zeroinitializer, [[ENTRY]] ], [ [[TMP23:%.*]], [[FOR_BODY]] ]
+; AVX-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 [[INDVARS_IV]]
+; AVX-NEXT: [[TMP1:%.*]] = load float, float* [[ARRAYIDX]], align 4
+; AVX-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; AVX-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[DEST:%.*]], i64 [[INDVARS_IV]]
+; AVX-NEXT: store float [[ACC1_056]], float* [[ARRAYIDX2]], align 4
+; AVX-NEXT: [[TMP2:%.*]] = extractelement <2 x float> [[TMP0]], i32 1
+; AVX-NEXT: [[TMP3:%.*]] = insertelement <2 x float> undef, float [[TMP2]], i32 0
+; AVX-NEXT: [[TMP4:%.*]] = extractelement <2 x float> [[TMP0]], i32 0
+; AVX-NEXT: [[TMP5:%.*]] = insertelement <2 x float> [[TMP3]], float [[TMP4]], i32 1
+; AVX-NEXT: [[TMP6:%.*]] = insertelement <2 x float> undef, float [[TMP1]], i32 0
+; AVX-NEXT: [[TMP7:%.*]] = insertelement <2 x float> [[TMP6]], float [[TMP1]], i32 1
+; AVX-NEXT: [[TMP8:%.*]] = fadd <2 x float> [[TMP5]], [[TMP7]]
+; AVX-NEXT: [[TMP9:%.*]] = fmul <2 x float> [[TMP0]], zeroinitializer
+; AVX-NEXT: [[TMP10:%.*]] = fadd <2 x float> [[TMP9]], [[TMP8]]
+; AVX-NEXT: [[TMP11:%.*]] = fcmp olt <2 x float> [[TMP10]], <float 1.000000e+00, float 1.000000e+00>
+; AVX-NEXT: [[TMP12:%.*]] = select <2 x i1> [[TMP11]], <2 x float> [[TMP10]], <2 x float> <float 1.000000e+00, float 1.000000e+00>
+; AVX-NEXT: [[TMP13:%.*]] = fcmp olt <2 x float> [[TMP12]], <float -1.000000e+00, float -1.000000e+00>
+; AVX-NEXT: [[TMP14:%.*]] = fmul <2 x float> [[TMP12]], zeroinitializer
+; AVX-NEXT: [[TMP15:%.*]] = select <2 x i1> [[TMP13]], <2 x float> <float -0.000000e+00, float -0.000000e+00>, <2 x float> [[TMP14]]
+; AVX-NEXT: [[TMP16:%.*]] = extractelement <2 x float> [[TMP15]], i32 0
+; AVX-NEXT: [[TMP17:%.*]] = extractelement <2 x float> [[TMP15]], i32 1
+; AVX-NEXT: [[ADD13]] = fadd float [[TMP16]], [[TMP17]]
+; AVX-NEXT: [[TMP18:%.*]] = insertelement <2 x float> undef, float [[TMP17]], i32 0
+; AVX-NEXT: [[TMP19:%.*]] = insertelement <2 x float> [[TMP18]], float [[ADD13]], i32 1
+; AVX-NEXT: [[TMP20:%.*]] = fcmp olt <2 x float> [[TMP19]], <float 1.000000e+00, float 1.000000e+00>
+; AVX-NEXT: [[TMP21:%.*]] = select <2 x i1> [[TMP20]], <2 x float> [[TMP19]], <2 x float> <float 1.000000e+00, float 1.000000e+00>
+; AVX-NEXT: [[TMP22:%.*]] = fcmp olt <2 x float> [[TMP21]], <float -1.000000e+00, float -1.000000e+00>
+; AVX-NEXT: [[TMP23]] = select <2 x i1> [[TMP22]], <2 x float> <float -1.000000e+00, float -1.000000e+00>, <2 x float> [[TMP21]]
+; AVX-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 32
+; AVX-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; AVX: for.end:
+; AVX-NEXT: ret void
+;
+entry:
+ br label %for.body
+
+for.body:
+ %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
+ %acc1.056 = phi float [ 0.000000e+00, %entry ], [ %add13, %for.body ]
+ %s1.055 = phi float [ 0.000000e+00, %entry ], [ %cond.i40, %for.body ]
+ %s0.054 = phi float [ 0.000000e+00, %entry ], [ %cond.i44, %for.body ]
+ %arrayidx = getelementptr inbounds float, float* %src, i64 %indvars.iv
+ %0 = load float, float* %arrayidx, align 4
+ %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+ %arrayidx2 = getelementptr inbounds float, float* %dest, i64 %indvars.iv
+ store float %acc1.056, float* %arrayidx2, align 4
+ %add = fadd float %s0.054, %0
+ %add3 = fadd float %s1.055, %0
+ %mul = fmul float %s0.054, 0.000000e+00
+ %add4 = fadd float %mul, %add3
+ %mul5 = fmul float %s1.055, 0.000000e+00
+ %add6 = fadd float %mul5, %add
+ %cmp.i = fcmp olt float %add6, 1.000000e+00
+ %cond.i = select i1 %cmp.i, float %add6, float 1.000000e+00
+ %cmp.i51 = fcmp olt float %cond.i, -1.000000e+00
+ %cmp.i49 = fcmp olt float %add4, 1.000000e+00
+ %cond.i50 = select i1 %cmp.i49, float %add4, float 1.000000e+00
+ %cmp.i47 = fcmp olt float %cond.i50, -1.000000e+00
+ %cond.i.op = fmul float %cond.i, 0.000000e+00
+ %mul10 = select i1 %cmp.i51, float -0.000000e+00, float %cond.i.op
+ %cond.i50.op = fmul float %cond.i50, 0.000000e+00
+ %mul11 = select i1 %cmp.i47, float -0.000000e+00, float %cond.i50.op
+ %add13 = fadd float %mul10, %mul11
+
+ ; The SLPVectorizer crashed in vectorizeChainsInBlock() because it tried
+ ; to access the second operand of the following cmp after the cmp itself
+ ; was already vectorized and deleted.
+ %cmp.i45 = fcmp olt float %add13, 1.000000e+00
+
+ %cond.i46 = select i1 %cmp.i45, float %add13, float 1.000000e+00
+ %cmp.i43 = fcmp olt float %cond.i46, -1.000000e+00
+ %cond.i44 = select i1 %cmp.i43, float -1.000000e+00, float %cond.i46
+ %cmp.i41 = fcmp olt float %mul11, 1.000000e+00
+ %cond.i42 = select i1 %cmp.i41, float %mul11, float 1.000000e+00
+ %cmp.i39 = fcmp olt float %cond.i42, -1.000000e+00
+ %cond.i40 = select i1 %cmp.i39, float -1.000000e+00, float %cond.i42
+ %exitcond = icmp eq i64 %indvars.iv.next, 32
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end:
+ ret void
+}
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_dequeue.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_dequeue.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_dequeue.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_dequeue.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,68 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+%"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731" = type { double*, double*, double*, double** }
+
+; Function Attrs: nounwind ssp uwtable
+define void @_ZSt6uniqueISt15_Deque_iteratorIdRdPdEET_S4_S4_(%"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731"* %__first, %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731"* nocapture %__last) {
+; CHECK-LABEL: @_ZSt6uniqueISt15_Deque_iteratorIdRdPdEET_S4_S4_(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[_M_CUR2_I_I:%.*]] = getelementptr inbounds %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731", %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731"* [[__FIRST:%.*]], i64 0, i32 0
+; CHECK-NEXT: [[TMP0:%.*]] = load double*, double** [[_M_CUR2_I_I]], align 8
+; CHECK-NEXT: [[_M_FIRST3_I_I:%.*]] = getelementptr inbounds %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731", %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731"* [[__FIRST]], i64 0, i32 1
+; CHECK-NEXT: [[_M_CUR2_I_I81:%.*]] = getelementptr inbounds %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731", %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731"* [[__LAST:%.*]], i64 0, i32 0
+; CHECK-NEXT: [[TMP1:%.*]] = load double*, double** [[_M_CUR2_I_I81]], align 8
+; CHECK-NEXT: [[_M_FIRST3_I_I83:%.*]] = getelementptr inbounds %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731", %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731"* [[__LAST]], i64 0, i32 1
+; CHECK-NEXT: [[TMP2:%.*]] = load double*, double** [[_M_FIRST3_I_I83]], align 8
+; CHECK-NEXT: br i1 undef, label [[_ZST13ADJACENT_FINDIST15_DEQUE_ITERATORIDRDPDEET_S4_S4__EXIT:%.*]], label [[WHILE_COND_I_PREHEADER:%.*]]
+; CHECK: while.cond.i.preheader:
+; CHECK-NEXT: br label [[WHILE_COND_I:%.*]]
+; CHECK: while.cond.i:
+; CHECK-NEXT: br i1 undef, label [[_ZST13ADJACENT_FINDIST15_DEQUE_ITERATORIDRDPDEET_S4_S4__EXIT]], label [[WHILE_BODY_I:%.*]]
+; CHECK: while.body.i:
+; CHECK-NEXT: br i1 undef, label [[_ZST13ADJACENT_FINDIST15_DEQUE_ITERATORIDRDPDEET_S4_S4__EXIT]], label [[WHILE_COND_I]]
+; CHECK: _ZSt13adjacent_findISt15_Deque_iteratorIdRdPdEET_S4_S4_.exit:
+; CHECK-NEXT: [[TMP3:%.*]] = phi double* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[TMP2]], [[WHILE_COND_I]] ], [ undef, [[WHILE_BODY_I]] ]
+; CHECK-NEXT: [[TMP4:%.*]] = phi double* [ [[TMP0]], [[ENTRY]] ], [ [[TMP1]], [[WHILE_COND_I]] ], [ undef, [[WHILE_BODY_I]] ]
+; CHECK-NEXT: store double* [[TMP4]], double** [[_M_CUR2_I_I]], align 8
+; CHECK-NEXT: store double* [[TMP3]], double** [[_M_FIRST3_I_I]], align 8
+; CHECK-NEXT: br i1 undef, label [[IF_THEN_I55:%.*]], label [[WHILE_COND:%.*]]
+; CHECK: if.then.i55:
+; CHECK-NEXT: br label [[WHILE_COND]]
+; CHECK: while.cond:
+; CHECK-NEXT: br label [[WHILE_COND]]
+;
+entry:
+ %_M_cur2.i.i = getelementptr inbounds %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731", %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731"* %__first, i64 0, i32 0
+ %0 = load double*, double** %_M_cur2.i.i, align 8
+ %_M_first3.i.i = getelementptr inbounds %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731", %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731"* %__first, i64 0, i32 1
+ %_M_cur2.i.i81 = getelementptr inbounds %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731", %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731"* %__last, i64 0, i32 0
+ %1 = load double*, double** %_M_cur2.i.i81, align 8
+ %_M_first3.i.i83 = getelementptr inbounds %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731", %"struct.std::_Deque_iterator.4.157.174.208.259.276.344.731"* %__last, i64 0, i32 1
+ %2 = load double*, double** %_M_first3.i.i83, align 8
+ br i1 undef, label %_ZSt13adjacent_findISt15_Deque_iteratorIdRdPdEET_S4_S4_.exit, label %while.cond.i.preheader
+
+while.cond.i.preheader: ; preds = %entry
+ br label %while.cond.i
+
+while.cond.i: ; preds = %while.body.i, %while.cond.i.preheader
+ br i1 undef, label %_ZSt13adjacent_findISt15_Deque_iteratorIdRdPdEET_S4_S4_.exit, label %while.body.i
+
+while.body.i: ; preds = %while.cond.i
+ br i1 undef, label %_ZSt13adjacent_findISt15_Deque_iteratorIdRdPdEET_S4_S4_.exit, label %while.cond.i
+
+_ZSt13adjacent_findISt15_Deque_iteratorIdRdPdEET_S4_S4_.exit: ; preds = %while.body.i, %while.cond.i, %entry
+ %3 = phi double* [ %2, %entry ], [ %2, %while.cond.i ], [ undef, %while.body.i ]
+ %4 = phi double* [ %0, %entry ], [ %1, %while.cond.i ], [ undef, %while.body.i ]
+ store double* %4, double** %_M_cur2.i.i, align 8
+ store double* %3, double** %_M_first3.i.i, align 8
+ br i1 undef, label %if.then.i55, label %while.cond
+
+if.then.i55: ; preds = %_ZSt13adjacent_findISt15_Deque_iteratorIdRdPdEET_S4_S4_.exit
+ br label %while.cond
+
+while.cond: ; preds = %while.cond, %if.then.i55, %_ZSt13adjacent_findISt15_Deque_iteratorIdRdPdEET_S4_S4_.exit
+ br label %while.cond
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_flop7.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_flop7.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_flop7.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_flop7.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,77 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+; Function Attrs: nounwind ssp uwtable
+define void @main() #0 {
+; CHECK-LABEL: @main(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 undef, label [[WHILE_BODY:%.*]], label [[WHILE_END:%.*]]
+; CHECK: while.body:
+; CHECK-NEXT: unreachable
+; CHECK: while.end:
+; CHECK-NEXT: br i1 undef, label [[FOR_END80:%.*]], label [[FOR_BODY75_LR_PH:%.*]]
+; CHECK: for.body75.lr.ph:
+; CHECK-NEXT: br label [[FOR_BODY75:%.*]]
+; CHECK: for.body75:
+; CHECK-NEXT: br label [[FOR_BODY75]]
+; CHECK: for.end80:
+; CHECK-NEXT: br i1 undef, label [[FOR_END300:%.*]], label [[FOR_BODY267_LR_PH:%.*]]
+; CHECK: for.body267.lr.ph:
+; CHECK-NEXT: br label [[FOR_BODY267:%.*]]
+; CHECK: for.body267:
+; CHECK-NEXT: [[S_71010:%.*]] = phi double [ 0.000000e+00, [[FOR_BODY267_LR_PH]] ], [ [[ADD297:%.*]], [[FOR_BODY267]] ]
+; CHECK-NEXT: [[MUL269:%.*]] = fmul double undef, undef
+; CHECK-NEXT: [[MUL270:%.*]] = fmul double [[MUL269]], [[MUL269]]
+; CHECK-NEXT: [[ADD282:%.*]] = fadd double undef, undef
+; CHECK-NEXT: [[MUL283:%.*]] = fmul double [[MUL269]], [[ADD282]]
+; CHECK-NEXT: [[ADD293:%.*]] = fadd double undef, undef
+; CHECK-NEXT: [[MUL294:%.*]] = fmul double [[MUL270]], [[ADD293]]
+; CHECK-NEXT: [[ADD295:%.*]] = fadd double undef, [[MUL294]]
+; CHECK-NEXT: [[DIV296:%.*]] = fdiv double [[MUL283]], [[ADD295]]
+; CHECK-NEXT: [[ADD297]] = fadd double [[S_71010]], [[DIV296]]
+; CHECK-NEXT: br i1 undef, label [[FOR_BODY267]], label [[FOR_END300]]
+; CHECK: for.end300:
+; CHECK-NEXT: unreachable
+;
+entry:
+ br i1 undef, label %while.body, label %while.end
+
+while.body: ; preds = %entry
+ unreachable
+
+while.end: ; preds = %entry
+ br i1 undef, label %for.end80, label %for.body75.lr.ph
+
+for.body75.lr.ph: ; preds = %while.end
+ br label %for.body75
+
+for.body75: ; preds = %for.body75, %for.body75.lr.ph
+ br label %for.body75
+
+for.end80: ; preds = %while.end
+ br i1 undef, label %for.end300, label %for.body267.lr.ph
+
+for.body267.lr.ph: ; preds = %for.end80
+ br label %for.body267
+
+for.body267: ; preds = %for.body267, %for.body267.lr.ph
+ %s.71010 = phi double [ 0.000000e+00, %for.body267.lr.ph ], [ %add297, %for.body267 ]
+ %mul269 = fmul double undef, undef
+ %mul270 = fmul double %mul269, %mul269
+ %add282 = fadd double undef, undef
+ %mul283 = fmul double %mul269, %add282
+ %add293 = fadd double undef, undef
+ %mul294 = fmul double %mul270, %add293
+ %add295 = fadd double undef, %mul294
+ %div296 = fdiv double %mul283, %add295
+ %add297 = fadd double %s.71010, %div296
+ br i1 undef, label %for.body267, label %for.end300
+
+for.end300: ; preds = %for.body267, %for.end80
+ unreachable
+}
+
+attributes #0 = { nounwind ssp uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-frame-pointer-elim-non-leaf"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "unsafe-fp-math"="false" "use-soft-float"="false" }
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_gep.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_gep.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_gep.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_gep.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,31 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-unknown-linux-gnu | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+ at a = common global i64* null, align 8
+
+; Function Attrs: nounwind uwtable
+define i32 @fn1() {
+; CHECK-LABEL: @fn1(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = load i64*, i64** @a, align 8
+; CHECK-NEXT: [[ADD_PTR:%.*]] = getelementptr inbounds i64, i64* [[TMP0]], i64 1
+; CHECK-NEXT: [[TMP1:%.*]] = ptrtoint i64* [[ADD_PTR]] to i64
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i64, i64* [[TMP0]], i64 2
+; CHECK-NEXT: store i64 [[TMP1]], i64* [[ARRAYIDX]], align 8
+; CHECK-NEXT: [[TMP2:%.*]] = ptrtoint i64* [[ARRAYIDX]] to i64
+; CHECK-NEXT: store i64 [[TMP2]], i64* [[ADD_PTR]], align 8
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %0 = load i64*, i64** @a, align 8
+ %add.ptr = getelementptr inbounds i64, i64* %0, i64 1
+ %1 = ptrtoint i64* %add.ptr to i64
+ %arrayidx = getelementptr inbounds i64, i64* %0, i64 2
+ store i64 %1, i64* %arrayidx, align 8
+ %2 = ptrtoint i64* %arrayidx to i64
+ store i64 %2, i64* %add.ptr, align 8
+ ret i32 undef
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_lencod.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_lencod.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_lencod.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_lencod.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,150 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+; Function Attrs: nounwind ssp uwtable
+define void @RCModelEstimator() {
+; CHECK-LABEL: @RCModelEstimator(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 undef, label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END_THREAD:%.*]]
+; CHECK: for.end.thread:
+; CHECK-NEXT: unreachable
+; CHECK: for.body.lr.ph:
+; CHECK-NEXT: br i1 undef, label [[FOR_END:%.*]], label [[FOR_BODY:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: br i1 undef, label [[FOR_END]], label [[FOR_BODY]]
+; CHECK: for.end:
+; CHECK-NEXT: br i1 undef, label [[FOR_BODY3:%.*]], label [[IF_END103:%.*]]
+; CHECK: for.cond14.preheader:
+; CHECK-NEXT: br i1 undef, label [[FOR_BODY16_LR_PH:%.*]], label [[IF_END103]]
+; CHECK: for.body16.lr.ph:
+; CHECK-NEXT: br label [[FOR_BODY16:%.*]]
+; CHECK: for.body3:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN7:%.*]], label [[FOR_INC11:%.*]]
+; CHECK: if.then7:
+; CHECK-NEXT: br label [[FOR_INC11]]
+; CHECK: for.inc11:
+; CHECK-NEXT: br i1 false, label [[FOR_COND14_PREHEADER:%.*]], label [[FOR_BODY3]]
+; CHECK: for.body16:
+; CHECK-NEXT: br i1 undef, label [[FOR_END39:%.*]], label [[FOR_BODY16]]
+; CHECK: for.end39:
+; CHECK-NEXT: br i1 undef, label [[IF_END103]], label [[FOR_COND45_PREHEADER:%.*]]
+; CHECK: for.cond45.preheader:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN88:%.*]], label [[IF_ELSE:%.*]]
+; CHECK: if.then88:
+; CHECK-NEXT: br label [[IF_END103]]
+; CHECK: if.else:
+; CHECK-NEXT: br label [[IF_END103]]
+; CHECK: if.end103:
+; CHECK-NEXT: ret void
+;
+entry:
+ br i1 undef, label %for.body.lr.ph, label %for.end.thread
+
+for.end.thread: ; preds = %entry
+ unreachable
+
+for.body.lr.ph: ; preds = %entry
+ br i1 undef, label %for.end, label %for.body
+
+for.body: ; preds = %for.body, %for.body.lr.ph
+ br i1 undef, label %for.end, label %for.body
+
+for.end: ; preds = %for.body, %for.body.lr.ph
+ br i1 undef, label %for.body3, label %if.end103
+
+for.cond14.preheader: ; preds = %for.inc11
+ br i1 undef, label %for.body16.lr.ph, label %if.end103
+
+for.body16.lr.ph: ; preds = %for.cond14.preheader
+ br label %for.body16
+
+for.body3: ; preds = %for.inc11, %for.end
+ br i1 undef, label %if.then7, label %for.inc11
+
+if.then7: ; preds = %for.body3
+ br label %for.inc11
+
+for.inc11: ; preds = %if.then7, %for.body3
+ br i1 false, label %for.cond14.preheader, label %for.body3
+
+for.body16: ; preds = %for.body16, %for.body16.lr.ph
+ br i1 undef, label %for.end39, label %for.body16
+
+for.end39: ; preds = %for.body16
+ br i1 undef, label %if.end103, label %for.cond45.preheader
+
+for.cond45.preheader: ; preds = %for.end39
+ br i1 undef, label %if.then88, label %if.else
+
+if.then88: ; preds = %for.cond45.preheader
+ %mul89 = fmul double 0.000000e+00, 0.000000e+00
+ %mul90 = fmul double 0.000000e+00, 0.000000e+00
+ %sub91 = fsub double %mul89, %mul90
+ %div92 = fdiv double %sub91, undef
+ %mul94 = fmul double 0.000000e+00, 0.000000e+00
+ %mul95 = fmul double 0.000000e+00, 0.000000e+00
+ %sub96 = fsub double %mul94, %mul95
+ %div97 = fdiv double %sub96, undef
+ br label %if.end103
+
+if.else: ; preds = %for.cond45.preheader
+ br label %if.end103
+
+if.end103: ; preds = %if.else, %if.then88, %for.end39, %for.cond14.preheader, %for.end
+ %0 = phi double [ 0.000000e+00, %for.end39 ], [ %div97, %if.then88 ], [ 0.000000e+00, %if.else ], [ 0.000000e+00, %for.cond14.preheader ], [ 0.000000e+00, %for.end ]
+ %1 = phi double [ undef, %for.end39 ], [ %div92, %if.then88 ], [ undef, %if.else ], [ 0.000000e+00, %for.cond14.preheader ], [ 0.000000e+00, %for.end ]
+ ret void
+}
+
+
+define void @intrapred_luma() {
+; CHECK-LABEL: @intrapred_luma(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[CONV153:%.*]] = trunc i32 undef to i16
+; CHECK-NEXT: [[ARRAYIDX154:%.*]] = getelementptr inbounds [13 x i16], [13 x i16]* undef, i64 0, i64 12
+; CHECK-NEXT: store i16 [[CONV153]], i16* [[ARRAYIDX154]], align 8
+; CHECK-NEXT: [[ARRAYIDX155:%.*]] = getelementptr inbounds [13 x i16], [13 x i16]* undef, i64 0, i64 11
+; CHECK-NEXT: store i16 [[CONV153]], i16* [[ARRAYIDX155]], align 2
+; CHECK-NEXT: [[ARRAYIDX156:%.*]] = getelementptr inbounds [13 x i16], [13 x i16]* undef, i64 0, i64 10
+; CHECK-NEXT: store i16 [[CONV153]], i16* [[ARRAYIDX156]], align 4
+; CHECK-NEXT: ret void
+;
+entry:
+ %conv153 = trunc i32 undef to i16
+ %arrayidx154 = getelementptr inbounds [13 x i16], [13 x i16]* undef, i64 0, i64 12
+ store i16 %conv153, i16* %arrayidx154, align 8
+ %arrayidx155 = getelementptr inbounds [13 x i16], [13 x i16]* undef, i64 0, i64 11
+ store i16 %conv153, i16* %arrayidx155, align 2
+ %arrayidx156 = getelementptr inbounds [13 x i16], [13 x i16]* undef, i64 0, i64 10
+ store i16 %conv153, i16* %arrayidx156, align 4
+ ret void
+}
+
+define fastcc void @dct36(double* %inbuf) {
+; CHECK-LABEL: @dct36(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[ARRAYIDX44:%.*]] = getelementptr inbounds double, double* [[INBUF:%.*]], i64 1
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[INBUF]] to <2 x double>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
+; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <2 x double> undef, double [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x double> [[TMP3]], double undef, i32 1
+; CHECK-NEXT: [[TMP5:%.*]] = fadd <2 x double> [[TMP1]], [[TMP4]]
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast double* [[ARRAYIDX44]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP5]], <2 x double>* [[TMP6]], align 8
+; CHECK-NEXT: ret void
+;
+entry:
+ %arrayidx41 = getelementptr inbounds double, double* %inbuf, i64 2
+ %arrayidx44 = getelementptr inbounds double, double* %inbuf, i64 1
+ %0 = load double, double* %arrayidx44, align 8
+ %add46 = fadd double %0, undef
+ store double %add46, double* %arrayidx41, align 8
+ %1 = load double, double* %inbuf, align 8
+ %add49 = fadd double %1, %0
+ store double %add49, double* %arrayidx44, align 8
+ ret void
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_mandeltext.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_mandeltext.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_mandeltext.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_mandeltext.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,180 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+define void @main() {
+; CHECK-LABEL: @main(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br label [[FOR_BODY:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: br label [[FOR_COND4_PREHEADER:%.*]]
+; CHECK: for.cond4.preheader:
+; CHECK-NEXT: br label [[FOR_BODY6:%.*]]
+; CHECK: for.body6:
+; CHECK-NEXT: br label [[FOR_BODY12:%.*]]
+; CHECK: for.body12:
+; CHECK-NEXT: [[FZIMG_069:%.*]] = phi double [ undef, [[FOR_BODY6]] ], [ [[ADD19:%.*]], [[IF_END:%.*]] ]
+; CHECK-NEXT: [[FZREAL_068:%.*]] = phi double [ undef, [[FOR_BODY6]] ], [ [[ADD20:%.*]], [[IF_END]] ]
+; CHECK-NEXT: [[MUL13:%.*]] = fmul double [[FZREAL_068]], [[FZREAL_068]]
+; CHECK-NEXT: [[MUL14:%.*]] = fmul double [[FZIMG_069]], [[FZIMG_069]]
+; CHECK-NEXT: [[ADD15:%.*]] = fadd double [[MUL13]], [[MUL14]]
+; CHECK-NEXT: [[CMP16:%.*]] = fcmp ogt double [[ADD15]], 4.000000e+00
+; CHECK-NEXT: br i1 [[CMP16]], label [[FOR_INC21:%.*]], label [[IF_END]]
+; CHECK: if.end:
+; CHECK-NEXT: [[MUL18:%.*]] = fmul double undef, [[FZIMG_069]]
+; CHECK-NEXT: [[ADD19]] = fadd double undef, [[MUL18]]
+; CHECK-NEXT: [[SUB:%.*]] = fsub double [[MUL13]], [[MUL14]]
+; CHECK-NEXT: [[ADD20]] = fadd double undef, [[SUB]]
+; CHECK-NEXT: br i1 undef, label [[FOR_BODY12]], label [[FOR_INC21]]
+; CHECK: for.inc21:
+; CHECK-NEXT: br i1 undef, label [[FOR_END23:%.*]], label [[FOR_BODY6]]
+; CHECK: for.end23:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN25:%.*]], label [[IF_THEN26:%.*]]
+; CHECK: if.then25:
+; CHECK-NEXT: br i1 undef, label [[FOR_END44:%.*]], label [[FOR_COND4_PREHEADER]]
+; CHECK: if.then26:
+; CHECK-NEXT: unreachable
+; CHECK: for.end44:
+; CHECK-NEXT: br i1 undef, label [[FOR_END48:%.*]], label [[FOR_BODY]]
+; CHECK: for.end48:
+; CHECK-NEXT: ret void
+;
+entry:
+ br label %for.body
+
+for.body: ; preds = %for.end44, %entry
+ br label %for.cond4.preheader
+
+for.cond4.preheader: ; preds = %if.then25, %for.body
+ br label %for.body6
+
+for.body6: ; preds = %for.inc21, %for.cond4.preheader
+ br label %for.body12
+
+for.body12: ; preds = %if.end, %for.body6
+ %fZImg.069 = phi double [ undef, %for.body6 ], [ %add19, %if.end ]
+ %fZReal.068 = phi double [ undef, %for.body6 ], [ %add20, %if.end ]
+ %mul13 = fmul double %fZReal.068, %fZReal.068
+ %mul14 = fmul double %fZImg.069, %fZImg.069
+ %add15 = fadd double %mul13, %mul14
+ %cmp16 = fcmp ogt double %add15, 4.000000e+00
+ br i1 %cmp16, label %for.inc21, label %if.end
+
+if.end: ; preds = %for.body12
+ %mul18 = fmul double undef, %fZImg.069
+ %add19 = fadd double undef, %mul18
+ %sub = fsub double %mul13, %mul14
+ %add20 = fadd double undef, %sub
+ br i1 undef, label %for.body12, label %for.inc21
+
+for.inc21: ; preds = %if.end, %for.body12
+ br i1 undef, label %for.end23, label %for.body6
+
+for.end23: ; preds = %for.inc21
+ br i1 undef, label %if.then25, label %if.then26
+
+if.then25: ; preds = %for.end23
+ br i1 undef, label %for.end44, label %for.cond4.preheader
+
+if.then26: ; preds = %for.end23
+ unreachable
+
+for.end44: ; preds = %if.then25
+ br i1 undef, label %for.end48, label %for.body
+
+for.end48: ; preds = %for.end44
+ ret void
+}
+
+%struct.hoge = type { double, double, double}
+
+define void @zot(%struct.hoge* %arg) {
+; CHECK-LABEL: @zot(
+; CHECK-NEXT: bb:
+; CHECK-NEXT: [[TMP:%.*]] = load double, double* undef, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load double, double* undef, align 8
+; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x double> undef, double [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x double> [[TMP0]], double [[TMP]], i32 1
+; CHECK-NEXT: [[TMP2:%.*]] = fsub <2 x double> [[TMP1]], undef
+; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_HOGE:%.*]], %struct.hoge* [[ARG:%.*]], i64 0, i32 1
+; CHECK-NEXT: [[TMP3:%.*]] = fmul <2 x double> [[TMP2]], undef
+; CHECK-NEXT: [[TMP4:%.*]] = fsub <2 x double> [[TMP3]], undef
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast double* [[TMP7]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP4]], <2 x double>* [[TMP5]], align 8
+; CHECK-NEXT: br i1 undef, label [[BB11:%.*]], label [[BB12:%.*]]
+; CHECK: bb11:
+; CHECK-NEXT: br label [[BB14:%.*]]
+; CHECK: bb12:
+; CHECK-NEXT: br label [[BB14]]
+; CHECK: bb14:
+; CHECK-NEXT: ret void
+;
+bb:
+ %tmp = load double, double* undef, align 8
+ %tmp1 = fsub double %tmp, undef
+ %tmp2 = load double, double* undef, align 8
+ %tmp3 = fsub double %tmp2, undef
+ %tmp4 = fmul double %tmp3, undef
+ %tmp5 = fmul double %tmp3, undef
+ %tmp6 = fsub double %tmp5, undef
+ %tmp7 = getelementptr inbounds %struct.hoge, %struct.hoge* %arg, i64 0, i32 1
+ store double %tmp6, double* %tmp7, align 8
+ %tmp8 = fmul double %tmp1, undef
+ %tmp9 = fsub double %tmp8, undef
+ %tmp10 = getelementptr inbounds %struct.hoge, %struct.hoge* %arg, i64 0, i32 2
+ store double %tmp9, double* %tmp10, align 8
+ br i1 undef, label %bb11, label %bb12
+
+bb11: ; preds = %bb
+ br label %bb14
+
+bb12: ; preds = %bb
+ %tmp13 = fmul double undef, %tmp2
+ br label %bb14
+
+bb14: ; preds = %bb12, %bb11
+ ret void
+}
+
+
+%struct.rc4_state.0.24 = type { i32, i32, [256 x i32] }
+
+define void @rc4_crypt(%struct.rc4_state.0.24* nocapture %s) {
+; CHECK-LABEL: @rc4_crypt(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[X1:%.*]] = getelementptr inbounds [[STRUCT_RC4_STATE_0_24:%.*]], %struct.rc4_state.0.24* [[S:%.*]], i64 0, i32 0
+; CHECK-NEXT: [[Y2:%.*]] = getelementptr inbounds [[STRUCT_RC4_STATE_0_24]], %struct.rc4_state.0.24* [[S]], i64 0, i32 1
+; CHECK-NEXT: br i1 undef, label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: [[CONV4:%.*]] = and i32 undef, 255
+; CHECK-NEXT: [[CONV7:%.*]] = and i32 undef, 255
+; CHECK-NEXT: br i1 undef, label [[FOR_END]], label [[FOR_BODY]]
+; CHECK: for.end:
+; CHECK-NEXT: [[X_0_LCSSA:%.*]] = phi i32 [ undef, [[ENTRY:%.*]] ], [ [[CONV4]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[Y_0_LCSSA:%.*]] = phi i32 [ undef, [[ENTRY]] ], [ [[CONV7]], [[FOR_BODY]] ]
+; CHECK-NEXT: store i32 [[X_0_LCSSA]], i32* [[X1]], align 4
+; CHECK-NEXT: store i32 [[Y_0_LCSSA]], i32* [[Y2]], align 4
+; CHECK-NEXT: ret void
+;
+entry:
+ %x1 = getelementptr inbounds %struct.rc4_state.0.24, %struct.rc4_state.0.24* %s, i64 0, i32 0
+ %y2 = getelementptr inbounds %struct.rc4_state.0.24, %struct.rc4_state.0.24* %s, i64 0, i32 1
+ br i1 undef, label %for.body, label %for.end
+
+for.body: ; preds = %for.body, %entry
+ %x.045 = phi i32 [ %conv4, %for.body ], [ undef, %entry ]
+ %conv4 = and i32 undef, 255
+ %conv7 = and i32 undef, 255
+ %idxprom842 = zext i32 %conv7 to i64
+ br i1 undef, label %for.end, label %for.body
+
+for.end: ; preds = %for.body, %entry
+ %x.0.lcssa = phi i32 [ undef, %entry ], [ %conv4, %for.body ]
+ %y.0.lcssa = phi i32 [ undef, %entry ], [ %conv7, %for.body ]
+ store i32 %x.0.lcssa, i32* %x1, align 4
+ store i32 %y.0.lcssa, i32* %y2, align 4
+ ret void
+}
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_netbsd_decompress.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_netbsd_decompress.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_netbsd_decompress.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_netbsd_decompress.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,66 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+%struct.DState = type { i32, i32 }
+
+ at b = common global %struct.DState zeroinitializer, align 4
+ at d = common global i32 0, align 4
+ at c = common global i32 0, align 4
+ at a = common global i32 0, align 4
+ at e = common global i32 0, align 4
+
+define i32 @fn1() {
+; CHECK-LABEL: @fn1(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* getelementptr inbounds (%struct.DState, %struct.DState* @b, i32 0, i32 0), align 4
+; CHECK-NEXT: [[TMP1:%.*]] = load i32, i32* getelementptr inbounds (%struct.DState, %struct.DState* @b, i32 0, i32 1), align 4
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, i32* @d, align 4
+; CHECK-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP2]], 0
+; CHECK-NEXT: br i1 [[COND]], label [[SW_BB:%.*]], label [[SAVE_STATE_AND_RETURN:%.*]]
+; CHECK: sw.bb:
+; CHECK-NEXT: [[TMP3:%.*]] = load i32, i32* @c, align 4
+; CHECK-NEXT: [[AND:%.*]] = and i32 [[TMP3]], 7
+; CHECK-NEXT: store i32 [[AND]], i32* @a, align 4
+; CHECK-NEXT: switch i32 [[AND]], label [[IF_END:%.*]] [
+; CHECK-NEXT: i32 7, label [[SAVE_STATE_AND_RETURN]]
+; CHECK-NEXT: i32 0, label [[SAVE_STATE_AND_RETURN]]
+; CHECK-NEXT: ]
+; CHECK: if.end:
+; CHECK-NEXT: br label [[SAVE_STATE_AND_RETURN]]
+; CHECK: save_state_and_return:
+; CHECK-NEXT: [[T_0:%.*]] = phi i32 [ 0, [[IF_END]] ], [ [[TMP0]], [[ENTRY:%.*]] ], [ [[TMP0]], [[SW_BB]] ], [ [[TMP0]], [[SW_BB]] ]
+; CHECK-NEXT: [[F_0:%.*]] = phi i32 [ 0, [[IF_END]] ], [ [[TMP1]], [[ENTRY]] ], [ 0, [[SW_BB]] ], [ 0, [[SW_BB]] ]
+; CHECK-NEXT: store i32 [[T_0]], i32* getelementptr inbounds (%struct.DState, %struct.DState* @b, i32 0, i32 0), align 4
+; CHECK-NEXT: store i32 [[F_0]], i32* getelementptr inbounds (%struct.DState, %struct.DState* @b, i32 0, i32 1), align 4
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %0 = load i32, i32* getelementptr inbounds (%struct.DState, %struct.DState* @b, i32 0, i32 0), align 4
+ %1 = load i32, i32* getelementptr inbounds (%struct.DState, %struct.DState* @b, i32 0, i32 1), align 4
+ %2 = load i32, i32* @d, align 4
+ %cond = icmp eq i32 %2, 0
+ br i1 %cond, label %sw.bb, label %save_state_and_return
+
+sw.bb: ; preds = %entry
+ %3 = load i32, i32* @c, align 4
+ %and = and i32 %3, 7
+ store i32 %and, i32* @a, align 4
+ switch i32 %and, label %if.end [
+ i32 7, label %save_state_and_return
+ i32 0, label %save_state_and_return
+ ]
+
+if.end: ; preds = %sw.bb
+ br label %save_state_and_return
+
+save_state_and_return: ; preds = %sw.bb, %sw.bb, %if.end, %entry
+ %t.0 = phi i32 [ 0, %if.end ], [ %0, %entry ], [ %0, %sw.bb ], [ %0, %sw.bb ]
+ %f.0 = phi i32 [ 0, %if.end ], [ %1, %entry ], [ 0, %sw.bb ], [ 0, %sw.bb ]
+ store i32 %t.0, i32* getelementptr inbounds (%struct.DState, %struct.DState* @b, i32 0, i32 0), align 4
+ store i32 %f.0, i32* getelementptr inbounds (%struct.DState, %struct.DState* @b, i32 0, i32 1), align 4
+ ret i32 undef
+}
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_scheduling.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_scheduling.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_scheduling.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_scheduling.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,79 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -disable-verify -slp-vectorizer -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-darwin13.3.0"
+
+define void @_foo(double %p1, double %p2, double %p3) #0 {
+; CHECK-LABEL: @_foo(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TAB1:%.*]] = alloca [256 x i32], align 16
+; CHECK-NEXT: [[TAB2:%.*]] = alloca [256 x i32], align 16
+; CHECK-NEXT: br label [[BB1:%.*]]
+; CHECK: bb1:
+; CHECK-NEXT: [[MUL19:%.*]] = fmul double [[P1:%.*]], 1.638400e+04
+; CHECK-NEXT: [[MUL20:%.*]] = fmul double [[P3:%.*]], 1.638400e+04
+; CHECK-NEXT: [[ADD:%.*]] = fadd double [[MUL20]], 8.192000e+03
+; CHECK-NEXT: [[MUL21:%.*]] = fmul double [[P2:%.*]], 1.638400e+04
+; CHECK-NEXT: br label [[FOR_BODY:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: [[INDVARS_IV266:%.*]] = phi i64 [ 0, [[BB1]] ], [ [[INDVARS_IV_NEXT267:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[T_0259:%.*]] = phi double [ 0.000000e+00, [[BB1]] ], [ [[ADD27:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[P3_ADDR_0258:%.*]] = phi double [ [[ADD]], [[BB1]] ], [ [[ADD28:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[VECINIT_I_I237:%.*]] = insertelement <2 x double> undef, double [[T_0259]], i32 0
+; CHECK-NEXT: [[X13:%.*]] = tail call i32 @_xfn(<2 x double> [[VECINIT_I_I237]])
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [256 x i32], [256 x i32]* [[TAB1]], i64 0, i64 [[INDVARS_IV266]]
+; CHECK-NEXT: store i32 [[X13]], i32* [[ARRAYIDX]], align 4, !tbaa !0
+; CHECK-NEXT: [[VECINIT_I_I:%.*]] = insertelement <2 x double> undef, double [[P3_ADDR_0258]], i32 0
+; CHECK-NEXT: [[X14:%.*]] = tail call i32 @_xfn(<2 x double> [[VECINIT_I_I]])
+; CHECK-NEXT: [[ARRAYIDX26:%.*]] = getelementptr inbounds [256 x i32], [256 x i32]* [[TAB2]], i64 0, i64 [[INDVARS_IV266]]
+; CHECK-NEXT: store i32 [[X14]], i32* [[ARRAYIDX26]], align 4, !tbaa !0
+; CHECK-NEXT: [[ADD27]] = fadd double [[MUL19]], [[T_0259]]
+; CHECK-NEXT: [[ADD28]] = fadd double [[MUL21]], [[P3_ADDR_0258]]
+; CHECK-NEXT: [[INDVARS_IV_NEXT267]] = add nuw nsw i64 [[INDVARS_IV266]], 1
+; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT267]], 256
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[RETURN:%.*]], label [[FOR_BODY]]
+; CHECK: return:
+; CHECK-NEXT: ret void
+;
+entry:
+ %tab1 = alloca [256 x i32], align 16
+ %tab2 = alloca [256 x i32], align 16
+ br label %bb1
+
+
+bb1:
+ %mul19 = fmul double %p1, 1.638400e+04
+ %mul20 = fmul double %p3, 1.638400e+04
+ %add = fadd double %mul20, 8.192000e+03
+ %mul21 = fmul double %p2, 1.638400e+04
+ ; The SLPVectorizer crashed when scheduling this block after it inserted an
+ ; insertelement instruction (during vectorizing the for.body block) at this position.
+ br label %for.body
+
+for.body:
+ %indvars.iv266 = phi i64 [ 0, %bb1 ], [ %indvars.iv.next267, %for.body ]
+ %t.0259 = phi double [ 0.000000e+00, %bb1 ], [ %add27, %for.body ]
+ %p3.addr.0258 = phi double [ %add, %bb1 ], [ %add28, %for.body ]
+ %vecinit.i.i237 = insertelement <2 x double> undef, double %t.0259, i32 0
+ %x13 = tail call i32 @_xfn(<2 x double> %vecinit.i.i237) #2
+ %arrayidx = getelementptr inbounds [256 x i32], [256 x i32]* %tab1, i64 0, i64 %indvars.iv266
+ store i32 %x13, i32* %arrayidx, align 4, !tbaa !4
+ %vecinit.i.i = insertelement <2 x double> undef, double %p3.addr.0258, i32 0
+ %x14 = tail call i32 @_xfn(<2 x double> %vecinit.i.i) #2
+ %arrayidx26 = getelementptr inbounds [256 x i32], [256 x i32]* %tab2, i64 0, i64 %indvars.iv266
+ store i32 %x14, i32* %arrayidx26, align 4, !tbaa !4
+ %add27 = fadd double %mul19, %t.0259
+ %add28 = fadd double %mul21, %p3.addr.0258
+ %indvars.iv.next267 = add nuw nsw i64 %indvars.iv266, 1
+ %exitcond = icmp eq i64 %indvars.iv.next267, 256
+ br i1 %exitcond, label %return, label %for.body
+
+return:
+ ret void
+}
+
+declare i32 @_xfn(<2 x double>) #4
+
+!3 = !{!"int", !4, i64 0}
+!4 = !{!3, !3, i64 0}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_sim4b1.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_sim4b1.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_sim4b1.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_sim4b1.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,191 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+%struct._exon_t.12.103.220.363.480.649.740.857.1039.1065.1078.1091.1117.1130.1156.1169.1195.1221.1234.1286.1299.1312.1338.1429.1455.1468.1494.1520.1884.1897.1975.2066.2105.2170.2171 = type { i32, i32, i32, i32, i32, i32, [8 x i8] }
+
+define void @SIM4() {
+; CHECK-LABEL: @SIM4(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 undef, label [[RETURN:%.*]], label [[LOR_LHS_FALSE:%.*]]
+; CHECK: lor.lhs.false:
+; CHECK-NEXT: br i1 undef, label [[RETURN]], label [[IF_END:%.*]]
+; CHECK: if.end:
+; CHECK-NEXT: br i1 undef, label [[FOR_END605:%.*]], label [[FOR_BODY_LR_PH:%.*]]
+; CHECK: for.body.lr.ph:
+; CHECK-NEXT: br label [[FOR_BODY:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: br i1 undef, label [[FOR_INC603:%.*]], label [[IF_END12:%.*]]
+; CHECK: if.end12:
+; CHECK-NEXT: br i1 undef, label [[LAND_LHS_TRUE:%.*]], label [[LAND_LHS_TRUE167:%.*]]
+; CHECK: land.lhs.true:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN17:%.*]], label [[LAND_LHS_TRUE167]]
+; CHECK: if.then17:
+; CHECK-NEXT: br i1 undef, label [[IF_END98:%.*]], label [[LAND_RHS_LR_PH:%.*]]
+; CHECK: land.rhs.lr.ph:
+; CHECK-NEXT: unreachable
+; CHECK: if.end98:
+; CHECK-NEXT: [[FROM299:%.*]] = getelementptr inbounds [[STRUCT__EXON_T_12_103_220_363_480_649_740_857_1039_1065_1078_1091_1117_1130_1156_1169_1195_1221_1234_1286_1299_1312_1338_1429_1455_1468_1494_1520_1884_1897_1975_2066_2105_2170_2171:%.*]], %struct._exon_t.12.103.220.363.480.649.740.857.1039.1065.1078.1091.1117.1130.1156.1169.1195.1221.1234.1286.1299.1312.1338.1429.1455.1468.1494.1520.1884.1897.1975.2066.2105.2170.2171* undef, i64 0, i32 1
+; CHECK-NEXT: br i1 undef, label [[LAND_LHS_TRUE167]], label [[IF_THEN103:%.*]]
+; CHECK: if.then103:
+; CHECK-NEXT: [[DOTSUB100:%.*]] = select i1 undef, i32 250, i32 undef
+; CHECK-NEXT: [[MUL114:%.*]] = shl nsw i32 [[DOTSUB100]], 2
+; CHECK-NEXT: [[FROM1115:%.*]] = getelementptr inbounds [[STRUCT__EXON_T_12_103_220_363_480_649_740_857_1039_1065_1078_1091_1117_1130_1156_1169_1195_1221_1234_1286_1299_1312_1338_1429_1455_1468_1494_1520_1884_1897_1975_2066_2105_2170_2171]], %struct._exon_t.12.103.220.363.480.649.740.857.1039.1065.1078.1091.1117.1130.1156.1169.1195.1221.1234.1286.1299.1312.1338.1429.1455.1468.1494.1520.1884.1897.1975.2066.2105.2170.2171* undef, i64 0, i32 0
+; CHECK-NEXT: [[COND125:%.*]] = select i1 undef, i32 undef, i32 [[MUL114]]
+; CHECK-NEXT: br label [[FOR_COND_I:%.*]]
+; CHECK: for.cond.i:
+; CHECK-NEXT: [[ROW_0_I:%.*]] = phi i32 [ undef, [[LAND_RHS_I874:%.*]] ], [ [[DOTSUB100]], [[IF_THEN103]] ]
+; CHECK-NEXT: [[COL_0_I:%.*]] = phi i32 [ undef, [[LAND_RHS_I874]] ], [ [[COND125]], [[IF_THEN103]] ]
+; CHECK-NEXT: br i1 undef, label [[LAND_RHS_I874]], label [[FOR_END_I:%.*]]
+; CHECK: land.rhs.i874:
+; CHECK-NEXT: br i1 undef, label [[FOR_COND_I]], label [[FOR_END_I]]
+; CHECK: for.end.i:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN_I:%.*]], label [[IF_END_I:%.*]]
+; CHECK: if.then.i:
+; CHECK-NEXT: [[ADD14_I:%.*]] = add nsw i32 [[ROW_0_I]], undef
+; CHECK-NEXT: [[ADD15_I:%.*]] = add nsw i32 [[COL_0_I]], undef
+; CHECK-NEXT: br label [[EXTEND_BW_EXIT:%.*]]
+; CHECK: if.end.i:
+; CHECK-NEXT: [[ADD16_I:%.*]] = add i32 [[COND125]], [[DOTSUB100]]
+; CHECK-NEXT: [[CMP26514_I:%.*]] = icmp slt i32 [[ADD16_I]], 0
+; CHECK-NEXT: br i1 [[CMP26514_I]], label [[FOR_END33_I:%.*]], label [[FOR_BODY28_LR_PH_I:%.*]]
+; CHECK: for.body28.lr.ph.i:
+; CHECK-NEXT: br label [[FOR_END33_I]]
+; CHECK: for.end33.i:
+; CHECK-NEXT: br i1 undef, label [[FOR_END58_I:%.*]], label [[FOR_BODY52_LR_PH_I:%.*]]
+; CHECK: for.body52.lr.ph.i:
+; CHECK-NEXT: br label [[FOR_END58_I]]
+; CHECK: for.end58.i:
+; CHECK-NEXT: br label [[WHILE_COND260_I:%.*]]
+; CHECK: while.cond260.i:
+; CHECK-NEXT: br i1 undef, label [[LAND_RHS263_I:%.*]], label [[WHILE_END275_I:%.*]]
+; CHECK: land.rhs263.i:
+; CHECK-NEXT: br i1 undef, label [[WHILE_COND260_I]], label [[WHILE_END275_I]]
+; CHECK: while.end275.i:
+; CHECK-NEXT: br label [[EXTEND_BW_EXIT]]
+; CHECK: extend_bw.exit:
+; CHECK-NEXT: [[ADD14_I1262:%.*]] = phi i32 [ [[ADD14_I]], [[IF_THEN_I]] ], [ undef, [[WHILE_END275_I]] ]
+; CHECK-NEXT: [[ADD15_I1261:%.*]] = phi i32 [ [[ADD15_I]], [[IF_THEN_I]] ], [ undef, [[WHILE_END275_I]] ]
+; CHECK-NEXT: br i1 false, label [[IF_THEN157:%.*]], label [[LAND_LHS_TRUE167]]
+; CHECK: if.then157:
+; CHECK-NEXT: [[ADD158:%.*]] = add nsw i32 [[ADD14_I1262]], 1
+; CHECK-NEXT: store i32 [[ADD158]], i32* [[FROM299]], align 4
+; CHECK-NEXT: [[ADD160:%.*]] = add nsw i32 [[ADD15_I1261]], 1
+; CHECK-NEXT: store i32 [[ADD160]], i32* [[FROM1115]], align 4
+; CHECK-NEXT: br label [[LAND_LHS_TRUE167]]
+; CHECK: land.lhs.true167:
+; CHECK-NEXT: unreachable
+; CHECK: for.inc603:
+; CHECK-NEXT: br i1 undef, label [[FOR_BODY]], label [[FOR_END605]]
+; CHECK: for.end605:
+; CHECK-NEXT: unreachable
+; CHECK: return:
+; CHECK-NEXT: ret void
+;
+entry:
+ br i1 undef, label %return, label %lor.lhs.false
+
+lor.lhs.false: ; preds = %entry
+ br i1 undef, label %return, label %if.end
+
+if.end: ; preds = %lor.lhs.false
+ br i1 undef, label %for.end605, label %for.body.lr.ph
+
+for.body.lr.ph: ; preds = %if.end
+ br label %for.body
+
+for.body: ; preds = %for.inc603, %for.body.lr.ph
+ br i1 undef, label %for.inc603, label %if.end12
+
+if.end12: ; preds = %for.body
+ br i1 undef, label %land.lhs.true, label %land.lhs.true167
+
+land.lhs.true: ; preds = %if.end12
+ br i1 undef, label %if.then17, label %land.lhs.true167
+
+if.then17: ; preds = %land.lhs.true
+ br i1 undef, label %if.end98, label %land.rhs.lr.ph
+
+land.rhs.lr.ph: ; preds = %if.then17
+ unreachable
+
+if.end98: ; preds = %if.then17
+ %from299 = getelementptr inbounds %struct._exon_t.12.103.220.363.480.649.740.857.1039.1065.1078.1091.1117.1130.1156.1169.1195.1221.1234.1286.1299.1312.1338.1429.1455.1468.1494.1520.1884.1897.1975.2066.2105.2170.2171, %struct._exon_t.12.103.220.363.480.649.740.857.1039.1065.1078.1091.1117.1130.1156.1169.1195.1221.1234.1286.1299.1312.1338.1429.1455.1468.1494.1520.1884.1897.1975.2066.2105.2170.2171* undef, i64 0, i32 1
+ br i1 undef, label %land.lhs.true167, label %if.then103
+
+if.then103: ; preds = %if.end98
+ %.sub100 = select i1 undef, i32 250, i32 undef
+ %mul114 = shl nsw i32 %.sub100, 2
+ %from1115 = getelementptr inbounds %struct._exon_t.12.103.220.363.480.649.740.857.1039.1065.1078.1091.1117.1130.1156.1169.1195.1221.1234.1286.1299.1312.1338.1429.1455.1468.1494.1520.1884.1897.1975.2066.2105.2170.2171, %struct._exon_t.12.103.220.363.480.649.740.857.1039.1065.1078.1091.1117.1130.1156.1169.1195.1221.1234.1286.1299.1312.1338.1429.1455.1468.1494.1520.1884.1897.1975.2066.2105.2170.2171* undef, i64 0, i32 0
+ %cond125 = select i1 undef, i32 undef, i32 %mul114
+ br label %for.cond.i
+
+for.cond.i: ; preds = %land.rhs.i874, %if.then103
+ %row.0.i = phi i32 [ undef, %land.rhs.i874 ], [ %.sub100, %if.then103 ]
+ %col.0.i = phi i32 [ undef, %land.rhs.i874 ], [ %cond125, %if.then103 ]
+ br i1 undef, label %land.rhs.i874, label %for.end.i
+
+land.rhs.i874: ; preds = %for.cond.i
+ br i1 undef, label %for.cond.i, label %for.end.i
+
+for.end.i: ; preds = %land.rhs.i874, %for.cond.i
+ br i1 undef, label %if.then.i, label %if.end.i
+
+if.then.i: ; preds = %for.end.i
+ %add14.i = add nsw i32 %row.0.i, undef
+ %add15.i = add nsw i32 %col.0.i, undef
+ br label %extend_bw.exit
+
+if.end.i: ; preds = %for.end.i
+ %add16.i = add i32 %cond125, %.sub100
+ %cmp26514.i = icmp slt i32 %add16.i, 0
+ br i1 %cmp26514.i, label %for.end33.i, label %for.body28.lr.ph.i
+
+for.body28.lr.ph.i: ; preds = %if.end.i
+ br label %for.end33.i
+
+for.end33.i: ; preds = %for.body28.lr.ph.i, %if.end.i
+ br i1 undef, label %for.end58.i, label %for.body52.lr.ph.i
+
+for.body52.lr.ph.i: ; preds = %for.end33.i
+ br label %for.end58.i
+
+for.end58.i: ; preds = %for.body52.lr.ph.i, %for.end33.i
+ br label %while.cond260.i
+
+while.cond260.i: ; preds = %land.rhs263.i, %for.end58.i
+ br i1 undef, label %land.rhs263.i, label %while.end275.i
+
+land.rhs263.i: ; preds = %while.cond260.i
+ br i1 undef, label %while.cond260.i, label %while.end275.i
+
+while.end275.i: ; preds = %land.rhs263.i, %while.cond260.i
+ br label %extend_bw.exit
+
+extend_bw.exit: ; preds = %while.end275.i, %if.then.i
+ %add14.i1262 = phi i32 [ %add14.i, %if.then.i ], [ undef, %while.end275.i ]
+ %add15.i1261 = phi i32 [ %add15.i, %if.then.i ], [ undef, %while.end275.i ]
+ br i1 false, label %if.then157, label %land.lhs.true167
+
+if.then157: ; preds = %extend_bw.exit
+ %add158 = add nsw i32 %add14.i1262, 1
+ store i32 %add158, i32* %from299, align 4
+ %add160 = add nsw i32 %add15.i1261, 1
+ store i32 %add160, i32* %from1115, align 4
+ br label %land.lhs.true167
+
+land.lhs.true167: ; preds = %if.then157, %extend_bw.exit, %if.end98, %land.lhs.true, %if.end12
+ unreachable
+
+for.inc603: ; preds = %for.body
+ br i1 undef, label %for.body, label %for.end605
+
+for.end605: ; preds = %for.inc603, %if.end
+ unreachable
+
+return: ; preds = %lor.lhs.false, %entry
+ ret void
+}
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_smallpt.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_smallpt.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_smallpt.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_smallpt.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,166 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+%struct.Ray.5.11.53.113.119.137.149.185.329.389.416 = type { %struct.Vec.0.6.48.108.114.132.144.180.324.384.414, %struct.Vec.0.6.48.108.114.132.144.180.324.384.414 }
+%struct.Vec.0.6.48.108.114.132.144.180.324.384.414 = type { double, double, double }
+
+; Function Attrs: ssp uwtable
+define void @main() #0 {
+; CHECK-LABEL: @main(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 undef, label [[COND_TRUE:%.*]], label [[COND_END:%.*]]
+; CHECK: cond.true:
+; CHECK-NEXT: unreachable
+; CHECK: cond.end:
+; CHECK-NEXT: br label [[INVOKE_CONT:%.*]]
+; CHECK: invoke.cont:
+; CHECK-NEXT: br i1 undef, label [[ARRAYCTOR_CONT:%.*]], label [[INVOKE_CONT]]
+; CHECK: arrayctor.cont:
+; CHECK-NEXT: [[AGG_TMP99208_SROA_0_0_IDX:%.*]] = getelementptr inbounds [[STRUCT_RAY_5_11_53_113_119_137_149_185_329_389_416:%.*]], %struct.Ray.5.11.53.113.119.137.149.185.329.389.416* undef, i64 0, i32 0, i32 0
+; CHECK-NEXT: [[AGG_TMP101211_SROA_0_0_IDX:%.*]] = getelementptr inbounds [[STRUCT_RAY_5_11_53_113_119_137_149_185_329_389_416]], %struct.Ray.5.11.53.113.119.137.149.185.329.389.416* undef, i64 0, i32 1, i32 0
+; CHECK-NEXT: br label [[FOR_COND36_PREHEADER:%.*]]
+; CHECK: for.cond36.preheader:
+; CHECK-NEXT: br i1 undef, label [[FOR_BODY42_LR_PH_US:%.*]], label [[_Z5CLAMPD_EXIT_1:%.*]]
+; CHECK: cond.false51.us:
+; CHECK-NEXT: unreachable
+; CHECK: cond.true48.us:
+; CHECK-NEXT: br i1 undef, label [[COND_TRUE63_US:%.*]], label [[COND_FALSE66_US:%.*]]
+; CHECK: cond.false66.us:
+; CHECK-NEXT: [[ADD_I276_US:%.*]] = fadd double 0.000000e+00, undef
+; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x double> undef, double [[ADD_I276_US]], i32 0
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x double> [[TMP0]], double undef, i32 1
+; CHECK-NEXT: [[TMP2:%.*]] = fadd <2 x double> [[TMP1]], <double 0.000000e+00, double 0xBFA5CC2D1960285F>
+; CHECK-NEXT: [[TMP3:%.*]] = fmul <2 x double> [[TMP2]], <double 1.400000e+02, double 1.400000e+02>
+; CHECK-NEXT: [[TMP4:%.*]] = fadd <2 x double> [[TMP3]], <double 5.000000e+01, double 5.200000e+01>
+; CHECK-NEXT: [[TMP5:%.*]] = fmul <2 x double> undef, [[TMP2]]
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast double* [[AGG_TMP99208_SROA_0_0_IDX]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP4]], <2 x double>* [[TMP6]], align 8
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast double* [[AGG_TMP101211_SROA_0_0_IDX]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP5]], <2 x double>* [[TMP7]], align 8
+; CHECK-NEXT: unreachable
+; CHECK: cond.true63.us:
+; CHECK-NEXT: unreachable
+; CHECK: for.body42.lr.ph.us:
+; CHECK-NEXT: br i1 undef, label [[COND_TRUE48_US:%.*]], label [[COND_FALSE51_US:%.*]]
+; CHECK: _Z5clampd.exit.1:
+; CHECK-NEXT: br label [[FOR_COND36_PREHEADER]]
+;
+entry:
+ br i1 undef, label %cond.true, label %cond.end
+
+cond.true: ; preds = %entry
+ unreachable
+
+cond.end: ; preds = %entry
+ br label %invoke.cont
+
+invoke.cont: ; preds = %invoke.cont, %cond.end
+ br i1 undef, label %arrayctor.cont, label %invoke.cont
+
+arrayctor.cont: ; preds = %invoke.cont
+ %agg.tmp99208.sroa.0.0.idx = getelementptr inbounds %struct.Ray.5.11.53.113.119.137.149.185.329.389.416, %struct.Ray.5.11.53.113.119.137.149.185.329.389.416* undef, i64 0, i32 0, i32 0
+ %agg.tmp99208.sroa.1.8.idx388 = getelementptr inbounds %struct.Ray.5.11.53.113.119.137.149.185.329.389.416, %struct.Ray.5.11.53.113.119.137.149.185.329.389.416* undef, i64 0, i32 0, i32 1
+ %agg.tmp101211.sroa.0.0.idx = getelementptr inbounds %struct.Ray.5.11.53.113.119.137.149.185.329.389.416, %struct.Ray.5.11.53.113.119.137.149.185.329.389.416* undef, i64 0, i32 1, i32 0
+ %agg.tmp101211.sroa.1.8.idx390 = getelementptr inbounds %struct.Ray.5.11.53.113.119.137.149.185.329.389.416, %struct.Ray.5.11.53.113.119.137.149.185.329.389.416* undef, i64 0, i32 1, i32 1
+ br label %for.cond36.preheader
+
+for.cond36.preheader: ; preds = %_Z5clampd.exit.1, %arrayctor.cont
+ br i1 undef, label %for.body42.lr.ph.us, label %_Z5clampd.exit.1
+
+cond.false51.us: ; preds = %for.body42.lr.ph.us
+ unreachable
+
+cond.true48.us: ; preds = %for.body42.lr.ph.us
+ br i1 undef, label %cond.true63.us, label %cond.false66.us
+
+cond.false66.us: ; preds = %cond.true48.us
+ %add.i276.us = fadd double 0.000000e+00, undef
+ %add.i264.us = fadd double %add.i276.us, 0.000000e+00
+ %add4.i267.us = fadd double undef, 0xBFA5CC2D1960285F
+ %mul.i254.us = fmul double %add.i264.us, 1.400000e+02
+ %mul2.i256.us = fmul double %add4.i267.us, 1.400000e+02
+ %add.i243.us = fadd double %mul.i254.us, 5.000000e+01
+ %add4.i246.us = fadd double %mul2.i256.us, 5.200000e+01
+ %mul.i.i.us = fmul double undef, %add.i264.us
+ %mul2.i.i.us = fmul double undef, %add4.i267.us
+ store double %add.i243.us, double* %agg.tmp99208.sroa.0.0.idx, align 8
+ store double %add4.i246.us, double* %agg.tmp99208.sroa.1.8.idx388, align 8
+ store double %mul.i.i.us, double* %agg.tmp101211.sroa.0.0.idx, align 8
+ store double %mul2.i.i.us, double* %agg.tmp101211.sroa.1.8.idx390, align 8
+ unreachable
+
+cond.true63.us: ; preds = %cond.true48.us
+ unreachable
+
+for.body42.lr.ph.us: ; preds = %for.cond36.preheader
+ br i1 undef, label %cond.true48.us, label %cond.false51.us
+
+_Z5clampd.exit.1: ; preds = %for.cond36.preheader
+ br label %for.cond36.preheader
+}
+
+
+%struct.Ray.5.11.53.95.137.191.197.203.239.257.263.269.275.281.287.293.383.437.443.455.461.599.601 = type { %struct.Vec.0.6.48.90.132.186.192.198.234.252.258.264.270.276.282.288.378.432.438.450.456.594.600, %struct.Vec.0.6.48.90.132.186.192.198.234.252.258.264.270.276.282.288.378.432.438.450.456.594.600 }
+%struct.Vec.0.6.48.90.132.186.192.198.234.252.258.264.270.276.282.288.378.432.438.450.456.594.600 = type { double, double, double }
+
+define void @_Z8radianceRK3RayiPt() #0 {
+; CHECK-LABEL: @_Z8radianceRK3RayiPt(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 undef, label [[IF_THEN78:%.*]], label [[IF_THEN38:%.*]]
+; CHECK: if.then38:
+; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x double> undef, double undef, i32 1
+; CHECK-NEXT: [[TMP1:%.*]] = fmul <2 x double> undef, [[TMP0]]
+; CHECK-NEXT: [[TMP2:%.*]] = fsub <2 x double> undef, [[TMP1]]
+; CHECK-NEXT: [[TMP3:%.*]] = fmul <2 x double> undef, [[TMP2]]
+; CHECK-NEXT: [[TMP4:%.*]] = fmul <2 x double> undef, [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = fadd <2 x double> undef, [[TMP4]]
+; CHECK-NEXT: [[TMP6:%.*]] = fadd <2 x double> undef, [[TMP5]]
+; CHECK-NEXT: [[TMP7:%.*]] = fmul <2 x double> undef, [[TMP6]]
+; CHECK-NEXT: [[AGG_TMP74663_SROA_0_0_IDX:%.*]] = getelementptr inbounds [[STRUCT_RAY_5_11_53_95_137_191_197_203_239_257_263_269_275_281_287_293_383_437_443_455_461_599_601:%.*]], %struct.Ray.5.11.53.95.137.191.197.203.239.257.263.269.275.281.287.293.383.437.443.455.461.599.601* undef, i64 0, i32 1, i32 0
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast double* [[AGG_TMP74663_SROA_0_0_IDX]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP7]], <2 x double>* [[TMP8]], align 8
+; CHECK-NEXT: br label [[RETURN:%.*]]
+; CHECK: if.then78:
+; CHECK-NEXT: br label [[RETURN]]
+; CHECK: return:
+; CHECK-NEXT: ret void
+;
+entry:
+ br i1 undef, label %if.then78, label %if.then38
+
+if.then38: ; preds = %entry
+ %mul.i.i790 = fmul double undef, undef
+ %mul3.i.i792 = fmul double undef, undef
+ %mul.i764 = fmul double undef, %mul3.i.i792
+ %mul4.i767 = fmul double undef, undef
+ %sub.i768 = fsub double %mul.i764, %mul4.i767
+ %mul6.i770 = fmul double undef, %mul.i.i790
+ %mul9.i772 = fmul double undef, %mul3.i.i792
+ %sub10.i773 = fsub double %mul6.i770, %mul9.i772
+ %mul.i736 = fmul double undef, %sub.i768
+ %mul2.i738 = fmul double undef, %sub10.i773
+ %mul.i727 = fmul double undef, %mul.i736
+ %mul2.i729 = fmul double undef, %mul2.i738
+ %add.i716 = fadd double undef, %mul.i727
+ %add4.i719 = fadd double undef, %mul2.i729
+ %add.i695 = fadd double undef, %add.i716
+ %add4.i698 = fadd double undef, %add4.i719
+ %mul.i.i679 = fmul double undef, %add.i695
+ %mul2.i.i680 = fmul double undef, %add4.i698
+ %agg.tmp74663.sroa.0.0.idx = getelementptr inbounds %struct.Ray.5.11.53.95.137.191.197.203.239.257.263.269.275.281.287.293.383.437.443.455.461.599.601, %struct.Ray.5.11.53.95.137.191.197.203.239.257.263.269.275.281.287.293.383.437.443.455.461.599.601* undef, i64 0, i32 1, i32 0
+ store double %mul.i.i679, double* %agg.tmp74663.sroa.0.0.idx, align 8
+ %agg.tmp74663.sroa.1.8.idx943 = getelementptr inbounds %struct.Ray.5.11.53.95.137.191.197.203.239.257.263.269.275.281.287.293.383.437.443.455.461.599.601, %struct.Ray.5.11.53.95.137.191.197.203.239.257.263.269.275.281.287.293.383.437.443.455.461.599.601* undef, i64 0, i32 1, i32 1
+ store double %mul2.i.i680, double* %agg.tmp74663.sroa.1.8.idx943, align 8
+ br label %return
+
+if.then78: ; preds = %entry
+ br label %return
+
+return: ; preds = %if.then78, %if.then38
+ ret void
+}
+
+attributes #0 = { ssp uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-frame-pointer-elim-non-leaf"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "unsafe-fp-math"="false" "use-soft-float"="false" }
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_vectorizeTree.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_vectorizeTree.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_vectorizeTree.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/crash_vectorizeTree.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,95 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -basicaa -slp-vectorizer -mtriple=x86_64-apple-macosx10.9.0 -mcpu=corei7-avx -S < %s | FileCheck %s
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.9.0"
+
+
+; This test used to crash because we were following phi chains incorrectly.
+; We used indices to get the incoming value of two phi nodes rather than
+; incoming block lookup.
+; This can give wrong results when the ordering of incoming
+; edges in the two phi nodes don't match.
+
+%0 = type { %1, %2 }
+%1 = type { double, double }
+%2 = type { double, double }
+
+
+;define fastcc void @bar() {
+define void @bar() {
+; CHECK-LABEL: @bar(
+; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[TMP0:%.*]], %0* undef, i64 0, i32 1, i32 0
+; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[TMP0]], %0* undef, i64 0, i32 1, i32 1
+; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[TMP0]], %0* undef, i64 0, i32 1, i32 0
+; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[TMP0]], %0* undef, i64 0, i32 1, i32 1
+; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[TMP0]], %0* undef, i64 0, i32 1, i32 0
+; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[TMP0]], %0* undef, i64 0, i32 1, i32 1
+; CHECK-NEXT: br label [[TMP7:%.*]]
+; CHECK: [[TMP8:%.*]] = phi <2 x double> [ <double 1.800000e+01, double 2.800000e+01>, [[TMP0]] ], [ [[TMP11:%.*]], [[TMP21:%.*]] ], [ [[TMP11]], [[TMP18:%.*]] ], [ [[TMP11]], [[TMP18]] ]
+; CHECK-NEXT: [[TMP9:%.*]] = bitcast double* [[TMP1]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP8]], <2 x double>* [[TMP9]], align 8
+; CHECK-NEXT: [[TMP10:%.*]] = bitcast double* [[TMP3]] to <2 x double>*
+; CHECK-NEXT: [[TMP11]] = load <2 x double>, <2 x double>* [[TMP10]], align 8
+; CHECK-NEXT: br i1 undef, label [[TMP12:%.*]], label [[TMP13:%.*]]
+; CHECK: ret void
+; CHECK: [[TMP14:%.*]] = bitcast double* [[TMP5]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP11]], <2 x double>* [[TMP14]], align 8
+; CHECK-NEXT: br i1 undef, label [[TMP15:%.*]], label [[TMP16:%.*]]
+; CHECK: br label [[TMP16]]
+; CHECK: br i1 undef, label [[TMP17:%.*]], label [[TMP18]]
+; CHECK: unreachable
+; CHECK: [[TMP19:%.*]] = extractelement <2 x double> [[TMP11]], i32 0
+; CHECK-NEXT: [[TMP20:%.*]] = extractelement <2 x double> [[TMP11]], i32 1
+; CHECK-NEXT: switch i32 undef, label [[TMP21]] [
+; CHECK-NEXT: i32 32, label [[TMP7]]
+; CHECK-NEXT: i32 103, label [[TMP7]]
+; CHECK-NEXT: ]
+; CHECK: br i1 undef, label [[TMP7]], label [[TMP22:%.*]]
+; CHECK: unreachable
+;
+ %1 = getelementptr inbounds %0, %0* undef, i64 0, i32 1, i32 0
+ %2 = getelementptr inbounds %0, %0* undef, i64 0, i32 1, i32 1
+ %3 = getelementptr inbounds %0, %0* undef, i64 0, i32 1, i32 0
+ %4 = getelementptr inbounds %0, %0* undef, i64 0, i32 1, i32 1
+ %5 = getelementptr inbounds %0, %0* undef, i64 0, i32 1, i32 0
+ %6 = getelementptr inbounds %0, %0* undef, i64 0, i32 1, i32 1
+ br label %7
+
+; <label>:7 ; preds = %18, %17, %17, %0
+ %8 = phi double [ 2.800000e+01, %0 ], [ %11, %18 ], [ %11, %17 ], [ %11, %17 ]
+ %9 = phi double [ 1.800000e+01, %0 ], [ %10, %18 ], [ %10, %17 ], [ %10, %17 ]
+ store double %9, double* %1, align 8
+ store double %8, double* %2, align 8
+ %10 = load double, double* %3, align 8
+ %11 = load double, double* %4, align 8
+ br i1 undef, label %12, label %13
+
+; <label>:12 ; preds = %7
+ ret void
+
+; <label>:13 ; preds = %7
+ store double %10, double* %5, align 8
+ store double %11, double* %6, align 8
+ br i1 undef, label %14, label %15
+
+; <label>:14 ; preds = %13
+ br label %15
+
+; <label>:15 ; preds = %14, %13
+ br i1 undef, label %16, label %17
+
+; <label>:16 ; preds = %15
+ unreachable
+
+; <label>:17 ; preds = %15
+ switch i32 undef, label %18 [
+ i32 32, label %7
+ i32 103, label %7
+ ]
+
+; <label>:18 ; preds = %17
+ br i1 undef, label %7, label %19
+
+; <label>:19 ; preds = %18
+ unreachable
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/cross_block_slp.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/cross_block_slp.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/cross_block_slp.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/cross_block_slp.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,66 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+; int foo(double *A, float *B, int g) {
+; float B0 = B[0];
+; float B1 = B[1]; <----- BasicBlock #1
+; B0 += 5;
+; B1 += 8;
+;
+; if (g) bar();
+;
+; A[0] += B0; <------- BasicBlock #3
+; A[1] += B1;
+; }
+
+
+define i32 @foo(double* nocapture %A, float* nocapture %B, i32 %g) {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast float* [[B:%.*]] to <2 x float>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x float>, <2 x float>* [[TMP0]], align 4
+; CHECK-NEXT: [[TMP2:%.*]] = fadd <2 x float> [[TMP1]], <float 5.000000e+00, float 8.000000e+00>
+; CHECK-NEXT: [[TOBOOL:%.*]] = icmp eq i32 [[G:%.*]], 0
+; CHECK-NEXT: br i1 [[TOBOOL]], label [[IF_END:%.*]], label [[IF_THEN:%.*]]
+; CHECK: if.then:
+; CHECK-NEXT: [[CALL:%.*]] = tail call i32 (...) @bar()
+; CHECK-NEXT: br label [[IF_END]]
+; CHECK: if.end:
+; CHECK-NEXT: [[TMP3:%.*]] = fpext <2 x float> [[TMP2]] to <2 x double>
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast double* [[A:%.*]] to <2 x double>*
+; CHECK-NEXT: [[TMP5:%.*]] = load <2 x double>, <2 x double>* [[TMP4]], align 8
+; CHECK-NEXT: [[TMP6:%.*]] = fadd <2 x double> [[TMP3]], [[TMP5]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast double* [[A]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP6]], <2 x double>* [[TMP7]], align 8
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %0 = load float, float* %B, align 4
+ %arrayidx1 = getelementptr inbounds float, float* %B, i64 1
+ %1 = load float, float* %arrayidx1, align 4
+ %add = fadd float %0, 5.000000e+00
+ %add2 = fadd float %1, 8.000000e+00
+ %tobool = icmp eq i32 %g, 0
+ br i1 %tobool, label %if.end, label %if.then
+
+if.then:
+ %call = tail call i32 (...) @bar()
+ br label %if.end
+
+if.end:
+ %conv = fpext float %add to double
+ %2 = load double, double* %A, align 8
+ %add4 = fadd double %conv, %2
+ store double %add4, double* %A, align 8
+ %conv5 = fpext float %add2 to double
+ %arrayidx6 = getelementptr inbounds double, double* %A, i64 1
+ %3 = load double, double* %arrayidx6, align 8
+ %add7 = fadd double %conv5, %3
+ store double %add7, double* %arrayidx6, align 8
+ ret i32 undef
+}
+
+declare i32 @bar(...)
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/cse.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/cse.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/cse.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/cse.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,363 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -S -mtriple=i386-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32-S128"
+target triple = "i386-apple-macosx10.8.0"
+
+;int test(double *G) {
+; G[0] = 1+G[5]*4;
+; G[1] = 6+G[6]*3;
+; G[2] = 7+G[5]*4;
+; G[3] = 8+G[6]*4;
+;}
+
+define i32 @test(double* nocapture %G) {
+; CHECK-LABEL: @test(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[G:%.*]], i64 5
+; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds double, double* [[G]], i64 6
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
+; CHECK-NEXT: [[TMP2:%.*]] = fmul <2 x double> [[TMP1]], <double 4.000000e+00, double 3.000000e+00>
+; CHECK-NEXT: [[TMP3:%.*]] = fadd <2 x double> [[TMP2]], <double 1.000000e+00, double 6.000000e+00>
+; CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds double, double* [[G]], i64 1
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast double* [[G]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP3]], <2 x double>* [[TMP4]], align 8
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x double> [[TMP2]], i32 0
+; CHECK-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds double, double* [[G]], i64 2
+; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x double> [[TMP1]], i32 1
+; CHECK-NEXT: [[MUL11:%.*]] = fmul double [[TMP6]], 4.000000e+00
+; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x double> undef, double [[TMP5]], i32 0
+; CHECK-NEXT: [[TMP8:%.*]] = insertelement <2 x double> [[TMP7]], double [[MUL11]], i32 1
+; CHECK-NEXT: [[TMP9:%.*]] = fadd <2 x double> [[TMP8]], <double 7.000000e+00, double 8.000000e+00>
+; CHECK-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds double, double* [[G]], i64 3
+; CHECK-NEXT: [[TMP10:%.*]] = bitcast double* [[ARRAYIDX9]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP9]], <2 x double>* [[TMP10]], align 8
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %arrayidx = getelementptr inbounds double, double* %G, i64 5
+ %0 = load double, double* %arrayidx, align 8
+ %mul = fmul double %0, 4.000000e+00
+ %add = fadd double %mul, 1.000000e+00
+ store double %add, double* %G, align 8
+ %arrayidx2 = getelementptr inbounds double, double* %G, i64 6
+ %1 = load double, double* %arrayidx2, align 8
+ %mul3 = fmul double %1, 3.000000e+00
+ %add4 = fadd double %mul3, 6.000000e+00
+ %arrayidx5 = getelementptr inbounds double, double* %G, i64 1
+ store double %add4, double* %arrayidx5, align 8
+ %add8 = fadd double %mul, 7.000000e+00
+ %arrayidx9 = getelementptr inbounds double, double* %G, i64 2
+ store double %add8, double* %arrayidx9, align 8
+ %mul11 = fmul double %1, 4.000000e+00
+ %add12 = fadd double %mul11, 8.000000e+00
+ %arrayidx13 = getelementptr inbounds double, double* %G, i64 3
+ store double %add12, double* %arrayidx13, align 8
+ ret i32 undef
+}
+
+;int foo(double *A, int n) {
+; A[0] = A[0] * 7.9 * n + 6.0;
+; A[1] = A[1] * 7.7 * n + 2.0;
+; A[2] = A[2] * 7.6 * n + 3.0;
+; A[3] = A[3] * 7.4 * n + 4.0;
+;}
+
+define i32 @foo(double* nocapture %A, i32 %n) {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[CONV:%.*]] = sitofp i32 [[N:%.*]] to double
+; CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds double, double* [[A:%.*]], i64 1
+; CHECK-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds double, double* [[A]], i64 2
+; CHECK-NEXT: [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[A]], i64 3
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[A]] to <4 x double>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* [[TMP0]], align 8
+; CHECK-NEXT: [[TMP2:%.*]] = fmul <4 x double> [[TMP1]], <double 7.900000e+00, double 7.700000e+00, double 7.600000e+00, double 7.400000e+00>
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x double> undef, double [[CONV]], i32 0
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <4 x double> [[TMP3]], double [[CONV]], i32 1
+; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x double> [[TMP4]], double [[CONV]], i32 2
+; CHECK-NEXT: [[TMP6:%.*]] = insertelement <4 x double> [[TMP5]], double [[CONV]], i32 3
+; CHECK-NEXT: [[TMP7:%.*]] = fmul <4 x double> [[TMP6]], [[TMP2]]
+; CHECK-NEXT: [[TMP8:%.*]] = fadd <4 x double> [[TMP7]], <double 6.000000e+00, double 2.000000e+00, double 3.000000e+00, double 4.000000e+00>
+; CHECK-NEXT: [[TMP9:%.*]] = bitcast double* [[A]] to <4 x double>*
+; CHECK-NEXT: store <4 x double> [[TMP8]], <4 x double>* [[TMP9]], align 8
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %0 = load double, double* %A, align 8
+ %mul = fmul double %0, 7.900000e+00
+ %conv = sitofp i32 %n to double
+ %mul1 = fmul double %conv, %mul
+ %add = fadd double %mul1, 6.000000e+00
+ store double %add, double* %A, align 8
+ %arrayidx3 = getelementptr inbounds double, double* %A, i64 1
+ %1 = load double, double* %arrayidx3, align 8
+ %mul4 = fmul double %1, 7.700000e+00
+ %mul6 = fmul double %conv, %mul4
+ %add7 = fadd double %mul6, 2.000000e+00
+ store double %add7, double* %arrayidx3, align 8
+ %arrayidx9 = getelementptr inbounds double, double* %A, i64 2
+ %2 = load double, double* %arrayidx9, align 8
+ %mul10 = fmul double %2, 7.600000e+00
+ %mul12 = fmul double %conv, %mul10
+ %add13 = fadd double %mul12, 3.000000e+00
+ store double %add13, double* %arrayidx9, align 8
+ %arrayidx15 = getelementptr inbounds double, double* %A, i64 3
+ %3 = load double, double* %arrayidx15, align 8
+ %mul16 = fmul double %3, 7.400000e+00
+ %mul18 = fmul double %conv, %mul16
+ %add19 = fadd double %mul18, 4.000000e+00
+ store double %add19, double* %arrayidx15, align 8
+ ret i32 undef
+}
+
+; int test2(double *G, int k) {
+; if (k) {
+; G[0] = 1+G[5]*4;
+; G[1] = 6+G[6]*3;
+; } else {
+; G[2] = 7+G[5]*4;
+; G[3] = 8+G[6]*3;
+; }
+; }
+
+; We can't merge the gather sequences because one does not dominate the other.
+
+define i32 @test2(double* nocapture %G, i32 %k) {
+; CHECK-LABEL: @test2(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[K:%.*]], 0
+; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds double, double* [[G:%.*]], i64 5
+; CHECK-NEXT: [[TMP3:%.*]] = load double, double* [[TMP2]], align 8
+; CHECK-NEXT: [[TMP4:%.*]] = fmul double [[TMP3]], 4.000000e+00
+; CHECK-NEXT: br i1 [[TMP1]], label [[TMP14:%.*]], label [[TMP5:%.*]]
+; CHECK: [[TMP6:%.*]] = getelementptr inbounds double, double* [[G]], i64 6
+; CHECK-NEXT: [[TMP7:%.*]] = load double, double* [[TMP6]], align 8
+; CHECK-NEXT: [[TMP8:%.*]] = fmul double [[TMP7]], 3.000000e+00
+; CHECK-NEXT: [[TMP9:%.*]] = insertelement <2 x double> undef, double [[TMP4]], i32 0
+; CHECK-NEXT: [[TMP10:%.*]] = insertelement <2 x double> [[TMP9]], double [[TMP8]], i32 1
+; CHECK-NEXT: [[TMP11:%.*]] = fadd <2 x double> [[TMP10]], <double 1.000000e+00, double 6.000000e+00>
+; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds double, double* [[G]], i64 1
+; CHECK-NEXT: [[TMP13:%.*]] = bitcast double* [[G]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP11]], <2 x double>* [[TMP13]], align 8
+; CHECK-NEXT: br label [[TMP24:%.*]]
+; CHECK: [[TMP15:%.*]] = getelementptr inbounds double, double* [[G]], i64 2
+; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds double, double* [[G]], i64 6
+; CHECK-NEXT: [[TMP17:%.*]] = load double, double* [[TMP16]], align 8
+; CHECK-NEXT: [[TMP18:%.*]] = fmul double [[TMP17]], 3.000000e+00
+; CHECK-NEXT: [[TMP19:%.*]] = insertelement <2 x double> undef, double [[TMP4]], i32 0
+; CHECK-NEXT: [[TMP20:%.*]] = insertelement <2 x double> [[TMP19]], double [[TMP18]], i32 1
+; CHECK-NEXT: [[TMP21:%.*]] = fadd <2 x double> [[TMP20]], <double 7.000000e+00, double 8.000000e+00>
+; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds double, double* [[G]], i64 3
+; CHECK-NEXT: [[TMP23:%.*]] = bitcast double* [[TMP15]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP21]], <2 x double>* [[TMP23]], align 8
+; CHECK-NEXT: br label [[TMP24]]
+; CHECK: ret i32 undef
+;
+ %1 = icmp eq i32 %k, 0
+ %2 = getelementptr inbounds double, double* %G, i64 5
+ %3 = load double, double* %2, align 8
+ %4 = fmul double %3, 4.000000e+00
+ br i1 %1, label %12, label %5
+
+; <label>:5 ; preds = %0
+ %6 = fadd double %4, 1.000000e+00
+ store double %6, double* %G, align 8
+ %7 = getelementptr inbounds double, double* %G, i64 6
+ %8 = load double, double* %7, align 8
+ %9 = fmul double %8, 3.000000e+00
+ %10 = fadd double %9, 6.000000e+00
+ %11 = getelementptr inbounds double, double* %G, i64 1
+ store double %10, double* %11, align 8
+ br label %20
+
+; <label>:12 ; preds = %0
+ %13 = fadd double %4, 7.000000e+00
+ %14 = getelementptr inbounds double, double* %G, i64 2
+ store double %13, double* %14, align 8
+ %15 = getelementptr inbounds double, double* %G, i64 6
+ %16 = load double, double* %15, align 8
+ %17 = fmul double %16, 3.000000e+00
+ %18 = fadd double %17, 8.000000e+00
+ %19 = getelementptr inbounds double, double* %G, i64 3
+ store double %18, double* %19, align 8
+ br label %20
+
+; <label>:20 ; preds = %12, %5
+ ret i32 undef
+}
+
+
+;int foo(double *A, int n) {
+; A[0] = A[0] * 7.9 * n + 6.0;
+; A[1] = A[1] * 7.9 * n + 6.0;
+; A[2] = A[2] * 7.9 * n + 6.0;
+; A[3] = A[3] * 7.9 * n + 6.0;
+;}
+
+define i32 @foo4(double* nocapture %A, i32 %n) {
+; CHECK-LABEL: @foo4(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[CONV:%.*]] = sitofp i32 [[N:%.*]] to double
+; CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds double, double* [[A:%.*]], i64 1
+; CHECK-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds double, double* [[A]], i64 2
+; CHECK-NEXT: [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[A]], i64 3
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[A]] to <4 x double>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* [[TMP0]], align 8
+; CHECK-NEXT: [[TMP2:%.*]] = fmul <4 x double> [[TMP1]], <double 7.900000e+00, double 7.900000e+00, double 7.900000e+00, double 7.900000e+00>
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x double> undef, double [[CONV]], i32 0
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <4 x double> [[TMP3]], double [[CONV]], i32 1
+; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x double> [[TMP4]], double [[CONV]], i32 2
+; CHECK-NEXT: [[TMP6:%.*]] = insertelement <4 x double> [[TMP5]], double [[CONV]], i32 3
+; CHECK-NEXT: [[TMP7:%.*]] = fmul <4 x double> [[TMP6]], [[TMP2]]
+; CHECK-NEXT: [[TMP8:%.*]] = fadd <4 x double> [[TMP7]], <double 6.000000e+00, double 6.000000e+00, double 6.000000e+00, double 6.000000e+00>
+; CHECK-NEXT: [[TMP9:%.*]] = bitcast double* [[A]] to <4 x double>*
+; CHECK-NEXT: store <4 x double> [[TMP8]], <4 x double>* [[TMP9]], align 8
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %0 = load double, double* %A, align 8
+ %mul = fmul double %0, 7.900000e+00
+ %conv = sitofp i32 %n to double
+ %mul1 = fmul double %conv, %mul
+ %add = fadd double %mul1, 6.000000e+00
+ store double %add, double* %A, align 8
+ %arrayidx3 = getelementptr inbounds double, double* %A, i64 1
+ %1 = load double, double* %arrayidx3, align 8
+ %mul4 = fmul double %1, 7.900000e+00
+ %mul6 = fmul double %conv, %mul4
+ %add7 = fadd double %mul6, 6.000000e+00
+ store double %add7, double* %arrayidx3, align 8
+ %arrayidx9 = getelementptr inbounds double, double* %A, i64 2
+ %2 = load double, double* %arrayidx9, align 8
+ %mul10 = fmul double %2, 7.900000e+00
+ %mul12 = fmul double %conv, %mul10
+ %add13 = fadd double %mul12, 6.000000e+00
+ store double %add13, double* %arrayidx9, align 8
+ %arrayidx15 = getelementptr inbounds double, double* %A, i64 3
+ %3 = load double, double* %arrayidx15, align 8
+ %mul16 = fmul double %3, 7.900000e+00
+ %mul18 = fmul double %conv, %mul16
+ %add19 = fadd double %mul18, 6.000000e+00
+ store double %add19, double* %arrayidx15, align 8
+ ret i32 undef
+}
+
+;int partial_mrg(double *A, int n) {
+; A[0] = A[0] * n;
+; A[1] = A[1] * n;
+; if (n < 4) return 0;
+; A[2] = A[2] * n;
+; A[3] = A[3] * (n+4);
+;}
+
+define i32 @partial_mrg(double* nocapture %A, i32 %n) {
+; CHECK-LABEL: @partial_mrg(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[CONV:%.*]] = sitofp i32 [[N:%.*]] to double
+; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds double, double* [[A:%.*]], i64 1
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[A]] to <2 x double>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
+; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x double> undef, double [[CONV]], i32 0
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <2 x double> [[TMP2]], double [[CONV]], i32 1
+; CHECK-NEXT: [[TMP4:%.*]] = fmul <2 x double> [[TMP3]], [[TMP1]]
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast double* [[A]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP4]], <2 x double>* [[TMP5]], align 8
+; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[N]], 4
+; CHECK-NEXT: br i1 [[CMP]], label [[RETURN:%.*]], label [[IF_END:%.*]]
+; CHECK: if.end:
+; CHECK-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds double, double* [[A]], i64 2
+; CHECK-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[A]], i64 3
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast double* [[ARRAYIDX7]] to <2 x double>*
+; CHECK-NEXT: [[TMP7:%.*]] = load <2 x double>, <2 x double>* [[TMP6]], align 8
+; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[N]], 4
+; CHECK-NEXT: [[CONV12:%.*]] = sitofp i32 [[ADD]] to double
+; CHECK-NEXT: [[TMP8:%.*]] = insertelement <2 x double> [[TMP2]], double [[CONV12]], i32 1
+; CHECK-NEXT: [[TMP9:%.*]] = fmul <2 x double> [[TMP8]], [[TMP7]]
+; CHECK-NEXT: [[TMP10:%.*]] = bitcast double* [[ARRAYIDX7]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP9]], <2 x double>* [[TMP10]], align 8
+; CHECK-NEXT: br label [[RETURN]]
+; CHECK: return:
+; CHECK-NEXT: ret i32 0
+;
+entry:
+ %0 = load double, double* %A, align 8
+ %conv = sitofp i32 %n to double
+ %mul = fmul double %conv, %0
+ store double %mul, double* %A, align 8
+ %arrayidx2 = getelementptr inbounds double, double* %A, i64 1
+ %1 = load double, double* %arrayidx2, align 8
+ %mul4 = fmul double %conv, %1
+ store double %mul4, double* %arrayidx2, align 8
+ %cmp = icmp slt i32 %n, 4
+ br i1 %cmp, label %return, label %if.end
+
+if.end: ; preds = %entry
+ %arrayidx7 = getelementptr inbounds double, double* %A, i64 2
+ %2 = load double, double* %arrayidx7, align 8
+ %mul9 = fmul double %conv, %2
+ store double %mul9, double* %arrayidx7, align 8
+ %arrayidx11 = getelementptr inbounds double, double* %A, i64 3
+ %3 = load double, double* %arrayidx11, align 8
+ %add = add nsw i32 %n, 4
+ %conv12 = sitofp i32 %add to double
+ %mul13 = fmul double %conv12, %3
+ store double %mul13, double* %arrayidx11, align 8
+ br label %return
+
+return: ; preds = %entry, %if.end
+ ret i32 0
+}
+
+%class.B.53.55 = type { %class.A.52.54, double }
+%class.A.52.54 = type { double, double, double }
+
+ at a = external global double, align 8
+
+define void @PR19646(%class.B.53.55* %this) {
+; CHECK-LABEL: @PR19646(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 undef, label [[IF_END13:%.*]], label [[IF_END13]]
+; CHECK: sw.epilog7:
+; CHECK-NEXT: [[DOTIN:%.*]] = getelementptr inbounds [[CLASS_B_53_55:%.*]], %class.B.53.55* [[THIS:%.*]], i64 0, i32 0, i32 1
+; CHECK-NEXT: [[TMP0:%.*]] = load double, double* [[DOTIN]], align 8
+; CHECK-NEXT: [[ADD:%.*]] = fadd double undef, 0.000000e+00
+; CHECK-NEXT: [[ADD6:%.*]] = fadd double [[ADD]], [[TMP0]]
+; CHECK-NEXT: [[TMP1:%.*]] = load double, double* @a, align 8
+; CHECK-NEXT: [[ADD8:%.*]] = fadd double [[TMP1]], 0.000000e+00
+; CHECK-NEXT: [[_DY:%.*]] = getelementptr inbounds [[CLASS_B_53_55]], %class.B.53.55* [[THIS]], i64 0, i32 0, i32 2
+; CHECK-NEXT: [[TMP2:%.*]] = load double, double* [[_DY]], align 8
+; CHECK-NEXT: [[ADD10:%.*]] = fadd double [[ADD8]], [[TMP2]]
+; CHECK-NEXT: br i1 undef, label [[IF_THEN12:%.*]], label [[IF_END13]]
+; CHECK: if.then12:
+; CHECK-NEXT: [[TMP3:%.*]] = load double, double* undef, align 8
+; CHECK-NEXT: br label [[IF_END13]]
+; CHECK: if.end13:
+; CHECK-NEXT: [[X_1:%.*]] = phi double [ 0.000000e+00, [[IF_THEN12]] ], [ [[ADD6]], [[SW_EPILOG7:%.*]] ], [ undef, [[ENTRY:%.*]] ], [ undef, [[ENTRY]] ]
+; CHECK-NEXT: [[B_0:%.*]] = phi double [ [[TMP3]], [[IF_THEN12]] ], [ [[ADD10]], [[SW_EPILOG7]] ], [ undef, [[ENTRY]] ], [ undef, [[ENTRY]] ]
+; CHECK-NEXT: unreachable
+;
+entry:
+ br i1 undef, label %if.end13, label %if.end13
+
+sw.epilog7: ; No predecessors!
+ %.in = getelementptr inbounds %class.B.53.55, %class.B.53.55* %this, i64 0, i32 0, i32 1
+ %0 = load double, double* %.in, align 8
+ %add = fadd double undef, 0.000000e+00
+ %add6 = fadd double %add, %0
+ %1 = load double, double* @a, align 8
+ %add8 = fadd double %1, 0.000000e+00
+ %_dy = getelementptr inbounds %class.B.53.55, %class.B.53.55* %this, i64 0, i32 0, i32 2
+ %2 = load double, double* %_dy, align 8
+ %add10 = fadd double %add8, %2
+ br i1 undef, label %if.then12, label %if.end13
+
+if.then12: ; preds = %sw.epilog7
+ %3 = load double, double* undef, align 8
+ br label %if.end13
+
+if.end13: ; preds = %if.then12, %sw.epilog7, %entry
+ %x.1 = phi double [ 0.000000e+00, %if.then12 ], [ %add6, %sw.epilog7 ], [ undef, %entry ], [ undef, %entry ]
+ %b.0 = phi double [ %3, %if.then12 ], [ %add10, %sw.epilog7 ], [ undef, %entry], [ undef, %entry ]
+ unreachable
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/ctlz.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/ctlz.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/ctlz.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/ctlz.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,905 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE2
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE42
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+ at src64 = common global [4 x i64] zeroinitializer, align 32
+ at dst64 = common global [4 x i64] zeroinitializer, align 32
+ at src32 = common global [8 x i32] zeroinitializer, align 32
+ at dst32 = common global [8 x i32] zeroinitializer, align 32
+ at src16 = common global [16 x i16] zeroinitializer, align 32
+ at dst16 = common global [16 x i16] zeroinitializer, align 32
+ at src8 = common global [32 x i8] zeroinitializer, align 32
+ at dst8 = common global [32 x i8] zeroinitializer, align 32
+
+declare i64 @llvm.ctlz.i64(i64, i1)
+declare i32 @llvm.ctlz.i32(i32, i1)
+declare i16 @llvm.ctlz.i16(i16, i1)
+declare i8 @llvm.ctlz.i8(i8, i1)
+
+;
+; CTLZ
+;
+
+define void @ctlz_2i64() #0 {
+; CHECK-LABEL: @ctlz_2i64(
+; CHECK-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+; CHECK-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+; CHECK-NEXT: [[CTLZ0:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD0]], i1 false)
+; CHECK-NEXT: [[CTLZ1:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD1]], i1 false)
+; CHECK-NEXT: store i64 [[CTLZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+; CHECK-NEXT: store i64 [[CTLZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+ %ctlz0 = call i64 @llvm.ctlz.i64(i64 %ld0, i1 0)
+ %ctlz1 = call i64 @llvm.ctlz.i64(i64 %ld1, i1 0)
+ store i64 %ctlz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+ store i64 %ctlz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+ ret void
+}
+
+define void @ctlz_4i64() #0 {
+; CHECK-LABEL: @ctlz_4i64(
+; CHECK-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+; CHECK-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+; CHECK-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+; CHECK-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+; CHECK-NEXT: [[CTLZ0:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD0]], i1 false)
+; CHECK-NEXT: [[CTLZ1:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD1]], i1 false)
+; CHECK-NEXT: [[CTLZ2:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD2]], i1 false)
+; CHECK-NEXT: [[CTLZ3:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD3]], i1 false)
+; CHECK-NEXT: store i64 [[CTLZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+; CHECK-NEXT: store i64 [[CTLZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+; CHECK-NEXT: store i64 [[CTLZ2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+; CHECK-NEXT: store i64 [[CTLZ3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+ %ld2 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+ %ld3 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+ %ctlz0 = call i64 @llvm.ctlz.i64(i64 %ld0, i1 0)
+ %ctlz1 = call i64 @llvm.ctlz.i64(i64 %ld1, i1 0)
+ %ctlz2 = call i64 @llvm.ctlz.i64(i64 %ld2, i1 0)
+ %ctlz3 = call i64 @llvm.ctlz.i64(i64 %ld3, i1 0)
+ store i64 %ctlz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+ store i64 %ctlz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+ store i64 %ctlz2, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+ store i64 %ctlz3, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+ ret void
+}
+
+define void @ctlz_4i32() #0 {
+; CHECK-LABEL: @ctlz_4i32(
+; CHECK-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+; CHECK-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+; CHECK-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+; CHECK-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+; CHECK-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 false)
+; CHECK-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 false)
+; CHECK-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 false)
+; CHECK-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 false)
+; CHECK-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+; CHECK-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+; CHECK-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+; CHECK-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+ %ctlz0 = call i32 @llvm.ctlz.i32(i32 %ld0, i1 0)
+ %ctlz1 = call i32 @llvm.ctlz.i32(i32 %ld1, i1 0)
+ %ctlz2 = call i32 @llvm.ctlz.i32(i32 %ld2, i1 0)
+ %ctlz3 = call i32 @llvm.ctlz.i32(i32 %ld3, i1 0)
+ store i32 %ctlz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+ store i32 %ctlz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+ store i32 %ctlz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+ store i32 %ctlz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+ ret void
+}
+
+define void @ctlz_8i32() #0 {
+; SSE-LABEL: @ctlz_8i32(
+; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+; SSE-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+; SSE-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+; SSE-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+; SSE-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+; SSE-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 false)
+; SSE-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 false)
+; SSE-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 false)
+; SSE-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 false)
+; SSE-NEXT: [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 false)
+; SSE-NEXT: [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 false)
+; SSE-NEXT: [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 false)
+; SSE-NEXT: [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 false)
+; SSE-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+; SSE-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+; SSE-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+; SSE-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+; SSE-NEXT: store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+; SSE-NEXT: store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+; SSE-NEXT: store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+; SSE-NEXT: store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+; SSE-NEXT: ret void
+;
+; AVX1-LABEL: @ctlz_8i32(
+; AVX1-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+; AVX1-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+; AVX1-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+; AVX1-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+; AVX1-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+; AVX1-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+; AVX1-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+; AVX1-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+; AVX1-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 false)
+; AVX1-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 false)
+; AVX1-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 false)
+; AVX1-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 false)
+; AVX1-NEXT: [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 false)
+; AVX1-NEXT: [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 false)
+; AVX1-NEXT: [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 false)
+; AVX1-NEXT: [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 false)
+; AVX1-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+; AVX1-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+; AVX1-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+; AVX1-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+; AVX1-NEXT: store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+; AVX1-NEXT: store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+; AVX1-NEXT: store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+; AVX1-NEXT: store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+; AVX1-NEXT: ret void
+;
+; AVX2-LABEL: @ctlz_8i32(
+; AVX2-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2
+; AVX2-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> [[TMP1]], i1 false)
+; AVX2-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2
+; AVX2-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+ %ld4 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+ %ld5 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+ %ld6 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+ %ld7 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+ %ctlz0 = call i32 @llvm.ctlz.i32(i32 %ld0, i1 0)
+ %ctlz1 = call i32 @llvm.ctlz.i32(i32 %ld1, i1 0)
+ %ctlz2 = call i32 @llvm.ctlz.i32(i32 %ld2, i1 0)
+ %ctlz3 = call i32 @llvm.ctlz.i32(i32 %ld3, i1 0)
+ %ctlz4 = call i32 @llvm.ctlz.i32(i32 %ld4, i1 0)
+ %ctlz5 = call i32 @llvm.ctlz.i32(i32 %ld5, i1 0)
+ %ctlz6 = call i32 @llvm.ctlz.i32(i32 %ld6, i1 0)
+ %ctlz7 = call i32 @llvm.ctlz.i32(i32 %ld7, i1 0)
+ store i32 %ctlz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+ store i32 %ctlz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+ store i32 %ctlz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+ store i32 %ctlz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+ store i32 %ctlz4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+ store i32 %ctlz5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+ store i32 %ctlz6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+ store i32 %ctlz7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+ ret void
+}
+
+define void @ctlz_8i16() #0 {
+; CHECK-LABEL: @ctlz_8i16(
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 false)
+; CHECK-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %ctlz0 = call i16 @llvm.ctlz.i16(i16 %ld0, i1 0)
+ %ctlz1 = call i16 @llvm.ctlz.i16(i16 %ld1, i1 0)
+ %ctlz2 = call i16 @llvm.ctlz.i16(i16 %ld2, i1 0)
+ %ctlz3 = call i16 @llvm.ctlz.i16(i16 %ld3, i1 0)
+ %ctlz4 = call i16 @llvm.ctlz.i16(i16 %ld4, i1 0)
+ %ctlz5 = call i16 @llvm.ctlz.i16(i16 %ld5, i1 0)
+ %ctlz6 = call i16 @llvm.ctlz.i16(i16 %ld6, i1 0)
+ %ctlz7 = call i16 @llvm.ctlz.i16(i16 %ld7, i1 0)
+ store i16 %ctlz0, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %ctlz1, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %ctlz2, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %ctlz3, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %ctlz4, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %ctlz5, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %ctlz6, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %ctlz7, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ ret void
+}
+
+define void @ctlz_16i16() #0 {
+; SSE-LABEL: @ctlz_16i16(
+; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 false)
+; SSE-NEXT: [[TMP4:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP2]], i1 false)
+; SSE-NEXT: store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @ctlz_16i16(
+; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> [[TMP1]], i1 false)
+; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2
+; AVX-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %ld8 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2
+ %ld9 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2
+ %ld10 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2
+ %ld11 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2
+ %ld12 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2
+ %ld13 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2
+ %ld14 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2
+ %ld15 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2
+ %ctlz0 = call i16 @llvm.ctlz.i16(i16 %ld0, i1 0)
+ %ctlz1 = call i16 @llvm.ctlz.i16(i16 %ld1, i1 0)
+ %ctlz2 = call i16 @llvm.ctlz.i16(i16 %ld2, i1 0)
+ %ctlz3 = call i16 @llvm.ctlz.i16(i16 %ld3, i1 0)
+ %ctlz4 = call i16 @llvm.ctlz.i16(i16 %ld4, i1 0)
+ %ctlz5 = call i16 @llvm.ctlz.i16(i16 %ld5, i1 0)
+ %ctlz6 = call i16 @llvm.ctlz.i16(i16 %ld6, i1 0)
+ %ctlz7 = call i16 @llvm.ctlz.i16(i16 %ld7, i1 0)
+ %ctlz8 = call i16 @llvm.ctlz.i16(i16 %ld8, i1 0)
+ %ctlz9 = call i16 @llvm.ctlz.i16(i16 %ld9, i1 0)
+ %ctlz10 = call i16 @llvm.ctlz.i16(i16 %ld10, i1 0)
+ %ctlz11 = call i16 @llvm.ctlz.i16(i16 %ld11, i1 0)
+ %ctlz12 = call i16 @llvm.ctlz.i16(i16 %ld12, i1 0)
+ %ctlz13 = call i16 @llvm.ctlz.i16(i16 %ld13, i1 0)
+ %ctlz14 = call i16 @llvm.ctlz.i16(i16 %ld14, i1 0)
+ %ctlz15 = call i16 @llvm.ctlz.i16(i16 %ld15, i1 0)
+ store i16 %ctlz0 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %ctlz1 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %ctlz2 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %ctlz3 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %ctlz4 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %ctlz5 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %ctlz6 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %ctlz7 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ store i16 %ctlz8 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2
+ store i16 %ctlz9 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2
+ store i16 %ctlz10, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2
+ store i16 %ctlz11, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2
+ store i16 %ctlz12, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2
+ store i16 %ctlz13, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2
+ store i16 %ctlz14, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2
+ store i16 %ctlz15, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2
+ ret void
+}
+
+define void @ctlz_16i8() #0 {
+; CHECK-LABEL: @ctlz_16i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 false)
+; CHECK-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %ctlz0 = call i8 @llvm.ctlz.i8(i8 %ld0, i1 0)
+ %ctlz1 = call i8 @llvm.ctlz.i8(i8 %ld1, i1 0)
+ %ctlz2 = call i8 @llvm.ctlz.i8(i8 %ld2, i1 0)
+ %ctlz3 = call i8 @llvm.ctlz.i8(i8 %ld3, i1 0)
+ %ctlz4 = call i8 @llvm.ctlz.i8(i8 %ld4, i1 0)
+ %ctlz5 = call i8 @llvm.ctlz.i8(i8 %ld5, i1 0)
+ %ctlz6 = call i8 @llvm.ctlz.i8(i8 %ld6, i1 0)
+ %ctlz7 = call i8 @llvm.ctlz.i8(i8 %ld7, i1 0)
+ %ctlz8 = call i8 @llvm.ctlz.i8(i8 %ld8, i1 0)
+ %ctlz9 = call i8 @llvm.ctlz.i8(i8 %ld9, i1 0)
+ %ctlz10 = call i8 @llvm.ctlz.i8(i8 %ld10, i1 0)
+ %ctlz11 = call i8 @llvm.ctlz.i8(i8 %ld11, i1 0)
+ %ctlz12 = call i8 @llvm.ctlz.i8(i8 %ld12, i1 0)
+ %ctlz13 = call i8 @llvm.ctlz.i8(i8 %ld13, i1 0)
+ %ctlz14 = call i8 @llvm.ctlz.i8(i8 %ld14, i1 0)
+ %ctlz15 = call i8 @llvm.ctlz.i8(i8 %ld15, i1 0)
+ store i8 %ctlz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %ctlz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %ctlz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %ctlz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %ctlz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %ctlz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %ctlz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %ctlz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %ctlz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %ctlz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %ctlz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %ctlz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %ctlz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %ctlz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %ctlz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %ctlz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ ret void
+}
+
+define void @ctlz_32i8() #0 {
+; CHECK-LABEL: @ctlz_32i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 false)
+; CHECK-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP2]], i1 false)
+; CHECK-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %ld16 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1
+ %ld17 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1
+ %ld18 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1
+ %ld19 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1
+ %ld20 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1
+ %ld21 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1
+ %ld22 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1
+ %ld23 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1
+ %ld24 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1
+ %ld25 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1
+ %ld26 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1
+ %ld27 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1
+ %ld28 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1
+ %ld29 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1
+ %ld30 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1
+ %ld31 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1
+ %ctlz0 = call i8 @llvm.ctlz.i8(i8 %ld0, i1 0)
+ %ctlz1 = call i8 @llvm.ctlz.i8(i8 %ld1, i1 0)
+ %ctlz2 = call i8 @llvm.ctlz.i8(i8 %ld2, i1 0)
+ %ctlz3 = call i8 @llvm.ctlz.i8(i8 %ld3, i1 0)
+ %ctlz4 = call i8 @llvm.ctlz.i8(i8 %ld4, i1 0)
+ %ctlz5 = call i8 @llvm.ctlz.i8(i8 %ld5, i1 0)
+ %ctlz6 = call i8 @llvm.ctlz.i8(i8 %ld6, i1 0)
+ %ctlz7 = call i8 @llvm.ctlz.i8(i8 %ld7, i1 0)
+ %ctlz8 = call i8 @llvm.ctlz.i8(i8 %ld8, i1 0)
+ %ctlz9 = call i8 @llvm.ctlz.i8(i8 %ld9, i1 0)
+ %ctlz10 = call i8 @llvm.ctlz.i8(i8 %ld10, i1 0)
+ %ctlz11 = call i8 @llvm.ctlz.i8(i8 %ld11, i1 0)
+ %ctlz12 = call i8 @llvm.ctlz.i8(i8 %ld12, i1 0)
+ %ctlz13 = call i8 @llvm.ctlz.i8(i8 %ld13, i1 0)
+ %ctlz14 = call i8 @llvm.ctlz.i8(i8 %ld14, i1 0)
+ %ctlz15 = call i8 @llvm.ctlz.i8(i8 %ld15, i1 0)
+ %ctlz16 = call i8 @llvm.ctlz.i8(i8 %ld16, i1 0)
+ %ctlz17 = call i8 @llvm.ctlz.i8(i8 %ld17, i1 0)
+ %ctlz18 = call i8 @llvm.ctlz.i8(i8 %ld18, i1 0)
+ %ctlz19 = call i8 @llvm.ctlz.i8(i8 %ld19, i1 0)
+ %ctlz20 = call i8 @llvm.ctlz.i8(i8 %ld20, i1 0)
+ %ctlz21 = call i8 @llvm.ctlz.i8(i8 %ld21, i1 0)
+ %ctlz22 = call i8 @llvm.ctlz.i8(i8 %ld22, i1 0)
+ %ctlz23 = call i8 @llvm.ctlz.i8(i8 %ld23, i1 0)
+ %ctlz24 = call i8 @llvm.ctlz.i8(i8 %ld24, i1 0)
+ %ctlz25 = call i8 @llvm.ctlz.i8(i8 %ld25, i1 0)
+ %ctlz26 = call i8 @llvm.ctlz.i8(i8 %ld26, i1 0)
+ %ctlz27 = call i8 @llvm.ctlz.i8(i8 %ld27, i1 0)
+ %ctlz28 = call i8 @llvm.ctlz.i8(i8 %ld28, i1 0)
+ %ctlz29 = call i8 @llvm.ctlz.i8(i8 %ld29, i1 0)
+ %ctlz30 = call i8 @llvm.ctlz.i8(i8 %ld30, i1 0)
+ %ctlz31 = call i8 @llvm.ctlz.i8(i8 %ld31, i1 0)
+ store i8 %ctlz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %ctlz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %ctlz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %ctlz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %ctlz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %ctlz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %ctlz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %ctlz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %ctlz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %ctlz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %ctlz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %ctlz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %ctlz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %ctlz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %ctlz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %ctlz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ store i8 %ctlz16, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1
+ store i8 %ctlz17, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1
+ store i8 %ctlz18, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1
+ store i8 %ctlz19, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1
+ store i8 %ctlz20, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1
+ store i8 %ctlz21, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1
+ store i8 %ctlz22, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1
+ store i8 %ctlz23, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1
+ store i8 %ctlz24, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1
+ store i8 %ctlz25, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1
+ store i8 %ctlz26, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1
+ store i8 %ctlz27, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1
+ store i8 %ctlz28, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1
+ store i8 %ctlz29, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1
+ store i8 %ctlz30, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1
+ store i8 %ctlz31, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1
+ ret void
+}
+
+;
+; CTLZ_ZERO_UNDEF
+;
+
+define void @ctlz_undef_2i64() #0 {
+; CHECK-LABEL: @ctlz_undef_2i64(
+; CHECK-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+; CHECK-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+; CHECK-NEXT: [[CTLZ0:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD0]], i1 true)
+; CHECK-NEXT: [[CTLZ1:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD1]], i1 true)
+; CHECK-NEXT: store i64 [[CTLZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+; CHECK-NEXT: store i64 [[CTLZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+ %ctlz0 = call i64 @llvm.ctlz.i64(i64 %ld0, i1 -1)
+ %ctlz1 = call i64 @llvm.ctlz.i64(i64 %ld1, i1 -1)
+ store i64 %ctlz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+ store i64 %ctlz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+ ret void
+}
+
+define void @ctlz_undef_4i64() #0 {
+; CHECK-LABEL: @ctlz_undef_4i64(
+; CHECK-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+; CHECK-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+; CHECK-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+; CHECK-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+; CHECK-NEXT: [[CTLZ0:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD0]], i1 true)
+; CHECK-NEXT: [[CTLZ1:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD1]], i1 true)
+; CHECK-NEXT: [[CTLZ2:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD2]], i1 true)
+; CHECK-NEXT: [[CTLZ3:%.*]] = call i64 @llvm.ctlz.i64(i64 [[LD3]], i1 true)
+; CHECK-NEXT: store i64 [[CTLZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+; CHECK-NEXT: store i64 [[CTLZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+; CHECK-NEXT: store i64 [[CTLZ2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+; CHECK-NEXT: store i64 [[CTLZ3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+ %ld2 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+ %ld3 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+ %ctlz0 = call i64 @llvm.ctlz.i64(i64 %ld0, i1 -1)
+ %ctlz1 = call i64 @llvm.ctlz.i64(i64 %ld1, i1 -1)
+ %ctlz2 = call i64 @llvm.ctlz.i64(i64 %ld2, i1 -1)
+ %ctlz3 = call i64 @llvm.ctlz.i64(i64 %ld3, i1 -1)
+ store i64 %ctlz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+ store i64 %ctlz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+ store i64 %ctlz2, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+ store i64 %ctlz3, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+ ret void
+}
+
+define void @ctlz_undef_4i32() #0 {
+; CHECK-LABEL: @ctlz_undef_4i32(
+; CHECK-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+; CHECK-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+; CHECK-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+; CHECK-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+; CHECK-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 true)
+; CHECK-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 true)
+; CHECK-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 true)
+; CHECK-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 true)
+; CHECK-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+; CHECK-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+; CHECK-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+; CHECK-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+ %ctlz0 = call i32 @llvm.ctlz.i32(i32 %ld0, i1 -1)
+ %ctlz1 = call i32 @llvm.ctlz.i32(i32 %ld1, i1 -1)
+ %ctlz2 = call i32 @llvm.ctlz.i32(i32 %ld2, i1 -1)
+ %ctlz3 = call i32 @llvm.ctlz.i32(i32 %ld3, i1 -1)
+ store i32 %ctlz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+ store i32 %ctlz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+ store i32 %ctlz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+ store i32 %ctlz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+ ret void
+}
+
+define void @ctlz_undef_8i32() #0 {
+; SSE-LABEL: @ctlz_undef_8i32(
+; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+; SSE-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+; SSE-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+; SSE-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+; SSE-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+; SSE-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 true)
+; SSE-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 true)
+; SSE-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 true)
+; SSE-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 true)
+; SSE-NEXT: [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 true)
+; SSE-NEXT: [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 true)
+; SSE-NEXT: [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 true)
+; SSE-NEXT: [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 true)
+; SSE-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+; SSE-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+; SSE-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+; SSE-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+; SSE-NEXT: store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+; SSE-NEXT: store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+; SSE-NEXT: store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+; SSE-NEXT: store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+; SSE-NEXT: ret void
+;
+; AVX1-LABEL: @ctlz_undef_8i32(
+; AVX1-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+; AVX1-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+; AVX1-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+; AVX1-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+; AVX1-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+; AVX1-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+; AVX1-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+; AVX1-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+; AVX1-NEXT: [[CTLZ0:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD0]], i1 true)
+; AVX1-NEXT: [[CTLZ1:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD1]], i1 true)
+; AVX1-NEXT: [[CTLZ2:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD2]], i1 true)
+; AVX1-NEXT: [[CTLZ3:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD3]], i1 true)
+; AVX1-NEXT: [[CTLZ4:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD4]], i1 true)
+; AVX1-NEXT: [[CTLZ5:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD5]], i1 true)
+; AVX1-NEXT: [[CTLZ6:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD6]], i1 true)
+; AVX1-NEXT: [[CTLZ7:%.*]] = call i32 @llvm.ctlz.i32(i32 [[LD7]], i1 true)
+; AVX1-NEXT: store i32 [[CTLZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+; AVX1-NEXT: store i32 [[CTLZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+; AVX1-NEXT: store i32 [[CTLZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+; AVX1-NEXT: store i32 [[CTLZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+; AVX1-NEXT: store i32 [[CTLZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+; AVX1-NEXT: store i32 [[CTLZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+; AVX1-NEXT: store i32 [[CTLZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+; AVX1-NEXT: store i32 [[CTLZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+; AVX1-NEXT: ret void
+;
+; AVX2-LABEL: @ctlz_undef_8i32(
+; AVX2-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2
+; AVX2-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> [[TMP1]], i1 true)
+; AVX2-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2
+; AVX2-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+ %ld4 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+ %ld5 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+ %ld6 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+ %ld7 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+ %ctlz0 = call i32 @llvm.ctlz.i32(i32 %ld0, i1 -1)
+ %ctlz1 = call i32 @llvm.ctlz.i32(i32 %ld1, i1 -1)
+ %ctlz2 = call i32 @llvm.ctlz.i32(i32 %ld2, i1 -1)
+ %ctlz3 = call i32 @llvm.ctlz.i32(i32 %ld3, i1 -1)
+ %ctlz4 = call i32 @llvm.ctlz.i32(i32 %ld4, i1 -1)
+ %ctlz5 = call i32 @llvm.ctlz.i32(i32 %ld5, i1 -1)
+ %ctlz6 = call i32 @llvm.ctlz.i32(i32 %ld6, i1 -1)
+ %ctlz7 = call i32 @llvm.ctlz.i32(i32 %ld7, i1 -1)
+ store i32 %ctlz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+ store i32 %ctlz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+ store i32 %ctlz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+ store i32 %ctlz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+ store i32 %ctlz4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+ store i32 %ctlz5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+ store i32 %ctlz6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+ store i32 %ctlz7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+ ret void
+}
+
+define void @ctlz_undef_8i16() #0 {
+; CHECK-LABEL: @ctlz_undef_8i16(
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 true)
+; CHECK-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %ctlz0 = call i16 @llvm.ctlz.i16(i16 %ld0, i1 -1)
+ %ctlz1 = call i16 @llvm.ctlz.i16(i16 %ld1, i1 -1)
+ %ctlz2 = call i16 @llvm.ctlz.i16(i16 %ld2, i1 -1)
+ %ctlz3 = call i16 @llvm.ctlz.i16(i16 %ld3, i1 -1)
+ %ctlz4 = call i16 @llvm.ctlz.i16(i16 %ld4, i1 -1)
+ %ctlz5 = call i16 @llvm.ctlz.i16(i16 %ld5, i1 -1)
+ %ctlz6 = call i16 @llvm.ctlz.i16(i16 %ld6, i1 -1)
+ %ctlz7 = call i16 @llvm.ctlz.i16(i16 %ld7, i1 -1)
+ store i16 %ctlz0, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %ctlz1, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %ctlz2, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %ctlz3, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %ctlz4, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %ctlz5, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %ctlz6, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %ctlz7, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ ret void
+}
+
+define void @ctlz_undef_16i16() #0 {
+; SSE-LABEL: @ctlz_undef_16i16(
+; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP1]], i1 true)
+; SSE-NEXT: [[TMP4:%.*]] = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> [[TMP2]], i1 true)
+; SSE-NEXT: store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @ctlz_undef_16i16(
+; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> [[TMP1]], i1 true)
+; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2
+; AVX-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %ld8 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2
+ %ld9 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2
+ %ld10 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2
+ %ld11 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2
+ %ld12 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2
+ %ld13 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2
+ %ld14 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2
+ %ld15 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2
+ %ctlz0 = call i16 @llvm.ctlz.i16(i16 %ld0, i1 -1)
+ %ctlz1 = call i16 @llvm.ctlz.i16(i16 %ld1, i1 -1)
+ %ctlz2 = call i16 @llvm.ctlz.i16(i16 %ld2, i1 -1)
+ %ctlz3 = call i16 @llvm.ctlz.i16(i16 %ld3, i1 -1)
+ %ctlz4 = call i16 @llvm.ctlz.i16(i16 %ld4, i1 -1)
+ %ctlz5 = call i16 @llvm.ctlz.i16(i16 %ld5, i1 -1)
+ %ctlz6 = call i16 @llvm.ctlz.i16(i16 %ld6, i1 -1)
+ %ctlz7 = call i16 @llvm.ctlz.i16(i16 %ld7, i1 -1)
+ %ctlz8 = call i16 @llvm.ctlz.i16(i16 %ld8, i1 -1)
+ %ctlz9 = call i16 @llvm.ctlz.i16(i16 %ld9, i1 -1)
+ %ctlz10 = call i16 @llvm.ctlz.i16(i16 %ld10, i1 -1)
+ %ctlz11 = call i16 @llvm.ctlz.i16(i16 %ld11, i1 -1)
+ %ctlz12 = call i16 @llvm.ctlz.i16(i16 %ld12, i1 -1)
+ %ctlz13 = call i16 @llvm.ctlz.i16(i16 %ld13, i1 -1)
+ %ctlz14 = call i16 @llvm.ctlz.i16(i16 %ld14, i1 -1)
+ %ctlz15 = call i16 @llvm.ctlz.i16(i16 %ld15, i1 -1)
+ store i16 %ctlz0 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %ctlz1 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %ctlz2 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %ctlz3 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %ctlz4 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %ctlz5 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %ctlz6 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %ctlz7 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ store i16 %ctlz8 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2
+ store i16 %ctlz9 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2
+ store i16 %ctlz10, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2
+ store i16 %ctlz11, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2
+ store i16 %ctlz12, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2
+ store i16 %ctlz13, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2
+ store i16 %ctlz14, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2
+ store i16 %ctlz15, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2
+ ret void
+}
+
+define void @ctlz_undef_16i8() #0 {
+; CHECK-LABEL: @ctlz_undef_16i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 true)
+; CHECK-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %ctlz0 = call i8 @llvm.ctlz.i8(i8 %ld0, i1 -1)
+ %ctlz1 = call i8 @llvm.ctlz.i8(i8 %ld1, i1 -1)
+ %ctlz2 = call i8 @llvm.ctlz.i8(i8 %ld2, i1 -1)
+ %ctlz3 = call i8 @llvm.ctlz.i8(i8 %ld3, i1 -1)
+ %ctlz4 = call i8 @llvm.ctlz.i8(i8 %ld4, i1 -1)
+ %ctlz5 = call i8 @llvm.ctlz.i8(i8 %ld5, i1 -1)
+ %ctlz6 = call i8 @llvm.ctlz.i8(i8 %ld6, i1 -1)
+ %ctlz7 = call i8 @llvm.ctlz.i8(i8 %ld7, i1 -1)
+ %ctlz8 = call i8 @llvm.ctlz.i8(i8 %ld8, i1 -1)
+ %ctlz9 = call i8 @llvm.ctlz.i8(i8 %ld9, i1 -1)
+ %ctlz10 = call i8 @llvm.ctlz.i8(i8 %ld10, i1 -1)
+ %ctlz11 = call i8 @llvm.ctlz.i8(i8 %ld11, i1 -1)
+ %ctlz12 = call i8 @llvm.ctlz.i8(i8 %ld12, i1 -1)
+ %ctlz13 = call i8 @llvm.ctlz.i8(i8 %ld13, i1 -1)
+ %ctlz14 = call i8 @llvm.ctlz.i8(i8 %ld14, i1 -1)
+ %ctlz15 = call i8 @llvm.ctlz.i8(i8 %ld15, i1 -1)
+ store i8 %ctlz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %ctlz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %ctlz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %ctlz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %ctlz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %ctlz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %ctlz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %ctlz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %ctlz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %ctlz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %ctlz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %ctlz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %ctlz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %ctlz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %ctlz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %ctlz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ ret void
+}
+
+define void @ctlz_undef_32i8() #0 {
+; CHECK-LABEL: @ctlz_undef_32i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP1]], i1 true)
+; CHECK-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> [[TMP2]], i1 true)
+; CHECK-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %ld16 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1
+ %ld17 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1
+ %ld18 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1
+ %ld19 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1
+ %ld20 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1
+ %ld21 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1
+ %ld22 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1
+ %ld23 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1
+ %ld24 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1
+ %ld25 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1
+ %ld26 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1
+ %ld27 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1
+ %ld28 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1
+ %ld29 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1
+ %ld30 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1
+ %ld31 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1
+ %ctlz0 = call i8 @llvm.ctlz.i8(i8 %ld0, i1 -1)
+ %ctlz1 = call i8 @llvm.ctlz.i8(i8 %ld1, i1 -1)
+ %ctlz2 = call i8 @llvm.ctlz.i8(i8 %ld2, i1 -1)
+ %ctlz3 = call i8 @llvm.ctlz.i8(i8 %ld3, i1 -1)
+ %ctlz4 = call i8 @llvm.ctlz.i8(i8 %ld4, i1 -1)
+ %ctlz5 = call i8 @llvm.ctlz.i8(i8 %ld5, i1 -1)
+ %ctlz6 = call i8 @llvm.ctlz.i8(i8 %ld6, i1 -1)
+ %ctlz7 = call i8 @llvm.ctlz.i8(i8 %ld7, i1 -1)
+ %ctlz8 = call i8 @llvm.ctlz.i8(i8 %ld8, i1 -1)
+ %ctlz9 = call i8 @llvm.ctlz.i8(i8 %ld9, i1 -1)
+ %ctlz10 = call i8 @llvm.ctlz.i8(i8 %ld10, i1 -1)
+ %ctlz11 = call i8 @llvm.ctlz.i8(i8 %ld11, i1 -1)
+ %ctlz12 = call i8 @llvm.ctlz.i8(i8 %ld12, i1 -1)
+ %ctlz13 = call i8 @llvm.ctlz.i8(i8 %ld13, i1 -1)
+ %ctlz14 = call i8 @llvm.ctlz.i8(i8 %ld14, i1 -1)
+ %ctlz15 = call i8 @llvm.ctlz.i8(i8 %ld15, i1 -1)
+ %ctlz16 = call i8 @llvm.ctlz.i8(i8 %ld16, i1 -1)
+ %ctlz17 = call i8 @llvm.ctlz.i8(i8 %ld17, i1 -1)
+ %ctlz18 = call i8 @llvm.ctlz.i8(i8 %ld18, i1 -1)
+ %ctlz19 = call i8 @llvm.ctlz.i8(i8 %ld19, i1 -1)
+ %ctlz20 = call i8 @llvm.ctlz.i8(i8 %ld20, i1 -1)
+ %ctlz21 = call i8 @llvm.ctlz.i8(i8 %ld21, i1 -1)
+ %ctlz22 = call i8 @llvm.ctlz.i8(i8 %ld22, i1 -1)
+ %ctlz23 = call i8 @llvm.ctlz.i8(i8 %ld23, i1 -1)
+ %ctlz24 = call i8 @llvm.ctlz.i8(i8 %ld24, i1 -1)
+ %ctlz25 = call i8 @llvm.ctlz.i8(i8 %ld25, i1 -1)
+ %ctlz26 = call i8 @llvm.ctlz.i8(i8 %ld26, i1 -1)
+ %ctlz27 = call i8 @llvm.ctlz.i8(i8 %ld27, i1 -1)
+ %ctlz28 = call i8 @llvm.ctlz.i8(i8 %ld28, i1 -1)
+ %ctlz29 = call i8 @llvm.ctlz.i8(i8 %ld29, i1 -1)
+ %ctlz30 = call i8 @llvm.ctlz.i8(i8 %ld30, i1 -1)
+ %ctlz31 = call i8 @llvm.ctlz.i8(i8 %ld31, i1 -1)
+ store i8 %ctlz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %ctlz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %ctlz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %ctlz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %ctlz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %ctlz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %ctlz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %ctlz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %ctlz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %ctlz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %ctlz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %ctlz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %ctlz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %ctlz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %ctlz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %ctlz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ store i8 %ctlz16, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1
+ store i8 %ctlz17, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1
+ store i8 %ctlz18, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1
+ store i8 %ctlz19, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1
+ store i8 %ctlz20, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1
+ store i8 %ctlz21, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1
+ store i8 %ctlz22, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1
+ store i8 %ctlz23, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1
+ store i8 %ctlz24, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1
+ store i8 %ctlz25, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1
+ store i8 %ctlz26, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1
+ store i8 %ctlz27, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1
+ store i8 %ctlz28, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1
+ store i8 %ctlz29, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1
+ store i8 %ctlz30, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1
+ store i8 %ctlz31, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1
+ ret void
+}
+
+attributes #0 = { nounwind }
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/ctpop.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/ctpop.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/ctpop.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/ctpop.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,512 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -mattr=+sse2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE2
+; RUN: opt < %s -mtriple=x86_64-unknown -mattr=+sse4.2,+popcnt -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE42
+; RUN: opt < %s -mtriple=x86_64-unknown -mattr=+avx,+popcnt -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown -mattr=+avx2,+popcnt -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+ at src64 = common global [4 x i64] zeroinitializer, align 32
+ at dst64 = common global [4 x i64] zeroinitializer, align 32
+ at src32 = common global [8 x i32] zeroinitializer, align 32
+ at dst32 = common global [8 x i32] zeroinitializer, align 32
+ at src16 = common global [16 x i16] zeroinitializer, align 32
+ at dst16 = common global [16 x i16] zeroinitializer, align 32
+ at src8 = common global [32 x i8] zeroinitializer, align 32
+ at dst8 = common global [32 x i8] zeroinitializer, align 32
+
+declare i64 @llvm.ctpop.i64(i64)
+declare i32 @llvm.ctpop.i32(i32)
+declare i16 @llvm.ctpop.i16(i16)
+declare i8 @llvm.ctpop.i8(i8)
+
+define void @ctpop_2i64() #0 {
+; CHECK-LABEL: @ctpop_2i64(
+; CHECK-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+; CHECK-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+; CHECK-NEXT: [[CTPOP0:%.*]] = call i64 @llvm.ctpop.i64(i64 [[LD0]])
+; CHECK-NEXT: [[CTPOP1:%.*]] = call i64 @llvm.ctpop.i64(i64 [[LD1]])
+; CHECK-NEXT: store i64 [[CTPOP0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+; CHECK-NEXT: store i64 [[CTPOP1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+ %ctpop0 = call i64 @llvm.ctpop.i64(i64 %ld0)
+ %ctpop1 = call i64 @llvm.ctpop.i64(i64 %ld1)
+ store i64 %ctpop0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+ store i64 %ctpop1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+ ret void
+}
+
+define void @ctpop_4i64() #0 {
+; SSE-LABEL: @ctpop_4i64(
+; SSE-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+; SSE-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+; SSE-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+; SSE-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+; SSE-NEXT: [[CTPOP0:%.*]] = call i64 @llvm.ctpop.i64(i64 [[LD0]])
+; SSE-NEXT: [[CTPOP1:%.*]] = call i64 @llvm.ctpop.i64(i64 [[LD1]])
+; SSE-NEXT: [[CTPOP2:%.*]] = call i64 @llvm.ctpop.i64(i64 [[LD2]])
+; SSE-NEXT: [[CTPOP3:%.*]] = call i64 @llvm.ctpop.i64(i64 [[LD3]])
+; SSE-NEXT: store i64 [[CTPOP0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+; SSE-NEXT: store i64 [[CTPOP1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+; SSE-NEXT: store i64 [[CTPOP2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+; SSE-NEXT: store i64 [[CTPOP3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+; SSE-NEXT: ret void
+;
+; AVX1-LABEL: @ctpop_4i64(
+; AVX1-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+; AVX1-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+; AVX1-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+; AVX1-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+; AVX1-NEXT: [[CTPOP0:%.*]] = call i64 @llvm.ctpop.i64(i64 [[LD0]])
+; AVX1-NEXT: [[CTPOP1:%.*]] = call i64 @llvm.ctpop.i64(i64 [[LD1]])
+; AVX1-NEXT: [[CTPOP2:%.*]] = call i64 @llvm.ctpop.i64(i64 [[LD2]])
+; AVX1-NEXT: [[CTPOP3:%.*]] = call i64 @llvm.ctpop.i64(i64 [[LD3]])
+; AVX1-NEXT: store i64 [[CTPOP0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+; AVX1-NEXT: store i64 [[CTPOP1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+; AVX1-NEXT: store i64 [[CTPOP2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+; AVX1-NEXT: store i64 [[CTPOP3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+; AVX1-NEXT: ret void
+;
+; AVX2-LABEL: @ctpop_4i64(
+; AVX2-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([4 x i64]* @src64 to <4 x i64>*), align 4
+; AVX2-NEXT: [[TMP2:%.*]] = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> [[TMP1]])
+; AVX2-NEXT: store <4 x i64> [[TMP2]], <4 x i64>* bitcast ([4 x i64]* @dst64 to <4 x i64>*), align 4
+; AVX2-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+ %ld2 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+ %ld3 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+ %ctpop0 = call i64 @llvm.ctpop.i64(i64 %ld0)
+ %ctpop1 = call i64 @llvm.ctpop.i64(i64 %ld1)
+ %ctpop2 = call i64 @llvm.ctpop.i64(i64 %ld2)
+ %ctpop3 = call i64 @llvm.ctpop.i64(i64 %ld3)
+ store i64 %ctpop0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+ store i64 %ctpop1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+ store i64 %ctpop2, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+ store i64 %ctpop3, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+ ret void
+}
+
+define void @ctpop_4i32() #0 {
+; SSE2-LABEL: @ctpop_4i32(
+; SSE2-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([8 x i32]* @src32 to <4 x i32>*), align 4
+; SSE2-NEXT: [[TMP2:%.*]] = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> [[TMP1]])
+; SSE2-NEXT: store <4 x i32> [[TMP2]], <4 x i32>* bitcast ([8 x i32]* @dst32 to <4 x i32>*), align 4
+; SSE2-NEXT: ret void
+;
+; SSE42-LABEL: @ctpop_4i32(
+; SSE42-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+; SSE42-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+; SSE42-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+; SSE42-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+; SSE42-NEXT: [[CTPOP0:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD0]])
+; SSE42-NEXT: [[CTPOP1:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD1]])
+; SSE42-NEXT: [[CTPOP2:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD2]])
+; SSE42-NEXT: [[CTPOP3:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD3]])
+; SSE42-NEXT: store i32 [[CTPOP0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+; SSE42-NEXT: store i32 [[CTPOP1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+; SSE42-NEXT: store i32 [[CTPOP2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+; SSE42-NEXT: store i32 [[CTPOP3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+; SSE42-NEXT: ret void
+;
+; AVX-LABEL: @ctpop_4i32(
+; AVX-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+; AVX-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+; AVX-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+; AVX-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+; AVX-NEXT: [[CTPOP0:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD0]])
+; AVX-NEXT: [[CTPOP1:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD1]])
+; AVX-NEXT: [[CTPOP2:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD2]])
+; AVX-NEXT: [[CTPOP3:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD3]])
+; AVX-NEXT: store i32 [[CTPOP0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+; AVX-NEXT: store i32 [[CTPOP1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+; AVX-NEXT: store i32 [[CTPOP2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+; AVX-NEXT: store i32 [[CTPOP3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+; AVX-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+ %ctpop0 = call i32 @llvm.ctpop.i32(i32 %ld0)
+ %ctpop1 = call i32 @llvm.ctpop.i32(i32 %ld1)
+ %ctpop2 = call i32 @llvm.ctpop.i32(i32 %ld2)
+ %ctpop3 = call i32 @llvm.ctpop.i32(i32 %ld3)
+ store i32 %ctpop0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+ store i32 %ctpop1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+ store i32 %ctpop2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+ store i32 %ctpop3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+ ret void
+}
+
+define void @ctpop_8i32() #0 {
+; SSE2-LABEL: @ctpop_8i32(
+; SSE2-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([8 x i32]* @src32 to <4 x i32>*), align 2
+; SSE2-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4) to <4 x i32>*), align 2
+; SSE2-NEXT: [[TMP3:%.*]] = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> [[TMP1]])
+; SSE2-NEXT: [[TMP4:%.*]] = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> [[TMP2]])
+; SSE2-NEXT: store <4 x i32> [[TMP3]], <4 x i32>* bitcast ([8 x i32]* @dst32 to <4 x i32>*), align 2
+; SSE2-NEXT: store <4 x i32> [[TMP4]], <4 x i32>* bitcast (i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4) to <4 x i32>*), align 2
+; SSE2-NEXT: ret void
+;
+; SSE42-LABEL: @ctpop_8i32(
+; SSE42-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+; SSE42-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+; SSE42-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+; SSE42-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+; SSE42-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+; SSE42-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+; SSE42-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+; SSE42-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+; SSE42-NEXT: [[CTPOP0:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD0]])
+; SSE42-NEXT: [[CTPOP1:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD1]])
+; SSE42-NEXT: [[CTPOP2:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD2]])
+; SSE42-NEXT: [[CTPOP3:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD3]])
+; SSE42-NEXT: [[CTPOP4:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD4]])
+; SSE42-NEXT: [[CTPOP5:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD5]])
+; SSE42-NEXT: [[CTPOP6:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD6]])
+; SSE42-NEXT: [[CTPOP7:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD7]])
+; SSE42-NEXT: store i32 [[CTPOP0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+; SSE42-NEXT: store i32 [[CTPOP1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+; SSE42-NEXT: store i32 [[CTPOP2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+; SSE42-NEXT: store i32 [[CTPOP3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+; SSE42-NEXT: store i32 [[CTPOP4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+; SSE42-NEXT: store i32 [[CTPOP5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+; SSE42-NEXT: store i32 [[CTPOP6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+; SSE42-NEXT: store i32 [[CTPOP7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+; SSE42-NEXT: ret void
+;
+; AVX1-LABEL: @ctpop_8i32(
+; AVX1-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+; AVX1-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+; AVX1-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+; AVX1-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+; AVX1-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+; AVX1-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+; AVX1-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+; AVX1-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+; AVX1-NEXT: [[CTPOP0:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD0]])
+; AVX1-NEXT: [[CTPOP1:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD1]])
+; AVX1-NEXT: [[CTPOP2:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD2]])
+; AVX1-NEXT: [[CTPOP3:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD3]])
+; AVX1-NEXT: [[CTPOP4:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD4]])
+; AVX1-NEXT: [[CTPOP5:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD5]])
+; AVX1-NEXT: [[CTPOP6:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD6]])
+; AVX1-NEXT: [[CTPOP7:%.*]] = call i32 @llvm.ctpop.i32(i32 [[LD7]])
+; AVX1-NEXT: store i32 [[CTPOP0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+; AVX1-NEXT: store i32 [[CTPOP1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+; AVX1-NEXT: store i32 [[CTPOP2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+; AVX1-NEXT: store i32 [[CTPOP3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+; AVX1-NEXT: store i32 [[CTPOP4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+; AVX1-NEXT: store i32 [[CTPOP5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+; AVX1-NEXT: store i32 [[CTPOP6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+; AVX1-NEXT: store i32 [[CTPOP7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+; AVX1-NEXT: ret void
+;
+; AVX2-LABEL: @ctpop_8i32(
+; AVX2-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2
+; AVX2-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> [[TMP1]])
+; AVX2-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2
+; AVX2-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+ %ld4 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+ %ld5 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+ %ld6 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+ %ld7 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+ %ctpop0 = call i32 @llvm.ctpop.i32(i32 %ld0)
+ %ctpop1 = call i32 @llvm.ctpop.i32(i32 %ld1)
+ %ctpop2 = call i32 @llvm.ctpop.i32(i32 %ld2)
+ %ctpop3 = call i32 @llvm.ctpop.i32(i32 %ld3)
+ %ctpop4 = call i32 @llvm.ctpop.i32(i32 %ld4)
+ %ctpop5 = call i32 @llvm.ctpop.i32(i32 %ld5)
+ %ctpop6 = call i32 @llvm.ctpop.i32(i32 %ld6)
+ %ctpop7 = call i32 @llvm.ctpop.i32(i32 %ld7)
+ store i32 %ctpop0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+ store i32 %ctpop1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+ store i32 %ctpop2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+ store i32 %ctpop3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+ store i32 %ctpop4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+ store i32 %ctpop5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+ store i32 %ctpop6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+ store i32 %ctpop7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+ ret void
+}
+
+define void @ctpop_8i16() #0 {
+; CHECK-LABEL: @ctpop_8i16(
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> [[TMP1]])
+; CHECK-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %ctpop0 = call i16 @llvm.ctpop.i16(i16 %ld0)
+ %ctpop1 = call i16 @llvm.ctpop.i16(i16 %ld1)
+ %ctpop2 = call i16 @llvm.ctpop.i16(i16 %ld2)
+ %ctpop3 = call i16 @llvm.ctpop.i16(i16 %ld3)
+ %ctpop4 = call i16 @llvm.ctpop.i16(i16 %ld4)
+ %ctpop5 = call i16 @llvm.ctpop.i16(i16 %ld5)
+ %ctpop6 = call i16 @llvm.ctpop.i16(i16 %ld6)
+ %ctpop7 = call i16 @llvm.ctpop.i16(i16 %ld7)
+ store i16 %ctpop0, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %ctpop1, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %ctpop2, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %ctpop3, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %ctpop4, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %ctpop5, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %ctpop6, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %ctpop7, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ ret void
+}
+
+define void @ctpop_16i16() #0 {
+; SSE-LABEL: @ctpop_16i16(
+; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> [[TMP1]])
+; SSE-NEXT: [[TMP4:%.*]] = call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> [[TMP2]])
+; SSE-NEXT: store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @ctpop_16i16(
+; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> [[TMP1]])
+; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2
+; AVX-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %ld8 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2
+ %ld9 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2
+ %ld10 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2
+ %ld11 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2
+ %ld12 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2
+ %ld13 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2
+ %ld14 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2
+ %ld15 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2
+ %ctpop0 = call i16 @llvm.ctpop.i16(i16 %ld0)
+ %ctpop1 = call i16 @llvm.ctpop.i16(i16 %ld1)
+ %ctpop2 = call i16 @llvm.ctpop.i16(i16 %ld2)
+ %ctpop3 = call i16 @llvm.ctpop.i16(i16 %ld3)
+ %ctpop4 = call i16 @llvm.ctpop.i16(i16 %ld4)
+ %ctpop5 = call i16 @llvm.ctpop.i16(i16 %ld5)
+ %ctpop6 = call i16 @llvm.ctpop.i16(i16 %ld6)
+ %ctpop7 = call i16 @llvm.ctpop.i16(i16 %ld7)
+ %ctpop8 = call i16 @llvm.ctpop.i16(i16 %ld8)
+ %ctpop9 = call i16 @llvm.ctpop.i16(i16 %ld9)
+ %ctpop10 = call i16 @llvm.ctpop.i16(i16 %ld10)
+ %ctpop11 = call i16 @llvm.ctpop.i16(i16 %ld11)
+ %ctpop12 = call i16 @llvm.ctpop.i16(i16 %ld12)
+ %ctpop13 = call i16 @llvm.ctpop.i16(i16 %ld13)
+ %ctpop14 = call i16 @llvm.ctpop.i16(i16 %ld14)
+ %ctpop15 = call i16 @llvm.ctpop.i16(i16 %ld15)
+ store i16 %ctpop0 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %ctpop1 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %ctpop2 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %ctpop3 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %ctpop4 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %ctpop5 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %ctpop6 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %ctpop7 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ store i16 %ctpop8 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2
+ store i16 %ctpop9 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2
+ store i16 %ctpop10, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2
+ store i16 %ctpop11, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2
+ store i16 %ctpop12, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2
+ store i16 %ctpop13, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2
+ store i16 %ctpop14, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2
+ store i16 %ctpop15, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2
+ ret void
+}
+
+define void @ctpop_16i8() #0 {
+; CHECK-LABEL: @ctpop_16i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> [[TMP1]])
+; CHECK-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %ctpop0 = call i8 @llvm.ctpop.i8(i8 %ld0)
+ %ctpop1 = call i8 @llvm.ctpop.i8(i8 %ld1)
+ %ctpop2 = call i8 @llvm.ctpop.i8(i8 %ld2)
+ %ctpop3 = call i8 @llvm.ctpop.i8(i8 %ld3)
+ %ctpop4 = call i8 @llvm.ctpop.i8(i8 %ld4)
+ %ctpop5 = call i8 @llvm.ctpop.i8(i8 %ld5)
+ %ctpop6 = call i8 @llvm.ctpop.i8(i8 %ld6)
+ %ctpop7 = call i8 @llvm.ctpop.i8(i8 %ld7)
+ %ctpop8 = call i8 @llvm.ctpop.i8(i8 %ld8)
+ %ctpop9 = call i8 @llvm.ctpop.i8(i8 %ld9)
+ %ctpop10 = call i8 @llvm.ctpop.i8(i8 %ld10)
+ %ctpop11 = call i8 @llvm.ctpop.i8(i8 %ld11)
+ %ctpop12 = call i8 @llvm.ctpop.i8(i8 %ld12)
+ %ctpop13 = call i8 @llvm.ctpop.i8(i8 %ld13)
+ %ctpop14 = call i8 @llvm.ctpop.i8(i8 %ld14)
+ %ctpop15 = call i8 @llvm.ctpop.i8(i8 %ld15)
+ store i8 %ctpop0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %ctpop1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %ctpop2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %ctpop3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %ctpop4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %ctpop5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %ctpop6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %ctpop7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %ctpop8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %ctpop9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %ctpop10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %ctpop11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %ctpop12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %ctpop13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %ctpop14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %ctpop15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ ret void
+}
+
+define void @ctpop_32i8() #0 {
+; CHECK-LABEL: @ctpop_32i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> [[TMP1]])
+; CHECK-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> [[TMP2]])
+; CHECK-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %ld16 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1
+ %ld17 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1
+ %ld18 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1
+ %ld19 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1
+ %ld20 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1
+ %ld21 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1
+ %ld22 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1
+ %ld23 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1
+ %ld24 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1
+ %ld25 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1
+ %ld26 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1
+ %ld27 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1
+ %ld28 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1
+ %ld29 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1
+ %ld30 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1
+ %ld31 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1
+ %ctpop0 = call i8 @llvm.ctpop.i8(i8 %ld0)
+ %ctpop1 = call i8 @llvm.ctpop.i8(i8 %ld1)
+ %ctpop2 = call i8 @llvm.ctpop.i8(i8 %ld2)
+ %ctpop3 = call i8 @llvm.ctpop.i8(i8 %ld3)
+ %ctpop4 = call i8 @llvm.ctpop.i8(i8 %ld4)
+ %ctpop5 = call i8 @llvm.ctpop.i8(i8 %ld5)
+ %ctpop6 = call i8 @llvm.ctpop.i8(i8 %ld6)
+ %ctpop7 = call i8 @llvm.ctpop.i8(i8 %ld7)
+ %ctpop8 = call i8 @llvm.ctpop.i8(i8 %ld8)
+ %ctpop9 = call i8 @llvm.ctpop.i8(i8 %ld9)
+ %ctpop10 = call i8 @llvm.ctpop.i8(i8 %ld10)
+ %ctpop11 = call i8 @llvm.ctpop.i8(i8 %ld11)
+ %ctpop12 = call i8 @llvm.ctpop.i8(i8 %ld12)
+ %ctpop13 = call i8 @llvm.ctpop.i8(i8 %ld13)
+ %ctpop14 = call i8 @llvm.ctpop.i8(i8 %ld14)
+ %ctpop15 = call i8 @llvm.ctpop.i8(i8 %ld15)
+ %ctpop16 = call i8 @llvm.ctpop.i8(i8 %ld16)
+ %ctpop17 = call i8 @llvm.ctpop.i8(i8 %ld17)
+ %ctpop18 = call i8 @llvm.ctpop.i8(i8 %ld18)
+ %ctpop19 = call i8 @llvm.ctpop.i8(i8 %ld19)
+ %ctpop20 = call i8 @llvm.ctpop.i8(i8 %ld20)
+ %ctpop21 = call i8 @llvm.ctpop.i8(i8 %ld21)
+ %ctpop22 = call i8 @llvm.ctpop.i8(i8 %ld22)
+ %ctpop23 = call i8 @llvm.ctpop.i8(i8 %ld23)
+ %ctpop24 = call i8 @llvm.ctpop.i8(i8 %ld24)
+ %ctpop25 = call i8 @llvm.ctpop.i8(i8 %ld25)
+ %ctpop26 = call i8 @llvm.ctpop.i8(i8 %ld26)
+ %ctpop27 = call i8 @llvm.ctpop.i8(i8 %ld27)
+ %ctpop28 = call i8 @llvm.ctpop.i8(i8 %ld28)
+ %ctpop29 = call i8 @llvm.ctpop.i8(i8 %ld29)
+ %ctpop30 = call i8 @llvm.ctpop.i8(i8 %ld30)
+ %ctpop31 = call i8 @llvm.ctpop.i8(i8 %ld31)
+ store i8 %ctpop0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %ctpop1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %ctpop2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %ctpop3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %ctpop4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %ctpop5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %ctpop6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %ctpop7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %ctpop8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %ctpop9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %ctpop10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %ctpop11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %ctpop12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %ctpop13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %ctpop14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %ctpop15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ store i8 %ctpop16, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1
+ store i8 %ctpop17, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1
+ store i8 %ctpop18, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1
+ store i8 %ctpop19, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1
+ store i8 %ctpop20, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1
+ store i8 %ctpop21, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1
+ store i8 %ctpop22, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1
+ store i8 %ctpop23, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1
+ store i8 %ctpop24, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1
+ store i8 %ctpop25, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1
+ store i8 %ctpop26, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1
+ store i8 %ctpop27, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1
+ store i8 %ctpop28, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1
+ store i8 %ctpop29, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1
+ store i8 %ctpop30, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1
+ store i8 %ctpop31, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1
+ ret void
+}
+
+attributes #0 = { nounwind }
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/cttz.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/cttz.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/cttz.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/cttz.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,905 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE2
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE --check-prefix=SSE42
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+ at src64 = common global [4 x i64] zeroinitializer, align 32
+ at dst64 = common global [4 x i64] zeroinitializer, align 32
+ at src32 = common global [8 x i32] zeroinitializer, align 32
+ at dst32 = common global [8 x i32] zeroinitializer, align 32
+ at src16 = common global [16 x i16] zeroinitializer, align 32
+ at dst16 = common global [16 x i16] zeroinitializer, align 32
+ at src8 = common global [32 x i8] zeroinitializer, align 32
+ at dst8 = common global [32 x i8] zeroinitializer, align 32
+
+declare i64 @llvm.cttz.i64(i64, i1)
+declare i32 @llvm.cttz.i32(i32, i1)
+declare i16 @llvm.cttz.i16(i16, i1)
+declare i8 @llvm.cttz.i8(i8, i1)
+
+;
+; CTTZ
+;
+
+define void @cttz_2i64() #0 {
+; CHECK-LABEL: @cttz_2i64(
+; CHECK-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+; CHECK-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+; CHECK-NEXT: [[CTTZ0:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD0]], i1 false)
+; CHECK-NEXT: [[CTTZ1:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD1]], i1 false)
+; CHECK-NEXT: store i64 [[CTTZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+; CHECK-NEXT: store i64 [[CTTZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+ %cttz0 = call i64 @llvm.cttz.i64(i64 %ld0, i1 0)
+ %cttz1 = call i64 @llvm.cttz.i64(i64 %ld1, i1 0)
+ store i64 %cttz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+ store i64 %cttz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+ ret void
+}
+
+define void @cttz_4i64() #0 {
+; CHECK-LABEL: @cttz_4i64(
+; CHECK-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+; CHECK-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+; CHECK-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+; CHECK-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+; CHECK-NEXT: [[CTTZ0:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD0]], i1 false)
+; CHECK-NEXT: [[CTTZ1:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD1]], i1 false)
+; CHECK-NEXT: [[CTTZ2:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD2]], i1 false)
+; CHECK-NEXT: [[CTTZ3:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD3]], i1 false)
+; CHECK-NEXT: store i64 [[CTTZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+; CHECK-NEXT: store i64 [[CTTZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+; CHECK-NEXT: store i64 [[CTTZ2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+; CHECK-NEXT: store i64 [[CTTZ3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+ %ld2 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+ %ld3 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+ %cttz0 = call i64 @llvm.cttz.i64(i64 %ld0, i1 0)
+ %cttz1 = call i64 @llvm.cttz.i64(i64 %ld1, i1 0)
+ %cttz2 = call i64 @llvm.cttz.i64(i64 %ld2, i1 0)
+ %cttz3 = call i64 @llvm.cttz.i64(i64 %ld3, i1 0)
+ store i64 %cttz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+ store i64 %cttz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+ store i64 %cttz2, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+ store i64 %cttz3, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+ ret void
+}
+
+define void @cttz_4i32() #0 {
+; CHECK-LABEL: @cttz_4i32(
+; CHECK-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+; CHECK-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+; CHECK-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+; CHECK-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+; CHECK-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 false)
+; CHECK-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 false)
+; CHECK-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 false)
+; CHECK-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 false)
+; CHECK-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+; CHECK-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+; CHECK-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+; CHECK-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+ %cttz0 = call i32 @llvm.cttz.i32(i32 %ld0, i1 0)
+ %cttz1 = call i32 @llvm.cttz.i32(i32 %ld1, i1 0)
+ %cttz2 = call i32 @llvm.cttz.i32(i32 %ld2, i1 0)
+ %cttz3 = call i32 @llvm.cttz.i32(i32 %ld3, i1 0)
+ store i32 %cttz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+ store i32 %cttz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+ store i32 %cttz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+ store i32 %cttz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+ ret void
+}
+
+define void @cttz_8i32() #0 {
+; SSE-LABEL: @cttz_8i32(
+; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+; SSE-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+; SSE-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+; SSE-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+; SSE-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+; SSE-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 false)
+; SSE-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 false)
+; SSE-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 false)
+; SSE-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 false)
+; SSE-NEXT: [[CTTZ4:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD4]], i1 false)
+; SSE-NEXT: [[CTTZ5:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD5]], i1 false)
+; SSE-NEXT: [[CTTZ6:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD6]], i1 false)
+; SSE-NEXT: [[CTTZ7:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD7]], i1 false)
+; SSE-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+; SSE-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+; SSE-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+; SSE-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+; SSE-NEXT: store i32 [[CTTZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+; SSE-NEXT: store i32 [[CTTZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+; SSE-NEXT: store i32 [[CTTZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+; SSE-NEXT: store i32 [[CTTZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+; SSE-NEXT: ret void
+;
+; AVX1-LABEL: @cttz_8i32(
+; AVX1-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+; AVX1-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+; AVX1-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+; AVX1-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+; AVX1-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+; AVX1-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+; AVX1-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+; AVX1-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+; AVX1-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 false)
+; AVX1-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 false)
+; AVX1-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 false)
+; AVX1-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 false)
+; AVX1-NEXT: [[CTTZ4:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD4]], i1 false)
+; AVX1-NEXT: [[CTTZ5:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD5]], i1 false)
+; AVX1-NEXT: [[CTTZ6:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD6]], i1 false)
+; AVX1-NEXT: [[CTTZ7:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD7]], i1 false)
+; AVX1-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+; AVX1-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+; AVX1-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+; AVX1-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+; AVX1-NEXT: store i32 [[CTTZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+; AVX1-NEXT: store i32 [[CTTZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+; AVX1-NEXT: store i32 [[CTTZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+; AVX1-NEXT: store i32 [[CTTZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+; AVX1-NEXT: ret void
+;
+; AVX2-LABEL: @cttz_8i32(
+; AVX2-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2
+; AVX2-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> [[TMP1]], i1 false)
+; AVX2-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2
+; AVX2-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+ %ld4 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+ %ld5 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+ %ld6 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+ %ld7 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+ %cttz0 = call i32 @llvm.cttz.i32(i32 %ld0, i1 0)
+ %cttz1 = call i32 @llvm.cttz.i32(i32 %ld1, i1 0)
+ %cttz2 = call i32 @llvm.cttz.i32(i32 %ld2, i1 0)
+ %cttz3 = call i32 @llvm.cttz.i32(i32 %ld3, i1 0)
+ %cttz4 = call i32 @llvm.cttz.i32(i32 %ld4, i1 0)
+ %cttz5 = call i32 @llvm.cttz.i32(i32 %ld5, i1 0)
+ %cttz6 = call i32 @llvm.cttz.i32(i32 %ld6, i1 0)
+ %cttz7 = call i32 @llvm.cttz.i32(i32 %ld7, i1 0)
+ store i32 %cttz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+ store i32 %cttz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+ store i32 %cttz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+ store i32 %cttz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+ store i32 %cttz4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+ store i32 %cttz5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+ store i32 %cttz6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+ store i32 %cttz7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+ ret void
+}
+
+define void @cttz_8i16() #0 {
+; CHECK-LABEL: @cttz_8i16(
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP1]], i1 false)
+; CHECK-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %cttz0 = call i16 @llvm.cttz.i16(i16 %ld0, i1 0)
+ %cttz1 = call i16 @llvm.cttz.i16(i16 %ld1, i1 0)
+ %cttz2 = call i16 @llvm.cttz.i16(i16 %ld2, i1 0)
+ %cttz3 = call i16 @llvm.cttz.i16(i16 %ld3, i1 0)
+ %cttz4 = call i16 @llvm.cttz.i16(i16 %ld4, i1 0)
+ %cttz5 = call i16 @llvm.cttz.i16(i16 %ld5, i1 0)
+ %cttz6 = call i16 @llvm.cttz.i16(i16 %ld6, i1 0)
+ %cttz7 = call i16 @llvm.cttz.i16(i16 %ld7, i1 0)
+ store i16 %cttz0, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %cttz1, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %cttz2, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %cttz3, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %cttz4, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %cttz5, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %cttz6, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %cttz7, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ ret void
+}
+
+define void @cttz_16i16() #0 {
+; SSE-LABEL: @cttz_16i16(
+; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP1]], i1 false)
+; SSE-NEXT: [[TMP4:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP2]], i1 false)
+; SSE-NEXT: store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @cttz_16i16(
+; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> [[TMP1]], i1 false)
+; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2
+; AVX-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %ld8 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2
+ %ld9 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2
+ %ld10 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2
+ %ld11 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2
+ %ld12 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2
+ %ld13 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2
+ %ld14 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2
+ %ld15 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2
+ %cttz0 = call i16 @llvm.cttz.i16(i16 %ld0, i1 0)
+ %cttz1 = call i16 @llvm.cttz.i16(i16 %ld1, i1 0)
+ %cttz2 = call i16 @llvm.cttz.i16(i16 %ld2, i1 0)
+ %cttz3 = call i16 @llvm.cttz.i16(i16 %ld3, i1 0)
+ %cttz4 = call i16 @llvm.cttz.i16(i16 %ld4, i1 0)
+ %cttz5 = call i16 @llvm.cttz.i16(i16 %ld5, i1 0)
+ %cttz6 = call i16 @llvm.cttz.i16(i16 %ld6, i1 0)
+ %cttz7 = call i16 @llvm.cttz.i16(i16 %ld7, i1 0)
+ %cttz8 = call i16 @llvm.cttz.i16(i16 %ld8, i1 0)
+ %cttz9 = call i16 @llvm.cttz.i16(i16 %ld9, i1 0)
+ %cttz10 = call i16 @llvm.cttz.i16(i16 %ld10, i1 0)
+ %cttz11 = call i16 @llvm.cttz.i16(i16 %ld11, i1 0)
+ %cttz12 = call i16 @llvm.cttz.i16(i16 %ld12, i1 0)
+ %cttz13 = call i16 @llvm.cttz.i16(i16 %ld13, i1 0)
+ %cttz14 = call i16 @llvm.cttz.i16(i16 %ld14, i1 0)
+ %cttz15 = call i16 @llvm.cttz.i16(i16 %ld15, i1 0)
+ store i16 %cttz0 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %cttz1 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %cttz2 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %cttz3 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %cttz4 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %cttz5 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %cttz6 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %cttz7 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ store i16 %cttz8 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2
+ store i16 %cttz9 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2
+ store i16 %cttz10, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2
+ store i16 %cttz11, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2
+ store i16 %cttz12, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2
+ store i16 %cttz13, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2
+ store i16 %cttz14, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2
+ store i16 %cttz15, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2
+ ret void
+}
+
+define void @cttz_16i8() #0 {
+; CHECK-LABEL: @cttz_16i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP1]], i1 false)
+; CHECK-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %cttz0 = call i8 @llvm.cttz.i8(i8 %ld0, i1 0)
+ %cttz1 = call i8 @llvm.cttz.i8(i8 %ld1, i1 0)
+ %cttz2 = call i8 @llvm.cttz.i8(i8 %ld2, i1 0)
+ %cttz3 = call i8 @llvm.cttz.i8(i8 %ld3, i1 0)
+ %cttz4 = call i8 @llvm.cttz.i8(i8 %ld4, i1 0)
+ %cttz5 = call i8 @llvm.cttz.i8(i8 %ld5, i1 0)
+ %cttz6 = call i8 @llvm.cttz.i8(i8 %ld6, i1 0)
+ %cttz7 = call i8 @llvm.cttz.i8(i8 %ld7, i1 0)
+ %cttz8 = call i8 @llvm.cttz.i8(i8 %ld8, i1 0)
+ %cttz9 = call i8 @llvm.cttz.i8(i8 %ld9, i1 0)
+ %cttz10 = call i8 @llvm.cttz.i8(i8 %ld10, i1 0)
+ %cttz11 = call i8 @llvm.cttz.i8(i8 %ld11, i1 0)
+ %cttz12 = call i8 @llvm.cttz.i8(i8 %ld12, i1 0)
+ %cttz13 = call i8 @llvm.cttz.i8(i8 %ld13, i1 0)
+ %cttz14 = call i8 @llvm.cttz.i8(i8 %ld14, i1 0)
+ %cttz15 = call i8 @llvm.cttz.i8(i8 %ld15, i1 0)
+ store i8 %cttz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %cttz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %cttz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %cttz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %cttz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %cttz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %cttz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %cttz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %cttz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %cttz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %cttz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %cttz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %cttz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %cttz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %cttz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %cttz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ ret void
+}
+
+define void @cttz_32i8() #0 {
+; CHECK-LABEL: @cttz_32i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP1]], i1 false)
+; CHECK-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP2]], i1 false)
+; CHECK-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %ld16 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1
+ %ld17 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1
+ %ld18 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1
+ %ld19 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1
+ %ld20 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1
+ %ld21 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1
+ %ld22 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1
+ %ld23 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1
+ %ld24 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1
+ %ld25 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1
+ %ld26 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1
+ %ld27 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1
+ %ld28 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1
+ %ld29 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1
+ %ld30 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1
+ %ld31 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1
+ %cttz0 = call i8 @llvm.cttz.i8(i8 %ld0, i1 0)
+ %cttz1 = call i8 @llvm.cttz.i8(i8 %ld1, i1 0)
+ %cttz2 = call i8 @llvm.cttz.i8(i8 %ld2, i1 0)
+ %cttz3 = call i8 @llvm.cttz.i8(i8 %ld3, i1 0)
+ %cttz4 = call i8 @llvm.cttz.i8(i8 %ld4, i1 0)
+ %cttz5 = call i8 @llvm.cttz.i8(i8 %ld5, i1 0)
+ %cttz6 = call i8 @llvm.cttz.i8(i8 %ld6, i1 0)
+ %cttz7 = call i8 @llvm.cttz.i8(i8 %ld7, i1 0)
+ %cttz8 = call i8 @llvm.cttz.i8(i8 %ld8, i1 0)
+ %cttz9 = call i8 @llvm.cttz.i8(i8 %ld9, i1 0)
+ %cttz10 = call i8 @llvm.cttz.i8(i8 %ld10, i1 0)
+ %cttz11 = call i8 @llvm.cttz.i8(i8 %ld11, i1 0)
+ %cttz12 = call i8 @llvm.cttz.i8(i8 %ld12, i1 0)
+ %cttz13 = call i8 @llvm.cttz.i8(i8 %ld13, i1 0)
+ %cttz14 = call i8 @llvm.cttz.i8(i8 %ld14, i1 0)
+ %cttz15 = call i8 @llvm.cttz.i8(i8 %ld15, i1 0)
+ %cttz16 = call i8 @llvm.cttz.i8(i8 %ld16, i1 0)
+ %cttz17 = call i8 @llvm.cttz.i8(i8 %ld17, i1 0)
+ %cttz18 = call i8 @llvm.cttz.i8(i8 %ld18, i1 0)
+ %cttz19 = call i8 @llvm.cttz.i8(i8 %ld19, i1 0)
+ %cttz20 = call i8 @llvm.cttz.i8(i8 %ld20, i1 0)
+ %cttz21 = call i8 @llvm.cttz.i8(i8 %ld21, i1 0)
+ %cttz22 = call i8 @llvm.cttz.i8(i8 %ld22, i1 0)
+ %cttz23 = call i8 @llvm.cttz.i8(i8 %ld23, i1 0)
+ %cttz24 = call i8 @llvm.cttz.i8(i8 %ld24, i1 0)
+ %cttz25 = call i8 @llvm.cttz.i8(i8 %ld25, i1 0)
+ %cttz26 = call i8 @llvm.cttz.i8(i8 %ld26, i1 0)
+ %cttz27 = call i8 @llvm.cttz.i8(i8 %ld27, i1 0)
+ %cttz28 = call i8 @llvm.cttz.i8(i8 %ld28, i1 0)
+ %cttz29 = call i8 @llvm.cttz.i8(i8 %ld29, i1 0)
+ %cttz30 = call i8 @llvm.cttz.i8(i8 %ld30, i1 0)
+ %cttz31 = call i8 @llvm.cttz.i8(i8 %ld31, i1 0)
+ store i8 %cttz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %cttz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %cttz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %cttz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %cttz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %cttz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %cttz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %cttz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %cttz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %cttz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %cttz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %cttz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %cttz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %cttz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %cttz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %cttz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ store i8 %cttz16, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1
+ store i8 %cttz17, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1
+ store i8 %cttz18, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1
+ store i8 %cttz19, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1
+ store i8 %cttz20, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1
+ store i8 %cttz21, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1
+ store i8 %cttz22, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1
+ store i8 %cttz23, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1
+ store i8 %cttz24, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1
+ store i8 %cttz25, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1
+ store i8 %cttz26, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1
+ store i8 %cttz27, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1
+ store i8 %cttz28, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1
+ store i8 %cttz29, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1
+ store i8 %cttz30, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1
+ store i8 %cttz31, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1
+ ret void
+}
+
+;
+; CTTZ_ZERO_UNDEF
+;
+
+define void @cttz_undef_2i64() #0 {
+; CHECK-LABEL: @cttz_undef_2i64(
+; CHECK-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+; CHECK-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+; CHECK-NEXT: [[CTTZ0:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD0]], i1 true)
+; CHECK-NEXT: [[CTTZ1:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD1]], i1 true)
+; CHECK-NEXT: store i64 [[CTTZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+; CHECK-NEXT: store i64 [[CTTZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8
+ %cttz0 = call i64 @llvm.cttz.i64(i64 %ld0, i1 -1)
+ %cttz1 = call i64 @llvm.cttz.i64(i64 %ld1, i1 -1)
+ store i64 %cttz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8
+ store i64 %cttz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8
+ ret void
+}
+
+define void @cttz_undef_4i64() #0 {
+; CHECK-LABEL: @cttz_undef_4i64(
+; CHECK-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+; CHECK-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+; CHECK-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+; CHECK-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+; CHECK-NEXT: [[CTTZ0:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD0]], i1 true)
+; CHECK-NEXT: [[CTTZ1:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD1]], i1 true)
+; CHECK-NEXT: [[CTTZ2:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD2]], i1 true)
+; CHECK-NEXT: [[CTTZ3:%.*]] = call i64 @llvm.cttz.i64(i64 [[LD3]], i1 true)
+; CHECK-NEXT: store i64 [[CTTZ0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+; CHECK-NEXT: store i64 [[CTTZ1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+; CHECK-NEXT: store i64 [[CTTZ2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+; CHECK-NEXT: store i64 [[CTTZ3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4
+ %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4
+ %ld2 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4
+ %ld3 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4
+ %cttz0 = call i64 @llvm.cttz.i64(i64 %ld0, i1 -1)
+ %cttz1 = call i64 @llvm.cttz.i64(i64 %ld1, i1 -1)
+ %cttz2 = call i64 @llvm.cttz.i64(i64 %ld2, i1 -1)
+ %cttz3 = call i64 @llvm.cttz.i64(i64 %ld3, i1 -1)
+ store i64 %cttz0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4
+ store i64 %cttz1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4
+ store i64 %cttz2, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4
+ store i64 %cttz3, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4
+ ret void
+}
+
+define void @cttz_undef_4i32() #0 {
+; CHECK-LABEL: @cttz_undef_4i32(
+; CHECK-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+; CHECK-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+; CHECK-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+; CHECK-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+; CHECK-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 true)
+; CHECK-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 true)
+; CHECK-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 true)
+; CHECK-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 true)
+; CHECK-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+; CHECK-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+; CHECK-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+; CHECK-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4
+ %cttz0 = call i32 @llvm.cttz.i32(i32 %ld0, i1 -1)
+ %cttz1 = call i32 @llvm.cttz.i32(i32 %ld1, i1 -1)
+ %cttz2 = call i32 @llvm.cttz.i32(i32 %ld2, i1 -1)
+ %cttz3 = call i32 @llvm.cttz.i32(i32 %ld3, i1 -1)
+ store i32 %cttz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4
+ store i32 %cttz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4
+ store i32 %cttz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4
+ store i32 %cttz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4
+ ret void
+}
+
+define void @cttz_undef_8i32() #0 {
+; SSE-LABEL: @cttz_undef_8i32(
+; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+; SSE-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+; SSE-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+; SSE-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+; SSE-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+; SSE-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 true)
+; SSE-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 true)
+; SSE-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 true)
+; SSE-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 true)
+; SSE-NEXT: [[CTTZ4:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD4]], i1 true)
+; SSE-NEXT: [[CTTZ5:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD5]], i1 true)
+; SSE-NEXT: [[CTTZ6:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD6]], i1 true)
+; SSE-NEXT: [[CTTZ7:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD7]], i1 true)
+; SSE-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+; SSE-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+; SSE-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+; SSE-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+; SSE-NEXT: store i32 [[CTTZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+; SSE-NEXT: store i32 [[CTTZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+; SSE-NEXT: store i32 [[CTTZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+; SSE-NEXT: store i32 [[CTTZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+; SSE-NEXT: ret void
+;
+; AVX1-LABEL: @cttz_undef_8i32(
+; AVX1-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+; AVX1-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+; AVX1-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+; AVX1-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+; AVX1-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+; AVX1-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+; AVX1-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+; AVX1-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+; AVX1-NEXT: [[CTTZ0:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD0]], i1 true)
+; AVX1-NEXT: [[CTTZ1:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD1]], i1 true)
+; AVX1-NEXT: [[CTTZ2:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD2]], i1 true)
+; AVX1-NEXT: [[CTTZ3:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD3]], i1 true)
+; AVX1-NEXT: [[CTTZ4:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD4]], i1 true)
+; AVX1-NEXT: [[CTTZ5:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD5]], i1 true)
+; AVX1-NEXT: [[CTTZ6:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD6]], i1 true)
+; AVX1-NEXT: [[CTTZ7:%.*]] = call i32 @llvm.cttz.i32(i32 [[LD7]], i1 true)
+; AVX1-NEXT: store i32 [[CTTZ0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+; AVX1-NEXT: store i32 [[CTTZ1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+; AVX1-NEXT: store i32 [[CTTZ2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+; AVX1-NEXT: store i32 [[CTTZ3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+; AVX1-NEXT: store i32 [[CTTZ4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+; AVX1-NEXT: store i32 [[CTTZ5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+; AVX1-NEXT: store i32 [[CTTZ6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+; AVX1-NEXT: store i32 [[CTTZ7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+; AVX1-NEXT: ret void
+;
+; AVX2-LABEL: @cttz_undef_8i32(
+; AVX2-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2
+; AVX2-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.cttz.v8i32(<8 x i32> [[TMP1]], i1 true)
+; AVX2-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2
+; AVX2-NEXT: ret void
+;
+ %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2
+ %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2
+ %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2
+ %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2
+ %ld4 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2
+ %ld5 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2
+ %ld6 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2
+ %ld7 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2
+ %cttz0 = call i32 @llvm.cttz.i32(i32 %ld0, i1 -1)
+ %cttz1 = call i32 @llvm.cttz.i32(i32 %ld1, i1 -1)
+ %cttz2 = call i32 @llvm.cttz.i32(i32 %ld2, i1 -1)
+ %cttz3 = call i32 @llvm.cttz.i32(i32 %ld3, i1 -1)
+ %cttz4 = call i32 @llvm.cttz.i32(i32 %ld4, i1 -1)
+ %cttz5 = call i32 @llvm.cttz.i32(i32 %ld5, i1 -1)
+ %cttz6 = call i32 @llvm.cttz.i32(i32 %ld6, i1 -1)
+ %cttz7 = call i32 @llvm.cttz.i32(i32 %ld7, i1 -1)
+ store i32 %cttz0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2
+ store i32 %cttz1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2
+ store i32 %cttz2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2
+ store i32 %cttz3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2
+ store i32 %cttz4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2
+ store i32 %cttz5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2
+ store i32 %cttz6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2
+ store i32 %cttz7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2
+ ret void
+}
+
+define void @cttz_undef_8i16() #0 {
+; CHECK-LABEL: @cttz_undef_8i16(
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP1]], i1 true)
+; CHECK-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %cttz0 = call i16 @llvm.cttz.i16(i16 %ld0, i1 -1)
+ %cttz1 = call i16 @llvm.cttz.i16(i16 %ld1, i1 -1)
+ %cttz2 = call i16 @llvm.cttz.i16(i16 %ld2, i1 -1)
+ %cttz3 = call i16 @llvm.cttz.i16(i16 %ld3, i1 -1)
+ %cttz4 = call i16 @llvm.cttz.i16(i16 %ld4, i1 -1)
+ %cttz5 = call i16 @llvm.cttz.i16(i16 %ld5, i1 -1)
+ %cttz6 = call i16 @llvm.cttz.i16(i16 %ld6, i1 -1)
+ %cttz7 = call i16 @llvm.cttz.i16(i16 %ld7, i1 -1)
+ store i16 %cttz0, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %cttz1, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %cttz2, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %cttz3, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %cttz4, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %cttz5, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %cttz6, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %cttz7, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ ret void
+}
+
+define void @cttz_undef_16i16() #0 {
+; SSE-LABEL: @cttz_undef_16i16(
+; SSE-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP1]], i1 true)
+; SSE-NEXT: [[TMP4:%.*]] = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> [[TMP2]], i1 true)
+; SSE-NEXT: store <8 x i16> [[TMP3]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2
+; SSE-NEXT: store <8 x i16> [[TMP4]], <8 x i16>* bitcast (i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8) to <8 x i16>*), align 2
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @cttz_undef_16i16(
+; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2
+; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.cttz.v16i16(<16 x i16> [[TMP1]], i1 true)
+; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2
+; AVX-NEXT: ret void
+;
+ %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2
+ %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2
+ %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2
+ %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2
+ %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2
+ %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2
+ %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2
+ %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2
+ %ld8 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2
+ %ld9 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2
+ %ld10 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2
+ %ld11 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2
+ %ld12 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2
+ %ld13 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2
+ %ld14 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2
+ %ld15 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2
+ %cttz0 = call i16 @llvm.cttz.i16(i16 %ld0, i1 -1)
+ %cttz1 = call i16 @llvm.cttz.i16(i16 %ld1, i1 -1)
+ %cttz2 = call i16 @llvm.cttz.i16(i16 %ld2, i1 -1)
+ %cttz3 = call i16 @llvm.cttz.i16(i16 %ld3, i1 -1)
+ %cttz4 = call i16 @llvm.cttz.i16(i16 %ld4, i1 -1)
+ %cttz5 = call i16 @llvm.cttz.i16(i16 %ld5, i1 -1)
+ %cttz6 = call i16 @llvm.cttz.i16(i16 %ld6, i1 -1)
+ %cttz7 = call i16 @llvm.cttz.i16(i16 %ld7, i1 -1)
+ %cttz8 = call i16 @llvm.cttz.i16(i16 %ld8, i1 -1)
+ %cttz9 = call i16 @llvm.cttz.i16(i16 %ld9, i1 -1)
+ %cttz10 = call i16 @llvm.cttz.i16(i16 %ld10, i1 -1)
+ %cttz11 = call i16 @llvm.cttz.i16(i16 %ld11, i1 -1)
+ %cttz12 = call i16 @llvm.cttz.i16(i16 %ld12, i1 -1)
+ %cttz13 = call i16 @llvm.cttz.i16(i16 %ld13, i1 -1)
+ %cttz14 = call i16 @llvm.cttz.i16(i16 %ld14, i1 -1)
+ %cttz15 = call i16 @llvm.cttz.i16(i16 %ld15, i1 -1)
+ store i16 %cttz0 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2
+ store i16 %cttz1 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2
+ store i16 %cttz2 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2
+ store i16 %cttz3 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2
+ store i16 %cttz4 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2
+ store i16 %cttz5 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2
+ store i16 %cttz6 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2
+ store i16 %cttz7 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2
+ store i16 %cttz8 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2
+ store i16 %cttz9 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2
+ store i16 %cttz10, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2
+ store i16 %cttz11, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2
+ store i16 %cttz12, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2
+ store i16 %cttz13, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2
+ store i16 %cttz14, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2
+ store i16 %cttz15, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2
+ ret void
+}
+
+define void @cttz_undef_16i8() #0 {
+; CHECK-LABEL: @cttz_undef_16i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP1]], i1 true)
+; CHECK-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %cttz0 = call i8 @llvm.cttz.i8(i8 %ld0, i1 -1)
+ %cttz1 = call i8 @llvm.cttz.i8(i8 %ld1, i1 -1)
+ %cttz2 = call i8 @llvm.cttz.i8(i8 %ld2, i1 -1)
+ %cttz3 = call i8 @llvm.cttz.i8(i8 %ld3, i1 -1)
+ %cttz4 = call i8 @llvm.cttz.i8(i8 %ld4, i1 -1)
+ %cttz5 = call i8 @llvm.cttz.i8(i8 %ld5, i1 -1)
+ %cttz6 = call i8 @llvm.cttz.i8(i8 %ld6, i1 -1)
+ %cttz7 = call i8 @llvm.cttz.i8(i8 %ld7, i1 -1)
+ %cttz8 = call i8 @llvm.cttz.i8(i8 %ld8, i1 -1)
+ %cttz9 = call i8 @llvm.cttz.i8(i8 %ld9, i1 -1)
+ %cttz10 = call i8 @llvm.cttz.i8(i8 %ld10, i1 -1)
+ %cttz11 = call i8 @llvm.cttz.i8(i8 %ld11, i1 -1)
+ %cttz12 = call i8 @llvm.cttz.i8(i8 %ld12, i1 -1)
+ %cttz13 = call i8 @llvm.cttz.i8(i8 %ld13, i1 -1)
+ %cttz14 = call i8 @llvm.cttz.i8(i8 %ld14, i1 -1)
+ %cttz15 = call i8 @llvm.cttz.i8(i8 %ld15, i1 -1)
+ store i8 %cttz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %cttz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %cttz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %cttz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %cttz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %cttz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %cttz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %cttz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %cttz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %cttz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %cttz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %cttz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %cttz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %cttz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %cttz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %cttz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ ret void
+}
+
+define void @cttz_undef_32i8() #0 {
+; CHECK-LABEL: @cttz_undef_32i8(
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP1]], i1 true)
+; CHECK-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> [[TMP2]], i1 true)
+; CHECK-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1
+; CHECK-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1
+; CHECK-NEXT: ret void
+;
+ %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1
+ %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1
+ %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1
+ %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1
+ %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1
+ %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1
+ %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1
+ %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1
+ %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1
+ %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1
+ %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1
+ %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1
+ %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1
+ %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1
+ %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1
+ %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1
+ %ld16 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1
+ %ld17 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1
+ %ld18 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1
+ %ld19 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1
+ %ld20 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1
+ %ld21 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1
+ %ld22 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1
+ %ld23 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1
+ %ld24 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1
+ %ld25 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1
+ %ld26 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1
+ %ld27 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1
+ %ld28 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1
+ %ld29 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1
+ %ld30 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1
+ %ld31 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1
+ %cttz0 = call i8 @llvm.cttz.i8(i8 %ld0, i1 -1)
+ %cttz1 = call i8 @llvm.cttz.i8(i8 %ld1, i1 -1)
+ %cttz2 = call i8 @llvm.cttz.i8(i8 %ld2, i1 -1)
+ %cttz3 = call i8 @llvm.cttz.i8(i8 %ld3, i1 -1)
+ %cttz4 = call i8 @llvm.cttz.i8(i8 %ld4, i1 -1)
+ %cttz5 = call i8 @llvm.cttz.i8(i8 %ld5, i1 -1)
+ %cttz6 = call i8 @llvm.cttz.i8(i8 %ld6, i1 -1)
+ %cttz7 = call i8 @llvm.cttz.i8(i8 %ld7, i1 -1)
+ %cttz8 = call i8 @llvm.cttz.i8(i8 %ld8, i1 -1)
+ %cttz9 = call i8 @llvm.cttz.i8(i8 %ld9, i1 -1)
+ %cttz10 = call i8 @llvm.cttz.i8(i8 %ld10, i1 -1)
+ %cttz11 = call i8 @llvm.cttz.i8(i8 %ld11, i1 -1)
+ %cttz12 = call i8 @llvm.cttz.i8(i8 %ld12, i1 -1)
+ %cttz13 = call i8 @llvm.cttz.i8(i8 %ld13, i1 -1)
+ %cttz14 = call i8 @llvm.cttz.i8(i8 %ld14, i1 -1)
+ %cttz15 = call i8 @llvm.cttz.i8(i8 %ld15, i1 -1)
+ %cttz16 = call i8 @llvm.cttz.i8(i8 %ld16, i1 -1)
+ %cttz17 = call i8 @llvm.cttz.i8(i8 %ld17, i1 -1)
+ %cttz18 = call i8 @llvm.cttz.i8(i8 %ld18, i1 -1)
+ %cttz19 = call i8 @llvm.cttz.i8(i8 %ld19, i1 -1)
+ %cttz20 = call i8 @llvm.cttz.i8(i8 %ld20, i1 -1)
+ %cttz21 = call i8 @llvm.cttz.i8(i8 %ld21, i1 -1)
+ %cttz22 = call i8 @llvm.cttz.i8(i8 %ld22, i1 -1)
+ %cttz23 = call i8 @llvm.cttz.i8(i8 %ld23, i1 -1)
+ %cttz24 = call i8 @llvm.cttz.i8(i8 %ld24, i1 -1)
+ %cttz25 = call i8 @llvm.cttz.i8(i8 %ld25, i1 -1)
+ %cttz26 = call i8 @llvm.cttz.i8(i8 %ld26, i1 -1)
+ %cttz27 = call i8 @llvm.cttz.i8(i8 %ld27, i1 -1)
+ %cttz28 = call i8 @llvm.cttz.i8(i8 %ld28, i1 -1)
+ %cttz29 = call i8 @llvm.cttz.i8(i8 %ld29, i1 -1)
+ %cttz30 = call i8 @llvm.cttz.i8(i8 %ld30, i1 -1)
+ %cttz31 = call i8 @llvm.cttz.i8(i8 %ld31, i1 -1)
+ store i8 %cttz0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1
+ store i8 %cttz1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1
+ store i8 %cttz2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1
+ store i8 %cttz3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1
+ store i8 %cttz4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1
+ store i8 %cttz5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1
+ store i8 %cttz6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1
+ store i8 %cttz7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1
+ store i8 %cttz8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1
+ store i8 %cttz9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1
+ store i8 %cttz10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1
+ store i8 %cttz11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1
+ store i8 %cttz12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1
+ store i8 %cttz13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1
+ store i8 %cttz14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1
+ store i8 %cttz15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1
+ store i8 %cttz16, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1
+ store i8 %cttz17, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1
+ store i8 %cttz18, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1
+ store i8 %cttz19, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1
+ store i8 %cttz20, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1
+ store i8 %cttz21, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1
+ store i8 %cttz22, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1
+ store i8 %cttz23, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1
+ store i8 %cttz24, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1
+ store i8 %cttz25, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1
+ store i8 %cttz26, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1
+ store i8 %cttz27, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1
+ store i8 %cttz28, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1
+ store i8 %cttz29, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1
+ store i8 %cttz30, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1
+ store i8 %cttz31, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1
+ ret void
+}
+
+attributes #0 = { nounwind }
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/cycle_dup.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/cycle_dup.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/cycle_dup.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/cycle_dup.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,76 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.9.0"
+
+; int foo(int *A) {
+; int r = A[0], g = A[1], b = A[2], a = A[3];
+; for (int i=0; i < A[13]; i++) {
+; r*=18; g*=19; b*=12; a *=9;
+; }
+; A[0] = r; A[1] = g; A[2] = b; A[3] = a;
+; }
+
+define i32 @foo(i32* nocapture %A) #0 {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast i32* [[A:%.*]] to <4 x i32>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* [[TMP0]], align 4
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 13
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, i32* [[ARRAYIDX4]], align 4
+; CHECK-NEXT: [[CMP24:%.*]] = icmp sgt i32 [[TMP2]], 0
+; CHECK-NEXT: br i1 [[CMP24]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: [[I_029:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT: [[TMP3:%.*]] = phi <4 x i32> [ [[TMP4:%.*]], [[FOR_BODY]] ], [ [[TMP1]], [[ENTRY]] ]
+; CHECK-NEXT: [[TMP4]] = mul nsw <4 x i32> [[TMP3]], <i32 18, i32 19, i32 12, i32 9>
+; CHECK-NEXT: [[INC]] = add nsw i32 [[I_029]], 1
+; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[INC]], [[TMP2]]
+; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_END]]
+; CHECK: for.end:
+; CHECK-NEXT: [[TMP5:%.*]] = phi <4 x i32> [ [[TMP1]], [[ENTRY]] ], [ [[TMP4]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast i32* [[A]] to <4 x i32>*
+; CHECK-NEXT: store <4 x i32> [[TMP5]], <4 x i32>* [[TMP6]], align 4
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %0 = load i32, i32* %A, align 4
+ %arrayidx1 = getelementptr inbounds i32, i32* %A, i64 1
+ %1 = load i32, i32* %arrayidx1, align 4
+ %arrayidx2 = getelementptr inbounds i32, i32* %A, i64 2
+ %2 = load i32, i32* %arrayidx2, align 4
+ %arrayidx3 = getelementptr inbounds i32, i32* %A, i64 3
+ %3 = load i32, i32* %arrayidx3, align 4
+ %arrayidx4 = getelementptr inbounds i32, i32* %A, i64 13
+ %4 = load i32, i32* %arrayidx4, align 4
+ %cmp24 = icmp sgt i32 %4, 0
+ br i1 %cmp24, label %for.body, label %for.end
+
+for.body: ; preds = %entry, %for.body
+ %i.029 = phi i32 [ %inc, %for.body ], [ 0, %entry ]
+ %a.028 = phi i32 [ %mul7, %for.body ], [ %3, %entry ]
+ %b.027 = phi i32 [ %mul6, %for.body ], [ %2, %entry ]
+ %g.026 = phi i32 [ %mul5, %for.body ], [ %1, %entry ]
+ %r.025 = phi i32 [ %mul, %for.body ], [ %0, %entry ]
+ %mul = mul nsw i32 %r.025, 18
+ %mul5 = mul nsw i32 %g.026, 19
+ %mul6 = mul nsw i32 %b.027, 12
+ %mul7 = mul nsw i32 %a.028, 9
+ %inc = add nsw i32 %i.029, 1
+ %cmp = icmp slt i32 %inc, %4
+ br i1 %cmp, label %for.body, label %for.end
+
+for.end: ; preds = %for.body, %entry
+ %a.0.lcssa = phi i32 [ %3, %entry ], [ %mul7, %for.body ]
+ %b.0.lcssa = phi i32 [ %2, %entry ], [ %mul6, %for.body ]
+ %g.0.lcssa = phi i32 [ %1, %entry ], [ %mul5, %for.body ]
+ %r.0.lcssa = phi i32 [ %0, %entry ], [ %mul, %for.body ]
+ store i32 %r.0.lcssa, i32* %A, align 4
+ store i32 %g.0.lcssa, i32* %arrayidx1, align 4
+ store i32 %b.0.lcssa, i32* %arrayidx2, align 4
+ store i32 %a.0.lcssa, i32* %arrayidx3, align 4
+ ret i32 undef
+}
+
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/debug_info.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/debug_info.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/debug_info.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/debug_info.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,101 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.7.0"
+
+; int depth(double *A, int m) {
+; double y0 = 0; double y1 = 1;
+; for (int i=0; i < m; i++) {
+; y0 = A[4];
+; y1 = A[5];
+; }
+; A[8] = y0; A[8+1] = y1;
+; }
+
+define i32 @depth(double* nocapture %A, i32 %m) #0 !dbg !4 {
+; CHECK-LABEL: @depth(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: call void @llvm.dbg.value(metadata double* [[A:%.*]], metadata !12, metadata !DIExpression()), !dbg !18
+; CHECK-NEXT: call void @llvm.dbg.value(metadata i32 [[M:%.*]], metadata !13, metadata !DIExpression()), !dbg !18
+; CHECK-NEXT: call void @llvm.dbg.value(metadata double 0.000000e+00, metadata !14, metadata !DIExpression()), !dbg !19
+; CHECK-NEXT: call void @llvm.dbg.value(metadata double 2.000000e-01, metadata !15, metadata !DIExpression()), !dbg !19
+; CHECK-NEXT: call void @llvm.dbg.value(metadata i32 0, metadata !16, metadata !DIExpression()), !dbg !20
+; CHECK-NEXT: [[CMP8:%.*]] = icmp sgt i32 [[M]], 0, !dbg !20
+; CHECK-NEXT: br i1 [[CMP8]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END:%.*]], !dbg !20
+; CHECK: for.body.lr.ph:
+; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[A]], i64 4, !dbg !21
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[ARRAYIDX]] to <2 x double>*, !dbg !21
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8, !dbg !21
+; CHECK-NEXT: br label [[FOR_END]], !dbg !20
+; CHECK: for.end:
+; CHECK-NEXT: [[TMP2:%.*]] = phi <2 x double> [ [[TMP1]], [[FOR_BODY_LR_PH]] ], [ <double 0.000000e+00, double 1.000000e+00>, [[ENTRY:%.*]] ]
+; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds double, double* [[A]], i64 8, !dbg !23
+; CHECK-NEXT: [[TMP3:%.*]] = bitcast double* [[ARRAYIDX2]] to <2 x double>*, !dbg !23
+; CHECK-NEXT: store <2 x double> [[TMP2]], <2 x double>* [[TMP3]], align 8, !dbg !23
+; CHECK-NEXT: ret i32 undef, !dbg !24
+;
+entry:
+ tail call void @llvm.dbg.value(metadata double* %A, i64 0, metadata !12, metadata !DIExpression()), !dbg !19
+ tail call void @llvm.dbg.value(metadata i32 %m, i64 0, metadata !13, metadata !DIExpression()), !dbg !19
+ tail call void @llvm.dbg.value(metadata double 0.0, i64 0, metadata !14, metadata !DIExpression()), !dbg !21
+ tail call void @llvm.dbg.value(metadata double 0.2, i64 0, metadata !15, metadata !DIExpression()), !dbg !21
+ tail call void @llvm.dbg.value(metadata i32 0, i64 0, metadata !16, metadata !DIExpression()), !dbg !23
+ %cmp8 = icmp sgt i32 %m, 0, !dbg !23
+ br i1 %cmp8, label %for.body.lr.ph, label %for.end, !dbg !23
+
+for.body.lr.ph: ; preds = %entry
+ %arrayidx = getelementptr inbounds double, double* %A, i64 4, !dbg !24
+ %0 = load double, double* %arrayidx, align 8, !dbg !24
+ %arrayidx1 = getelementptr inbounds double, double* %A, i64 5, !dbg !29
+ %1 = load double, double* %arrayidx1, align 8, !dbg !29
+ br label %for.end, !dbg !23
+
+for.end: ; preds = %for.body.lr.ph, %entry
+ %y1.0.lcssa = phi double [ %1, %for.body.lr.ph ], [ 1.000000e+00, %entry ]
+ %y0.0.lcssa = phi double [ %0, %for.body.lr.ph ], [ 0.000000e+00, %entry ]
+ %arrayidx2 = getelementptr inbounds double, double* %A, i64 8, !dbg !30
+ store double %y0.0.lcssa, double* %arrayidx2, align 8, !dbg !30
+ %arrayidx3 = getelementptr inbounds double, double* %A, i64 9, !dbg !30
+ store double %y1.0.lcssa, double* %arrayidx3, align 8, !dbg !30
+ ret i32 undef, !dbg !31
+}
+
+; Function Attrs: nounwind readnone
+declare void @llvm.dbg.value(metadata, i64, metadata, metadata) #1
+
+attributes #0 = { nounwind ssp uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf"="true" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nounwind readnone }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!18, !32}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, producer: "clang version 3.4 (trunk 187335) (llvm/trunk 187335:187340M)", isOptimized: true, emissionKind: FullDebug, file: !1, enums: !2, retainedTypes: !2, globals: !2, imports: !2)
+!1 = !DIFile(filename: "file.c", directory: "/Users/nadav")
+!2 = !{}
+!4 = distinct !DISubprogram(name: "depth", line: 1, isLocal: false, isDefinition: true, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: true, unit: !0, scopeLine: 1, file: !1, scope: !5, type: !6, retainedNodes: !11)
+!5 = !DIFile(filename: "file.c", directory: "/Users/nadav")
+!6 = !DISubroutineType(types: !7)
+!7 = !{!8, !9, !8}
+!8 = !DIBasicType(tag: DW_TAG_base_type, name: "int", size: 32, align: 32, encoding: DW_ATE_signed)
+!9 = !DIDerivedType(tag: DW_TAG_pointer_type, size: 64, align: 64, baseType: !10)
+!10 = !DIBasicType(tag: DW_TAG_base_type, name: "double", size: 64, align: 64, encoding: DW_ATE_float)
+!11 = !{!12, !13, !14, !15, !16}
+!12 = !DILocalVariable(name: "A", line: 1, arg: 1, scope: !4, file: !5, type: !9)
+!13 = !DILocalVariable(name: "m", line: 1, arg: 2, scope: !4, file: !5, type: !8)
+!14 = !DILocalVariable(name: "y0", line: 2, scope: !4, file: !5, type: !10)
+!15 = !DILocalVariable(name: "y1", line: 2, scope: !4, file: !5, type: !10)
+!16 = !DILocalVariable(name: "i", line: 3, scope: !17, file: !5, type: !8)
+!17 = distinct !DILexicalBlock(line: 3, column: 0, file: !1, scope: !4)
+!18 = !{i32 2, !"Dwarf Version", i32 2}
+!19 = !DILocation(line: 1, scope: !4)
+!20 = !{double 0.000000e+00}
+!21 = !DILocation(line: 2, scope: !4)
+!22 = !{double 1.000000e+00}
+!23 = !DILocation(line: 3, scope: !17)
+!24 = !DILocation(line: 4, scope: !25)
+!25 = distinct !DILexicalBlock(line: 3, column: 0, file: !1, scope: !17)
+!29 = !DILocation(line: 5, scope: !25)
+!30 = !DILocation(line: 7, scope: !4)
+!31 = !DILocation(line: 8, scope: !4)
+!32 = !{i32 1, !"Debug Info Version", i32 3}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/diamond.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/diamond.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/diamond.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/diamond.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,157 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+; int foo(int * restrict B, int * restrict A, int n, int m) {
+; B[0] = n * A[0] + m * A[0];
+; B[1] = n * A[1] + m * A[1];
+; B[2] = n * A[2] + m * A[2];
+; B[3] = n * A[3] + m * A[3];
+; return 0;
+; }
+
+define i32 @foo(i32* noalias nocapture %B, i32* noalias nocapture %A, i32 %n, i32 %m) #0 {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[MUL238:%.*]] = add i32 [[M:%.*]], [[N:%.*]]
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 1
+; CHECK-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 1
+; CHECK-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 2
+; CHECK-NEXT: [[ARRAYIDX15:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 2
+; CHECK-NEXT: [[ARRAYIDX16:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 3
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast i32* [[A]] to <4 x i32>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* [[TMP0]], align 4
+; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> undef, i32 [[MUL238]], i32 0
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[MUL238]], i32 1
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <4 x i32> [[TMP3]], i32 [[MUL238]], i32 2
+; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x i32> [[TMP4]], i32 [[MUL238]], i32 3
+; CHECK-NEXT: [[TMP6:%.*]] = mul <4 x i32> [[TMP1]], [[TMP5]]
+; CHECK-NEXT: [[ARRAYIDX21:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 3
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast i32* [[B]] to <4 x i32>*
+; CHECK-NEXT: store <4 x i32> [[TMP6]], <4 x i32>* [[TMP7]], align 4
+; CHECK-NEXT: ret i32 0
+;
+entry:
+ %0 = load i32, i32* %A, align 4
+ %mul238 = add i32 %m, %n
+ %add = mul i32 %0, %mul238
+ store i32 %add, i32* %B, align 4
+ %arrayidx4 = getelementptr inbounds i32, i32* %A, i64 1
+ %1 = load i32, i32* %arrayidx4, align 4
+ %add8 = mul i32 %1, %mul238
+ %arrayidx9 = getelementptr inbounds i32, i32* %B, i64 1
+ store i32 %add8, i32* %arrayidx9, align 4
+ %arrayidx10 = getelementptr inbounds i32, i32* %A, i64 2
+ %2 = load i32, i32* %arrayidx10, align 4
+ %add14 = mul i32 %2, %mul238
+ %arrayidx15 = getelementptr inbounds i32, i32* %B, i64 2
+ store i32 %add14, i32* %arrayidx15, align 4
+ %arrayidx16 = getelementptr inbounds i32, i32* %A, i64 3
+ %3 = load i32, i32* %arrayidx16, align 4
+ %add20 = mul i32 %3, %mul238
+ %arrayidx21 = getelementptr inbounds i32, i32* %B, i64 3
+ store i32 %add20, i32* %arrayidx21, align 4
+ ret i32 0
+}
+
+
+; int extr_user(int * restrict B, int * restrict A, int n, int m) {
+; B[0] = n * A[0] + m * A[0];
+; B[1] = n * A[1] + m * A[1];
+; B[2] = n * A[2] + m * A[2];
+; B[3] = n * A[3] + m * A[3];
+; return A[0];
+; }
+
+define i32 @extr_user(i32* noalias nocapture %B, i32* noalias nocapture %A, i32 %n, i32 %m) {
+; CHECK-LABEL: @extr_user(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[MUL238:%.*]] = add i32 [[M:%.*]], [[N:%.*]]
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 1
+; CHECK-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 1
+; CHECK-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 2
+; CHECK-NEXT: [[ARRAYIDX15:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 2
+; CHECK-NEXT: [[ARRAYIDX16:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 3
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast i32* [[A]] to <4 x i32>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* [[TMP0]], align 4
+; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> undef, i32 [[MUL238]], i32 0
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[MUL238]], i32 1
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <4 x i32> [[TMP3]], i32 [[MUL238]], i32 2
+; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x i32> [[TMP4]], i32 [[MUL238]], i32 3
+; CHECK-NEXT: [[TMP6:%.*]] = mul <4 x i32> [[TMP1]], [[TMP5]]
+; CHECK-NEXT: [[ARRAYIDX21:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 3
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast i32* [[B]] to <4 x i32>*
+; CHECK-NEXT: store <4 x i32> [[TMP6]], <4 x i32>* [[TMP7]], align 4
+; CHECK-NEXT: [[TMP8:%.*]] = extractelement <4 x i32> [[TMP1]], i32 0
+; CHECK-NEXT: ret i32 [[TMP8]]
+;
+entry:
+ %0 = load i32, i32* %A, align 4
+ %mul238 = add i32 %m, %n
+ %add = mul i32 %0, %mul238
+ store i32 %add, i32* %B, align 4
+ %arrayidx4 = getelementptr inbounds i32, i32* %A, i64 1
+ %1 = load i32, i32* %arrayidx4, align 4
+ %add8 = mul i32 %1, %mul238
+ %arrayidx9 = getelementptr inbounds i32, i32* %B, i64 1
+ store i32 %add8, i32* %arrayidx9, align 4
+ %arrayidx10 = getelementptr inbounds i32, i32* %A, i64 2
+ %2 = load i32, i32* %arrayidx10, align 4
+ %add14 = mul i32 %2, %mul238
+ %arrayidx15 = getelementptr inbounds i32, i32* %B, i64 2
+ store i32 %add14, i32* %arrayidx15, align 4
+ %arrayidx16 = getelementptr inbounds i32, i32* %A, i64 3
+ %3 = load i32, i32* %arrayidx16, align 4
+ %add20 = mul i32 %3, %mul238
+ %arrayidx21 = getelementptr inbounds i32, i32* %B, i64 3
+ store i32 %add20, i32* %arrayidx21, align 4
+ ret i32 %0 ;<--------- This value has multiple users
+}
+
+; In this example we have an external user that is not the first element in the vector.
+define i32 @extr_user1(i32* noalias nocapture %B, i32* noalias nocapture %A, i32 %n, i32 %m) {
+; CHECK-LABEL: @extr_user1(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[MUL238:%.*]] = add i32 [[M:%.*]], [[N:%.*]]
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 1
+; CHECK-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 1
+; CHECK-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 2
+; CHECK-NEXT: [[ARRAYIDX15:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 2
+; CHECK-NEXT: [[ARRAYIDX16:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 3
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast i32* [[A]] to <4 x i32>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* [[TMP0]], align 4
+; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> undef, i32 [[MUL238]], i32 0
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[MUL238]], i32 1
+; CHECK-NEXT: [[TMP4:%.*]] = insertelement <4 x i32> [[TMP3]], i32 [[MUL238]], i32 2
+; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x i32> [[TMP4]], i32 [[MUL238]], i32 3
+; CHECK-NEXT: [[TMP6:%.*]] = mul <4 x i32> [[TMP1]], [[TMP5]]
+; CHECK-NEXT: [[ARRAYIDX21:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 3
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast i32* [[B]] to <4 x i32>*
+; CHECK-NEXT: store <4 x i32> [[TMP6]], <4 x i32>* [[TMP7]], align 4
+; CHECK-NEXT: [[TMP8:%.*]] = extractelement <4 x i32> [[TMP1]], i32 1
+; CHECK-NEXT: ret i32 [[TMP8]]
+;
+entry:
+ %0 = load i32, i32* %A, align 4
+ %mul238 = add i32 %m, %n
+ %add = mul i32 %0, %mul238
+ store i32 %add, i32* %B, align 4
+ %arrayidx4 = getelementptr inbounds i32, i32* %A, i64 1
+ %1 = load i32, i32* %arrayidx4, align 4
+ %add8 = mul i32 %1, %mul238
+ %arrayidx9 = getelementptr inbounds i32, i32* %B, i64 1
+ store i32 %add8, i32* %arrayidx9, align 4
+ %arrayidx10 = getelementptr inbounds i32, i32* %A, i64 2
+ %2 = load i32, i32* %arrayidx10, align 4
+ %add14 = mul i32 %2, %mul238
+ %arrayidx15 = getelementptr inbounds i32, i32* %B, i64 2
+ store i32 %add14, i32* %arrayidx15, align 4
+ %arrayidx16 = getelementptr inbounds i32, i32* %A, i64 3
+ %3 = load i32, i32* %arrayidx16, align 4
+ %add20 = mul i32 %3, %mul238
+ %arrayidx21 = getelementptr inbounds i32, i32* %B, i64 3
+ store i32 %add20, i32* %arrayidx21, align 4
+ ret i32 %1 ;<--------- This value has multiple users
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/external_user.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/external_user.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/external_user.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/external_user.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,132 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+; double foo(double * restrict b, double * restrict a, int n, int m) {
+; double r=a[1];
+; double g=a[0];
+; double x;
+; for (int i=0; i < 100; i++) {
+; r += 10;
+; g += 10;
+; r *= 4;
+; g *= 4;
+; x = g; <----- external user!
+; r += 4;
+; g += 4;
+; }
+; b[0] = g;
+; b[1] = r;
+;
+; return x; <-- must extract here!
+; }
+
+define double @ext_user(double* noalias nocapture %B, double* noalias nocapture %A, i32 %n, i32 %m) {
+; CHECK-LABEL: @ext_user(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[A:%.*]] to <2 x double>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8
+; CHECK-NEXT: br label [[FOR_BODY:%.*]]
+; CHECK: for.body:
+; CHECK-NEXT: [[I_020:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[TMP2:%.*]] = phi <2 x double> [ [[TMP1]], [[ENTRY]] ], [ [[TMP5:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT: [[TMP3:%.*]] = fadd <2 x double> [[TMP2]], <double 1.000000e+01, double 1.000000e+01>
+; CHECK-NEXT: [[TMP4:%.*]] = fmul <2 x double> [[TMP3]], <double 4.000000e+00, double 4.000000e+00>
+; CHECK-NEXT: [[TMP5]] = fadd <2 x double> [[TMP4]], <double 4.000000e+00, double 4.000000e+00>
+; CHECK-NEXT: [[INC]] = add nsw i32 [[I_020]], 1
+; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 100
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; CHECK: for.end:
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast double* [[B:%.*]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP5]], <2 x double>* [[TMP6]], align 8
+; CHECK-NEXT: [[TMP7:%.*]] = extractelement <2 x double> [[TMP4]], i32 0
+; CHECK-NEXT: ret double [[TMP7]]
+;
+entry:
+ %arrayidx = getelementptr inbounds double, double* %A, i64 1
+ %0 = load double, double* %arrayidx, align 8
+ %1 = load double, double* %A, align 8
+ br label %for.body
+
+for.body: ; preds = %for.body, %entry
+ %i.020 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
+ %G.019 = phi double [ %1, %entry ], [ %add5, %for.body ]
+ %R.018 = phi double [ %0, %entry ], [ %add4, %for.body ]
+ %add = fadd double %R.018, 1.000000e+01
+ %add2 = fadd double %G.019, 1.000000e+01
+ %mul = fmul double %add, 4.000000e+00
+ %mul3 = fmul double %add2, 4.000000e+00
+ %add4 = fadd double %mul, 4.000000e+00
+ %add5 = fadd double %mul3, 4.000000e+00
+ %inc = add nsw i32 %i.020, 1
+ %exitcond = icmp eq i32 %inc, 100
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.body
+ store double %add5, double* %B, align 8
+ %arrayidx7 = getelementptr inbounds double, double* %B, i64 1
+ store double %add4, double* %arrayidx7, align 8
+ ret double %mul3
+}
+
+; A need-to-gather entry cannot be an external use of the scalar element.
+; Instead the insertelement instructions of the need-to-gather entry are the
+; external users.
+; This test would assert because we would keep the scalar fpext and fadd alive.
+; PR18129
+
+define i32 @needtogather(double *noalias %a, i32 *noalias %b, float * noalias %c,
+; CHECK-LABEL: @needtogather(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[D:%.*]], align 4
+; CHECK-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
+; CHECK-NEXT: [[TMP1:%.*]] = load float, float* [[C:%.*]]
+; CHECK-NEXT: [[SUB:%.*]] = fsub float 0.000000e+00, [[TMP1]]
+; CHECK-NEXT: [[MUL:%.*]] = fmul float [[SUB]], 0.000000e+00
+; CHECK-NEXT: [[ADD:%.*]] = fadd float [[CONV]], [[MUL]]
+; CHECK-NEXT: [[CONV1:%.*]] = fpext float [[ADD]] to double
+; CHECK-NEXT: [[SUB3:%.*]] = fsub float 1.000000e+00, [[TMP1]]
+; CHECK-NEXT: [[MUL4:%.*]] = fmul float [[SUB3]], 0.000000e+00
+; CHECK-NEXT: [[ADD5:%.*]] = fadd float [[CONV]], [[MUL4]]
+; CHECK-NEXT: [[CONV6:%.*]] = fpext float [[ADD5]] to double
+; CHECK-NEXT: [[TOBOOL:%.*]] = fcmp une float [[ADD]], 0.000000e+00
+; CHECK-NEXT: br i1 [[TOBOOL]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
+; CHECK: if.then:
+; CHECK-NEXT: br label [[IF_END]]
+; CHECK: if.end:
+; CHECK-NEXT: [[STOREMERGE:%.*]] = phi double [ [[CONV6]], [[IF_THEN]] ], [ [[CONV1]], [[ENTRY:%.*]] ]
+; CHECK-NEXT: [[E_0:%.*]] = phi double [ [[CONV1]], [[IF_THEN]] ], [ [[CONV6]], [[ENTRY]] ]
+; CHECK-NEXT: store double [[STOREMERGE]], double* [[A:%.*]], align 8
+; CHECK-NEXT: [[CONV7:%.*]] = fptosi double [[E_0]] to i32
+; CHECK-NEXT: store i32 [[CONV7]], i32* [[B:%.*]], align 4
+; CHECK-NEXT: ret i32 undef
+;
+ i32 * noalias %d) {
+entry:
+ %0 = load i32, i32* %d, align 4
+ %conv = sitofp i32 %0 to float
+ %1 = load float, float* %c
+ %sub = fsub float 0.000000e+00, %1
+ %mul = fmul float %sub, 0.000000e+00
+ %add = fadd float %conv, %mul
+ %conv1 = fpext float %add to double
+ %sub3 = fsub float 1.000000e+00, %1
+ %mul4 = fmul float %sub3, 0.000000e+00
+ %add5 = fadd float %conv, %mul4
+ %conv6 = fpext float %add5 to double
+ %tobool = fcmp une float %add, 0.000000e+00
+ br i1 %tobool, label %if.then, label %if.end
+
+if.then:
+ br label %if.end
+
+if.end:
+ %storemerge = phi double [ %conv6, %if.then ], [ %conv1, %entry ]
+ %e.0 = phi double [ %conv1, %if.then ], [ %conv6, %entry ]
+ store double %storemerge, double* %a, align 8
+ %conv7 = fptosi double %e.0 to i32
+ store i32 %conv7, i32* %b, align 4
+ ret i32 undef
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/external_user_jumbled_load.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/external_user_jumbled_load.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/external_user_jumbled_load.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/external_user_jumbled_load.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,43 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -S -mtriple=x86_64-unknown -mattr=+avx -slp-vectorizer | FileCheck %s
+
+ at array = external global [20 x [13 x i32]]
+
+define void @hoge(i64 %idx, <4 x i32>* %sink) {
+; CHECK-LABEL: @hoge(
+; CHECK-NEXT: bb:
+; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds [20 x [13 x i32]], [20 x [13 x i32]]* @array, i64 0, i64 [[IDX:%.*]], i64 5
+; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds [20 x [13 x i32]], [20 x [13 x i32]]* @array, i64 0, i64 [[IDX]], i64 6
+; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [20 x [13 x i32]], [20 x [13 x i32]]* @array, i64 0, i64 [[IDX]], i64 7
+; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds [20 x [13 x i32]], [20 x [13 x i32]]* @array, i64 0, i64 [[IDX]], i64 8
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast i32* [[TMP0]] to <4 x i32>*
+; CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[TMP4]], align 4
+; CHECK-NEXT: [[REORDER_SHUFFLE:%.*]] = shufflevector <4 x i32> [[TMP5]], <4 x i32> undef, <4 x i32> <i32 1, i32 2, i32 3, i32 0>
+; CHECK-NEXT: [[TMP6:%.*]] = extractelement <4 x i32> [[REORDER_SHUFFLE]], i32 0
+; CHECK-NEXT: [[TMP7:%.*]] = insertelement <4 x i32> undef, i32 [[TMP6]], i32 0
+; CHECK-NEXT: [[TMP8:%.*]] = extractelement <4 x i32> [[REORDER_SHUFFLE]], i32 1
+; CHECK-NEXT: [[TMP9:%.*]] = insertelement <4 x i32> [[TMP7]], i32 [[TMP8]], i32 1
+; CHECK-NEXT: [[TMP10:%.*]] = extractelement <4 x i32> [[REORDER_SHUFFLE]], i32 2
+; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x i32> [[TMP9]], i32 [[TMP10]], i32 2
+; CHECK-NEXT: [[TMP12:%.*]] = extractelement <4 x i32> [[REORDER_SHUFFLE]], i32 3
+; CHECK-NEXT: [[TMP13:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP12]], i32 3
+; CHECK-NEXT: store <4 x i32> [[TMP13]], <4 x i32>* [[SINK:%.*]]
+; CHECK-NEXT: ret void
+;
+bb:
+ %0 = getelementptr inbounds [20 x [13 x i32]], [20 x [13 x i32]]* @array, i64 0, i64 %idx, i64 5
+ %1 = getelementptr inbounds [20 x [13 x i32]], [20 x [13 x i32]]* @array, i64 0, i64 %idx, i64 6
+ %2 = getelementptr inbounds [20 x [13 x i32]], [20 x [13 x i32]]* @array, i64 0, i64 %idx, i64 7
+ %3 = getelementptr inbounds [20 x [13 x i32]], [20 x [13 x i32]]* @array, i64 0, i64 %idx, i64 8
+ %4 = load i32, i32* %1, align 4
+ %5 = insertelement <4 x i32> undef, i32 %4, i32 0
+ %6 = load i32, i32* %2, align 4
+ %7 = insertelement <4 x i32> %5, i32 %6, i32 1
+ %8 = load i32, i32* %3, align 4
+ %9 = insertelement <4 x i32> %7, i32 %8, i32 2
+ %10 = load i32, i32* %0, align 4
+ %11 = insertelement <4 x i32> %9, i32 %10, i32 3
+ store <4 x i32> %11, <4 x i32>* %sink
+ ret void
+}
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/extract-shuffle.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/extract-shuffle.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/extract-shuffle.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/extract-shuffle.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,22 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -slp-vectorizer -S -o - -mtriple=x86_64-unknown-linux -mcpu=bdver2 -slp-schedule-budget=1 | FileCheck %s
+
+define <2 x i8> @g(<2 x i8> %x, <2 x i8> %y) {
+; CHECK-LABEL: @g(
+; CHECK-NEXT: [[X0:%.*]] = extractelement <2 x i8> [[X:%.*]], i32 0
+; CHECK-NEXT: [[Y1:%.*]] = extractelement <2 x i8> [[Y:%.*]], i32 1
+; CHECK-NEXT: [[X0X0:%.*]] = mul i8 [[X0]], [[X0]]
+; CHECK-NEXT: [[Y1Y1:%.*]] = mul i8 [[Y1]], [[Y1]]
+; CHECK-NEXT: [[INS1:%.*]] = insertelement <2 x i8> undef, i8 [[X0X0]], i32 0
+; CHECK-NEXT: [[INS2:%.*]] = insertelement <2 x i8> [[INS1]], i8 [[Y1Y1]], i32 1
+; CHECK-NEXT: ret <2 x i8> [[INS2]]
+;
+ %x0 = extractelement <2 x i8> %x, i32 0
+ %y1 = extractelement <2 x i8> %y, i32 1
+ %x0x0 = mul i8 %x0, %x0
+ %y1y1 = mul i8 %y1, %y1
+ %ins1 = insertelement <2 x i8> undef, i8 %x0x0, i32 0
+ %ins2 = insertelement <2 x i8> %ins1, i8 %y1y1, i32 1
+ ret <2 x i8> %ins2
+}
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/extract.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/extract.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/extract.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/extract.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,79 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+define void @fextr(double* %ptr) {
+; CHECK-LABEL: @fextr(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[LD:%.*]] = load <2 x double>, <2 x double>* undef
+; CHECK-NEXT: [[P0:%.*]] = getelementptr inbounds double, double* [[PTR:%.*]], i64 0
+; CHECK-NEXT: [[TMP0:%.*]] = fadd <2 x double> [[LD]], <double 0.000000e+00, double 1.100000e+00>
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast double* [[P0]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP0]], <2 x double>* [[TMP1]], align 4
+; CHECK-NEXT: ret void
+;
+entry:
+ %LD = load <2 x double>, <2 x double>* undef
+ %V0 = extractelement <2 x double> %LD, i32 0
+ %V1 = extractelement <2 x double> %LD, i32 1
+ %P0 = getelementptr inbounds double, double* %ptr, i64 0
+ %P1 = getelementptr inbounds double, double* %ptr, i64 1
+ %A0 = fadd double %V0, 0.0
+ %A1 = fadd double %V1, 1.1
+ store double %A0, double* %P0, align 4
+ store double %A1, double* %P1, align 4
+ ret void
+}
+
+define void @fextr1(double* %ptr) {
+; CHECK-LABEL: @fextr1(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[LD:%.*]] = load <2 x double>, <2 x double>* undef
+; CHECK-NEXT: [[REORDER_SHUFFLE:%.*]] = shufflevector <2 x double> [[LD]], <2 x double> undef, <2 x i32> <i32 1, i32 0>
+; CHECK-NEXT: [[P1:%.*]] = getelementptr inbounds double, double* [[PTR:%.*]], i64 0
+; CHECK-NEXT: [[TMP0:%.*]] = fadd <2 x double> [[REORDER_SHUFFLE]], <double 3.400000e+00, double 1.200000e+00>
+; CHECK-NEXT: [[TMP1:%.*]] = bitcast double* [[P1]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP0]], <2 x double>* [[TMP1]], align 4
+; CHECK-NEXT: ret void
+;
+entry:
+ %LD = load <2 x double>, <2 x double>* undef
+ %V0 = extractelement <2 x double> %LD, i32 0
+ %V1 = extractelement <2 x double> %LD, i32 1
+ %P0 = getelementptr inbounds double, double* %ptr, i64 1 ; <--- incorrect order
+ %P1 = getelementptr inbounds double, double* %ptr, i64 0
+ %A0 = fadd double %V0, 1.2
+ %A1 = fadd double %V1, 3.4
+ store double %A0, double* %P0, align 4
+ store double %A1, double* %P1, align 4
+ ret void
+}
+
+define void @fextr2(double* %ptr) {
+; CHECK-LABEL: @fextr2(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[LD:%.*]] = load <4 x double>, <4 x double>* undef
+; CHECK-NEXT: [[V0:%.*]] = extractelement <4 x double> [[LD]], i32 0
+; CHECK-NEXT: [[V1:%.*]] = extractelement <4 x double> [[LD]], i32 1
+; CHECK-NEXT: [[P0:%.*]] = getelementptr inbounds double, double* [[PTR:%.*]], i64 0
+; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x double> undef, double [[V0]], i32 0
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x double> [[TMP0]], double [[V1]], i32 1
+; CHECK-NEXT: [[TMP2:%.*]] = fadd <2 x double> [[TMP1]], <double 5.500000e+00, double 6.600000e+00>
+; CHECK-NEXT: [[TMP3:%.*]] = bitcast double* [[P0]] to <2 x double>*
+; CHECK-NEXT: store <2 x double> [[TMP2]], <2 x double>* [[TMP3]], align 4
+; CHECK-NEXT: ret void
+;
+entry:
+ %LD = load <4 x double>, <4 x double>* undef
+ %V0 = extractelement <4 x double> %LD, i32 0 ; <--- invalid size.
+ %V1 = extractelement <4 x double> %LD, i32 1
+ %P0 = getelementptr inbounds double, double* %ptr, i64 0
+ %P1 = getelementptr inbounds double, double* %ptr, i64 1
+ %A0 = fadd double %V0, 5.5
+ %A1 = fadd double %V1, 6.6
+ store double %A0, double* %P0, align 4
+ store double %A1, double* %P1, align 4
+ ret void
+}
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/extract_in_tree_user.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/extract_in_tree_user.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/extract_in_tree_user.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/extract_in_tree_user.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,101 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -S -mtriple=i386-apple-macosx10.9.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+
+ at a = common global i64* null, align 8
+
+; Function Attrs: nounwind ssp uwtable
+define i32 @fn1() {
+; CHECK-LABEL: @fn1(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = load i64*, i64** @a, align 8
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x i64*> undef, i64* [[TMP0]], i32 0
+; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x i64*> [[TMP1]], i64* [[TMP0]], i32 1
+; CHECK-NEXT: [[TMP3:%.*]] = getelementptr i64, <2 x i64*> [[TMP2]], <2 x i64> <i64 11, i64 56>
+; CHECK-NEXT: [[TMP4:%.*]] = ptrtoint <2 x i64*> [[TMP3]] to <2 x i64>
+; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i64, i64* [[TMP0]], i64 12
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i64*> [[TMP3]], i32 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast i64* [[TMP5]] to <2 x i64>*
+; CHECK-NEXT: store <2 x i64> [[TMP4]], <2 x i64>* [[TMP6]], align 8
+; CHECK-NEXT: ret i32 undef
+;
+entry:
+ %0 = load i64*, i64** @a, align 8
+ %add.ptr = getelementptr inbounds i64, i64* %0, i64 11
+ %1 = ptrtoint i64* %add.ptr to i64
+ store i64 %1, i64* %add.ptr, align 8
+ %add.ptr1 = getelementptr inbounds i64, i64* %0, i64 56
+ %2 = ptrtoint i64* %add.ptr1 to i64
+ %arrayidx2 = getelementptr inbounds i64, i64* %0, i64 12
+ store i64 %2, i64* %arrayidx2, align 8
+ ret i32 undef
+}
+
+
+declare float @llvm.powi.f32(float, i32)
+define void @fn2(i32* %a, i32* %b, float* %c) {
+; CHECK-LABEL: @fn2(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 1
+; CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i32 1
+; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 2
+; CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i32 2
+; CHECK-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 3
+; CHECK-NEXT: [[TMP0:%.*]] = bitcast i32* [[A]] to <4 x i32>*
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* [[TMP0]], align 4
+; CHECK-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds i32, i32* [[B]], i32 3
+; CHECK-NEXT: [[TMP2:%.*]] = bitcast i32* [[B]] to <4 x i32>*
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* [[TMP2]], align 4
+; CHECK-NEXT: [[TMP4:%.*]] = add <4 x i32> [[TMP1]], [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = sitofp <4 x i32> [[TMP4]] to <4 x float>
+; CHECK-NEXT: [[TMP6:%.*]] = extractelement <4 x i32> [[TMP4]], i32 0
+; CHECK-NEXT: [[TMP7:%.*]] = call <4 x float> @llvm.powi.v4f32(<4 x float> [[TMP5]], i32 [[TMP6]])
+; CHECK-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[C:%.*]], i32 1
+; CHECK-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[C]], i32 2
+; CHECK-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[C]], i32 3
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast float* [[C]] to <4 x float>*
+; CHECK-NEXT: store <4 x float> [[TMP7]], <4 x float>* [[TMP8]], align 4
+; CHECK-NEXT: ret void
+;
+entry:
+ %i0 = load i32, i32* %a, align 4
+ %i1 = load i32, i32* %b, align 4
+ %add1 = add i32 %i0, %i1
+ %fp1 = sitofp i32 %add1 to float
+ %call1 = tail call float @llvm.powi.f32(float %fp1,i32 %add1) nounwind readnone
+
+ %arrayidx2 = getelementptr inbounds i32, i32* %a, i32 1
+ %i2 = load i32, i32* %arrayidx2, align 4
+ %arrayidx3 = getelementptr inbounds i32, i32* %b, i32 1
+ %i3 = load i32, i32* %arrayidx3, align 4
+ %add2 = add i32 %i2, %i3
+ %fp2 = sitofp i32 %add2 to float
+ %call2 = tail call float @llvm.powi.f32(float %fp2,i32 %add1) nounwind readnone
+
+ %arrayidx4 = getelementptr inbounds i32, i32* %a, i32 2
+ %i4 = load i32, i32* %arrayidx4, align 4
+ %arrayidx5 = getelementptr inbounds i32, i32* %b, i32 2
+ %i5 = load i32, i32* %arrayidx5, align 4
+ %add3 = add i32 %i4, %i5
+ %fp3 = sitofp i32 %add3 to float
+ %call3 = tail call float @llvm.powi.f32(float %fp3,i32 %add1) nounwind readnone
+
+ %arrayidx6 = getelementptr inbounds i32, i32* %a, i32 3
+ %i6 = load i32, i32* %arrayidx6, align 4
+ %arrayidx7 = getelementptr inbounds i32, i32* %b, i32 3
+ %i7 = load i32, i32* %arrayidx7, align 4
+ %add4 = add i32 %i6, %i7
+ %fp4 = sitofp i32 %add4 to float
+ %call4 = tail call float @llvm.powi.f32(float %fp4,i32 %add1) nounwind readnone
+
+ store float %call1, float* %c, align 4
+ %arrayidx8 = getelementptr inbounds float, float* %c, i32 1
+ store float %call2, float* %arrayidx8, align 4
+ %arrayidx9 = getelementptr inbounds float, float* %c, i32 2
+ store float %call3, float* %arrayidx9, align 4
+ %arrayidx10 = getelementptr inbounds float, float* %c, i32 3
+ store float %call4, float* %arrayidx10, align 4
+ ret void
+
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/extractcost.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/extractcost.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/extractcost.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/extractcost.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,46 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+define i32 @foo(i32* nocapture %A, i32 %n, i32 %m) {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = insertelement <4 x i32> undef, i32 [[N:%.*]], i32 0
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x i32> [[TMP0]], i32 [[N]], i32 1
+; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x i32> [[TMP1]], i32 [[N]], i32 2
+; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x i32> [[TMP2]], i32 [[N]], i32 3
+; CHECK-NEXT: [[TMP4:%.*]] = mul nsw <4 x i32> [[TMP3]], <i32 5, i32 9, i32 3, i32 10>
+; CHECK-NEXT: [[TMP5:%.*]] = shl <4 x i32> [[TMP3]], <i32 5, i32 9, i32 3, i32 10>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <4 x i32> [[TMP4]], <4 x i32> [[TMP5]], <4 x i32> <i32 0, i32 1, i32 6, i32 3>
+; CHECK-NEXT: [[TMP7:%.*]] = add nsw <4 x i32> [[TMP6]], <i32 9, i32 9, i32 9, i32 9>
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast i32* [[A:%.*]] to <4 x i32>*
+; CHECK-NEXT: store <4 x i32> [[TMP7]], <4 x i32>* [[TMP8]], align 4
+; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i32> [[TMP7]], i32 0
+; CHECK-NEXT: [[EXTERNALUSE1:%.*]] = add nsw i32 [[TMP9]], [[M:%.*]]
+; CHECK-NEXT: [[EXTERNALUSE2:%.*]] = mul nsw i32 [[TMP9]], [[M]]
+; CHECK-NEXT: [[ADD10:%.*]] = add nsw i32 [[EXTERNALUSE1]], [[EXTERNALUSE2]]
+; CHECK-NEXT: ret i32 [[ADD10]]
+;
+entry:
+ %mul = mul nsw i32 %n, 5
+ %add = add nsw i32 %mul, 9
+ store i32 %add, i32* %A, align 4
+ %mul1 = mul nsw i32 %n, 9
+ %add2 = add nsw i32 %mul1, 9
+ %arrayidx3 = getelementptr inbounds i32, i32* %A, i64 1
+ store i32 %add2, i32* %arrayidx3, align 4
+ %mul4 = shl i32 %n, 3
+ %add5 = add nsw i32 %mul4, 9
+ %arrayidx6 = getelementptr inbounds i32, i32* %A, i64 2
+ store i32 %add5, i32* %arrayidx6, align 4
+ %mul7 = mul nsw i32 %n, 10
+ %add8 = add nsw i32 %mul7, 9
+ %arrayidx9 = getelementptr inbounds i32, i32* %A, i64 3
+ store i32 %add8, i32* %arrayidx9, align 4
+ %externaluse1 = add nsw i32 %add, %m
+ %externaluse2 = mul nsw i32 %add, %m ; we should add the extract cost only once and the store will be vectorized
+ %add10 = add nsw i32 %externaluse1, %externaluse2
+ ret i32 %add10
+}
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/extractelement.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/extractelement.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/extractelement.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/extractelement.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,61 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -slp-vectorizer -S -mtriple=x86_64-unknown-linux -march=core-avx2 | FileCheck %s
+; RUN: opt < %s -slp-vectorizer -S -mtriple=x86_64-unknown-linux -march=core-avx2 -slp-threshold=-1 -slp-vectorize-hor-store | FileCheck %s --check-prefix=THRESH1
+; RUN: opt < %s -slp-vectorizer -S -mtriple=x86_64-unknown-linux -march=core-avx2 -slp-threshold=-2 -slp-vectorize-hor-store | FileCheck %s --check-prefix=THRESH2
+
+ at a = global float 0.000000e+00, align 4
+
+define float @f(<2 x float> %x) {
+; CHECK-LABEL: @f(
+; CHECK-NEXT: [[TMP1:%.*]] = fmul <2 x float> [[X:%.*]], [[X]]
+; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x float> [[TMP1]], i32 0
+; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x float> [[TMP1]], i32 1
+; CHECK-NEXT: [[ADD:%.*]] = fadd float [[TMP2]], [[TMP3]]
+; CHECK-NEXT: ret float [[ADD]]
+;
+ %x0 = extractelement <2 x float> %x, i32 0
+ %x1 = extractelement <2 x float> %x, i32 1
+ %x0x0 = fmul float %x0, %x0
+ %x1x1 = fmul float %x1, %x1
+ %add = fadd float %x0x0, %x1x1
+ ret float %add
+}
+
+define float @f_used_out_of_tree(<2 x float> %x) {
+; THRESH2-LABEL: @f_used_out_of_tree(
+; THRESH2-NEXT: [[TMP1:%.*]] = extractelement <2 x float> [[X:%.*]], i32 0
+; THRESH2-NEXT: [[TMP2:%.*]] = fmul <2 x float> [[X]], [[X]]
+; THRESH2-NEXT: [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 0
+; THRESH2-NEXT: [[TMP4:%.*]] = extractelement <2 x float> [[TMP2]], i32 1
+; THRESH2-NEXT: [[ADD:%.*]] = fadd float [[TMP3]], [[TMP4]]
+; THRESH2-NEXT: store float [[ADD]], float* @a
+; THRESH2-NEXT: ret float [[TMP1]]
+;
+ %x0 = extractelement <2 x float> %x, i32 0
+ %x1 = extractelement <2 x float> %x, i32 1
+ %x0x0 = fmul float %x0, %x0
+ %x1x1 = fmul float %x1, %x1
+ %add = fadd float %x0x0, %x1x1
+ store float %add, float* @a
+ ret float %x0
+}
+
+define float @f_used_twice_in_tree(<2 x float> %x) {
+; THRESH1-LABEL: @f_used_twice_in_tree(
+; THRESH1-NEXT: [[TMP1:%.*]] = extractelement <2 x float> [[X:%.*]], i32 1
+; THRESH1-NEXT: [[TMP2:%.*]] = insertelement <2 x float> undef, float [[TMP1]], i32 0
+; THRESH1-NEXT: [[TMP3:%.*]] = insertelement <2 x float> [[TMP2]], float [[TMP1]], i32 1
+; THRESH1-NEXT: [[TMP4:%.*]] = fmul <2 x float> [[X]], [[TMP3]]
+; THRESH1-NEXT: [[TMP5:%.*]] = extractelement <2 x float> [[TMP4]], i32 0
+; THRESH1-NEXT: [[TMP6:%.*]] = extractelement <2 x float> [[TMP4]], i32 1
+; THRESH1-NEXT: [[ADD:%.*]] = fadd float [[TMP5]], [[TMP6]]
+; THRESH1-NEXT: ret float [[ADD]]
+;
+ %x0 = extractelement <2 x float> %x, i32 0
+ %x1 = extractelement <2 x float> %x, i32 1
+ %x0x0 = fmul float %x0, %x1
+ %x1x1 = fmul float %x1, %x1
+ %add = fadd float %x0x0, %x1x1
+ ret float %add
+}
+
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/fabs-cost-softfp.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/fabs-cost-softfp.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/fabs-cost-softfp.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/fabs-cost-softfp.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,44 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; Regression test from https://bugs.llvm.org/show_bug.cgi?id=39168
+; Based on code from `compiler-rt/lib/builtins/multc3.c`
+; On plaforms where fp128 lowers to an interger type (soft-fp) we
+; shouldn't be calling isFAbsFree() on the legalized type.
+
+; RUN: opt -slp-vectorizer -slp-threshold=-10 -S %s | FileCheck %s
+
+target triple = "i686-unknown-linux-gnu"
+
+define void @vectorize_fp128(fp128 %c, fp128 %d) #0 {
+; CHECK-LABEL: @vectorize_fp128(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x fp128> undef, fp128 [[C:%.*]], i32 0
+; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x fp128> [[TMP0]], fp128 [[D:%.*]], i32 1
+; CHECK-NEXT: [[TMP2:%.*]] = call <2 x fp128> @llvm.fabs.v2f128(<2 x fp128> [[TMP1]])
+; CHECK-NEXT: [[TMP3:%.*]] = fcmp oeq <2 x fp128> [[TMP2]], <fp128 0xL00000000000000007FFF000000000000, fp128 0xL00000000000000007FFF000000000000>
+; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i32 0
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i1> [[TMP3]], i32 1
+; CHECK-NEXT: [[OR_COND39:%.*]] = or i1 [[TMP4]], [[TMP5]]
+; CHECK-NEXT: br i1 [[OR_COND39]], label [[IF_THEN13:%.*]], label [[IF_END24:%.*]]
+; CHECK: if.then13:
+; CHECK-NEXT: unreachable
+; CHECK: if.end24:
+; CHECK-NEXT: ret void
+;
+entry:
+ %0 = tail call fp128 @llvm.fabs.f128(fp128 %c)
+ %cmpinf10 = fcmp oeq fp128 %0, 0xL00000000000000007FFF000000000000
+ %1 = tail call fp128 @llvm.fabs.f128(fp128 %d)
+ %cmpinf12 = fcmp oeq fp128 %1, 0xL00000000000000007FFF000000000000
+ %or.cond39 = or i1 %cmpinf10, %cmpinf12
+ br i1 %or.cond39, label %if.then13, label %if.end24
+
+if.then13: ; preds = %entry
+ unreachable
+
+if.end24: ; preds = %entry
+ ret void
+}
+
+declare fp128 @llvm.fabs.f128(fp128)
+
+attributes #0 = { "target-features"="+fxsr,+mmx,+sse,+sse2,+x87" }
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/fabs.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/fabs.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/fabs.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/fabs.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,275 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=bdver1 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skylake-avx512 -mattr=-prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX512
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skylake-avx512 -mattr=+prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+ at src64 = common global [8 x double] zeroinitializer, align 64
+ at src32 = common global [16 x float] zeroinitializer, align 64
+ at dst64 = common global [8 x double] zeroinitializer, align 64
+ at dst32 = common global [16 x float] zeroinitializer, align 64
+
+declare float @llvm.fabs.f32(float)
+declare double @llvm.fabs.f64(double)
+
+;
+; FABS
+;
+
+define void @fabs_2f64() #0 {
+; CHECK-LABEL: @fabs_2f64(
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = call <2 x double> @llvm.fabs.v2f64(<2 x double> [[TMP1]])
+; CHECK-NEXT: store <2 x double> [[TMP2]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8
+; CHECK-NEXT: ret void
+;
+ %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
+ %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
+ %fabs0 = call double @llvm.fabs.f64(double %a0)
+ %fabs1 = call double @llvm.fabs.f64(double %a1)
+ store double %fabs0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8
+ store double %fabs1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
+ ret void
+}
+
+define void @fabs_4f64() #0 {
+; SSE-LABEL: @fabs_4f64(
+; SSE-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8
+; SSE-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8
+; SSE-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.fabs.v2f64(<2 x double> [[TMP1]])
+; SSE-NEXT: [[TMP4:%.*]] = call <2 x double> @llvm.fabs.v2f64(<2 x double> [[TMP2]])
+; SSE-NEXT: store <2 x double> [[TMP3]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8
+; SSE-NEXT: store <2 x double> [[TMP4]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @fabs_4f64(
+; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
+; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[TMP1]])
+; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
+; AVX-NEXT: ret void
+;
+ %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
+ %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
+ %a2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
+ %a3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
+ %fabs0 = call double @llvm.fabs.f64(double %a0)
+ %fabs1 = call double @llvm.fabs.f64(double %a1)
+ %fabs2 = call double @llvm.fabs.f64(double %a2)
+ %fabs3 = call double @llvm.fabs.f64(double %a3)
+ store double %fabs0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8
+ store double %fabs1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
+ store double %fabs2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8
+ store double %fabs3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
+ ret void
+}
+
+define void @fabs_8f64() #0 {
+; SSE-LABEL: @fabs_8f64(
+; SSE-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 4
+; SSE-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 4
+; SSE-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <2 x double>*), align 4
+; SSE-NEXT: [[TMP4:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6) to <2 x double>*), align 4
+; SSE-NEXT: [[TMP5:%.*]] = call <2 x double> @llvm.fabs.v2f64(<2 x double> [[TMP1]])
+; SSE-NEXT: [[TMP6:%.*]] = call <2 x double> @llvm.fabs.v2f64(<2 x double> [[TMP2]])
+; SSE-NEXT: [[TMP7:%.*]] = call <2 x double> @llvm.fabs.v2f64(<2 x double> [[TMP3]])
+; SSE-NEXT: [[TMP8:%.*]] = call <2 x double> @llvm.fabs.v2f64(<2 x double> [[TMP4]])
+; SSE-NEXT: store <2 x double> [[TMP5]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 4
+; SSE-NEXT: store <2 x double> [[TMP6]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 4
+; SSE-NEXT: store <2 x double> [[TMP7]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <2 x double>*), align 4
+; SSE-NEXT: store <2 x double> [[TMP8]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6) to <2 x double>*), align 4
+; SSE-NEXT: ret void
+;
+; AVX256-LABEL: @fabs_8f64(
+; AVX256-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 4
+; AVX256-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 4
+; AVX256-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[TMP1]])
+; AVX256-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[TMP2]])
+; AVX256-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 4
+; AVX256-NEXT: store <4 x double> [[TMP4]], <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <4 x double>*), align 4
+; AVX256-NEXT: ret void
+;
+; AVX512-LABEL: @fabs_8f64(
+; AVX512-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 4
+; AVX512-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.fabs.v8f64(<8 x double> [[TMP1]])
+; AVX512-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 4
+; AVX512-NEXT: ret void
+;
+ %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 4
+ %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 4
+ %a2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 4
+ %a3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 4
+ %a4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 4
+ %a5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 4
+ %a6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 4
+ %a7 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 7), align 4
+ %fabs0 = call double @llvm.fabs.f64(double %a0)
+ %fabs1 = call double @llvm.fabs.f64(double %a1)
+ %fabs2 = call double @llvm.fabs.f64(double %a2)
+ %fabs3 = call double @llvm.fabs.f64(double %a3)
+ %fabs4 = call double @llvm.fabs.f64(double %a4)
+ %fabs5 = call double @llvm.fabs.f64(double %a5)
+ %fabs6 = call double @llvm.fabs.f64(double %a6)
+ %fabs7 = call double @llvm.fabs.f64(double %a7)
+ store double %fabs0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 4
+ store double %fabs1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 4
+ store double %fabs2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 4
+ store double %fabs3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 4
+ store double %fabs4, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 4
+ store double %fabs5, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 4
+ store double %fabs6, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 4
+ store double %fabs7, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 4
+ ret void
+}
+
+define void @fabs_4f32() #0 {
+; CHECK-LABEL: @fabs_4f32(
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4
+; CHECK-NEXT: [[TMP2:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP1]])
+; CHECK-NEXT: store <4 x float> [[TMP2]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
+; CHECK-NEXT: ret void
+;
+ %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
+ %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
+ %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
+ %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
+ %fabs0 = call float @llvm.fabs.f32(float %a0)
+ %fabs1 = call float @llvm.fabs.f32(float %a1)
+ %fabs2 = call float @llvm.fabs.f32(float %a2)
+ %fabs3 = call float @llvm.fabs.f32(float %a3)
+ store float %fabs0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 4
+ store float %fabs1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+ store float %fabs2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 4
+ store float %fabs3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
+ ret void
+}
+
+define void @fabs_8f32() #0 {
+; SSE-LABEL: @fabs_8f32(
+; SSE-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4
+; SSE-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP1]])
+; SSE-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP2]])
+; SSE-NEXT: store <4 x float> [[TMP3]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
+; SSE-NEXT: store <4 x float> [[TMP4]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @fabs_8f32(
+; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
+; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.fabs.v8f32(<8 x float> [[TMP1]])
+; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
+; AVX-NEXT: ret void
+;
+ %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
+ %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
+ %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
+ %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
+ %a4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4), align 4
+ %a5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5), align 4
+ %a6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6), align 4
+ %a7 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 7), align 4
+ %fabs0 = call float @llvm.fabs.f32(float %a0)
+ %fabs1 = call float @llvm.fabs.f32(float %a1)
+ %fabs2 = call float @llvm.fabs.f32(float %a2)
+ %fabs3 = call float @llvm.fabs.f32(float %a3)
+ %fabs4 = call float @llvm.fabs.f32(float %a4)
+ %fabs5 = call float @llvm.fabs.f32(float %a5)
+ %fabs6 = call float @llvm.fabs.f32(float %a6)
+ %fabs7 = call float @llvm.fabs.f32(float %a7)
+ store float %fabs0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 4
+ store float %fabs1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+ store float %fabs2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 4
+ store float %fabs3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
+ store float %fabs4, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4
+ store float %fabs5, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
+ store float %fabs6, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4
+ store float %fabs7, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
+ ret void
+}
+
+define void @fabs_16f32() #0 {
+; SSE-LABEL: @fabs_16f32(
+; SSE-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4
+; SSE-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP3:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP4:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 12) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP1]])
+; SSE-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP2]])
+; SSE-NEXT: [[TMP7:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP3]])
+; SSE-NEXT: [[TMP8:%.*]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> [[TMP4]])
+; SSE-NEXT: store <4 x float> [[TMP5]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
+; SSE-NEXT: store <4 x float> [[TMP6]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4
+; SSE-NEXT: store <4 x float> [[TMP7]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <4 x float>*), align 4
+; SSE-NEXT: store <4 x float> [[TMP8]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12) to <4 x float>*), align 4
+; SSE-NEXT: ret void
+;
+; AVX256-LABEL: @fabs_16f32(
+; AVX256-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
+; AVX256-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
+; AVX256-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.fabs.v8f32(<8 x float> [[TMP1]])
+; AVX256-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.fabs.v8f32(<8 x float> [[TMP2]])
+; AVX256-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
+; AVX256-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
+; AVX256-NEXT: ret void
+;
+; AVX512-LABEL: @fabs_16f32(
+; AVX512-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
+; AVX512-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.fabs.v16f32(<16 x float> [[TMP1]])
+; AVX512-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
+; AVX512-NEXT: ret void
+;
+ %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
+ %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
+ %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
+ %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
+ %a4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4), align 4
+ %a5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5), align 4
+ %a6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6), align 4
+ %a7 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 7), align 4
+ %a8 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8), align 4
+ %a9 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 9), align 4
+ %a10 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 10), align 4
+ %a11 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 11), align 4
+ %a12 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 12), align 4
+ %a13 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 13), align 4
+ %a14 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 14), align 4
+ %a15 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 15), align 4
+ %fabs0 = call float @llvm.fabs.f32(float %a0 )
+ %fabs1 = call float @llvm.fabs.f32(float %a1 )
+ %fabs2 = call float @llvm.fabs.f32(float %a2 )
+ %fabs3 = call float @llvm.fabs.f32(float %a3 )
+ %fabs4 = call float @llvm.fabs.f32(float %a4 )
+ %fabs5 = call float @llvm.fabs.f32(float %a5 )
+ %fabs6 = call float @llvm.fabs.f32(float %a6 )
+ %fabs7 = call float @llvm.fabs.f32(float %a7 )
+ %fabs8 = call float @llvm.fabs.f32(float %a8 )
+ %fabs9 = call float @llvm.fabs.f32(float %a9 )
+ %fabs10 = call float @llvm.fabs.f32(float %a10)
+ %fabs11 = call float @llvm.fabs.f32(float %a11)
+ %fabs12 = call float @llvm.fabs.f32(float %a12)
+ %fabs13 = call float @llvm.fabs.f32(float %a13)
+ %fabs14 = call float @llvm.fabs.f32(float %a14)
+ %fabs15 = call float @llvm.fabs.f32(float %a15)
+ store float %fabs0 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 4
+ store float %fabs1 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+ store float %fabs2 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 4
+ store float %fabs3 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
+ store float %fabs4 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4
+ store float %fabs5 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
+ store float %fabs6 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4
+ store float %fabs7 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
+ store float %fabs8 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8), align 4
+ store float %fabs9 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 9), align 4
+ store float %fabs10, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 10), align 4
+ store float %fabs11, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4
+ store float %fabs12, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4
+ store float %fabs13, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4
+ store float %fabs14, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4
+ store float %fabs15, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4
+ ret void
+}
+
+attributes #0 = { nounwind }
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/fcopysign.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/fcopysign.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/fcopysign.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/fcopysign.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,343 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=bdver1 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skylake-avx512 -mattr=-prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX512
+; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skylake-avx512 -mattr=+prefer-256-bit -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+ at srcA64 = common global [8 x double] zeroinitializer, align 64
+ at srcB64 = common global [8 x double] zeroinitializer, align 64
+ at srcC64 = common global [8 x double] zeroinitializer, align 64
+ at srcA32 = common global [16 x float] zeroinitializer, align 64
+ at srcB32 = common global [16 x float] zeroinitializer, align 64
+ at srcC32 = common global [16 x float] zeroinitializer, align 64
+ at dst64 = common global [8 x double] zeroinitializer, align 64
+ at dst32 = common global [16 x float] zeroinitializer, align 64
+
+declare float @llvm.copysign.f32(float, float)
+declare double @llvm.copysign.f64(double, double)
+
+;
+; CHECK
+;
+
+define void @fcopysign_2f64() #0 {
+; CHECK-LABEL: @fcopysign_2f64(
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @srcA64 to <2 x double>*), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @srcB64 to <2 x double>*), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.copysign.v2f64(<2 x double> [[TMP1]], <2 x double> [[TMP2]])
+; CHECK-NEXT: store <2 x double> [[TMP3]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8
+; CHECK-NEXT: ret void
+;
+ %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 0), align 8
+ %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 1), align 8
+ %b0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 0), align 8
+ %b1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 1), align 8
+ %fcopysign0 = call double @llvm.copysign.f64(double %a0, double %b0)
+ %fcopysign1 = call double @llvm.copysign.f64(double %a1, double %b1)
+ store double %fcopysign0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8
+ store double %fcopysign1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
+ ret void
+}
+
+define void @fcopysign_4f64() #0 {
+; SSE-LABEL: @fcopysign_4f64(
+; SSE-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @srcA64 to <2 x double>*), align 8
+; SSE-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 2) to <2 x double>*), align 8
+; SSE-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @srcB64 to <2 x double>*), align 8
+; SSE-NEXT: [[TMP4:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 2) to <2 x double>*), align 8
+; SSE-NEXT: [[TMP5:%.*]] = call <2 x double> @llvm.copysign.v2f64(<2 x double> [[TMP1]], <2 x double> [[TMP3]])
+; SSE-NEXT: [[TMP6:%.*]] = call <2 x double> @llvm.copysign.v2f64(<2 x double> [[TMP2]], <2 x double> [[TMP4]])
+; SSE-NEXT: store <2 x double> [[TMP5]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8
+; SSE-NEXT: store <2 x double> [[TMP6]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @fcopysign_4f64(
+; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @srcA64 to <4 x double>*), align 8
+; AVX-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @srcB64 to <4 x double>*), align 8
+; AVX-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.copysign.v4f64(<4 x double> [[TMP1]], <4 x double> [[TMP2]])
+; AVX-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
+; AVX-NEXT: ret void
+;
+ %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 0), align 8
+ %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 1), align 8
+ %a2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 2), align 8
+ %a3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 3), align 8
+ %b0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 0), align 8
+ %b1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 1), align 8
+ %b2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 2), align 8
+ %b3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 3), align 8
+ %fcopysign0 = call double @llvm.copysign.f64(double %a0, double %b0)
+ %fcopysign1 = call double @llvm.copysign.f64(double %a1, double %b1)
+ %fcopysign2 = call double @llvm.copysign.f64(double %a2, double %b2)
+ %fcopysign3 = call double @llvm.copysign.f64(double %a3, double %b3)
+ store double %fcopysign0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8
+ store double %fcopysign1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
+ store double %fcopysign2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8
+ store double %fcopysign3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
+ ret void
+}
+
+define void @fcopysign_8f64() #0 {
+; SSE-LABEL: @fcopysign_8f64(
+; SSE-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @srcA64 to <2 x double>*), align 4
+; SSE-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 2) to <2 x double>*), align 4
+; SSE-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 4) to <2 x double>*), align 4
+; SSE-NEXT: [[TMP4:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 6) to <2 x double>*), align 4
+; SSE-NEXT: [[TMP5:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @srcB64 to <2 x double>*), align 4
+; SSE-NEXT: [[TMP6:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 2) to <2 x double>*), align 4
+; SSE-NEXT: [[TMP7:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 4) to <2 x double>*), align 4
+; SSE-NEXT: [[TMP8:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 6) to <2 x double>*), align 4
+; SSE-NEXT: [[TMP9:%.*]] = call <2 x double> @llvm.copysign.v2f64(<2 x double> [[TMP1]], <2 x double> [[TMP5]])
+; SSE-NEXT: [[TMP10:%.*]] = call <2 x double> @llvm.copysign.v2f64(<2 x double> [[TMP2]], <2 x double> [[TMP6]])
+; SSE-NEXT: [[TMP11:%.*]] = call <2 x double> @llvm.copysign.v2f64(<2 x double> [[TMP3]], <2 x double> [[TMP7]])
+; SSE-NEXT: [[TMP12:%.*]] = call <2 x double> @llvm.copysign.v2f64(<2 x double> [[TMP4]], <2 x double> [[TMP8]])
+; SSE-NEXT: store <2 x double> [[TMP9]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 4
+; SSE-NEXT: store <2 x double> [[TMP10]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 4
+; SSE-NEXT: store <2 x double> [[TMP11]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <2 x double>*), align 4
+; SSE-NEXT: store <2 x double> [[TMP12]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6) to <2 x double>*), align 4
+; SSE-NEXT: ret void
+;
+; AVX256-LABEL: @fcopysign_8f64(
+; AVX256-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @srcA64 to <4 x double>*), align 4
+; AVX256-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 4) to <4 x double>*), align 4
+; AVX256-NEXT: [[TMP3:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @srcB64 to <4 x double>*), align 4
+; AVX256-NEXT: [[TMP4:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 4) to <4 x double>*), align 4
+; AVX256-NEXT: [[TMP5:%.*]] = call <4 x double> @llvm.copysign.v4f64(<4 x double> [[TMP1]], <4 x double> [[TMP3]])
+; AVX256-NEXT: [[TMP6:%.*]] = call <4 x double> @llvm.copysign.v4f64(<4 x double> [[TMP2]], <4 x double> [[TMP4]])
+; AVX256-NEXT: store <4 x double> [[TMP5]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 4
+; AVX256-NEXT: store <4 x double> [[TMP6]], <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <4 x double>*), align 4
+; AVX256-NEXT: ret void
+;
+; AVX512-LABEL: @fcopysign_8f64(
+; AVX512-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @srcA64 to <8 x double>*), align 4
+; AVX512-NEXT: [[TMP2:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @srcB64 to <8 x double>*), align 4
+; AVX512-NEXT: [[TMP3:%.*]] = call <8 x double> @llvm.copysign.v8f64(<8 x double> [[TMP1]], <8 x double> [[TMP2]])
+; AVX512-NEXT: store <8 x double> [[TMP3]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 4
+; AVX512-NEXT: ret void
+;
+ %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 0), align 4
+ %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 1), align 4
+ %a2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 2), align 4
+ %a3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 3), align 4
+ %a4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 4), align 4
+ %a5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 5), align 4
+ %a6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 6), align 4
+ %a7 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcA64, i32 0, i64 7), align 4
+ %b0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 0), align 4
+ %b1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 1), align 4
+ %b2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 2), align 4
+ %b3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 3), align 4
+ %b4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 4), align 4
+ %b5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 5), align 4
+ %b6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 6), align 4
+ %b7 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @srcB64, i32 0, i64 7), align 4
+ %fcopysign0 = call double @llvm.copysign.f64(double %a0, double %b0)
+ %fcopysign1 = call double @llvm.copysign.f64(double %a1, double %b1)
+ %fcopysign2 = call double @llvm.copysign.f64(double %a2, double %b2)
+ %fcopysign3 = call double @llvm.copysign.f64(double %a3, double %b3)
+ %fcopysign4 = call double @llvm.copysign.f64(double %a4, double %b4)
+ %fcopysign5 = call double @llvm.copysign.f64(double %a5, double %b5)
+ %fcopysign6 = call double @llvm.copysign.f64(double %a6, double %b6)
+ %fcopysign7 = call double @llvm.copysign.f64(double %a7, double %b7)
+ store double %fcopysign0, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 4
+ store double %fcopysign1, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 4
+ store double %fcopysign2, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 4
+ store double %fcopysign3, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 4
+ store double %fcopysign4, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 4
+ store double %fcopysign5, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 4
+ store double %fcopysign6, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 4
+ store double %fcopysign7, double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 4
+ ret void
+}
+
+define void @fcopysign_4f32() #0 {
+; CHECK-LABEL: @fcopysign_4f32(
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @srcA32 to <4 x float>*), align 4
+; CHECK-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @srcB32 to <4 x float>*), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.copysign.v4f32(<4 x float> [[TMP1]], <4 x float> [[TMP2]])
+; CHECK-NEXT: store <4 x float> [[TMP3]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
+; CHECK-NEXT: ret void
+;
+ %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 0), align 4
+ %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 1), align 4
+ %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 2), align 4
+ %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 3), align 4
+ %b0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 0), align 4
+ %b1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 1), align 4
+ %b2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 2), align 4
+ %b3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 3), align 4
+ %fcopysign0 = call float @llvm.copysign.f32(float %a0, float %b0)
+ %fcopysign1 = call float @llvm.copysign.f32(float %a1, float %b1)
+ %fcopysign2 = call float @llvm.copysign.f32(float %a2, float %b2)
+ %fcopysign3 = call float @llvm.copysign.f32(float %a3, float %b3)
+ store float %fcopysign0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 4
+ store float %fcopysign1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+ store float %fcopysign2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 4
+ store float %fcopysign3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
+ ret void
+}
+
+define void @fcopysign_8f32() #0 {
+; SSE-LABEL: @fcopysign_8f32(
+; SSE-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @srcA32 to <4 x float>*), align 4
+; SSE-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 4) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP3:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @srcB32 to <4 x float>*), align 4
+; SSE-NEXT: [[TMP4:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 4) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.copysign.v4f32(<4 x float> [[TMP1]], <4 x float> [[TMP3]])
+; SSE-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.copysign.v4f32(<4 x float> [[TMP2]], <4 x float> [[TMP4]])
+; SSE-NEXT: store <4 x float> [[TMP5]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
+; SSE-NEXT: store <4 x float> [[TMP6]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4
+; SSE-NEXT: ret void
+;
+; AVX-LABEL: @fcopysign_8f32(
+; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @srcA32 to <8 x float>*), align 4
+; AVX-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @srcB32 to <8 x float>*), align 4
+; AVX-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.copysign.v8f32(<8 x float> [[TMP1]], <8 x float> [[TMP2]])
+; AVX-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
+; AVX-NEXT: ret void
+;
+ %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 0), align 4
+ %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 1), align 4
+ %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 2), align 4
+ %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 3), align 4
+ %a4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 4), align 4
+ %a5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 5), align 4
+ %a6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 6), align 4
+ %a7 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 7), align 4
+ %b0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 0), align 4
+ %b1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 1), align 4
+ %b2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 2), align 4
+ %b3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 3), align 4
+ %b4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 4), align 4
+ %b5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 5), align 4
+ %b6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 6), align 4
+ %b7 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 7), align 4
+ %fcopysign0 = call float @llvm.copysign.f32(float %a0, float %b0)
+ %fcopysign1 = call float @llvm.copysign.f32(float %a1, float %b1)
+ %fcopysign2 = call float @llvm.copysign.f32(float %a2, float %b2)
+ %fcopysign3 = call float @llvm.copysign.f32(float %a3, float %b3)
+ %fcopysign4 = call float @llvm.copysign.f32(float %a4, float %b4)
+ %fcopysign5 = call float @llvm.copysign.f32(float %a5, float %b5)
+ %fcopysign6 = call float @llvm.copysign.f32(float %a6, float %b6)
+ %fcopysign7 = call float @llvm.copysign.f32(float %a7, float %b7)
+ store float %fcopysign0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 4
+ store float %fcopysign1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+ store float %fcopysign2, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 4
+ store float %fcopysign3, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
+ store float %fcopysign4, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4
+ store float %fcopysign5, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
+ store float %fcopysign6, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4
+ store float %fcopysign7, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
+ ret void
+}
+
+define void @fcopysign_16f32() #0 {
+; SSE-LABEL: @fcopysign_16f32(
+; SSE-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @srcA32 to <4 x float>*), align 4
+; SSE-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 4) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP3:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 8) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP4:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 12) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP5:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @srcB32 to <4 x float>*), align 4
+; SSE-NEXT: [[TMP6:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 4) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP7:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 8) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP8:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 12) to <4 x float>*), align 4
+; SSE-NEXT: [[TMP9:%.*]] = call <4 x float> @llvm.copysign.v4f32(<4 x float> [[TMP1]], <4 x float> [[TMP5]])
+; SSE-NEXT: [[TMP10:%.*]] = call <4 x float> @llvm.copysign.v4f32(<4 x float> [[TMP2]], <4 x float> [[TMP6]])
+; SSE-NEXT: [[TMP11:%.*]] = call <4 x float> @llvm.copysign.v4f32(<4 x float> [[TMP3]], <4 x float> [[TMP7]])
+; SSE-NEXT: [[TMP12:%.*]] = call <4 x float> @llvm.copysign.v4f32(<4 x float> [[TMP4]], <4 x float> [[TMP8]])
+; SSE-NEXT: store <4 x float> [[TMP9]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
+; SSE-NEXT: store <4 x float> [[TMP10]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4
+; SSE-NEXT: store <4 x float> [[TMP11]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <4 x float>*), align 4
+; SSE-NEXT: store <4 x float> [[TMP12]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12) to <4 x float>*), align 4
+; SSE-NEXT: ret void
+;
+; AVX256-LABEL: @fcopysign_16f32(
+; AVX256-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @srcA32 to <8 x float>*), align 4
+; AVX256-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 8) to <8 x float>*), align 4
+; AVX256-NEXT: [[TMP3:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @srcB32 to <8 x float>*), align 4
+; AVX256-NEXT: [[TMP4:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 8) to <8 x float>*), align 4
+; AVX256-NEXT: [[TMP5:%.*]] = call <8 x float> @llvm.copysign.v8f32(<8 x float> [[TMP1]], <8 x float> [[TMP3]])
+; AVX256-NEXT: [[TMP6:%.*]] = call <8 x float> @llvm.copysign.v8f32(<8 x float> [[TMP2]], <8 x float> [[TMP4]])
+; AVX256-NEXT: store <8 x float> [[TMP5]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
+; AVX256-NEXT: store <8 x float> [[TMP6]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
+; AVX256-NEXT: ret void
+;
+; AVX512-LABEL: @fcopysign_16f32(
+; AVX512-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @srcA32 to <16 x float>*), align 4
+; AVX512-NEXT: [[TMP2:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @srcB32 to <16 x float>*), align 4
+; AVX512-NEXT: [[TMP3:%.*]] = call <16 x float> @llvm.copysign.v16f32(<16 x float> [[TMP1]], <16 x float> [[TMP2]])
+; AVX512-NEXT: store <16 x float> [[TMP3]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
+; AVX512-NEXT: ret void
+;
+ %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 0), align 4
+ %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 1), align 4
+ %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 2), align 4
+ %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 3), align 4
+ %a4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 4), align 4
+ %a5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 5), align 4
+ %a6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 6), align 4
+ %a7 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 7), align 4
+ %a8 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 8), align 4
+ %a9 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 9), align 4
+ %a10 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 10), align 4
+ %a11 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 11), align 4
+ %a12 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 12), align 4
+ %a13 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 13), align 4
+ %a14 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 14), align 4
+ %a15 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcA32, i32 0, i64 15), align 4
+ %b0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 0), align 4
+ %b1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 1), align 4
+ %b2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 2), align 4
+ %b3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 3), align 4
+ %b4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 4), align 4
+ %b5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 5), align 4
+ %b6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 6), align 4
+ %b7 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 7), align 4
+ %b8 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 8), align 4
+ %b9 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 9), align 4
+ %b10 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 10), align 4
+ %b11 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 11), align 4
+ %b12 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 12), align 4
+ %b13 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 13), align 4
+ %b14 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 14), align 4
+ %b15 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @srcB32, i32 0, i64 15), align 4
+ %fcopysign0 = call float @llvm.copysign.f32(float %a0 , float %b0 )
+ %fcopysign1 = call float @llvm.copysign.f32(float %a1 , float %b1 )
+ %fcopysign2 = call float @llvm.copysign.f32(float %a2 , float %b2 )
+ %fcopysign3 = call float @llvm.copysign.f32(float %a3 , float %b3 )
+ %fcopysign4 = call float @llvm.copysign.f32(float %a4 , float %b4 )
+ %fcopysign5 = call float @llvm.copysign.f32(float %a5 , float %b5 )
+ %fcopysign6 = call float @llvm.copysign.f32(float %a6 , float %b6 )
+ %fcopysign7 = call float @llvm.copysign.f32(float %a7 , float %b7 )
+ %fcopysign8 = call float @llvm.copysign.f32(float %a8 , float %b8 )
+ %fcopysign9 = call float @llvm.copysign.f32(float %a9 , float %b9 )
+ %fcopysign10 = call float @llvm.copysign.f32(float %a10, float %b10)
+ %fcopysign11 = call float @llvm.copysign.f32(float %a11, float %b11)
+ %fcopysign12 = call float @llvm.copysign.f32(float %a12, float %b12)
+ %fcopysign13 = call float @llvm.copysign.f32(float %a13, float %b13)
+ %fcopysign14 = call float @llvm.copysign.f32(float %a14, float %b14)
+ %fcopysign15 = call float @llvm.copysign.f32(float %a15, float %b15)
+ store float %fcopysign0 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 4
+ store float %fcopysign1 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+ store float %fcopysign2 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 2), align 4
+ store float %fcopysign3 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
+ store float %fcopysign4 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4
+ store float %fcopysign5 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
+ store float %fcopysign6 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4
+ store float %fcopysign7 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
+ store float %fcopysign8 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8), align 4
+ store float %fcopysign9 , float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 9), align 4
+ store float %fcopysign10, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 10), align 4
+ store float %fcopysign11, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4
+ store float %fcopysign12, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4
+ store float %fcopysign13, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4
+ store float %fcopysign14, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4
+ store float %fcopysign15, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4
+ ret void
+}
+
+attributes #0 = { nounwind }
Added: llvm/trunk/test/Transforms/SLPVectorizer/X86/flag.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/SLPVectorizer/X86/flag.ll?rev=358552&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/SLPVectorizer/X86/flag.ll (added)
+++ llvm/trunk/test/Transforms/SLPVectorizer/X86/flag.ll Tue Apr 16 21:52:47 2019
@@ -0,0 +1,87 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -basicaa -slp-vectorizer -slp-threshold=1000 -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+; Check that the command line flag works.
+define i32 @rollable(i32* noalias nocapture %in, i32* noalias nocapture %out, i64 %n) {
+; CHECK-LABEL: @rollable(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i64 [[N:%.*]], 0
+; CHECK-NEXT: br i1 [[TMP1]], label [[DOT_CRIT_EDGE:%.*]], label [[DOTLR_PH:%.*]]
+; CHECK: .lr.ph:
+; CHECK-NEXT: [[I_019:%.*]] = phi i64 [ [[TMP26:%.*]], [[DOTLR_PH]] ], [ 0, [[TMP0:%.*]] ]
+; CHECK-NEXT: [[TMP2:%.*]] = shl i64 [[I_019]], 2
+; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i64 [[TMP2]]
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
+; CHECK-NEXT: [[TMP5:%.*]] = or i64 [[TMP2]], 1
+; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[IN]], i64 [[TMP5]]
+; CHECK-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
+; CHECK-NEXT: [[TMP8:%.*]] = or i64 [[TMP2]], 2
+; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[IN]], i64 [[TMP8]]
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
+; CHECK-NEXT: [[TMP11:%.*]] = or i64 [[TMP2]], 3
+; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[IN]], i64 [[TMP11]]
+; CHECK-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
+; CHECK-NEXT: [[TMP14:%.*]] = mul i32 [[TMP4]], 7
+; CHECK-NEXT: [[TMP15:%.*]] = add i32 [[TMP14]], 7
+; CHECK-NEXT: [[TMP16:%.*]] = mul i32 [[TMP7]], 7
+; CHECK-NEXT: [[TMP17:%.*]] = add i32 [[TMP16]], 14
+; CHECK-NEXT: [[TMP18:%.*]] = mul i32 [[TMP10]], 7
+; CHECK-NEXT: [[TMP19:%.*]] = add i32 [[TMP18]], 21
+; CHECK-NEXT: [[TMP20:%.*]] = mul i32 [[TMP13]], 7
+; CHECK-NEXT: [[TMP21:%.*]] = add i32 [[TMP20]], 28
+; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i64 [[TMP2]]
+; CHECK-NEXT: store i32 [[TMP15]], i32* [[TMP22]], align 4
+; CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[OUT]], i64 [[TMP5]]
+; CHECK-NEXT: store i32 [[TMP17]], i32* [[TMP23]], align 4
+; CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[OUT]], i64 [[TMP8]]
+; CHECK-NEXT: store i32 [[TMP19]], i32* [[TMP24]], align 4
+; CHECK-NEXT: [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[OUT]], i64 [[TMP11]]
+; CHECK-NEXT: store i32 [[TMP21]], i32* [[TMP25]], align 4
+; CHECK-NEXT: [[TMP26]] = add i64 [[I_019]], 1
+; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[TMP26]], [[N]]
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[DOT_CRIT_EDGE]], label [[DOTLR_PH]]
+; CHECK: ._crit_edge:
+; CHECK-NEXT: ret i32 undef
+;
+ %1 = icmp eq i64 %n, 0
+ br i1 %1, label %._crit_edge, label %.lr.ph
+
+.lr.ph: ; preds = %0, %.lr.ph
+ %i.019 = phi i64 [ %26, %.lr.ph ], [ 0, %0 ]
+ %2 = shl i64 %i.019, 2
+ %3 = getelementptr inbounds i32, i32* %in, i64 %2
+ %4 = load i32, i32* %3, align 4
+ %5 = or i64 %2, 1
+ %6 = getelementptr inbounds i32, i32* %in, i64 %5
+ %7 = load i32, i32* %6, align 4
+ %8 = or i64 %2, 2
+ %9 = getelementptr inbounds i32, i32* %in, i64 %8
+ %10 = load i32, i32* %9, align 4
+ %11 = or i64 %2, 3
+ %12 = getelementptr inbounds i32, i32* %in, i64 %11
+ %13 = load i32, i32* %12, align 4
+ %14 = mul i32 %4, 7
+ %15 = add i32 %14, 7
+ %16 = mul i32 %7, 7
+ %17 = add i32 %16, 14
+ %18 = mul i32 %10, 7
+ %19 = add i32 %18, 21
+ %20 = mul i32 %13, 7
+ %21 = add i32 %20, 28
+ %22 = getelementptr inbounds i32, i32* %out, i64 %2
+ store i32 %15, i32* %22, align 4
+ %23 = getelementptr inbounds i32, i32* %out, i64 %5
+ store i32 %17, i32* %23, align 4
+ %24 = getelementptr inbounds i32, i32* %out, i64 %8
+ store i32 %19, i32* %24, align 4
+ %25 = getelementptr inbounds i32, i32* %out, i64 %11
+ store i32 %21, i32* %25, align 4
+ %26 = add i64 %i.019, 1
+ %exitcond = icmp eq i64 %26, %n
+ br i1 %exitcond, label %._crit_edge, label %.lr.ph
+
+._crit_edge: ; preds = %.lr.ph, %0
+ ret i32 undef
+}
More information about the llvm-commits
mailing list