Google Git
Sign in
llvm / llvm-project / refs/heads/main / . / llvm / test / Transforms / SLPVectorizer / X86 / pr47629.ll
blob: cf380f04a6939877805fab713c0eabd6205ac5b7 [file] [log] [blame]
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
; RUN: opt < %s -passes=slp-vectorizer -S -mtriple=x86_64-unknown-linux -mattr=+sse2 | FileCheck %s --check-prefixes=SSE
; RUN: opt < %s -passes=slp-vectorizer -S -mtriple=x86_64-unknown-linux -mattr=+avx | FileCheck %s --check-prefixes=AVX
; RUN: opt < %s -passes=slp-vectorizer -S -mtriple=x86_64-unknown-linux -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2
; RUN: opt < %s -passes=slp-vectorizer -S -mtriple=x86_64-unknown-linux -mattr=+avx512f | FileCheck %s --check-prefixes=AVX512F
; RUN: opt < %s -passes=slp-vectorizer -S -mtriple=x86_64-unknown-linux -mattr=+avx512vl | FileCheck %s --check-prefixes=AVX512VL
define void @gather_load(ptr noalias nocapture %0, ptr noalias nocapture readonly %1) {
; SSE-LABEL: define void @gather_load(
; SSE-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0:[0-9]+]] {
; SSE-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 1
; SSE-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP1]], align 4, !tbaa [[SHORT_TBAA0:![0-9]+]]
; SSE-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 1
; SSE-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 11
; SSE-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 2
; SSE-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 4
; SSE-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 3
; SSE-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP3]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP16:%.*]] = insertelement <4 x i32> undef, i32 [[TMP4]], i32 0
; SSE-NEXT: [[TMP17:%.*]] = insertelement <4 x i32> [[TMP16]], i32 [[TMP7]], i32 1
; SSE-NEXT: [[TMP15:%.*]] = insertelement <4 x i32> [[TMP17]], i32 [[TMP10]], i32 2
; SSE-NEXT: [[TMP13:%.*]] = insertelement <4 x i32> [[TMP15]], i32 [[TMP12]], i32 3
; SSE-NEXT: [[TMP14:%.*]] = add nsw <4 x i32> [[TMP13]], <i32 1, i32 2, i32 3, i32 4>
; SSE-NEXT: store <4 x i32> [[TMP14]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: ret void
;
; AVX-LABEL: define void @gather_load(
; AVX-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0:[0-9]+]] {
; AVX-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 1
; AVX-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP1]], align 4, !tbaa [[SHORT_TBAA0:![0-9]+]]
; AVX-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 1
; AVX-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 11
; AVX-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 2
; AVX-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 4
; AVX-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 3
; AVX-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP3]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP16:%.*]] = insertelement <4 x i32> undef, i32 [[TMP4]], i32 0
; AVX-NEXT: [[TMP17:%.*]] = insertelement <4 x i32> [[TMP16]], i32 [[TMP7]], i32 1
; AVX-NEXT: [[TMP15:%.*]] = insertelement <4 x i32> [[TMP17]], i32 [[TMP10]], i32 2
; AVX-NEXT: [[TMP13:%.*]] = insertelement <4 x i32> [[TMP15]], i32 [[TMP12]], i32 3
; AVX-NEXT: [[TMP14:%.*]] = add nsw <4 x i32> [[TMP13]], <i32 1, i32 2, i32 3, i32 4>
; AVX-NEXT: store <4 x i32> [[TMP14]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: ret void
;
; AVX2-LABEL: define void @gather_load(
; AVX2-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0:[0-9]+]] {
; AVX2-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 1
; AVX2-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP1]], align 4, !tbaa [[SHORT_TBAA0:![0-9]+]]
; AVX2-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 1
; AVX2-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 11
; AVX2-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 2
; AVX2-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 4
; AVX2-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 3
; AVX2-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP3]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP16:%.*]] = insertelement <4 x i32> undef, i32 [[TMP4]], i32 0
; AVX2-NEXT: [[TMP17:%.*]] = insertelement <4 x i32> [[TMP16]], i32 [[TMP7]], i32 1
; AVX2-NEXT: [[TMP15:%.*]] = insertelement <4 x i32> [[TMP17]], i32 [[TMP10]], i32 2
; AVX2-NEXT: [[TMP13:%.*]] = insertelement <4 x i32> [[TMP15]], i32 [[TMP12]], i32 3
; AVX2-NEXT: [[TMP14:%.*]] = add nsw <4 x i32> [[TMP13]], <i32 1, i32 2, i32 3, i32 4>
; AVX2-NEXT: store <4 x i32> [[TMP14]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: ret void
;
; AVX512F-LABEL: define void @gather_load(
; AVX512F-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0:[0-9]+]] {
; AVX512F-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 1
; AVX512F-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP1]], align 4, !tbaa [[SHORT_TBAA0:![0-9]+]]
; AVX512F-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 1
; AVX512F-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 11
; AVX512F-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 2
; AVX512F-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 4
; AVX512F-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 3
; AVX512F-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP3]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: [[TMP16:%.*]] = insertelement <4 x i32> undef, i32 [[TMP4]], i32 0
; AVX512F-NEXT: [[TMP17:%.*]] = insertelement <4 x i32> [[TMP16]], i32 [[TMP7]], i32 1
; AVX512F-NEXT: [[TMP15:%.*]] = insertelement <4 x i32> [[TMP17]], i32 [[TMP10]], i32 2
; AVX512F-NEXT: [[TMP13:%.*]] = insertelement <4 x i32> [[TMP15]], i32 [[TMP12]], i32 3
; AVX512F-NEXT: [[TMP14:%.*]] = add nsw <4 x i32> [[TMP13]], <i32 1, i32 2, i32 3, i32 4>
; AVX512F-NEXT: store <4 x i32> [[TMP14]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: ret void
;
; AVX512VL-LABEL: define void @gather_load(
; AVX512VL-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0:[0-9]+]] {
; AVX512VL-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 1
; AVX512VL-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP1]], align 4, !tbaa [[SHORT_TBAA0:![0-9]+]]
; AVX512VL-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 1
; AVX512VL-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 11
; AVX512VL-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 2
; AVX512VL-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 4
; AVX512VL-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 3
; AVX512VL-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP3]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: [[TMP16:%.*]] = insertelement <4 x i32> undef, i32 [[TMP4]], i32 0
; AVX512VL-NEXT: [[TMP17:%.*]] = insertelement <4 x i32> [[TMP16]], i32 [[TMP7]], i32 1
; AVX512VL-NEXT: [[TMP15:%.*]] = insertelement <4 x i32> [[TMP17]], i32 [[TMP10]], i32 2
; AVX512VL-NEXT: [[TMP13:%.*]] = insertelement <4 x i32> [[TMP15]], i32 [[TMP12]], i32 3
; AVX512VL-NEXT: [[TMP14:%.*]] = add nsw <4 x i32> [[TMP13]], <i32 1, i32 2, i32 3, i32 4>
; AVX512VL-NEXT: store <4 x i32> [[TMP14]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: ret void
;
%3 = getelementptr inbounds i32, ptr %1, i64 1
%4 = load i32, ptr %1, align 4, !tbaa !2
%5 = getelementptr inbounds i32, ptr %0, i64 1
%6 = getelementptr inbounds i32, ptr %1, i64 11
%7 = load i32, ptr %6, align 4, !tbaa !2
%8 = getelementptr inbounds i32, ptr %0, i64 2
%9 = getelementptr inbounds i32, ptr %1, i64 4
%10 = load i32, ptr %9, align 4, !tbaa !2
%11 = getelementptr inbounds i32, ptr %0, i64 3
%12 = load i32, ptr %3, align 4, !tbaa !2
%13 = insertelement <4 x i32> undef, i32 %4, i32 0
%14 = insertelement <4 x i32> %13, i32 %7, i32 1
%15 = insertelement <4 x i32> %14, i32 %10, i32 2
%16 = insertelement <4 x i32> %15, i32 %12, i32 3
%17 = add nsw <4 x i32> %16, <i32 1, i32 2, i32 3, i32 4>
store <4 x i32> %17, ptr %0, align 4, !tbaa !2
ret void
}
define void @gather_load_2(ptr noalias nocapture %0, ptr noalias nocapture readonly %1) {
; SSE-LABEL: define void @gather_load_2(
; SSE-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; SSE-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 1
; SSE-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP5:%.*]] = add nsw i32 [[TMP4]], 1
; SSE-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 1
; SSE-NEXT: store i32 [[TMP5]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 10
; SSE-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP9:%.*]] = add nsw i32 [[TMP8]], 2
; SSE-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 2
; SSE-NEXT: store i32 [[TMP9]], ptr [[TMP6]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 3
; SSE-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP13:%.*]] = add nsw i32 [[TMP12]], 3
; SSE-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 3
; SSE-NEXT: store i32 [[TMP13]], ptr [[TMP10]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 5
; SSE-NEXT: [[TMP16:%.*]] = load i32, ptr [[TMP15]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP17:%.*]] = add nsw i32 [[TMP16]], 4
; SSE-NEXT: store i32 [[TMP17]], ptr [[TMP14]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: ret void
;
; AVX-LABEL: define void @gather_load_2(
; AVX-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 1
; AVX-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 10
; AVX-NEXT: [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 3
; AVX-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 5
; AVX-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP11:%.*]] = insertelement <4 x i32> poison, i32 [[TMP4]], i32 0
; AVX-NEXT: [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP6]], i32 1
; AVX-NEXT: [[TMP13:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP8]], i32 2
; AVX-NEXT: [[TMP14:%.*]] = insertelement <4 x i32> [[TMP13]], i32 [[TMP10]], i32 3
; AVX-NEXT: [[TMP15:%.*]] = add nsw <4 x i32> [[TMP14]], <i32 1, i32 2, i32 3, i32 4>
; AVX-NEXT: store <4 x i32> [[TMP15]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: ret void
;
; AVX2-LABEL: define void @gather_load_2(
; AVX2-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX2-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 1
; AVX2-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 10
; AVX2-NEXT: [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 3
; AVX2-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 5
; AVX2-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP11:%.*]] = insertelement <4 x i32> poison, i32 [[TMP4]], i32 0
; AVX2-NEXT: [[TMP12:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP6]], i32 1
; AVX2-NEXT: [[TMP13:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP8]], i32 2
; AVX2-NEXT: [[TMP14:%.*]] = insertelement <4 x i32> [[TMP13]], i32 [[TMP10]], i32 3
; AVX2-NEXT: [[TMP15:%.*]] = add nsw <4 x i32> [[TMP14]], <i32 1, i32 2, i32 3, i32 4>
; AVX2-NEXT: store <4 x i32> [[TMP15]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: ret void
;
; AVX512F-LABEL: define void @gather_load_2(
; AVX512F-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX512F-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 1
; AVX512F-NEXT: [[TMP4:%.*]] = call <10 x i32> @llvm.masked.load.v10i32.p0(ptr [[TMP3]], i32 4, <10 x i1> <i1 true, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 false, i1 true>, <10 x i32> poison), !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: [[TMP5:%.*]] = shufflevector <10 x i32> [[TMP4]], <10 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 9>
; AVX512F-NEXT: [[TMP6:%.*]] = add nsw <4 x i32> [[TMP5]], <i32 1, i32 3, i32 4, i32 2>
; AVX512F-NEXT: [[TMP7:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> poison, <4 x i32> <i32 0, i32 3, i32 1, i32 2>
; AVX512F-NEXT: store <4 x i32> [[TMP7]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: ret void
;
; AVX512VL-LABEL: define void @gather_load_2(
; AVX512VL-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX512VL-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 1
; AVX512VL-NEXT: [[TMP4:%.*]] = call <10 x i32> @llvm.masked.load.v10i32.p0(ptr [[TMP3]], i32 4, <10 x i1> <i1 true, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 false, i1 true>, <10 x i32> poison), !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: [[TMP5:%.*]] = shufflevector <10 x i32> [[TMP4]], <10 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 9>
; AVX512VL-NEXT: [[TMP6:%.*]] = add nsw <4 x i32> [[TMP5]], <i32 1, i32 3, i32 4, i32 2>
; AVX512VL-NEXT: [[TMP7:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> poison, <4 x i32> <i32 0, i32 3, i32 1, i32 2>
; AVX512VL-NEXT: store <4 x i32> [[TMP7]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: ret void
;
%3 = getelementptr inbounds i32, ptr %1, i64 1
%4 = load i32, ptr %3, align 4, !tbaa !2
%5 = add nsw i32 %4, 1
%6 = getelementptr inbounds i32, ptr %0, i64 1
store i32 %5, ptr %0, align 4, !tbaa !2
%7 = getelementptr inbounds i32, ptr %1, i64 10
%8 = load i32, ptr %7, align 4, !tbaa !2
%9 = add nsw i32 %8, 2
%10 = getelementptr inbounds i32, ptr %0, i64 2
store i32 %9, ptr %6, align 4, !tbaa !2
%11 = getelementptr inbounds i32, ptr %1, i64 3
%12 = load i32, ptr %11, align 4, !tbaa !2
%13 = add nsw i32 %12, 3
%14 = getelementptr inbounds i32, ptr %0, i64 3
store i32 %13, ptr %10, align 4, !tbaa !2
%15 = getelementptr inbounds i32, ptr %1, i64 5
%16 = load i32, ptr %15, align 4, !tbaa !2
%17 = add nsw i32 %16, 4
store i32 %17, ptr %14, align 4, !tbaa !2
ret void
}
define void @gather_load_3(ptr noalias nocapture %0, ptr noalias nocapture readonly %1) {
; SSE-LABEL: define void @gather_load_3(
; SSE-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; SSE-NEXT: [[TMP3:%.*]] = load i32, ptr [[TMP1]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP4:%.*]] = add i32 [[TMP3]], 1
; SSE-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 1
; SSE-NEXT: store i32 [[TMP4]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 11
; SSE-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP8:%.*]] = add i32 [[TMP7]], 2
; SSE-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 2
; SSE-NEXT: store i32 [[TMP8]], ptr [[TMP5]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 4
; SSE-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP12:%.*]] = add i32 [[TMP11]], 3
; SSE-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 3
; SSE-NEXT: store i32 [[TMP12]], ptr [[TMP9]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 15
; SSE-NEXT: [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP16:%.*]] = add i32 [[TMP15]], 4
; SSE-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 4
; SSE-NEXT: store i32 [[TMP16]], ptr [[TMP13]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 18
; SSE-NEXT: [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP20:%.*]] = add i32 [[TMP19]], 1
; SSE-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 5
; SSE-NEXT: store i32 [[TMP20]], ptr [[TMP17]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 9
; SSE-NEXT: [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP24:%.*]] = add i32 [[TMP23]], 2
; SSE-NEXT: [[TMP25:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 6
; SSE-NEXT: store i32 [[TMP24]], ptr [[TMP21]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP26:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 6
; SSE-NEXT: [[TMP27:%.*]] = load i32, ptr [[TMP26]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP28:%.*]] = add i32 [[TMP27]], 3
; SSE-NEXT: [[TMP29:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 7
; SSE-NEXT: store i32 [[TMP28]], ptr [[TMP25]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP30:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 21
; SSE-NEXT: [[TMP31:%.*]] = load i32, ptr [[TMP30]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP32:%.*]] = add i32 [[TMP31]], 4
; SSE-NEXT: store i32 [[TMP32]], ptr [[TMP29]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: ret void
;
; AVX-LABEL: define void @gather_load_3(
; AVX-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX-NEXT: [[TMP3:%.*]] = load i32, ptr [[TMP1]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 11
; AVX-NEXT: [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 4
; AVX-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 15
; AVX-NEXT: [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 18
; AVX-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 9
; AVX-NEXT: [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 6
; AVX-NEXT: [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 21
; AVX-NEXT: [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP18:%.*]] = insertelement <8 x i32> poison, i32 [[TMP3]], i32 0
; AVX-NEXT: [[TMP19:%.*]] = insertelement <8 x i32> [[TMP18]], i32 [[TMP5]], i32 1
; AVX-NEXT: [[TMP20:%.*]] = insertelement <8 x i32> [[TMP19]], i32 [[TMP7]], i32 2
; AVX-NEXT: [[TMP21:%.*]] = insertelement <8 x i32> [[TMP20]], i32 [[TMP9]], i32 3
; AVX-NEXT: [[TMP22:%.*]] = insertelement <8 x i32> [[TMP21]], i32 [[TMP11]], i32 4
; AVX-NEXT: [[TMP23:%.*]] = insertelement <8 x i32> [[TMP22]], i32 [[TMP13]], i32 5
; AVX-NEXT: [[TMP24:%.*]] = insertelement <8 x i32> [[TMP23]], i32 [[TMP15]], i32 6
; AVX-NEXT: [[TMP25:%.*]] = insertelement <8 x i32> [[TMP24]], i32 [[TMP17]], i32 7
; AVX-NEXT: [[TMP26:%.*]] = add <8 x i32> [[TMP25]], <i32 1, i32 2, i32 3, i32 4, i32 1, i32 2, i32 3, i32 4>
; AVX-NEXT: store <8 x i32> [[TMP26]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: ret void
;
; AVX2-LABEL: define void @gather_load_3(
; AVX2-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX2-NEXT: [[TMP3:%.*]] = call <22 x i32> @llvm.masked.load.v22i32.p0(ptr [[TMP1]], i32 4, <22 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true>, <22 x i32> poison), !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP4:%.*]] = shufflevector <22 x i32> [[TMP3]], <22 x i32> poison, <8 x i32> <i32 0, i32 4, i32 6, i32 9, i32 11, i32 15, i32 18, i32 21>
; AVX2-NEXT: [[TMP5:%.*]] = add <8 x i32> [[TMP4]], <i32 1, i32 3, i32 3, i32 2, i32 2, i32 4, i32 1, i32 4>
; AVX2-NEXT: [[TMP6:%.*]] = shufflevector <8 x i32> [[TMP5]], <8 x i32> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 6, i32 3, i32 2, i32 7>
; AVX2-NEXT: store <8 x i32> [[TMP6]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: ret void
;
; AVX512F-LABEL: define void @gather_load_3(
; AVX512F-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX512F-NEXT: [[TMP3:%.*]] = call <22 x i32> @llvm.masked.load.v22i32.p0(ptr [[TMP1]], i32 4, <22 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true>, <22 x i32> poison), !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: [[TMP4:%.*]] = shufflevector <22 x i32> [[TMP3]], <22 x i32> poison, <8 x i32> <i32 0, i32 4, i32 6, i32 9, i32 11, i32 15, i32 18, i32 21>
; AVX512F-NEXT: [[TMP5:%.*]] = add <8 x i32> [[TMP4]], <i32 1, i32 3, i32 3, i32 2, i32 2, i32 4, i32 1, i32 4>
; AVX512F-NEXT: [[TMP6:%.*]] = shufflevector <8 x i32> [[TMP5]], <8 x i32> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 6, i32 3, i32 2, i32 7>
; AVX512F-NEXT: store <8 x i32> [[TMP6]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: ret void
;
; AVX512VL-LABEL: define void @gather_load_3(
; AVX512VL-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX512VL-NEXT: [[TMP3:%.*]] = call <22 x i32> @llvm.masked.load.v22i32.p0(ptr [[TMP1]], i32 4, <22 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true>, <22 x i32> poison), !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: [[TMP4:%.*]] = shufflevector <22 x i32> [[TMP3]], <22 x i32> poison, <8 x i32> <i32 0, i32 4, i32 6, i32 9, i32 11, i32 15, i32 18, i32 21>
; AVX512VL-NEXT: [[TMP5:%.*]] = add <8 x i32> [[TMP4]], <i32 1, i32 3, i32 3, i32 2, i32 2, i32 4, i32 1, i32 4>
; AVX512VL-NEXT: [[TMP6:%.*]] = shufflevector <8 x i32> [[TMP5]], <8 x i32> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 6, i32 3, i32 2, i32 7>
; AVX512VL-NEXT: store <8 x i32> [[TMP6]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: ret void
;
%3 = load i32, ptr %1, align 4, !tbaa !2
%4 = add i32 %3, 1
%5 = getelementptr inbounds i32, ptr %0, i64 1
store i32 %4, ptr %0, align 4, !tbaa !2
%6 = getelementptr inbounds i32, ptr %1, i64 11
%7 = load i32, ptr %6, align 4, !tbaa !2
%8 = add i32 %7, 2
%9 = getelementptr inbounds i32, ptr %0, i64 2
store i32 %8, ptr %5, align 4, !tbaa !2
%10 = getelementptr inbounds i32, ptr %1, i64 4
%11 = load i32, ptr %10, align 4, !tbaa !2
%12 = add i32 %11, 3
%13 = getelementptr inbounds i32, ptr %0, i64 3
store i32 %12, ptr %9, align 4, !tbaa !2
%14 = getelementptr inbounds i32, ptr %1, i64 15
%15 = load i32, ptr %14, align 4, !tbaa !2
%16 = add i32 %15, 4
%17 = getelementptr inbounds i32, ptr %0, i64 4
store i32 %16, ptr %13, align 4, !tbaa !2
%18 = getelementptr inbounds i32, ptr %1, i64 18
%19 = load i32, ptr %18, align 4, !tbaa !2
%20 = add i32 %19, 1
%21 = getelementptr inbounds i32, ptr %0, i64 5
store i32 %20, ptr %17, align 4, !tbaa !2
%22 = getelementptr inbounds i32, ptr %1, i64 9
%23 = load i32, ptr %22, align 4, !tbaa !2
%24 = add i32 %23, 2
%25 = getelementptr inbounds i32, ptr %0, i64 6
store i32 %24, ptr %21, align 4, !tbaa !2
%26 = getelementptr inbounds i32, ptr %1, i64 6
%27 = load i32, ptr %26, align 4, !tbaa !2
%28 = add i32 %27, 3
%29 = getelementptr inbounds i32, ptr %0, i64 7
store i32 %28, ptr %25, align 4, !tbaa !2
%30 = getelementptr inbounds i32, ptr %1, i64 21
%31 = load i32, ptr %30, align 4, !tbaa !2
%32 = add i32 %31, 4
store i32 %32, ptr %29, align 4, !tbaa !2
ret void
}
define void @gather_load_4(ptr noalias nocapture %t0, ptr noalias nocapture readonly %t1) {
; SSE-LABEL: define void @gather_load_4(
; SSE-SAME: ptr noalias captures(none) [[T0:%.*]], ptr noalias readonly captures(none) [[T1:%.*]]) #[[ATTR0]] {
; SSE-NEXT: [[T5:%.*]] = getelementptr inbounds i32, ptr [[T0]], i64 1
; SSE-NEXT: [[T6:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 11
; SSE-NEXT: [[T9:%.*]] = getelementptr inbounds i32, ptr [[T0]], i64 2
; SSE-NEXT: [[T10:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 4
; SSE-NEXT: [[T13:%.*]] = getelementptr inbounds i32, ptr [[T0]], i64 3
; SSE-NEXT: [[T14:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 15
; SSE-NEXT: [[T17:%.*]] = getelementptr inbounds i32, ptr [[T0]], i64 4
; SSE-NEXT: [[T18:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 18
; SSE-NEXT: [[T21:%.*]] = getelementptr inbounds i32, ptr [[T0]], i64 5
; SSE-NEXT: [[T22:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 9
; SSE-NEXT: [[T25:%.*]] = getelementptr inbounds i32, ptr [[T0]], i64 6
; SSE-NEXT: [[T26:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 6
; SSE-NEXT: [[T29:%.*]] = getelementptr inbounds i32, ptr [[T0]], i64 7
; SSE-NEXT: [[T30:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 21
; SSE-NEXT: [[T3:%.*]] = load i32, ptr [[T1]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[T7:%.*]] = load i32, ptr [[T6]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[T11:%.*]] = load i32, ptr [[T10]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[T15:%.*]] = load i32, ptr [[T14]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[T19:%.*]] = load i32, ptr [[T18]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[T23:%.*]] = load i32, ptr [[T22]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[T27:%.*]] = load i32, ptr [[T26]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[T31:%.*]] = load i32, ptr [[T30]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[T4:%.*]] = add i32 [[T3]], 1
; SSE-NEXT: [[T8:%.*]] = add i32 [[T7]], 2
; SSE-NEXT: [[T12:%.*]] = add i32 [[T11]], 3
; SSE-NEXT: [[T16:%.*]] = add i32 [[T15]], 4
; SSE-NEXT: [[T20:%.*]] = add i32 [[T19]], 1
; SSE-NEXT: [[T24:%.*]] = add i32 [[T23]], 2
; SSE-NEXT: [[T28:%.*]] = add i32 [[T27]], 3
; SSE-NEXT: [[T32:%.*]] = add i32 [[T31]], 4
; SSE-NEXT: store i32 [[T4]], ptr [[T0]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: store i32 [[T8]], ptr [[T5]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: store i32 [[T12]], ptr [[T9]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: store i32 [[T16]], ptr [[T13]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: store i32 [[T20]], ptr [[T17]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: store i32 [[T24]], ptr [[T21]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: store i32 [[T28]], ptr [[T25]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: store i32 [[T32]], ptr [[T29]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: ret void
;
; AVX-LABEL: define void @gather_load_4(
; AVX-SAME: ptr noalias captures(none) [[T0:%.*]], ptr noalias readonly captures(none) [[T1:%.*]]) #[[ATTR0]] {
; AVX-NEXT: [[T6:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 11
; AVX-NEXT: [[T10:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 4
; AVX-NEXT: [[T14:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 15
; AVX-NEXT: [[T18:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 18
; AVX-NEXT: [[T22:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 9
; AVX-NEXT: [[T26:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 6
; AVX-NEXT: [[T30:%.*]] = getelementptr inbounds i32, ptr [[T1]], i64 21
; AVX-NEXT: [[T3:%.*]] = load i32, ptr [[T1]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[T7:%.*]] = load i32, ptr [[T6]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[T11:%.*]] = load i32, ptr [[T10]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[T15:%.*]] = load i32, ptr [[T14]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[T19:%.*]] = load i32, ptr [[T18]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[T23:%.*]] = load i32, ptr [[T22]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[T27:%.*]] = load i32, ptr [[T26]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[T31:%.*]] = load i32, ptr [[T30]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP1:%.*]] = insertelement <8 x i32> poison, i32 [[T3]], i32 0
; AVX-NEXT: [[TMP2:%.*]] = insertelement <8 x i32> [[TMP1]], i32 [[T7]], i32 1
; AVX-NEXT: [[TMP3:%.*]] = insertelement <8 x i32> [[TMP2]], i32 [[T11]], i32 2
; AVX-NEXT: [[TMP4:%.*]] = insertelement <8 x i32> [[TMP3]], i32 [[T15]], i32 3
; AVX-NEXT: [[TMP5:%.*]] = insertelement <8 x i32> [[TMP4]], i32 [[T19]], i32 4
; AVX-NEXT: [[TMP6:%.*]] = insertelement <8 x i32> [[TMP5]], i32 [[T23]], i32 5
; AVX-NEXT: [[TMP7:%.*]] = insertelement <8 x i32> [[TMP6]], i32 [[T27]], i32 6
; AVX-NEXT: [[TMP8:%.*]] = insertelement <8 x i32> [[TMP7]], i32 [[T31]], i32 7
; AVX-NEXT: [[TMP9:%.*]] = add <8 x i32> [[TMP8]], <i32 1, i32 2, i32 3, i32 4, i32 1, i32 2, i32 3, i32 4>
; AVX-NEXT: store <8 x i32> [[TMP9]], ptr [[T0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: ret void
;
; AVX2-LABEL: define void @gather_load_4(
; AVX2-SAME: ptr noalias captures(none) [[T0:%.*]], ptr noalias readonly captures(none) [[T1:%.*]]) #[[ATTR0]] {
; AVX2-NEXT: [[TMP1:%.*]] = call <22 x i32> @llvm.masked.load.v22i32.p0(ptr [[T1]], i32 4, <22 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true>, <22 x i32> poison), !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP2:%.*]] = shufflevector <22 x i32> [[TMP1]], <22 x i32> poison, <8 x i32> <i32 0, i32 4, i32 6, i32 9, i32 11, i32 15, i32 18, i32 21>
; AVX2-NEXT: [[TMP3:%.*]] = add <8 x i32> [[TMP2]], <i32 1, i32 3, i32 3, i32 2, i32 2, i32 4, i32 1, i32 4>
; AVX2-NEXT: [[TMP4:%.*]] = shufflevector <8 x i32> [[TMP3]], <8 x i32> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 6, i32 3, i32 2, i32 7>
; AVX2-NEXT: store <8 x i32> [[TMP4]], ptr [[T0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: ret void
;
; AVX512F-LABEL: define void @gather_load_4(
; AVX512F-SAME: ptr noalias captures(none) [[T0:%.*]], ptr noalias readonly captures(none) [[T1:%.*]]) #[[ATTR0]] {
; AVX512F-NEXT: [[TMP1:%.*]] = call <22 x i32> @llvm.masked.load.v22i32.p0(ptr [[T1]], i32 4, <22 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true>, <22 x i32> poison), !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: [[TMP2:%.*]] = shufflevector <22 x i32> [[TMP1]], <22 x i32> poison, <8 x i32> <i32 0, i32 4, i32 6, i32 9, i32 11, i32 15, i32 18, i32 21>
; AVX512F-NEXT: [[TMP3:%.*]] = add <8 x i32> [[TMP2]], <i32 1, i32 3, i32 3, i32 2, i32 2, i32 4, i32 1, i32 4>
; AVX512F-NEXT: [[TMP4:%.*]] = shufflevector <8 x i32> [[TMP3]], <8 x i32> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 6, i32 3, i32 2, i32 7>
; AVX512F-NEXT: store <8 x i32> [[TMP4]], ptr [[T0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: ret void
;
; AVX512VL-LABEL: define void @gather_load_4(
; AVX512VL-SAME: ptr noalias captures(none) [[T0:%.*]], ptr noalias readonly captures(none) [[T1:%.*]]) #[[ATTR0]] {
; AVX512VL-NEXT: [[TMP1:%.*]] = call <22 x i32> @llvm.masked.load.v22i32.p0(ptr [[T1]], i32 4, <22 x i1> <i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true>, <22 x i32> poison), !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: [[TMP2:%.*]] = shufflevector <22 x i32> [[TMP1]], <22 x i32> poison, <8 x i32> <i32 0, i32 4, i32 6, i32 9, i32 11, i32 15, i32 18, i32 21>
; AVX512VL-NEXT: [[TMP3:%.*]] = add <8 x i32> [[TMP2]], <i32 1, i32 3, i32 3, i32 2, i32 2, i32 4, i32 1, i32 4>
; AVX512VL-NEXT: [[TMP4:%.*]] = shufflevector <8 x i32> [[TMP3]], <8 x i32> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 6, i32 3, i32 2, i32 7>
; AVX512VL-NEXT: store <8 x i32> [[TMP4]], ptr [[T0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: ret void
;
%t5 = getelementptr inbounds i32, ptr %t0, i64 1
%t6 = getelementptr inbounds i32, ptr %t1, i64 11
%t9 = getelementptr inbounds i32, ptr %t0, i64 2
%t10 = getelementptr inbounds i32, ptr %t1, i64 4
%t13 = getelementptr inbounds i32, ptr %t0, i64 3
%t14 = getelementptr inbounds i32, ptr %t1, i64 15
%t17 = getelementptr inbounds i32, ptr %t0, i64 4
%t18 = getelementptr inbounds i32, ptr %t1, i64 18
%t21 = getelementptr inbounds i32, ptr %t0, i64 5
%t22 = getelementptr inbounds i32, ptr %t1, i64 9
%t25 = getelementptr inbounds i32, ptr %t0, i64 6
%t26 = getelementptr inbounds i32, ptr %t1, i64 6
%t29 = getelementptr inbounds i32, ptr %t0, i64 7
%t30 = getelementptr inbounds i32, ptr %t1, i64 21
%t3 = load i32, ptr %t1, align 4, !tbaa !2
%t7 = load i32, ptr %t6, align 4, !tbaa !2
%t11 = load i32, ptr %t10, align 4, !tbaa !2
%t15 = load i32, ptr %t14, align 4, !tbaa !2
%t19 = load i32, ptr %t18, align 4, !tbaa !2
%t23 = load i32, ptr %t22, align 4, !tbaa !2
%t27 = load i32, ptr %t26, align 4, !tbaa !2
%t31 = load i32, ptr %t30, align 4, !tbaa !2
%t4 = add i32 %t3, 1
%t8 = add i32 %t7, 2
%t12 = add i32 %t11, 3
%t16 = add i32 %t15, 4
%t20 = add i32 %t19, 1
%t24 = add i32 %t23, 2
%t28 = add i32 %t27, 3
%t32 = add i32 %t31, 4
store i32 %t4, ptr %t0, align 4, !tbaa !2
store i32 %t8, ptr %t5, align 4, !tbaa !2
store i32 %t12, ptr %t9, align 4, !tbaa !2
store i32 %t16, ptr %t13, align 4, !tbaa !2
store i32 %t20, ptr %t17, align 4, !tbaa !2
store i32 %t24, ptr %t21, align 4, !tbaa !2
store i32 %t28, ptr %t25, align 4, !tbaa !2
store i32 %t32, ptr %t29, align 4, !tbaa !2
ret void
}
define void @gather_load_div(ptr noalias nocapture %0, ptr noalias nocapture readonly %1) {
; SSE-LABEL: define void @gather_load_div(
; SSE-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; SSE-NEXT: [[TMP3:%.*]] = load float, ptr [[TMP1]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 10
; SSE-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 13
; SSE-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 3
; SSE-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 44
; SSE-NEXT: [[TMP8:%.*]] = load float, ptr [[TMP7]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP0]], i64 4
; SSE-NEXT: [[TMP10:%.*]] = load <2 x float>, ptr [[TMP6]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP11:%.*]] = load <2 x float>, ptr [[TMP4]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP12:%.*]] = load <2 x float>, ptr [[TMP5]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP13:%.*]] = insertelement <4 x float> poison, float [[TMP3]], i32 0
; SSE-NEXT: [[TMP14:%.*]] = shufflevector <2 x float> [[TMP11]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
; SSE-NEXT: [[TMP15:%.*]] = shufflevector <4 x float> [[TMP13]], <4 x float> [[TMP14]], <4 x i32> <i32 0, i32 4, i32 poison, i32 poison>
; SSE-NEXT: [[TMP16:%.*]] = shufflevector <2 x float> [[TMP10]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
; SSE-NEXT: [[TMP17:%.*]] = shufflevector <4 x float> [[TMP15]], <4 x float> [[TMP16]], <4 x i32> <i32 0, i32 1, i32 4, i32 poison>
; SSE-NEXT: [[TMP18:%.*]] = shufflevector <2 x float> [[TMP12]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
; SSE-NEXT: [[TMP19:%.*]] = shufflevector <4 x float> [[TMP17]], <4 x float> [[TMP18]], <4 x i32> <i32 0, i32 1, i32 2, i32 5>
; SSE-NEXT: [[TMP20:%.*]] = shufflevector <2 x float> [[TMP10]], <2 x float> [[TMP12]], <4 x i32> <i32 1, i32 2, i32 poison, i32 poison>
; SSE-NEXT: [[TMP21:%.*]] = shufflevector <4 x float> [[TMP20]], <4 x float> [[TMP14]], <4 x i32> <i32 0, i32 1, i32 5, i32 poison>
; SSE-NEXT: [[TMP22:%.*]] = insertelement <4 x float> [[TMP21]], float [[TMP8]], i32 3
; SSE-NEXT: [[TMP23:%.*]] = fdiv <4 x float> [[TMP19]], [[TMP22]]
; SSE-NEXT: store <4 x float> [[TMP23]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP24:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 17
; SSE-NEXT: [[TMP25:%.*]] = load float, ptr [[TMP24]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 33
; SSE-NEXT: [[TMP27:%.*]] = load float, ptr [[TMP26]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP28:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 8
; SSE-NEXT: [[TMP29:%.*]] = load float, ptr [[TMP28]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP30:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 30
; SSE-NEXT: [[TMP31:%.*]] = load float, ptr [[TMP30]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP32:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 5
; SSE-NEXT: [[TMP33:%.*]] = load float, ptr [[TMP32]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP34:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 27
; SSE-NEXT: [[TMP35:%.*]] = load float, ptr [[TMP34]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP36:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 20
; SSE-NEXT: [[TMP37:%.*]] = load float, ptr [[TMP36]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 23
; SSE-NEXT: [[TMP39:%.*]] = load float, ptr [[TMP38]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: [[TMP40:%.*]] = insertelement <4 x float> poison, float [[TMP25]], i32 0
; SSE-NEXT: [[TMP41:%.*]] = insertelement <4 x float> [[TMP40]], float [[TMP29]], i32 1
; SSE-NEXT: [[TMP42:%.*]] = insertelement <4 x float> [[TMP41]], float [[TMP33]], i32 2
; SSE-NEXT: [[TMP43:%.*]] = insertelement <4 x float> [[TMP42]], float [[TMP37]], i32 3
; SSE-NEXT: [[TMP44:%.*]] = insertelement <4 x float> poison, float [[TMP27]], i32 0
; SSE-NEXT: [[TMP45:%.*]] = insertelement <4 x float> [[TMP44]], float [[TMP31]], i32 1
; SSE-NEXT: [[TMP46:%.*]] = insertelement <4 x float> [[TMP45]], float [[TMP35]], i32 2
; SSE-NEXT: [[TMP47:%.*]] = insertelement <4 x float> [[TMP46]], float [[TMP39]], i32 3
; SSE-NEXT: [[TMP48:%.*]] = fdiv <4 x float> [[TMP43]], [[TMP47]]
; SSE-NEXT: store <4 x float> [[TMP48]], ptr [[TMP9]], align 4, !tbaa [[SHORT_TBAA0]]
; SSE-NEXT: ret void
;
; AVX-LABEL: define void @gather_load_div(
; AVX-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX-NEXT: [[TMP3:%.*]] = load float, ptr [[TMP1]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 10
; AVX-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 13
; AVX-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 3
; AVX-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 44
; AVX-NEXT: [[TMP8:%.*]] = load float, ptr [[TMP7]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 17
; AVX-NEXT: [[TMP10:%.*]] = load float, ptr [[TMP9]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 33
; AVX-NEXT: [[TMP12:%.*]] = load float, ptr [[TMP11]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 8
; AVX-NEXT: [[TMP14:%.*]] = load float, ptr [[TMP13]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 30
; AVX-NEXT: [[TMP16:%.*]] = load float, ptr [[TMP15]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 5
; AVX-NEXT: [[TMP18:%.*]] = load float, ptr [[TMP17]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 27
; AVX-NEXT: [[TMP20:%.*]] = load float, ptr [[TMP19]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 20
; AVX-NEXT: [[TMP22:%.*]] = load float, ptr [[TMP21]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP23:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 23
; AVX-NEXT: [[TMP24:%.*]] = load float, ptr [[TMP23]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP25:%.*]] = load <2 x float>, ptr [[TMP6]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP26:%.*]] = load <2 x float>, ptr [[TMP4]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP27:%.*]] = load <2 x float>, ptr [[TMP5]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: [[TMP28:%.*]] = insertelement <8 x float> poison, float [[TMP3]], i32 0
; AVX-NEXT: [[TMP29:%.*]] = shufflevector <2 x float> [[TMP26]], <2 x float> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: [[TMP30:%.*]] = shufflevector <8 x float> [[TMP28]], <8 x float> [[TMP29]], <8 x i32> <i32 0, i32 8, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: [[TMP31:%.*]] = shufflevector <2 x float> [[TMP25]], <2 x float> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: [[TMP32:%.*]] = shufflevector <8 x float> [[TMP30]], <8 x float> [[TMP31]], <8 x i32> <i32 0, i32 1, i32 8, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: [[TMP33:%.*]] = shufflevector <2 x float> [[TMP27]], <2 x float> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: [[TMP34:%.*]] = shufflevector <8 x float> [[TMP32]], <8 x float> [[TMP33]], <8 x i32> <i32 0, i32 1, i32 2, i32 9, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: [[TMP35:%.*]] = insertelement <8 x float> [[TMP34]], float [[TMP10]], i32 4
; AVX-NEXT: [[TMP36:%.*]] = insertelement <8 x float> [[TMP35]], float [[TMP14]], i32 5
; AVX-NEXT: [[TMP37:%.*]] = insertelement <8 x float> [[TMP36]], float [[TMP18]], i32 6
; AVX-NEXT: [[TMP38:%.*]] = insertelement <8 x float> [[TMP37]], float [[TMP22]], i32 7
; AVX-NEXT: [[TMP39:%.*]] = shufflevector <2 x float> [[TMP25]], <2 x float> [[TMP27]], <8 x i32> <i32 1, i32 2, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: [[TMP40:%.*]] = shufflevector <8 x float> [[TMP39]], <8 x float> [[TMP29]], <8 x i32> <i32 0, i32 1, i32 9, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX-NEXT: [[TMP41:%.*]] = insertelement <8 x float> [[TMP40]], float [[TMP8]], i32 3
; AVX-NEXT: [[TMP42:%.*]] = insertelement <8 x float> [[TMP41]], float [[TMP12]], i32 4
; AVX-NEXT: [[TMP43:%.*]] = insertelement <8 x float> [[TMP42]], float [[TMP16]], i32 5
; AVX-NEXT: [[TMP44:%.*]] = insertelement <8 x float> [[TMP43]], float [[TMP20]], i32 6
; AVX-NEXT: [[TMP45:%.*]] = insertelement <8 x float> [[TMP44]], float [[TMP24]], i32 7
; AVX-NEXT: [[TMP46:%.*]] = fdiv <8 x float> [[TMP38]], [[TMP45]]
; AVX-NEXT: store <8 x float> [[TMP46]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX-NEXT: ret void
;
; AVX2-LABEL: define void @gather_load_div(
; AVX2-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX2-NEXT: [[TMP3:%.*]] = load float, ptr [[TMP1]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 10
; AVX2-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 13
; AVX2-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 3
; AVX2-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 44
; AVX2-NEXT: [[TMP8:%.*]] = load float, ptr [[TMP7]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 17
; AVX2-NEXT: [[TMP10:%.*]] = load float, ptr [[TMP9]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 33
; AVX2-NEXT: [[TMP12:%.*]] = load float, ptr [[TMP11]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 8
; AVX2-NEXT: [[TMP14:%.*]] = load float, ptr [[TMP13]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 30
; AVX2-NEXT: [[TMP16:%.*]] = load float, ptr [[TMP15]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 5
; AVX2-NEXT: [[TMP18:%.*]] = load float, ptr [[TMP17]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 27
; AVX2-NEXT: [[TMP20:%.*]] = load float, ptr [[TMP19]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 20
; AVX2-NEXT: [[TMP22:%.*]] = load float, ptr [[TMP21]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP23:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 23
; AVX2-NEXT: [[TMP24:%.*]] = load float, ptr [[TMP23]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP25:%.*]] = load <2 x float>, ptr [[TMP6]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP26:%.*]] = load <2 x float>, ptr [[TMP4]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP27:%.*]] = load <2 x float>, ptr [[TMP5]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: [[TMP28:%.*]] = insertelement <8 x float> poison, float [[TMP3]], i32 0
; AVX2-NEXT: [[TMP29:%.*]] = shufflevector <2 x float> [[TMP26]], <2 x float> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX2-NEXT: [[TMP30:%.*]] = shufflevector <8 x float> [[TMP28]], <8 x float> [[TMP29]], <8 x i32> <i32 0, i32 8, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX2-NEXT: [[TMP31:%.*]] = shufflevector <2 x float> [[TMP25]], <2 x float> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX2-NEXT: [[TMP32:%.*]] = shufflevector <8 x float> [[TMP30]], <8 x float> [[TMP31]], <8 x i32> <i32 0, i32 1, i32 8, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX2-NEXT: [[TMP33:%.*]] = shufflevector <2 x float> [[TMP27]], <2 x float> poison, <8 x i32> <i32 0, i32 1, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX2-NEXT: [[TMP34:%.*]] = shufflevector <8 x float> [[TMP32]], <8 x float> [[TMP33]], <8 x i32> <i32 0, i32 1, i32 2, i32 9, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX2-NEXT: [[TMP35:%.*]] = insertelement <8 x float> [[TMP34]], float [[TMP10]], i32 4
; AVX2-NEXT: [[TMP36:%.*]] = insertelement <8 x float> [[TMP35]], float [[TMP14]], i32 5
; AVX2-NEXT: [[TMP37:%.*]] = insertelement <8 x float> [[TMP36]], float [[TMP18]], i32 6
; AVX2-NEXT: [[TMP38:%.*]] = insertelement <8 x float> [[TMP37]], float [[TMP22]], i32 7
; AVX2-NEXT: [[TMP39:%.*]] = shufflevector <2 x float> [[TMP25]], <2 x float> [[TMP27]], <8 x i32> <i32 1, i32 2, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX2-NEXT: [[TMP40:%.*]] = shufflevector <8 x float> [[TMP39]], <8 x float> [[TMP29]], <8 x i32> <i32 0, i32 1, i32 9, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
; AVX2-NEXT: [[TMP41:%.*]] = insertelement <8 x float> [[TMP40]], float [[TMP8]], i32 3
; AVX2-NEXT: [[TMP42:%.*]] = insertelement <8 x float> [[TMP41]], float [[TMP12]], i32 4
; AVX2-NEXT: [[TMP43:%.*]] = insertelement <8 x float> [[TMP42]], float [[TMP16]], i32 5
; AVX2-NEXT: [[TMP44:%.*]] = insertelement <8 x float> [[TMP43]], float [[TMP20]], i32 6
; AVX2-NEXT: [[TMP45:%.*]] = insertelement <8 x float> [[TMP44]], float [[TMP24]], i32 7
; AVX2-NEXT: [[TMP46:%.*]] = fdiv <8 x float> [[TMP38]], [[TMP45]]
; AVX2-NEXT: store <8 x float> [[TMP46]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX2-NEXT: ret void
;
; AVX512F-LABEL: define void @gather_load_div(
; AVX512F-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX512F-NEXT: [[TMP3:%.*]] = call <45 x float> @llvm.masked.load.v45f32.p0(ptr [[TMP1]], i32 4, <45 x i1> <i1 true, i1 false, i1 false, i1 true, i1 true, i1 true, i1 false, i1 false, i1 true, i1 false, i1 true, i1 true, i1 false, i1 true, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false, i1 true>, <45 x float> poison), !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: [[TMP4:%.*]] = shufflevector <45 x float> [[TMP3]], <45 x float> poison, <16 x i32> <i32 0, i32 3, i32 4, i32 5, i32 8, i32 10, i32 11, i32 13, i32 14, i32 17, i32 20, i32 23, i32 27, i32 30, i32 33, i32 44>
; AVX512F-NEXT: [[TMP7:%.*]] = shufflevector <45 x float> [[TMP3]], <45 x float> poison, <8 x i32> <i32 0, i32 3, i32 5, i32 8, i32 10, i32 14, i32 17, i32 20>
; AVX512F-NEXT: [[TMP8:%.*]] = shufflevector <45 x float> [[TMP3]], <45 x float> poison, <8 x i32> <i32 4, i32 11, i32 27, i32 30, i32 13, i32 44, i32 33, i32 23>
; AVX512F-NEXT: [[TMP9:%.*]] = fdiv <8 x float> [[TMP7]], [[TMP8]]
; AVX512F-NEXT: [[TMP10:%.*]] = shufflevector <8 x float> [[TMP9]], <8 x float> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 6, i32 3, i32 2, i32 7>
; AVX512F-NEXT: store <8 x float> [[TMP10]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512F-NEXT: ret void
;
; AVX512VL-LABEL: define void @gather_load_div(
; AVX512VL-SAME: ptr noalias captures(none) [[TMP0:%.*]], ptr noalias readonly captures(none) [[TMP1:%.*]]) #[[ATTR0]] {
; AVX512VL-NEXT: [[TMP3:%.*]] = call <45 x float> @llvm.masked.load.v45f32.p0(ptr [[TMP1]], i32 4, <45 x i1> <i1 true, i1 false, i1 false, i1 true, i1 true, i1 true, i1 false, i1 false, i1 true, i1 false, i1 true, i1 true, i1 false, i1 true, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 true, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false, i1 false, i1 true>, <45 x float> poison), !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: [[TMP4:%.*]] = shufflevector <45 x float> [[TMP3]], <45 x float> poison, <16 x i32> <i32 0, i32 3, i32 4, i32 5, i32 8, i32 10, i32 11, i32 13, i32 14, i32 17, i32 20, i32 23, i32 27, i32 30, i32 33, i32 44>
; AVX512VL-NEXT: [[TMP7:%.*]] = shufflevector <45 x float> [[TMP3]], <45 x float> poison, <8 x i32> <i32 0, i32 3, i32 5, i32 8, i32 10, i32 14, i32 17, i32 20>
; AVX512VL-NEXT: [[TMP8:%.*]] = shufflevector <45 x float> [[TMP3]], <45 x float> poison, <8 x i32> <i32 4, i32 11, i32 27, i32 30, i32 13, i32 44, i32 33, i32 23>
; AVX512VL-NEXT: [[TMP9:%.*]] = fdiv <8 x float> [[TMP7]], [[TMP8]]
; AVX512VL-NEXT: [[TMP10:%.*]] = shufflevector <8 x float> [[TMP9]], <8 x float> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 6, i32 3, i32 2, i32 7>
; AVX512VL-NEXT: store <8 x float> [[TMP10]], ptr [[TMP0]], align 4, !tbaa [[SHORT_TBAA0]]
; AVX512VL-NEXT: ret void
;
%3 = load float, ptr %1, align 4, !tbaa !2
%4 = getelementptr inbounds float, ptr %1, i64 4
%5 = load float, ptr %4, align 4, !tbaa !2
%6 = fdiv float %3, %5
%7 = getelementptr inbounds float, ptr %0, i64 1
store float %6, ptr %0, align 4, !tbaa !2
%8 = getelementptr inbounds float, ptr %1, i64 10
%9 = load float, ptr %8, align 4, !tbaa !2
%10 = getelementptr inbounds float, ptr %1, i64 13
%11 = load float, ptr %10, align 4, !tbaa !2
%12 = fdiv float %9, %11
%13 = getelementptr inbounds float, ptr %0, i64 2
store float %12, ptr %7, align 4, !tbaa !2
%14 = getelementptr inbounds float, ptr %1, i64 3
%15 = load float, ptr %14, align 4, !tbaa !2
%16 = getelementptr inbounds float, ptr %1, i64 11
%17 = load float, ptr %16, align 4, !tbaa !2
%18 = fdiv float %15, %17
%19 = getelementptr inbounds float, ptr %0, i64 3
store float %18, ptr %13, align 4, !tbaa !2
%20 = getelementptr inbounds float, ptr %1, i64 14
%21 = load float, ptr %20, align 4, !tbaa !2
%22 = getelementptr inbounds float, ptr %1, i64 44
%23 = load float, ptr %22, align 4, !tbaa !2
%24 = fdiv float %21, %23
%25 = getelementptr inbounds float, ptr %0, i64 4
store float %24, ptr %19, align 4, !tbaa !2
%26 = getelementptr inbounds float, ptr %1, i64 17
%27 = load float, ptr %26, align 4, !tbaa !2
%28 = getelementptr inbounds float, ptr %1, i64 33
%29 = load float, ptr %28, align 4, !tbaa !2
%30 = fdiv float %27, %29
%31 = getelementptr inbounds float, ptr %0, i64 5
store float %30, ptr %25, align 4, !tbaa !2
%32 = getelementptr inbounds float, ptr %1, i64 8
%33 = load float, ptr %32, align 4, !tbaa !2
%34 = getelementptr inbounds float, ptr %1, i64 30
%35 = load float, ptr %34, align 4, !tbaa !2
%36 = fdiv float %33, %35
%37 = getelementptr inbounds float, ptr %0, i64 6
store float %36, ptr %31, align 4, !tbaa !2
%38 = getelementptr inbounds float, ptr %1, i64 5
%39 = load float, ptr %38, align 4, !tbaa !2
%40 = getelementptr inbounds float, ptr %1, i64 27
%41 = load float, ptr %40, align 4, !tbaa !2
%42 = fdiv float %39, %41
%43 = getelementptr inbounds float, ptr %0, i64 7
store float %42, ptr %37, align 4, !tbaa !2
%44 = getelementptr inbounds float, ptr %1, i64 20
%45 = load float, ptr %44, align 4, !tbaa !2
%46 = getelementptr inbounds float, ptr %1, i64 23
%47 = load float, ptr %46, align 4, !tbaa !2
%48 = fdiv float %45, %47
store float %48, ptr %43, align 4, !tbaa !2
ret void
}
!2 = !{!3, !3, i64 0}
!3 = !{!"short", !4, i64 0}
!4 = !{!"omnipotent char", !5, i64 0}
!5 = !{!"Simple C++ TBAA"}
;.
; SSE: [[SHORT_TBAA0]] = !{[[META1:![0-9]+]], [[META1]], i64 0}
; SSE: [[META1]] = !{!"short", [[META2:![0-9]+]], i64 0}
; SSE: [[META2]] = !{!"omnipotent char", [[META3:![0-9]+]], i64 0}
; SSE: [[META3]] = !{!"Simple C++ TBAA"}
;.
; AVX: [[SHORT_TBAA0]] = !{[[META1:![0-9]+]], [[META1]], i64 0}
; AVX: [[META1]] = !{!"short", [[META2:![0-9]+]], i64 0}
; AVX: [[META2]] = !{!"omnipotent char", [[META3:![0-9]+]], i64 0}
; AVX: [[META3]] = !{!"Simple C++ TBAA"}
;.
; AVX2: [[SHORT_TBAA0]] = !{[[META1:![0-9]+]], [[META1]], i64 0}
; AVX2: [[META1]] = !{!"short", [[META2:![0-9]+]], i64 0}
; AVX2: [[META2]] = !{!"omnipotent char", [[META3:![0-9]+]], i64 0}
; AVX2: [[META3]] = !{!"Simple C++ TBAA"}
;.
; AVX512F: [[SHORT_TBAA0]] = !{[[META1:![0-9]+]], [[META1]], i64 0}
; AVX512F: [[META1]] = !{!"short", [[META2:![0-9]+]], i64 0}
; AVX512F: [[META2]] = !{!"omnipotent char", [[META3:![0-9]+]], i64 0}
; AVX512F: [[META3]] = !{!"Simple C++ TBAA"}
;.
; AVX512VL: [[SHORT_TBAA0]] = !{[[META1:![0-9]+]], [[META1]], i64 0}
; AVX512VL: [[META1]] = !{!"short", [[META2:![0-9]+]], i64 0}
; AVX512VL: [[META2]] = !{!"omnipotent char", [[META3:![0-9]+]], i64 0}
; AVX512VL: [[META3]] = !{!"Simple C++ TBAA"}
;.