caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
a78861fc55d18046989ff4d624a037e9181da170
clang-p2996/llvm/test/Transforms/SLPVectorizer/X86/pr46983.ll
Alexey Bataev d74e42acd2 [SLP]Attempt to vectorize long stores, if short one failed. 
We can try to vectorize long store sequences, if short ones were
unsuccessful because of the non-profitable vectorization. It should not
increase compile time significantly (stores are sorted already,
complexity is n x log n), but vectorize extra code.

Metric: size..text

Program                                                                         size..text
                                                                                results     results0    diff
         test-suite :: External/SPEC/CINT2006/400.perlbench/400.perlbench.test  1088012.00  1088236.00  0.0%
                  test-suite :: SingleSource/UnitTests/matrix-types-spec.test   480396.00   480476.00  0.0%
          test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   664613.00   664661.00  0.0%
         test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   664613.00   664661.00  0.0%
        test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test  2041105.00  2040961.00 -0.0%
                 test-suite :: MultiSource/Applications/JM/lencod/lencod.test   836563.00   836387.00 -0.0%
                 test-suite :: MultiSource/Benchmarks/7zip/7zip-benchmark.test  1035100.00  1032140.00 -0.3%

In all benchmarks extra code gets vectorized

Reviewers: RKSimon

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/88563
2024-04-26 06:53:44 -07:00

161 lines
7.0 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -passes=slp-vectorizer,instcombine -S -mtriple=x86_64-unknown-linux-gnu -mattr=+sse2 -slp-threshold=-1 | FileCheck %s
; RUN: opt < %s -passes=slp-vectorizer,instcombine -S -mtriple=x86_64-unknown-linux-gnu -mattr=+sse4.2 | FileCheck %s
; RUN: opt < %s -passes=slp-vectorizer,instcombine -S -mtriple=x86_64-unknown-linux-gnu -mattr=+avx | FileCheck %s
; RUN: opt < %s -passes=slp-vectorizer,instcombine -S -mtriple=x86_64-unknown-linux-gnu -mattr=+avx2 | FileCheck %s
; RUN: opt < %s -passes=slp-vectorizer,instcombine -S -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512bw,+avx512vl | FileCheck %s
define void @store_i32(ptr nocapture %0, i32 %1, i32 %2) {
; CHECK-LABEL: @store_i32(
; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[TMP0:%.*]], align 4, !tbaa [[TBAA0:![0-9]+]]
; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1:%.*]], i64 0
; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <4 x i32> [[TMP5]], <4 x i32> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP7:%.*]] = mul <4 x i32> [[TMP4]], [[TMP6]]
; CHECK-NEXT: [[TMP8:%.*]] = lshr <4 x i32> [[TMP7]], <i32 15, i32 15, i32 15, i32 15>
; CHECK-NEXT: [[TMP9:%.*]] = call <4 x i32> @llvm.umin.v4i32(<4 x i32> [[TMP8]], <4 x i32> <i32 255, i32 255, i32 255, i32 255>)
; CHECK-NEXT: store <4 x i32> [[TMP9]], ptr [[TMP0]], align 4, !tbaa [[TBAA0]]
; CHECK-NEXT: ret void
;
%4 = load i32, ptr %0, align 4, !tbaa !2
%5 = mul i32 %4, %1
%6 = lshr i32 %5, 15
%7 = icmp ult i32 %6, 255
%8 = select i1 %7, i32 %6, i32 255
store i32 %8, ptr %0, align 4, !tbaa !2
%9 = getelementptr inbounds i32, ptr %0, i64 1
%10 = load i32, ptr %9, align 4, !tbaa !2
%11 = mul i32 %10, %1
%12 = lshr i32 %11, 15
%13 = icmp ult i32 %12, 255
%14 = select i1 %13, i32 %12, i32 255
store i32 %14, ptr %9, align 4, !tbaa !2
%15 = getelementptr inbounds i32, ptr %0, i64 2
%16 = load i32, ptr %15, align 4, !tbaa !2
%17 = mul i32 %16, %1
%18 = lshr i32 %17, 15
%19 = icmp ult i32 %18, 255
%20 = select i1 %19, i32 %18, i32 255
store i32 %20, ptr %15, align 4, !tbaa !2
%21 = getelementptr inbounds i32, ptr %0, i64 3
%22 = load i32, ptr %21, align 4, !tbaa !2
%23 = mul i32 %22, %1
%24 = lshr i32 %23, 15
%25 = icmp ult i32 %24, 255
%26 = select i1 %25, i32 %24, i32 255
store i32 %26, ptr %21, align 4, !tbaa !2
ret void
}
define void @store_i8(ptr nocapture %0, i32 %1, i32 %2) {
; CHECK-LABEL: @store_i8(
; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i8>, ptr [[TMP0:%.*]], align 1, !tbaa [[TBAA4:![0-9]+]]
; CHECK-NEXT: [[TMP5:%.*]] = zext <4 x i8> [[TMP4]] to <4 x i32>
; CHECK-NEXT: [[TMP6:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1:%.*]], i64 0
; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP8:%.*]] = mul <4 x i32> [[TMP7]], [[TMP5]]
; CHECK-NEXT: [[TMP9:%.*]] = lshr <4 x i32> [[TMP8]], <i32 15, i32 15, i32 15, i32 15>
; CHECK-NEXT: [[TMP10:%.*]] = call <4 x i32> @llvm.umin.v4i32(<4 x i32> [[TMP9]], <4 x i32> <i32 255, i32 255, i32 255, i32 255>)
; CHECK-NEXT: [[TMP11:%.*]] = trunc nuw <4 x i32> [[TMP10]] to <4 x i8>
; CHECK-NEXT: store <4 x i8> [[TMP11]], ptr [[TMP0]], align 1, !tbaa [[TBAA4]]
; CHECK-NEXT: ret void
;
%4 = load i8, ptr %0, align 1, !tbaa !6
%5 = zext i8 %4 to i32
%6 = mul i32 %5, %1
%7 = lshr i32 %6, 15
%8 = icmp ult i32 %7, 255
%9 = select i1 %8, i32 %7, i32 255
%10 = trunc i32 %9 to i8
store i8 %10, ptr %0, align 1, !tbaa !6
%11 = getelementptr inbounds i8, ptr %0, i64 1
%12 = load i8, ptr %11, align 1, !tbaa !6
%13 = zext i8 %12 to i32
%14 = mul i32 %13, %1
%15 = lshr i32 %14, 15
%16 = icmp ult i32 %15, 255
%17 = select i1 %16, i32 %15, i32 255
%18 = trunc i32 %17 to i8
store i8 %18, ptr %11, align 1, !tbaa !6
%19 = getelementptr inbounds i8, ptr %0, i64 2
%20 = load i8, ptr %19, align 1, !tbaa !6
%21 = zext i8 %20 to i32
%22 = mul i32 %21, %1
%23 = lshr i32 %22, 15
%24 = icmp ult i32 %23, 255
%25 = select i1 %24, i32 %23, i32 255
%26 = trunc i32 %25 to i8
store i8 %26, ptr %19, align 1, !tbaa !6
%27 = getelementptr inbounds i8, ptr %0, i64 3
%28 = load i8, ptr %27, align 1, !tbaa !6
%29 = zext i8 %28 to i32
%30 = mul i32 %29, %1
%31 = lshr i32 %30, 15
%32 = icmp ult i32 %31, 255
%33 = select i1 %32, i32 %31, i32 255
%34 = trunc i32 %33 to i8
store i8 %34, ptr %27, align 1, !tbaa !6
ret void
}
define void @store_i64(ptr nocapture %0, i32 %1, i32 %2) {
; CHECK-LABEL: @store_i64(
; CHECK-NEXT: [[TMP4:%.*]] = zext i32 [[TMP1:%.*]] to i64
; CHECK-NEXT: [[TMP5:%.*]] = load <4 x i64>, ptr [[TMP0:%.*]], align 8, !tbaa [[TBAA5:![0-9]+]]
; CHECK-NEXT: [[TMP6:%.*]] = insertelement <4 x i64> poison, i64 [[TMP4]], i64 0
; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <4 x i64> [[TMP6]], <4 x i64> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP8:%.*]] = mul <4 x i64> [[TMP5]], [[TMP7]]
; CHECK-NEXT: [[TMP9:%.*]] = lshr <4 x i64> [[TMP8]], <i64 15, i64 15, i64 15, i64 15>
; CHECK-NEXT: [[TMP10:%.*]] = trunc <4 x i64> [[TMP9]] to <4 x i32>
; CHECK-NEXT: [[TMP11:%.*]] = icmp ult <4 x i32> [[TMP10]], <i32 255, i32 255, i32 255, i32 255>
; CHECK-NEXT: [[TMP12:%.*]] = and <4 x i64> [[TMP9]], <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
; CHECK-NEXT: [[TMP13:%.*]] = select <4 x i1> [[TMP11]], <4 x i64> [[TMP12]], <4 x i64> <i64 255, i64 255, i64 255, i64 255>
; CHECK-NEXT: store <4 x i64> [[TMP13]], ptr [[TMP0]], align 8, !tbaa [[TBAA5]]
; CHECK-NEXT: ret void
;
%4 = zext i32 %1 to i64
%5 = load i64, ptr %0, align 8, !tbaa !7
%6 = mul i64 %5, %4
%7 = lshr i64 %6, 15
%8 = trunc i64 %7 to i32
%9 = icmp ult i32 %8, 255
%10 = and i64 %7, 4294967295
%11 = select i1 %9, i64 %10, i64 255
store i64 %11, ptr %0, align 8, !tbaa !7
%12 = getelementptr inbounds i64, ptr %0, i64 1
%13 = load i64, ptr %12, align 8, !tbaa !7
%14 = mul i64 %13, %4
%15 = lshr i64 %14, 15
%16 = trunc i64 %15 to i32
%17 = icmp ult i32 %16, 255
%18 = and i64 %15, 4294967295
%19 = select i1 %17, i64 %18, i64 255
store i64 %19, ptr %12, align 8, !tbaa !7
%20 = getelementptr inbounds i64, ptr %0, i64 2
%21 = load i64, ptr %20, align 8, !tbaa !7
%22 = mul i64 %21, %4
%23 = lshr i64 %22, 15
%24 = trunc i64 %23 to i32
%25 = icmp ult i32 %24, 255
%26 = and i64 %23, 4294967295
%27 = select i1 %25, i64 %26, i64 255
store i64 %27, ptr %20, align 8, !tbaa !7
%28 = getelementptr inbounds i64, ptr %0, i64 3
%29 = load i64, ptr %28, align 8, !tbaa !7
%30 = mul i64 %29, %4
%31 = lshr i64 %30, 15
%32 = trunc i64 %31 to i32
%33 = icmp ult i32 %32, 255
%34 = and i64 %31, 4294967295
%35 = select i1 %33, i64 %34, i64 255
store i64 %35, ptr %28, align 8, !tbaa !7
ret void
}
!2 = !{!3, !3, i64 0}
!3 = !{!"int", !4, i64 0}
!4 = !{!"omnipotent char", !5, i64 0}
!5 = !{!"Simple C++ TBAA"}
!6 = !{!4, !4, i64 0}
!7 = !{!8, !8, i64 0}
!8 = !{!"long", !4, i64 0}
Reference in New Issue View Git Blame Copy Permalink