caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
12ccdd67aaf56bd05ab3375d81c58dfd5a2db02c
clang-p2996/llvm/test/Transforms/SLPVectorizer/X86/crash_cmpop.ll
Alexey Bataev 2ac5ebedea [SLP]Do not emit extract elements for insertelements users, replace with shuffles directly. 
SLP vectorizer emits extracts for externally used vectorized scalars and
estimates the cost for each such extract. But in many cases these
scalars are input for insertelement instructions, forming buildvector,
and instead of extractelement/insertelement pair we can emit/cost
estimate shuffle(s) cost and generate series of shuffles, which can be
further optimized.

Tested using test-suite (+SPEC2017), the tests passed, SLP was able to
generate/vectorize more instructions in many cases and it allowed to reduce
number of re-vectorization attempts (where we could try to vectorize
buildector insertelements again and again).

Differential Revision: https://reviews.llvm.org/D107966
2022-05-23 07:06:45 -07:00

96 lines
5.6 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -basic-aa -slp-vectorizer -S | FileCheck %s
; RUN: opt < %s -basic-aa -slp-vectorizer -S -mattr=+avx | FileCheck %s
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) {
; CHECK-LABEL: @testfunc(
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[ACC1_056:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[ADD13:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = phi <2 x float> [ zeroinitializer, [[ENTRY]] ], [ [[TMP19:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 [[INDVARS_IV]]
; CHECK-NEXT: [[TMP1:%.*]] = load float, float* [[ARRAYIDX]], align 4
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[DEST:%.*]], i64 [[INDVARS_IV]]
; CHECK-NEXT: store float [[ACC1_056]], float* [[ARRAYIDX2]], align 4
; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x float> poison, float [[TMP1]], i32 0
; CHECK-NEXT: [[TMP3:%.*]] = insertelement <2 x float> [[TMP2]], float [[TMP1]], i32 1
; CHECK-NEXT: [[TMP4:%.*]] = fadd <2 x float> [[TMP0]], [[TMP3]]
; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <2 x float> [[TMP4]], <2 x float> poison, <2 x i32> <i32 1, i32 0>
; CHECK-NEXT: [[TMP5:%.*]] = fmul <2 x float> [[TMP0]], zeroinitializer
; CHECK-NEXT: [[TMP6:%.*]] = fadd <2 x float> [[TMP5]], [[SHUFFLE]]
; CHECK-NEXT: [[TMP7:%.*]] = fcmp olt <2 x float> [[TMP6]], <float 1.000000e+00, float 1.000000e+00>
; CHECK-NEXT: [[TMP8:%.*]] = select <2 x i1> [[TMP7]], <2 x float> [[TMP6]], <2 x float> <float 1.000000e+00, float 1.000000e+00>
; CHECK-NEXT: [[TMP9:%.*]] = fcmp olt <2 x float> [[TMP8]], <float -1.000000e+00, float -1.000000e+00>
; CHECK-NEXT: [[TMP10:%.*]] = fmul <2 x float> [[TMP8]], zeroinitializer
; CHECK-NEXT: [[TMP11:%.*]] = select <2 x i1> [[TMP9]], <2 x float> <float -0.000000e+00, float -0.000000e+00>, <2 x float> [[TMP10]]
; CHECK-NEXT: [[TMP12:%.*]] = extractelement <2 x float> [[TMP11]], i32 0
; CHECK-NEXT: [[TMP13:%.*]] = extractelement <2 x float> [[TMP11]], i32 1
; CHECK-NEXT: [[ADD13]] = fadd float [[TMP12]], [[TMP13]]
; CHECK-NEXT: [[TMP14:%.*]] = shufflevector <2 x float> [[TMP11]], <2 x float> poison, <2 x i32> <i32 1, i32 undef>
; CHECK-NEXT: [[TMP15:%.*]] = insertelement <2 x float> [[TMP14]], float [[ADD13]], i32 1
; CHECK-NEXT: [[TMP16:%.*]] = fcmp olt <2 x float> [[TMP15]], <float 1.000000e+00, float 1.000000e+00>
; CHECK-NEXT: [[TMP17:%.*]] = select <2 x i1> [[TMP16]], <2 x float> [[TMP15]], <2 x float> <float 1.000000e+00, float 1.000000e+00>
; CHECK-NEXT: [[TMP18:%.*]] = fcmp olt <2 x float> [[TMP17]], <float -1.000000e+00, float -1.000000e+00>
; CHECK-NEXT: [[TMP19]] = select <2 x i1> [[TMP18]], <2 x float> <float -1.000000e+00, float -1.000000e+00>, <2 x float> [[TMP17]]
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 32
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
; CHECK: for.end:
; CHECK-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
}
Reference in New Issue View Git Blame Copy Permalink