This is an archive of the discontinued LLVM Phabricator instance.

Differential D74431

[X86] Don't disable code in combineHorizontalPredicateResult just because we have avx512
ClosedPublic

Authored by craig.topper on Feb 11 2020, 11:31 AM.

Details

Reviewers
RKSimon
spatel
Commits
rG846d0ac43eb9: [X86] Don't disable code in combineHorizontalPredicateResult just because we…
Summary

We aren't doing a good job of optimizing AVX512 outside of this code. So remove the bail out for AVX512 and replace with a FIXME. This at least gets us the AVX2 codegen.

Diff Detail

Event Timeline

craig.topper created this revision.Feb 11 2020, 11:31 AM
Herald added a project: Restricted Project. · View Herald TranscriptFeb 11 2020, 11:31 AM
Herald added a subscriber: hiraditya. · View Herald Transcript
spatel accepted this revision.Feb 11 2020, 12:38 PM

LGTM

llvm/test/CodeGen/X86/vector-compare-all_of.ll
31–35

This and the similar v2i64 case may warrant revisiting. We assumed movmsk was generally cheaper even on a 2-way reduction, but it depends how recent uarch have implemented it.

This revision is now accepted and ready to land.Feb 11 2020, 12:38 PM
Closed by commit rG846d0ac43eb9: [X86] Don't disable code in combineHorizontalPredicateResult just because we… (authored by craig.topper). · Explain WhyFeb 11 2020, 2:40 PM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
llvm/
lib/
Target/
X86/
5 lines
test/
CodeGen/
X86/
197 lines
165 lines

Diff 244011

llvm/lib/Target/X86/X86ISelLowering.cpp

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 37,353 Lines • ▼ Show 20 Lines if (TLI.isTypeLegal(MatchVT)) {
} }
EVT MovmskVT = EVT::getIntegerVT(*DAG.getContext(), NumElts); EVT MovmskVT = EVT::getIntegerVT(*DAG.getContext(), NumElts);
Movmsk = combineBitcastvxi1(DAG, MovmskVT, Match, DL, Subtarget); Movmsk = combineBitcastvxi1(DAG, MovmskVT, Match, DL, Subtarget);
} }
if (!Movmsk) if (!Movmsk)
return SDValue(); return SDValue();
Movmsk = DAG.getZExtOrTrunc(Movmsk, DL, NumElts > 32 ? MVT::i64 : MVT::i32); Movmsk = DAG.getZExtOrTrunc(Movmsk, DL, NumElts > 32 ? MVT::i64 : MVT::i32);
} else { } else {
// Bail with AVX512VL (which uses predicate registers). // FIXME: Better handling of k-registers or 512-bit vectors?
if (Subtarget.hasVLX())
return SDValue();
unsigned MatchSizeInBits = Match.getValueSizeInBits(); unsigned MatchSizeInBits = Match.getValueSizeInBits();
if (!(MatchSizeInBits == 128 || if (!(MatchSizeInBits == 128 ||
(MatchSizeInBits == 256 && Subtarget.hasAVX()))) (MatchSizeInBits == 256 && Subtarget.hasAVX())))
return SDValue(); return SDValue();
// Make sure this isn't a vector of 1 element. The perf win from using // Make sure this isn't a vector of 1 element. The perf win from using
// MOVMSK diminishes with less elements in the reduction, but it is // MOVMSK diminishes with less elements in the reduction, but it is
// generally better to get the comparison over to the GPRs as soon as // generally better to get the comparison over to the GPRs as soon as
▲ Show 20 Lines • Show All 10,575 Lines • Show Last 20 Lines

llvm/test/CodeGen/X86/vector-compare-all_of.ll

Show All 22 Lines
; AVX-NEXT: cmpl $3, %ecx ; AVX-NEXT: cmpl $3, %ecx
; AVX-NEXT: sete %al ; AVX-NEXT: sete %al
; AVX-NEXT: negq %rax ; AVX-NEXT: negq %rax
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v2f64_sext: ; AVX512-LABEL: test_v2f64_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltpd %xmm0, %xmm1, %xmm0 ; AVX512-NEXT: vcmpltpd %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskpd %xmm0, %ecx
; AVX512-NEXT: vandpd %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vmovq %xmm0, %rax ; AVX512-NEXT: cmpl $3, %ecx
; AVX512-NEXT: sete %al
; AVX512-NEXT: negq %rax
spatelUnsubmitted
Not Done
ReplyInline Actions

This and the similar v2i64 case may warrant revisiting. We assumed movmsk was generally cheaper even on a 2-way reduction, but it depends how recent uarch have implemented it.

spatel: This and the similar v2i64 case may warrant revisiting. We assumed movmsk was generally cheaper…
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <2 x double> %a0, %a1 %c = fcmp ogt <2 x double> %a0, %a1
%s = sext <2 x i1> %c to <2 x i64> %s = sext <2 x i1> %c to <2 x i64>
%1 = shufflevector <2 x i64> %s, <2 x i64> undef, <2 x i32> <i32 1, i32 undef> %1 = shufflevector <2 x i64> %s, <2 x i64> undef, <2 x i32> <i32 1, i32 undef>
%2 = and <2 x i64> %s, %1 %2 = and <2 x i64> %s, %1
%3 = extractelement <2 x i64> %2, i32 0 %3 = extractelement <2 x i64> %2, i32 0
ret i64 %3 ret i64 %3
} }
Show All 20 Lines
; AVX-NEXT: sete %al ; AVX-NEXT: sete %al
; AVX-NEXT: negq %rax ; AVX-NEXT: negq %rax
; AVX-NEXT: vzeroupper ; AVX-NEXT: vzeroupper
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v4f64_sext: ; AVX512-LABEL: test_v4f64_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltpd %ymm0, %ymm1, %ymm0 ; AVX512-NEXT: vcmpltpd %ymm0, %ymm1, %ymm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vmovmskpd %ymm0, %ecx
; AVX512-NEXT: vandpd %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: cmpl $15, %ecx
; AVX512-NEXT: vandpd %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vmovq %xmm0, %rax ; AVX512-NEXT: negq %rax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <4 x double> %a0, %a1 %c = fcmp ogt <4 x double> %a0, %a1
%s = sext <4 x i1> %c to <4 x i64> %s = sext <4 x i1> %c to <4 x i64>
%1 = shufflevector <4 x i64> %s, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i64> %s, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = and <4 x i64> %s, %1 %2 = and <4 x i64> %s, %1
%3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = and <4 x i64> %2, %3 %4 = and <4 x i64> %2, %3
Show All 27 Lines
; AVX-NEXT: vzeroupper ; AVX-NEXT: vzeroupper
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v4f64_legal_sext: ; AVX512-LABEL: test_v4f64_legal_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltpd %ymm0, %ymm1, %k1 ; AVX512-NEXT: vcmpltpd %ymm0, %ymm1, %k1
; AVX512-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; AVX512-NEXT: vmovdqa32 %xmm0, %xmm0 {%k1} {z} ; AVX512-NEXT: vmovdqa32 %xmm0, %xmm0 {%k1} {z}
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskps %xmm0, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: cmpl $15, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vmovd %xmm0, %eax ; AVX512-NEXT: negq %rax
; AVX512-NEXT: cltq
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <4 x double> %a0, %a1 %c = fcmp ogt <4 x double> %a0, %a1
%s = sext <4 x i1> %c to <4 x i32> %s = sext <4 x i1> %c to <4 x i32>
%1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = and <4 x i32> %s, %1 %2 = and <4 x i32> %s, %1
%3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = and <4 x i32> %2, %3 %4 = and <4 x i32> %2, %3
Show All 21 Lines
; AVX-NEXT: cmpl $15, %ecx ; AVX-NEXT: cmpl $15, %ecx
; AVX-NEXT: sete %al ; AVX-NEXT: sete %al
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v4f32_sext: ; AVX512-LABEL: test_v4f32_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltps %xmm0, %xmm1, %xmm0 ; AVX512-NEXT: vcmpltps %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskps %xmm0, %ecx
; AVX512-NEXT: vandps %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: cmpl $15, %ecx
; AVX512-NEXT: vandps %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vmovd %xmm0, %eax ; AVX512-NEXT: negl %eax
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <4 x float> %a0, %a1 %c = fcmp ogt <4 x float> %a0, %a1
%s = sext <4 x i1> %c to <4 x i32> %s = sext <4 x i1> %c to <4 x i32>
%1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = and <4 x i32> %s, %1 %2 = and <4 x i32> %s, %1
%3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = and <4 x i32> %2, %3 %4 = and <4 x i32> %2, %3
%5 = extractelement <4 x i32> %4, i32 0 %5 = extractelement <4 x i32> %4, i32 0
Show All 22 Lines
; AVX-NEXT: sete %al ; AVX-NEXT: sete %al
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: vzeroupper ; AVX-NEXT: vzeroupper
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v8f32_sext: ; AVX512-LABEL: test_v8f32_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltps %ymm0, %ymm1, %ymm0 ; AVX512-NEXT: vcmpltps %ymm0, %ymm1, %ymm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vmovmskps %ymm0, %ecx
; AVX512-NEXT: vandps %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: cmpl $255, %ecx
; AVX512-NEXT: vandps %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vandps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <8 x float> %a0, %a1 %c = fcmp ogt <8 x float> %a0, %a1
%s = sext <8 x i1> %c to <8 x i32> %s = sext <8 x i1> %c to <8 x i32>
%1 = shufflevector <8 x i32> %s, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <8 x i32> %s, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = and <8 x i32> %s, %1 %2 = and <8 x i32> %s, %1
%3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = and <8 x i32> %2, %3 %4 = and <8 x i32> %2, %3
Show All 28 Lines
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: vzeroupper ; AVX-NEXT: vzeroupper
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v8f32_legal_sext: ; AVX512-LABEL: test_v8f32_legal_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltps %ymm0, %ymm1, %k0 ; AVX512-NEXT: vcmpltps %ymm0, %ymm1, %k0
; AVX512-NEXT: vpmovm2w %k0, %xmm0 ; AVX512-NEXT: vpmovm2w %k0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vpmovmskb %xmm0, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: cmpl $65535, %ecx # imm = 0xFFFF
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: cwtl
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <8 x float> %a0, %a1 %c = fcmp ogt <8 x float> %a0, %a1
%s = sext <8 x i1> %c to <8 x i16> %s = sext <8 x i1> %c to <8 x i16>
%1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = and <8 x i16> %s, %1 %2 = and <8 x i16> %s, %1
%3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = and <8 x i16> %2, %3 %4 = and <8 x i16> %2, %3
Show All 23 Lines
; AVX-NEXT: cmpl $3, %ecx ; AVX-NEXT: cmpl $3, %ecx
; AVX-NEXT: sete %al ; AVX-NEXT: sete %al
; AVX-NEXT: negq %rax ; AVX-NEXT: negq %rax
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v2i64_sext: ; AVX512-LABEL: test_v2i64_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskpd %xmm0, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vmovq %xmm0, %rax ; AVX512-NEXT: cmpl $3, %ecx
; AVX512-NEXT: sete %al
; AVX512-NEXT: negq %rax
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <2 x i64> %a0, %a1 %c = icmp sgt <2 x i64> %a0, %a1
%s = sext <2 x i1> %c to <2 x i64> %s = sext <2 x i1> %c to <2 x i64>
%1 = shufflevector <2 x i64> %s, <2 x i64> undef, <2 x i32> <i32 1, i32 undef> %1 = shufflevector <2 x i64> %s, <2 x i64> undef, <2 x i32> <i32 1, i32 undef>
%2 = and <2 x i64> %s, %1 %2 = and <2 x i64> %s, %1
%3 = extractelement <2 x i64> %2, i32 0 %3 = extractelement <2 x i64> %2, i32 0
ret i64 %3 ret i64 %3
} }
Show All 35 Lines
; AVX2-NEXT: sete %al ; AVX2-NEXT: sete %al
; AVX2-NEXT: negq %rax ; AVX2-NEXT: negq %rax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v4i64_sext: ; AVX512-LABEL: test_v4i64_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 ; AVX512-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vmovmskpd %ymm0, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: cmpl $15, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vmovq %xmm0, %rax ; AVX512-NEXT: negq %rax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <4 x i64> %a0, %a1 %c = icmp sgt <4 x i64> %a0, %a1
%s = sext <4 x i1> %c to <4 x i64> %s = sext <4 x i1> %c to <4 x i64>
%1 = shufflevector <4 x i64> %s, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i64> %s, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = and <4 x i64> %s, %1 %2 = and <4 x i64> %s, %1
%3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = and <4 x i64> %2, %3 %4 = and <4 x i64> %2, %3
▲ Show 20 Lines • Show All 42 Lines • ▼ Show 20 Lines
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v4i64_legal_sext: ; AVX512-LABEL: test_v4i64_legal_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtq %ymm1, %ymm0, %k1 ; AVX512-NEXT: vpcmpgtq %ymm1, %ymm0, %k1
; AVX512-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; AVX512-NEXT: vmovdqa32 %xmm0, %xmm0 {%k1} {z} ; AVX512-NEXT: vmovdqa32 %xmm0, %xmm0 {%k1} {z}
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskps %xmm0, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: cmpl $15, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vmovd %xmm0, %eax ; AVX512-NEXT: negq %rax
; AVX512-NEXT: cltq
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <4 x i64> %a0, %a1 %c = icmp sgt <4 x i64> %a0, %a1
%s = sext <4 x i1> %c to <4 x i32> %s = sext <4 x i1> %c to <4 x i32>
%1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = and <4 x i32> %s, %1 %2 = and <4 x i32> %s, %1
%3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = and <4 x i32> %2, %3 %4 = and <4 x i32> %2, %3
Show All 21 Lines
; AVX-NEXT: cmpl $15, %ecx ; AVX-NEXT: cmpl $15, %ecx
; AVX-NEXT: sete %al ; AVX-NEXT: sete %al
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v4i32_sext: ; AVX512-LABEL: test_v4i32_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskps %xmm0, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: cmpl $15, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vmovd %xmm0, %eax ; AVX512-NEXT: negl %eax
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <4 x i32> %a0, %a1 %c = icmp sgt <4 x i32> %a0, %a1
%s = sext <4 x i1> %c to <4 x i32> %s = sext <4 x i1> %c to <4 x i32>
%1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = and <4 x i32> %s, %1 %2 = and <4 x i32> %s, %1
%3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = and <4 x i32> %2, %3 %4 = and <4 x i32> %2, %3
%5 = extractelement <4 x i32> %4, i32 0 %5 = extractelement <4 x i32> %4, i32 0
Show All 37 Lines
; AVX2-NEXT: sete %al ; AVX2-NEXT: sete %al
; AVX2-NEXT: negl %eax ; AVX2-NEXT: negl %eax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v8i32_sext: ; AVX512-LABEL: test_v8i32_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 ; AVX512-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vmovmskps %ymm0, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: cmpl $255, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <8 x i32> %a0, %a1 %c = icmp sgt <8 x i32> %a0, %a1
%s = sext <8 x i1> %c to <8 x i32> %s = sext <8 x i1> %c to <8 x i32>
%1 = shufflevector <8 x i32> %s, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <8 x i32> %s, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = and <8 x i32> %s, %1 %2 = and <8 x i32> %s, %1
%3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = and <8 x i32> %2, %3 %4 = and <8 x i32> %2, %3
▲ Show 20 Lines • Show All 43 Lines • ▼ Show 20 Lines
; AVX2-NEXT: negl %eax ; AVX2-NEXT: negl %eax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v8i32_legal_sext: ; AVX512-LABEL: test_v8i32_legal_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtd %ymm1, %ymm0, %k0 ; AVX512-NEXT: vpcmpgtd %ymm1, %ymm0, %k0
; AVX512-NEXT: vpmovm2w %k0, %xmm0 ; AVX512-NEXT: vpmovm2w %k0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vpmovmskb %xmm0, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: cmpl $65535, %ecx # imm = 0xFFFF
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: cwtl
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <8 x i32> %a0, %a1 %c = icmp sgt <8 x i32> %a0, %a1
%s = sext <8 x i1> %c to <8 x i16> %s = sext <8 x i1> %c to <8 x i16>
%1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = and <8 x i16> %s, %1 %2 = and <8 x i16> %s, %1
%3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = and <8 x i16> %2, %3 %4 = and <8 x i16> %2, %3
Show All 25 Lines
; AVX-NEXT: sete %al ; AVX-NEXT: sete %al
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: # kill: def $ax killed $ax killed $eax ; AVX-NEXT: # kill: def $ax killed $ax killed $eax
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v8i16_sext: ; AVX512-LABEL: test_v8i16_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vpmovmskb %xmm0, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: cmpl $65535, %ecx # imm = 0xFFFF
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax ; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <8 x i16> %a0, %a1 %c = icmp sgt <8 x i16> %a0, %a1
%s = sext <8 x i1> %c to <8 x i16> %s = sext <8 x i1> %c to <8 x i16>
%1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = and <8 x i16> %s, %1 %2 = and <8 x i16> %s, %1
%3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = and <8 x i16> %2, %3 %4 = and <8 x i16> %2, %3
▲ Show 20 Lines • Show All 43 Lines • ▼ Show 20 Lines
; AVX2-NEXT: negl %eax ; AVX2-NEXT: negl %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax ; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v16i16_sext: ; AVX512-LABEL: test_v16i16_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 ; AVX512-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vpmovmskb %ymm0, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: xorl %eax, %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: cmpl $-1, %ecx
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: sete %al
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax ; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <16 x i16> %a0, %a1 %c = icmp sgt <16 x i16> %a0, %a1
%s = sext <16 x i1> %c to <16 x i16> %s = sext <16 x i1> %c to <16 x i16>
%1 = shufflevector <16 x i16> %s, <16 x i16> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <16 x i16> %s, <16 x i16> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = and <16 x i16> %s, %1 %2 = and <16 x i16> %s, %1
%3 = shufflevector <16 x i16> %2, <16 x i16> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
▲ Show 20 Lines • Show All 49 Lines • ▼ Show 20 Lines
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax ; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v16i16_legal_sext: ; AVX512-LABEL: test_v16i16_legal_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtw %ymm1, %ymm0, %k0 ; AVX512-NEXT: vpcmpgtw %ymm1, %ymm0, %k0
; AVX512-NEXT: vpmovm2b %k0, %xmm0 ; AVX512-NEXT: vpmovm2b %k0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vpmovmskb %xmm0, %eax
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: cmpl $65535, %eax # imm = 0xFFFF
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: sete %al
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negb %al
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpextrb $0, %xmm0, %eax
; AVX512-NEXT: movsbl %al, %eax ; AVX512-NEXT: movsbl %al, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax ; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <16 x i16> %a0, %a1 %c = icmp sgt <16 x i16> %a0, %a1
%s = sext <16 x i1> %c to <16 x i8> %s = sext <16 x i1> %c to <16 x i8>
%1 = shufflevector <16 x i8> %s, <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <16 x i8> %s, <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = and <16 x i8> %s, %1 %2 = and <16 x i8> %s, %1
Show All 25 Lines
; AVX-NEXT: cmpl $65535, %eax # imm = 0xFFFF ; AVX-NEXT: cmpl $65535, %eax # imm = 0xFFFF
; AVX-NEXT: sete %al ; AVX-NEXT: sete %al
; AVX-NEXT: negb %al ; AVX-NEXT: negb %al
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v16i8_sext: ; AVX512-LABEL: test_v16i8_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vpmovmskb %xmm0, %eax
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: cmpl $65535, %eax # imm = 0xFFFF
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: sete %al
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negb %al
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpextrb $0, %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <16 x i8> %a0, %a1 %c = icmp sgt <16 x i8> %a0, %a1
%s = sext <16 x i1> %c to <16 x i8> %s = sext <16 x i1> %c to <16 x i8>
%1 = shufflevector <16 x i8> %s, <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <16 x i8> %s, <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = and <16 x i8> %s, %1 %2 = and <16 x i8> %s, %1
%3 = shufflevector <16 x i8> %2, <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <16 x i8> %2, <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = and <16 x i8> %2, %3 %4 = and <16 x i8> %2, %3
%5 = shufflevector <16 x i8> %4, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %5 = shufflevector <16 x i8> %4, <16 x i8> undef, <16 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
Show All 38 Lines
; AVX2-NEXT: sete %al ; AVX2-NEXT: sete %al
; AVX2-NEXT: negb %al ; AVX2-NEXT: negb %al
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v32i8_sext: ; AVX512-LABEL: test_v32i8_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 ; AVX512-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vpmovmskb %ymm0, %eax
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: cmpl $-1, %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: sete %al
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negb %al
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX512-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpextrb $0, %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <32 x i8> %a0, %a1 %c = icmp sgt <32 x i8> %a0, %a1
%s = sext <32 x i1> %c to <32 x i8> %s = sext <32 x i1> %c to <32 x i8>
%1 = shufflevector <32 x i8> %s, <32 x i8> undef, <32 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <32 x i8> %s, <32 x i8> undef, <32 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = and <32 x i8> %s, %1 %2 = and <32 x i8> %s, %1
%3 = shufflevector <32 x i8> %2, <32 x i8> undef, <32 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <32 x i8> %2, <32 x i8> undef, <32 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = and <32 x i8> %2, %3 %4 = and <32 x i8> %2, %3
▲ Show 20 Lines • Show All 513 Lines • Show Last 20 Lines

llvm/test/CodeGen/X86/vector-compare-any_of.ll

Show All 18 Lines
; AVX-NEXT: vmovmskpd %xmm0, %eax ; AVX-NEXT: vmovmskpd %xmm0, %eax
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: sbbq %rax, %rax ; AVX-NEXT: sbbq %rax, %rax
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v2f64_sext: ; AVX512-LABEL: test_v2f64_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltpd %xmm0, %xmm1, %xmm0 ; AVX512-NEXT: vcmpltpd %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskpd %xmm0, %eax
; AVX512-NEXT: vorpd %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vmovq %xmm0, %rax ; AVX512-NEXT: sbbq %rax, %rax
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <2 x double> %a0, %a1 %c = fcmp ogt <2 x double> %a0, %a1
%s = sext <2 x i1> %c to <2 x i64> %s = sext <2 x i1> %c to <2 x i64>
%1 = shufflevector <2 x i64> %s, <2 x i64> undef, <2 x i32> <i32 1, i32 undef> %1 = shufflevector <2 x i64> %s, <2 x i64> undef, <2 x i32> <i32 1, i32 undef>
%2 = or <2 x i64> %s, %1 %2 = or <2 x i64> %s, %1
%3 = extractelement <2 x i64> %2, i32 0 %3 = extractelement <2 x i64> %2, i32 0
ret i64 %3 ret i64 %3
} }
Show All 16 Lines
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: sbbq %rax, %rax ; AVX-NEXT: sbbq %rax, %rax
; AVX-NEXT: vzeroupper ; AVX-NEXT: vzeroupper
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v4f64_sext: ; AVX512-LABEL: test_v4f64_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltpd %ymm0, %ymm1, %ymm0 ; AVX512-NEXT: vcmpltpd %ymm0, %ymm1, %ymm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vmovmskpd %ymm0, %eax
; AVX512-NEXT: vorpd %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: sbbq %rax, %rax
; AVX512-NEXT: vorpd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovq %xmm0, %rax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <4 x double> %a0, %a1 %c = fcmp ogt <4 x double> %a0, %a1
%s = sext <4 x i1> %c to <4 x i64> %s = sext <4 x i1> %c to <4 x i64>
%1 = shufflevector <4 x i64> %s, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i64> %s, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = or <4 x i64> %s, %1 %2 = or <4 x i64> %s, %1
%3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = or <4 x i64> %2, %3 %4 = or <4 x i64> %2, %3
Show All 23 Lines
; AVX-NEXT: vzeroupper ; AVX-NEXT: vzeroupper
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v4f64_legal_sext: ; AVX512-LABEL: test_v4f64_legal_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltpd %ymm0, %ymm1, %k1 ; AVX512-NEXT: vcmpltpd %ymm0, %ymm1, %k1
; AVX512-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; AVX512-NEXT: vmovdqa32 %xmm0, %xmm0 {%k1} {z} ; AVX512-NEXT: vmovdqa32 %xmm0, %xmm0 {%k1} {z}
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskps %xmm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: sbbq %rax, %rax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: cltq
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <4 x double> %a0, %a1 %c = fcmp ogt <4 x double> %a0, %a1
%s = sext <4 x i1> %c to <4 x i32> %s = sext <4 x i1> %c to <4 x i32>
%1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = or <4 x i32> %s, %1 %2 = or <4 x i32> %s, %1
%3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = or <4 x i32> %2, %3 %4 = or <4 x i32> %2, %3
Show All 17 Lines
; AVX-NEXT: vmovmskps %xmm0, %eax ; AVX-NEXT: vmovmskps %xmm0, %eax
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: sbbl %eax, %eax ; AVX-NEXT: sbbl %eax, %eax
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v4f32_sext: ; AVX512-LABEL: test_v4f32_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltps %xmm0, %xmm1, %xmm0 ; AVX512-NEXT: vcmpltps %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskps %xmm0, %eax
; AVX512-NEXT: vorps %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: sbbl %eax, %eax
; AVX512-NEXT: vorps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <4 x float> %a0, %a1 %c = fcmp ogt <4 x float> %a0, %a1
%s = sext <4 x i1> %c to <4 x i32> %s = sext <4 x i1> %c to <4 x i32>
%1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = or <4 x i32> %s, %1 %2 = or <4 x i32> %s, %1
%3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = or <4 x i32> %2, %3 %4 = or <4 x i32> %2, %3
%5 = extractelement <4 x i32> %4, i32 0 %5 = extractelement <4 x i32> %4, i32 0
Show All 18 Lines
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: sbbl %eax, %eax ; AVX-NEXT: sbbl %eax, %eax
; AVX-NEXT: vzeroupper ; AVX-NEXT: vzeroupper
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v8f32_sext: ; AVX512-LABEL: test_v8f32_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltps %ymm0, %ymm1, %ymm0 ; AVX512-NEXT: vcmpltps %ymm0, %ymm1, %ymm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vmovmskps %ymm0, %eax
; AVX512-NEXT: vorps %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: sbbl %eax, %eax
; AVX512-NEXT: vorps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX512-NEXT: vorps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <8 x float> %a0, %a1 %c = fcmp ogt <8 x float> %a0, %a1
%s = sext <8 x i1> %c to <8 x i32> %s = sext <8 x i1> %c to <8 x i32>
%1 = shufflevector <8 x i32> %s, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <8 x i32> %s, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = or <8 x i32> %s, %1 %2 = or <8 x i32> %s, %1
%3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = or <8 x i32> %2, %3 %4 = or <8 x i32> %2, %3
Show All 24 Lines
; AVX-NEXT: sbbl %eax, %eax ; AVX-NEXT: sbbl %eax, %eax
; AVX-NEXT: vzeroupper ; AVX-NEXT: vzeroupper
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v8f32_legal_sext: ; AVX512-LABEL: test_v8f32_legal_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vcmpltps %ymm0, %ymm1, %k0 ; AVX512-NEXT: vcmpltps %ymm0, %ymm1, %k0
; AVX512-NEXT: vpmovm2w %k0, %xmm0 ; AVX512-NEXT: vpmovm2w %k0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vpmovmskb %xmm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: sbbl %eax, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: cwtl
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = fcmp ogt <8 x float> %a0, %a1 %c = fcmp ogt <8 x float> %a0, %a1
%s = sext <8 x i1> %c to <8 x i16> %s = sext <8 x i1> %c to <8 x i16>
%1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = or <8 x i16> %s, %1 %2 = or <8 x i16> %s, %1
%3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = or <8 x i16> %2, %3 %4 = or <8 x i16> %2, %3
Show All 19 Lines
; AVX-NEXT: vmovmskpd %xmm0, %eax ; AVX-NEXT: vmovmskpd %xmm0, %eax
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: sbbq %rax, %rax ; AVX-NEXT: sbbq %rax, %rax
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v2i64_sext: ; AVX512-LABEL: test_v2i64_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskpd %xmm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vmovq %xmm0, %rax ; AVX512-NEXT: sbbq %rax, %rax
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <2 x i64> %a0, %a1 %c = icmp sgt <2 x i64> %a0, %a1
%s = sext <2 x i1> %c to <2 x i64> %s = sext <2 x i1> %c to <2 x i64>
%1 = shufflevector <2 x i64> %s, <2 x i64> undef, <2 x i32> <i32 1, i32 undef> %1 = shufflevector <2 x i64> %s, <2 x i64> undef, <2 x i32> <i32 1, i32 undef>
%2 = or <2 x i64> %s, %1 %2 = or <2 x i64> %s, %1
%3 = extractelement <2 x i64> %2, i32 0 %3 = extractelement <2 x i64> %2, i32 0
ret i64 %3 ret i64 %3
} }
Show All 29 Lines
; AVX2-NEXT: negl %eax ; AVX2-NEXT: negl %eax
; AVX2-NEXT: sbbq %rax, %rax ; AVX2-NEXT: sbbq %rax, %rax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v4i64_sext: ; AVX512-LABEL: test_v4i64_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 ; AVX512-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vmovmskpd %ymm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: sbbq %rax, %rax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovq %xmm0, %rax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <4 x i64> %a0, %a1 %c = icmp sgt <4 x i64> %a0, %a1
%s = sext <4 x i1> %c to <4 x i64> %s = sext <4 x i1> %c to <4 x i64>
%1 = shufflevector <4 x i64> %s, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i64> %s, <4 x i64> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = or <4 x i64> %s, %1 %2 = or <4 x i64> %s, %1
%3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = or <4 x i64> %2, %3 %4 = or <4 x i64> %2, %3
Show All 36 Lines
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v4i64_legal_sext: ; AVX512-LABEL: test_v4i64_legal_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtq %ymm1, %ymm0, %k1 ; AVX512-NEXT: vpcmpgtq %ymm1, %ymm0, %k1
; AVX512-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; AVX512-NEXT: vmovdqa32 %xmm0, %xmm0 {%k1} {z} ; AVX512-NEXT: vmovdqa32 %xmm0, %xmm0 {%k1} {z}
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskps %xmm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: sbbq %rax, %rax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: cltq
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <4 x i64> %a0, %a1 %c = icmp sgt <4 x i64> %a0, %a1
%s = sext <4 x i1> %c to <4 x i32> %s = sext <4 x i1> %c to <4 x i32>
%1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = or <4 x i32> %s, %1 %2 = or <4 x i32> %s, %1
%3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = or <4 x i32> %2, %3 %4 = or <4 x i32> %2, %3
Show All 17 Lines
; AVX-NEXT: vmovmskps %xmm0, %eax ; AVX-NEXT: vmovmskps %xmm0, %eax
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: sbbl %eax, %eax ; AVX-NEXT: sbbl %eax, %eax
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v4i32_sext: ; AVX512-LABEL: test_v4i32_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vmovmskps %xmm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: sbbl %eax, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <4 x i32> %a0, %a1 %c = icmp sgt <4 x i32> %a0, %a1
%s = sext <4 x i1> %c to <4 x i32> %s = sext <4 x i1> %c to <4 x i32>
%1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> %1 = shufflevector <4 x i32> %s, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
%2 = or <4 x i32> %s, %1 %2 = or <4 x i32> %s, %1
%3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef> %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = or <4 x i32> %2, %3 %4 = or <4 x i32> %2, %3
%5 = extractelement <4 x i32> %4, i32 0 %5 = extractelement <4 x i32> %4, i32 0
Show All 31 Lines
; AVX2-NEXT: negl %eax ; AVX2-NEXT: negl %eax
; AVX2-NEXT: sbbl %eax, %eax ; AVX2-NEXT: sbbl %eax, %eax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v8i32_sext: ; AVX512-LABEL: test_v8i32_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 ; AVX512-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vmovmskps %ymm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: sbbl %eax, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <8 x i32> %a0, %a1 %c = icmp sgt <8 x i32> %a0, %a1
%s = sext <8 x i1> %c to <8 x i32> %s = sext <8 x i1> %c to <8 x i32>
%1 = shufflevector <8 x i32> %s, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <8 x i32> %s, <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = or <8 x i32> %s, %1 %2 = or <8 x i32> %s, %1
%3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = or <8 x i32> %2, %3 %4 = or <8 x i32> %2, %3
Show All 37 Lines
; AVX2-NEXT: sbbl %eax, %eax ; AVX2-NEXT: sbbl %eax, %eax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v8i32_legal_sext: ; AVX512-LABEL: test_v8i32_legal_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtd %ymm1, %ymm0, %k0 ; AVX512-NEXT: vpcmpgtd %ymm1, %ymm0, %k0
; AVX512-NEXT: vpmovm2w %k0, %xmm0 ; AVX512-NEXT: vpmovm2w %k0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vpmovmskb %xmm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: sbbl %eax, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: cwtl
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <8 x i32> %a0, %a1 %c = icmp sgt <8 x i32> %a0, %a1
%s = sext <8 x i1> %c to <8 x i16> %s = sext <8 x i1> %c to <8 x i16>
%1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = or <8 x i16> %s, %1 %2 = or <8 x i16> %s, %1
%3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = or <8 x i16> %2, %3 %4 = or <8 x i16> %2, %3
Show All 21 Lines
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: sbbl %eax, %eax ; AVX-NEXT: sbbl %eax, %eax
; AVX-NEXT: # kill: def $ax killed $ax killed $eax ; AVX-NEXT: # kill: def $ax killed $ax killed $eax
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v8i16_sext: ; AVX512-LABEL: test_v8i16_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vpmovmskb %xmm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: sbbl %eax, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax ; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <8 x i16> %a0, %a1 %c = icmp sgt <8 x i16> %a0, %a1
%s = sext <8 x i1> %c to <8 x i16> %s = sext <8 x i1> %c to <8 x i16>
%1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <8 x i16> %s, <8 x i16> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = or <8 x i16> %s, %1 %2 = or <8 x i16> %s, %1
%3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = or <8 x i16> %2, %3 %4 = or <8 x i16> %2, %3
Show All 37 Lines
; AVX2-NEXT: sbbl %eax, %eax ; AVX2-NEXT: sbbl %eax, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax ; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v16i16_sext: ; AVX512-LABEL: test_v16i16_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 ; AVX512-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vpmovmskb %ymm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: sbbl %eax, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax ; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <16 x i16> %a0, %a1 %c = icmp sgt <16 x i16> %a0, %a1
%s = sext <16 x i1> %c to <16 x i16> %s = sext <16 x i1> %c to <16 x i16>
%1 = shufflevector <16 x i16> %s, <16 x i16> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <16 x i16> %s, <16 x i16> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = or <16 x i16> %s, %1 %2 = or <16 x i16> %s, %1
%3 = shufflevector <16 x i16> %2, <16 x i16> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
▲ Show 20 Lines • Show All 43 Lines • ▼ Show 20 Lines
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax ; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v16i16_legal_sext: ; AVX512-LABEL: test_v16i16_legal_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtw %ymm1, %ymm0, %k0 ; AVX512-NEXT: vpcmpgtw %ymm1, %ymm0, %k0
; AVX512-NEXT: vpmovm2b %k0, %xmm0 ; AVX512-NEXT: vpmovm2b %k0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vpmovmskb %xmm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: sbbl %eax, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpextrb $0, %xmm0, %eax
; AVX512-NEXT: movsbl %al, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax ; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <16 x i16> %a0, %a1 %c = icmp sgt <16 x i16> %a0, %a1
%s = sext <16 x i1> %c to <16 x i8> %s = sext <16 x i1> %c to <16 x i8>
%1 = shufflevector <16 x i8> %s, <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <16 x i8> %s, <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = or <16 x i8> %s, %1 %2 = or <16 x i8> %s, %1
%3 = shufflevector <16 x i8> %2, <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <16 x i8> %2, <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
Show All 24 Lines
; AVX-NEXT: negl %eax ; AVX-NEXT: negl %eax
; AVX-NEXT: sbbl %eax, %eax ; AVX-NEXT: sbbl %eax, %eax
; AVX-NEXT: # kill: def $al killed $al killed $eax ; AVX-NEXT: # kill: def $al killed $al killed $eax
; AVX-NEXT: retq ; AVX-NEXT: retq
; ;
; AVX512-LABEL: test_v16i8_sext: ; AVX512-LABEL: test_v16i8_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: vpmovmskb %xmm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3] ; AVX512-NEXT: sbbl %eax, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpextrb $0, %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax ; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <16 x i8> %a0, %a1 %c = icmp sgt <16 x i8> %a0, %a1
%s = sext <16 x i1> %c to <16 x i8> %s = sext <16 x i1> %c to <16 x i8>
%1 = shufflevector <16 x i8> %s, <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <16 x i8> %s, <16 x i8> undef, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = or <16 x i8> %s, %1 %2 = or <16 x i8> %s, %1
%3 = shufflevector <16 x i8> %2, <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <16 x i8> %2, <16 x i8> undef, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%4 = or <16 x i8> %2, %3 %4 = or <16 x i8> %2, %3
Show All 39 Lines
; AVX2-NEXT: sbbl %eax, %eax ; AVX2-NEXT: sbbl %eax, %eax
; AVX2-NEXT: # kill: def $al killed $al killed $eax ; AVX2-NEXT: # kill: def $al killed $al killed $eax
; AVX2-NEXT: vzeroupper ; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq ; AVX2-NEXT: retq
; ;
; AVX512-LABEL: test_v32i8_sext: ; AVX512-LABEL: test_v32i8_sext:
; AVX512: # %bb.0: ; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 ; AVX512-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1 ; AVX512-NEXT: vpmovmskb %ymm0, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0 ; AVX512-NEXT: negl %eax
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] ; AVX512-NEXT: sbbl %eax, %eax
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpextrb $0, %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax ; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: vzeroupper ; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%c = icmp sgt <32 x i8> %a0, %a1 %c = icmp sgt <32 x i8> %a0, %a1
%s = sext <32 x i1> %c to <32 x i8> %s = sext <32 x i1> %c to <32 x i8>
%1 = shufflevector <32 x i8> %s, <32 x i8> undef, <32 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %1 = shufflevector <32 x i8> %s, <32 x i8> undef, <32 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
%2 = or <32 x i8> %s, %1 %2 = or <32 x i8> %s, %1
%3 = shufflevector <32 x i8> %2, <32 x i8> undef, <32 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> %3 = shufflevector <32 x i8> %2, <32 x i8> undef, <32 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
▲ Show 20 Lines • Show All 512 Lines • Show Last 20 Lines