This is an archive of the discontinued LLVM Phabricator instance.

Differential D53671

[X86] Remove some uarch tuning flags from KNL that look to have been inherited from SNB/IVB incorrectly
ClosedPublic

Authored by craig.topper on Oct 24 2018, 2:13 PM.

Details

Reviewers
RKSimon
spatel
Commits
rG5d787ac4becc: [X86] Remove some uarch tuning flags from KNL that look to have been inherited…
rL345285: [X86] Remove some uarch tuning flags from KNL that look to have been inherited…
Summary

KNL is based on a modified Silvermont core so I don't think these features apply. I think the LEA flag is probably also wrong, but I'm less sure as I barely understand the 3 LEA flags we have currently.

Diff Detail

Event Timeline

craig.topper created this revision.Oct 24 2018, 2:13 PM
RKSimon accepted this revision.Oct 25 2018, 6:11 AM

LGTM

This revision is now accepted and ready to land.Oct 25 2018, 6:11 AM
Closed by commit rL345285: [X86] Remove some uarch tuning flags from KNL that look to have been inherited… (authored by ctopper). · Explain WhyOct 25 2018, 10:31 AM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
lib/
Target/
X86/
4 lines
test/
CodeGen/
X86/
4 lines

Diff 170981

lib/Target/X86/X86.td

Show First 20 Lines • Show All 817 Lines • ▼ Show 20 Linesdef KNLFeatures : ProcessorFeatures<[], [
FeatureCMPXCHG16B, FeatureCMPXCHG16B,
FeaturePOPCNT, FeaturePOPCNT,
FeatureSlowDivide64, FeatureSlowDivide64,
FeaturePCLMUL, FeaturePCLMUL,
FeatureXSAVE, FeatureXSAVE,
FeatureXSAVEOPT, FeatureXSAVEOPT,
FeatureLAHFSAHF, FeatureLAHFSAHF,
FeatureSlow3OpsLEA, FeatureSlow3OpsLEA,
FeatureFastScalarFSQRT,
FeatureFastSHLDRotate,
FeatureSlowIncDec, FeatureSlowIncDec,
FeatureMergeToThreeWayBranch,
FeatureMacroFusion,
FeatureRDRAND, FeatureRDRAND,
FeatureF16C, FeatureF16C,
FeatureFSGSBase, FeatureFSGSBase,
FeatureAVX512, FeatureAVX512,
FeatureERI, FeatureERI,
FeatureCDI, FeatureCDI,
FeaturePFI, FeaturePFI,
FeaturePREFETCHWT1, FeaturePREFETCHWT1,
▲ Show 20 Lines • Show All 435 Lines • Show Last 20 Lines

test/CodeGen/X86/avx512-select.ll

Show All 10 Lines
; X86-NEXT: # %bb.1: ; X86-NEXT: # %bb.1:
; X86-NEXT: vmovdqa64 %zmm0, %zmm1 ; X86-NEXT: vmovdqa64 %zmm0, %zmm1
; X86-NEXT: .LBB0_2: ; X86-NEXT: .LBB0_2:
; X86-NEXT: vpxorq %zmm1, %zmm0, %zmm0 ; X86-NEXT: vpxorq %zmm1, %zmm0, %zmm0
; X86-NEXT: retl ; X86-NEXT: retl
; ;
; X64-LABEL: select00: ; X64-LABEL: select00:
; X64: # %bb.0: ; X64: # %bb.0:
; X64-NEXT: vpxor %xmm1, %xmm1, %xmm1
; X64-NEXT: cmpl $255, %edi ; X64-NEXT: cmpl $255, %edi
; X64-NEXT: vpxor %xmm1, %xmm1, %xmm1
; X64-NEXT: je .LBB0_2 ; X64-NEXT: je .LBB0_2
; X64-NEXT: # %bb.1: ; X64-NEXT: # %bb.1:
; X64-NEXT: vmovdqa64 %zmm0, %zmm1 ; X64-NEXT: vmovdqa64 %zmm0, %zmm1
; X64-NEXT: .LBB0_2: ; X64-NEXT: .LBB0_2:
; X64-NEXT: vpxorq %zmm1, %zmm0, %zmm0 ; X64-NEXT: vpxorq %zmm1, %zmm0, %zmm0
; X64-NEXT: retq ; X64-NEXT: retq
%cmpres = icmp eq i32 %a, 255 %cmpres = icmp eq i32 %a, 255
%selres = select i1 %cmpres, <16 x i32> zeroinitializer, <16 x i32> %b %selres = select i1 %cmpres, <16 x i32> zeroinitializer, <16 x i32> %b
Show All 10 Lines
; X86-NEXT: # %bb.1: ; X86-NEXT: # %bb.1:
; X86-NEXT: vmovdqa64 %zmm0, %zmm1 ; X86-NEXT: vmovdqa64 %zmm0, %zmm1
; X86-NEXT: .LBB1_2: ; X86-NEXT: .LBB1_2:
; X86-NEXT: vpxorq %zmm1, %zmm0, %zmm0 ; X86-NEXT: vpxorq %zmm1, %zmm0, %zmm0
; X86-NEXT: retl ; X86-NEXT: retl
; ;
; X64-LABEL: select01: ; X64-LABEL: select01:
; X64: # %bb.0: ; X64: # %bb.0:
; X64-NEXT: vpxor %xmm1, %xmm1, %xmm1
; X64-NEXT: cmpl $255, %edi ; X64-NEXT: cmpl $255, %edi
; X64-NEXT: vpxor %xmm1, %xmm1, %xmm1
; X64-NEXT: je .LBB1_2 ; X64-NEXT: je .LBB1_2
; X64-NEXT: # %bb.1: ; X64-NEXT: # %bb.1:
; X64-NEXT: vmovdqa64 %zmm0, %zmm1 ; X64-NEXT: vmovdqa64 %zmm0, %zmm1
; X64-NEXT: .LBB1_2: ; X64-NEXT: .LBB1_2:
; X64-NEXT: vpxorq %zmm1, %zmm0, %zmm0 ; X64-NEXT: vpxorq %zmm1, %zmm0, %zmm0
; X64-NEXT: retq ; X64-NEXT: retq
%cmpres = icmp eq i32 %a, 255 %cmpres = icmp eq i32 %a, 255
%selres = select i1 %cmpres, <8 x i64> zeroinitializer, <8 x i64> %b %selres = select i1 %cmpres, <8 x i64> zeroinitializer, <8 x i64> %b
▲ Show 20 Lines • Show All 292 Lines • Show Last 20 Lines