Skip to content

[AArch64] Enable UseFixedOverScalableIfEqualCost for more Cortex-x cpus. #122807

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jan 20, 2025
Merged

[AArch64] Enable UseFixedOverScalableIfEqualCost for more Cortex-x cpus. #122807

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e3b1bb1
[AArch64] Enable UseFixedOverScalableIfEqualCost for more Cortex-x cpus.
davemgreen Jan 13, 2025
c9009fd
Cleanup test
davemgreen Jan 15, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions llvm/lib/Target/AArch64/AArch64Processors.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -229,6 +229,7 @@ def TuneX2 : SubtargetFeature<"cortex-x2", "ARMProcFamily", "CortexX2",
FeatureALULSLFast,
FeaturePostRAScheduler,
FeatureEnableSelectOptimize,
FeatureUseFixedOverScalableIfEqualCost,
FeaturePredictableSelectIsExpensive]>;

def TuneX3 : SubtargetFeature<"cortex-x3", "ARMProcFamily", "CortexX3",
Expand All @@ -238,6 +239,7 @@ def TuneX3 : SubtargetFeature<"cortex-x3", "ARMProcFamily", "CortexX3",
FeatureFuseAES,
FeaturePostRAScheduler,
FeatureEnableSelectOptimize,
FeatureUseFixedOverScalableIfEqualCost,
FeaturePredictableSelectIsExpensive]>;

def TuneX4 : SubtargetFeature<"cortex-x4", "ARMProcFamily", "CortexX4",
Expand All @@ -247,6 +249,7 @@ def TuneX4 : SubtargetFeature<"cortex-x4", "ARMProcFamily", "CortexX4",
FeatureFuseAES,
FeaturePostRAScheduler,
FeatureEnableSelectOptimize,
FeatureUseFixedOverScalableIfEqualCost,
FeaturePredictableSelectIsExpensive]>;

def TuneX925 : SubtargetFeature<"cortex-x925", "ARMProcFamily",
Expand All @@ -256,6 +259,7 @@ def TuneX925 : SubtargetFeature<"cortex-x925", "ARMProcFamily",
FeatureFuseAES,
FeaturePostRAScheduler,
FeatureEnableSelectOptimize,
FeatureUseFixedOverScalableIfEqualCost,
FeaturePredictableSelectIsExpensive]>;

def TuneA64FX : SubtargetFeature<"a64fx", "ARMProcFamily", "A64FX",
Expand Down
43 changes: 12 additions & 31 deletions llvm/test/Transforms/LoopVectorize/AArch64/prefer-fixed-if-equal-to-scalable.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,14 +1,21 @@
; RUN: opt -S < %s -passes=loop-vectorize -force-target-instruction-cost=1 | FileCheck %s
; RUN: opt -S < %s -passes=loop-vectorize -mcpu=generic -mattr=+sve2 | FileCheck %s --check-prefix=CHECK-GENERIC
; RUN: opt -S < %s -passes=loop-vectorize -mcpu=neoverse-v2 | FileCheck %s --check-prefix=CHECK-PREFERFIXED
; RUN: opt -S < %s -passes=loop-vectorize -mcpu=cortex-x2 | FileCheck %s --check-prefix=CHECK-PREFERFIXED
; RUN: opt -S < %s -passes=loop-vectorize -mcpu=cortex-x3 | FileCheck %s --check-prefix=CHECK-PREFERFIXED
; RUN: opt -S < %s -passes=loop-vectorize -mcpu=cortex-x4 | FileCheck %s --check-prefix=CHECK-PREFERFIXED
; RUN: opt -S < %s -passes=loop-vectorize -mcpu=cortex-x925 | FileCheck %s --check-prefix=CHECK-PREFERFIXED

target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128-Fn32"
target triple = "aarch64-unknown-linux-gnu"

@a = dso_local local_unnamed_addr global [32000 x float] zeroinitializer, align 64
@b = dso_local local_unnamed_addr global [32000 x float] zeroinitializer, align 64

define void @NeoverseV2() #0 {
; CHECK-LABEL: define void @NeoverseV2(
; CHECK: store <4 x float>
define void @test() #0 {
; CHECK-GENERIC-LABEL: define void @test(
; CHECK-GENERIC: store <vscale x 4 x float>
; CHECK-PREFERFIXED-LABEL: define void @test(
; CHECK-PREFERFIXED: store <4 x float>
;
entry:
br label %for.body
Expand All @@ -31,30 +38,4 @@ for.body:
br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
}

define void @GenericCPU() #1 {
; CHECK-LABEL: define void @GenericCPU(
; CHECK: store <vscale x 4 x float>
;
entry:
br label %for.body

for.cond.cleanup:
ret void

for.body:
%indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
%arrayidx = getelementptr inbounds [32000 x float], ptr @a, i64 0, i64 %indvars.iv
%0 = load float, ptr %arrayidx, align 4
%arrayidx2 = getelementptr inbounds [32000 x float], ptr @b, i64 0, i64 %indvars.iv
%1 = load float, ptr %arrayidx2, align 4
%add = fadd fast float %1, %0
%2 = add nuw nsw i64 %indvars.iv, 16000
%arrayidx5 = getelementptr inbounds [32000 x float], ptr @a, i64 0, i64 %2
store float %add, ptr %arrayidx5, align 4
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
%exitcond.not = icmp eq i64 %indvars.iv.next, 16000
br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
}

attributes #0 = { vscale_range(1,16) "target-cpu"="neoverse-v2" "target-features"="+sve,+sve2,+v9a" }
attributes #1 = { vscale_range(1,16) "target-cpu"="generic" "target-features"="+sve,+v9a" }
attributes #0 = { vscale_range(1,16) }
Loading