This is an archive of the discontinued LLVM Phabricator instance.

Differential D149050

[Vectorize] Pre-commit tests for D148905
ClosedPublic

Authored by foad on Apr 24 2023, 3:08 AM.

Details

Reviewers
foad
fhahn
Commits
rG3237497d01e8: [Vectorize] Pre-commit tests for D148905

Diff Detail

Event Timeline

foad created this revision.Apr 24 2023, 3:08 AM
Herald added a project: Restricted Project. · View Herald TranscriptApr 24 2023, 3:08 AM
Herald added a subscriber: StephenFan. · View Herald Transcript
foad requested review of this revision.Apr 24 2023, 3:08 AM
Herald added a project: Restricted Project. · View Herald TranscriptApr 24 2023, 3:08 AM
Herald added subscribers: llvm-commits, • pcwang-thead. · View Herald Transcript
foad accepted this revision.Apr 24 2023, 3:09 AM
foad added a child revision: D148905: [Vectorize] Fix vectorization, scalarization and folding of llvm.is.fpclass.

Self-approving

This revision is now accepted and ready to land.Apr 24 2023, 3:09 AM
fhahn accepted this revision.Apr 24 2023, 3:10 AM
fhahn added a subscriber: fhahn.

LGTM, thanks!

llvm/test/Transforms/LoopVectorize/is_fpclass.ll
25

might be a bit more readable if this is at the end of the function, named exit.

bb2 -> loop for better readability

foad updated this revision to Diff 516344.Apr 24 2023, 3:25 AM

Better readability.

foad marked an inline comment as done.Apr 24 2023, 3:26 AM
Harbormaster completed remote builds in B227679: Diff 516344.Apr 24 2023, 4:13 AM
This revision was landed with ongoing or failed builds.Apr 24 2023, 5:42 AM
Closed by commit rG3237497d01e8: [Vectorize] Pre-commit tests for D148905 (authored by foad). · Explain Why
This revision was automatically updated to reflect the committed changes.
foad added a commit: rG3237497d01e8: [Vectorize] Pre-commit tests for D148905.

Revision Contents

PathSize
llvm/
test/
Transforms/
LoopVectorize/
39 lines
SLPVectorizer/
24 lines
Scalarizer/
12 lines

Diff 516369

llvm/test/Transforms/LoopVectorize/is_fpclass.ll

  • This file was added.
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
; RUN: opt < %s -passes=loop-vectorize -force-vector-width=2 -S | FileCheck %s
define void @d() {
; CHECK-LABEL: define void @d() {
; CHECK-NEXT: entry:
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[I:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[I7:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[I3:%.*]] = load float, ptr null, align 4
; CHECK-NEXT: [[I4:%.*]] = getelementptr float, ptr @d, i64 [[I]]
; CHECK-NEXT: [[I5:%.*]] = tail call i1 @llvm.is.fpclass.f32(float 0.000000e+00, i32 0)
; CHECK-NEXT: [[I6:%.*]] = select i1 [[I5]], float 0.000000e+00, float 0.000000e+00
; CHECK-NEXT: store float [[I6]], ptr [[I4]], align 4
; CHECK-NEXT: [[I7]] = add i64 [[I]], 1
; CHECK-NEXT: [[I8:%.*]] = icmp eq i64 [[I7]], 0
; CHECK-NEXT: br i1 [[I8]], label [[EXIT:%.*]], label [[LOOP]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
br label %loop
loop:
%i = phi i64 [ 0, %entry ], [ %i7, %loop ]
fhahnUnsubmitted
Done
ReplyInline Actions

might be a bit more readable if this is at the end of the function, named exit.

bb2 -> loop for better readability

fhahn: might be a bit more readable if this is at the end of the function, named `exit`. `bb2` ->…
%i3 = load float, ptr null, align 4
%i4 = getelementptr float, ptr @d, i64 %i
%i5 = tail call i1 @llvm.is.fpclass.f32(float 0.0, i32 0)
%i6 = select i1 %i5, float 0.0, float 0.0
store float %i6, ptr %i4, align 4
%i7 = add i64 %i, 1
%i8 = icmp eq i64 %i7, 0
br i1 %i8, label %exit, label %loop
exit:
ret void
}
declare i1 @llvm.is.fpclass.f32(float, i32 immarg)

llvm/test/Transforms/SLPVectorizer/is_fpclass.ll

  • This file was added.
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
; RUN: opt -S -passes=slp-vectorizer < %s | FileCheck %s
define <2 x i1> @scalarize_is_fpclass(<2 x float> %x) {
; CHECK-LABEL: define <2 x i1> @scalarize_is_fpclass
; CHECK-SAME: (<2 x float> [[X:%.*]]) {
; CHECK-NEXT: [[X_I0:%.*]] = extractelement <2 x float> [[X]], i32 0
; CHECK-NEXT: [[ISFPCLASS_I0:%.*]] = call i1 @llvm.is.fpclass.f32(float [[X_I0]], i32 123)
; CHECK-NEXT: [[X_I1:%.*]] = extractelement <2 x float> [[X]], i32 1
; CHECK-NEXT: [[ISFPCLASS_I1:%.*]] = call i1 @llvm.is.fpclass.f32(float [[X_I1]], i32 123)
; CHECK-NEXT: [[ISFPCLASS_UPTO0:%.*]] = insertelement <2 x i1> poison, i1 [[ISFPCLASS_I0]], i32 0
; CHECK-NEXT: [[ISFPCLASS:%.*]] = insertelement <2 x i1> [[ISFPCLASS_UPTO0]], i1 [[ISFPCLASS_I1]], i32 1
; CHECK-NEXT: ret <2 x i1> [[ISFPCLASS]]
;
%x.i0 = extractelement <2 x float> %x, i32 0
%isfpclass.i0 = call i1 @llvm.is.fpclass.f32(float %x.i0, i32 123)
%x.i1 = extractelement <2 x float> %x, i32 1
%isfpclass.i1 = call i1 @llvm.is.fpclass.f32(float %x.i1, i32 123)
%isfpclass.upto0 = insertelement <2 x i1> poison, i1 %isfpclass.i0, i32 0
%isfpclass = insertelement <2 x i1> %isfpclass.upto0, i1 %isfpclass.i1, i32 1
ret <2 x i1> %isfpclass
}
declare i1 @llvm.is.fpclass.f32(float, i32 immarg)

llvm/test/Transforms/Scalarizer/intrinsics.ll

Show All 22 Lines
; Binary int plus constant scalar operand ; Binary int plus constant scalar operand
declare <2 x i32> @llvm.smul.fix.sat.v2i32(<2 x i32>, <2 x i32>, i32) declare <2 x i32> @llvm.smul.fix.sat.v2i32(<2 x i32>, <2 x i32>, i32)
declare <2 x i32> @llvm.umul.fix.sat.v2i32(<2 x i32>, <2 x i32>, i32) declare <2 x i32> @llvm.umul.fix.sat.v2i32(<2 x i32>, <2 x i32>, i32)
declare <2 x i32> @llvm.fptosi.sat.v2i32.v2f32(<2 x float>) declare <2 x i32> @llvm.fptosi.sat.v2i32.v2f32(<2 x float>)
declare <2 x i32> @llvm.fptoui.sat.v2i32.v2f32(<2 x float>) declare <2 x i32> @llvm.fptoui.sat.v2i32.v2f32(<2 x float>)
; Bool return type, overloaded on fp operand type
declare <2 x i1> @llvm.is.fpclass(<2 x float>, i32)
define <2 x float> @scalarize_sqrt_v2f32(<2 x float> %x) #0 { define <2 x float> @scalarize_sqrt_v2f32(<2 x float> %x) #0 {
; CHECK-LABEL: @scalarize_sqrt_v2f32( ; CHECK-LABEL: @scalarize_sqrt_v2f32(
; CHECK-NEXT: [[X_I0:%.*]] = extractelement <2 x float> [[X:%.*]], i32 0 ; CHECK-NEXT: [[X_I0:%.*]] = extractelement <2 x float> [[X:%.*]], i32 0
; CHECK-NEXT: [[SQRT_I0:%.*]] = call float @llvm.sqrt.f32(float [[X_I0]]) ; CHECK-NEXT: [[SQRT_I0:%.*]] = call float @llvm.sqrt.f32(float [[X_I0]])
; CHECK-NEXT: [[X_I1:%.*]] = extractelement <2 x float> [[X]], i32 1 ; CHECK-NEXT: [[X_I1:%.*]] = extractelement <2 x float> [[X]], i32 1
; CHECK-NEXT: [[SQRT_I1:%.*]] = call float @llvm.sqrt.f32(float [[X_I1]]) ; CHECK-NEXT: [[SQRT_I1:%.*]] = call float @llvm.sqrt.f32(float [[X_I1]])
; CHECK-NEXT: [[SQRT_UPTO0:%.*]] = insertelement <2 x float> poison, float [[SQRT_I0]], i32 0 ; CHECK-NEXT: [[SQRT_UPTO0:%.*]] = insertelement <2 x float> poison, float [[SQRT_I0]], i32 0
▲ Show 20 Lines • Show All 162 Lines • ▼ Show 20 Lines
; CHECK-NEXT: [[SAT_I1:%.*]] = call i32 @llvm.fptoui.sat.i32.f32(float [[X_I1]]) ; CHECK-NEXT: [[SAT_I1:%.*]] = call i32 @llvm.fptoui.sat.i32.f32(float [[X_I1]])
; CHECK-NEXT: [[SAT_UPTO0:%.*]] = insertelement <2 x i32> poison, i32 [[SAT_I0]], i32 0 ; CHECK-NEXT: [[SAT_UPTO0:%.*]] = insertelement <2 x i32> poison, i32 [[SAT_I0]], i32 0
; CHECK-NEXT: [[SAT:%.*]] = insertelement <2 x i32> [[SAT_UPTO0]], i32 [[SAT_I1]], i32 1 ; CHECK-NEXT: [[SAT:%.*]] = insertelement <2 x i32> [[SAT_UPTO0]], i32 [[SAT_I1]], i32 1
; CHECK-NEXT: ret <2 x i32> [[SAT]] ; CHECK-NEXT: ret <2 x i32> [[SAT]]
; ;
%sat = call <2 x i32> @llvm.fptoui.sat.v2i32.v2f32(<2 x float> %x) %sat = call <2 x i32> @llvm.fptoui.sat.v2i32.v2f32(<2 x float> %x)
ret <2 x i32> %sat ret <2 x i32> %sat
} }
define <2 x i1> @scalarize_is_fpclass(<2 x float> %x) #0 {
; CHECK-LABEL: @scalarize_is_fpclass(
; CHECK-NEXT: [[ISFPCLASS:%.*]] = call <2 x i1> @llvm.is.fpclass.v2f32(<2 x float> [[X:%.*]], i32 123)
; CHECK-NEXT: ret <2 x i1> [[ISFPCLASS]]
;
%isfpclass = call <2 x i1> @llvm.is.fpclass(<2 x float> %x, i32 123)
ret <2 x i1> %isfpclass
}