-
Notifications
You must be signed in to change notification settings - Fork 10.8k
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.
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
[InstCombine] Fold fpto{s|u}i non-norm
to zero
#85569
Merged
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
dtcxzyw
changed the title
[InstCombine] Fold \
[InstCombine] Fold Mar 17, 2024
fpto{s|u}i non-norm\
to zerofpto{s|u}i non-norm
to zero
@llvm/pr-subscribers-llvm-transforms Author: Yingwei Zheng (dtcxzyw) ChangesThis patch enables more optimization after canonicalizing Alive2: https://alive2.llvm.org/ce/z/ASM8tQ Full diff: https://github.com/llvm/llvm-project/pull/85569.diff 2 Files Affected:
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
index 381cd858d26293..b19da1e075819c 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -1923,10 +1923,26 @@ Instruction *InstCombinerImpl::foldItoFPtoI(CastInst &FI) {
return replaceInstUsesWith(FI, X);
}
+static Instruction *foldFPtoI(Instruction &FI, InstCombiner &IC) {
+ // fpto{u/s}i non-norm --> 0
+ FPClassTest Mask =
+ FI.getOpcode() == Instruction::FPToUI ? fcPosNormal : fcNormal;
+ KnownFPClass FPClass =
+ computeKnownFPClass(FI.getOperand(0), Mask, /*Depth=*/0,
+ IC.getSimplifyQuery().getWithInstruction(&FI));
+ if (FPClass.isKnownNever(Mask))
+ return IC.replaceInstUsesWith(FI, ConstantInt::getNullValue(FI.getType()));
+
+ return nullptr;
+}
+
Instruction *InstCombinerImpl::visitFPToUI(FPToUIInst &FI) {
if (Instruction *I = foldItoFPtoI(FI))
return I;
+ if (Instruction *I = foldFPtoI(FI, *this))
+ return I;
+
return commonCastTransforms(FI);
}
@@ -1934,6 +1950,9 @@ Instruction *InstCombinerImpl::visitFPToSI(FPToSIInst &FI) {
if (Instruction *I = foldItoFPtoI(FI))
return I;
+ if (Instruction *I = foldFPtoI(FI, *this))
+ return I;
+
return commonCastTransforms(FI);
}
diff --git a/llvm/test/Transforms/InstCombine/fpcast.ll b/llvm/test/Transforms/InstCombine/fpcast.ll
index 88ca556f4d8f8b..7eae44b6abfbec 100644
--- a/llvm/test/Transforms/InstCombine/fpcast.ll
+++ b/llvm/test/Transforms/InstCombine/fpcast.ll
@@ -347,3 +347,64 @@ define double @masked_uint_to_fpext3(i32 %x) {
%r = fpext float %f to double
ret double %r
}
+
+define i32 @fptosi_nonnorm(float nofpclass(norm) %x) {
+; CHECK-LABEL: @fptosi_nonnorm(
+; CHECK-NEXT: ret i32 0
+;
+ %ret = fptosi float %x to i32
+ ret i32 %ret
+}
+
+define i32 @fptoui_nonnorm(float nofpclass(pnorm) %x) {
+; CHECK-LABEL: @fptoui_nonnorm(
+; CHECK-NEXT: ret i32 0
+;
+ %ret = fptoui float %x to i32
+ ret i32 %ret
+}
+
+define i32 @fptosi_nonnorm_copysign(float %x) {
+; CHECK-LABEL: @fptosi_nonnorm_copysign(
+; CHECK-NEXT: ret i32 0
+;
+ %val = call float @llvm.copysign.f32(float 0.0, float %x)
+ %ret = fptosi float %val to i32
+ ret i32 %ret
+}
+
+define i32 @fptosi_nonnorm_fmul(float %x) {
+; CHECK-LABEL: @fptosi_nonnorm_fmul(
+; CHECK-NEXT: [[SEL:%.*]] = fmul float [[X:%.*]], 0.000000e+00
+; CHECK-NEXT: [[RET:%.*]] = fptosi float [[SEL]] to i32
+; CHECK-NEXT: ret i32 [[RET]]
+;
+ %sel = fmul float %x, 0.000000e+00
+ %ret = fptosi float %sel to i32
+ ret i32 %ret
+}
+
+define i32 @fptosi_select(i1 %cond) {
+; CHECK-LABEL: @fptosi_select(
+; CHECK-NEXT: [[RET:%.*]] = select i1 [[COND:%.*]], i32 1, i32 -1
+; CHECK-NEXT: ret i32 [[RET]]
+;
+ %sel = select i1 %cond, float 1.0, float -1.0
+ %ret = fptosi float %sel to i32
+ ret i32 %ret
+}
+
+define i32 @mul_pos_zero_convert(i32 %a) {
+; CHECK-LABEL: @mul_pos_zero_convert(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[FP:%.*]] = sitofp i32 [[A:%.*]] to float
+; CHECK-NEXT: [[RET:%.*]] = fmul float [[FP]], 0.000000e+00
+; CHECK-NEXT: [[CONV:%.*]] = fptosi float [[RET]] to i32
+; CHECK-NEXT: ret i32 [[CONV]]
+;
+entry:
+ %fp = sitofp i32 %a to float
+ %ret = fmul float %fp, 0.000000e+00
+ %conv = fptosi float %ret to i32
+ ret i32 %conv
+}
|
dtcxzyw
added a commit
to dtcxzyw/llvm-opt-benchmark
that referenced
this pull request
Mar 17, 2024
arsenm
reviewed
Mar 18, 2024
dtcxzyw
force-pushed
the
perf/fptoi-nonnorm
branch
from
March 19, 2024 08:28
929941c
to
442fe33
Compare
arsenm
approved these changes
Mar 19, 2024
chencha3
pushed a commit
to chencha3/llvm-project
that referenced
this pull request
Mar 23, 2024
This patch enables more optimization after canonicalizing `fmul X, 0.0` into a copysign. I decide to implement this fold in InstCombine because `computeKnownFPClass` may be expensive. Alive2: https://alive2.llvm.org/ce/z/ASM8tQ
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This patch enables more optimization after canonicalizing
fmul X, 0.0
into a copysign.I decide to implement this fold in InstCombine because
computeKnownFPClass
may be expensive.Alive2: https://alive2.llvm.org/ce/z/ASM8tQ