Skip to content

[InstCombine] Relax guard against FP min/max in select fold #143144

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

Merged
merged 3 commits into from
Jun 7, 2025

Conversation

Acthinks
Copy link
Contributor

@Acthinks Acthinks commented Jun 6, 2025

FCmp's commutativity predicates do not work with min/max semantics

Closes #142711

FCmp's equality and commutativity predicates do not work with min/max semantics

Closes llvm#142711
@Acthinks Acthinks requested a review from nikic as a code owner June 6, 2025 14:00
@llvmbot llvmbot added llvm:instcombine Covers the InstCombine, InstSimplify and AggressiveInstCombine passes llvm:transforms labels Jun 6, 2025
@llvmbot
Copy link
Member

llvmbot commented Jun 6, 2025

@llvm/pr-subscribers-llvm-transforms

Author: Acthink Yang (Acthink-Yang)

Changes

FCmp's equality and commutativity predicates do not work with min/max semantics

Closes #142711


Full diff: https://github.com/llvm/llvm-project/pull/143144.diff

2 Files Affected:

  • (modified) llvm/lib/Transforms/InstCombine/InstructionCombining.cpp (+2-1)
  • (modified) llvm/test/Transforms/InstCombine/fcmp-select.ll (+11)
diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
index 439a86d951a83..c85abd9b26d61 100644
--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -1729,7 +1729,8 @@ Instruction *InstCombinerImpl::FoldOpIntoSelect(Instruction &Op, SelectInst *SI,
   if (auto *CI = dyn_cast<FCmpInst>(SI->getCondition())) {
     if (CI->hasOneUse()) {
       Value *Op0 = CI->getOperand(0), *Op1 = CI->getOperand(1);
-      if ((TV == Op0 && FV == Op1) || (FV == Op0 && TV == Op1))
+      if (((TV == Op0 && FV == Op1) || (FV == Op0 && TV == Op1)) &&
+          !CI->isEquality() && !CI->isCommutative())
         return nullptr;
     }
   }
diff --git a/llvm/test/Transforms/InstCombine/fcmp-select.ll b/llvm/test/Transforms/InstCombine/fcmp-select.ll
index 408bc1cdc268f..bf08fbfc27302 100644
--- a/llvm/test/Transforms/InstCombine/fcmp-select.ll
+++ b/llvm/test/Transforms/InstCombine/fcmp-select.ll
@@ -268,3 +268,14 @@ define i1 @test_fcmp_select_var_const_unordered(double %x, double %y) {
   %cmp2 = fcmp ugt double %sel, 0x3E80000000000000
   ret i1 %cmp2
 }
+
+define i1 @test_fcmp_ord_select_fcmp_oeq_var_const(double %x) {
+; CHECK-LABEL:    @test_fcmp_ord_select_fcmp_oeq_var_const(
+; CHECK-NEXT:     [[CMP1:%.*]] = fcmp oeq double [[X:%.*]], 0.000000e+00
+; CHECK-NEXT:     ret i1 [[CMP1]]
+;
+  %cmp1 = fcmp ord double %x, 0.000000e+00
+  %sel = select i1 %cmp1, double %x, double 0.000000e+00
+  %cmp2 = fcmp oeq double %sel, 1.000000e+00
+  ret i1 %cmp2
+}

Co-authored-by: Yingwei Zheng <[email protected]>
Copy link
Member

@dtcxzyw dtcxzyw left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM w/ a nit.

@dtcxzyw dtcxzyw merged commit ca4dfca into llvm:main Jun 7, 2025
7 checks passed
@Acthinks Acthinks deleted the instrcombine-fcmp branch June 7, 2025 06:50
rorth pushed a commit to rorth/llvm-project that referenced this pull request Jun 11, 2025
)

FCmp's commutativity predicates do not work with min/max semantics

Closes llvm#142711
DhruvSrivastavaX pushed a commit to DhruvSrivastavaX/lldb-for-aix that referenced this pull request Jun 12, 2025
)

FCmp's commutativity predicates do not work with min/max semantics

Closes llvm#142711
tomtor pushed a commit to tomtor/llvm-project that referenced this pull request Jun 14, 2025
)

FCmp's commutativity predicates do not work with min/max semantics

Closes llvm#142711
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
llvm:instcombine Covers the InstCombine, InstSimplify and AggressiveInstCombine passes llvm:transforms
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Missed Optimization: fold fcmp ord + select before fcmp oeq (x, C) into a single fcmp oeq (x, C)
4 participants