[Analysis][LV] Handle FMinNum/FMaxNum in RecurrenceDescriptor::getOpcode #152794
Conversation
Teach getOpcode to return FCmp for FMinNum/FMaxNum. Without these cases, passes that query the opcode for minNum/maxNum reductions can hit llvm_unreachable. Add a unit test to cover both enums.
|
Thank you for submitting a Pull Request (PR) to the LLVM Project! This PR will be automatically labeled and the relevant teams will be notified. If you wish to, you can add reviewers by using the "Reviewers" section on this page. If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers. If you have further questions, they may be answered by the LLVM GitHub User Guide. You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums. |
llvmbot
added
the
llvm:analysis
|
@llvm/pr-subscribers-llvm-analysis Author: Uyiosa Iyekekpolor (uyoyo0) ChangesModify Without these cases, querying the opcode for minNum/maxNum reductions can hit Full diff: https://github.com/llvm/llvm-project/pull/152794.diff 2 Files Affected:
diff --git a/llvm/lib/Analysis/IVDescriptors.cpp b/llvm/lib/Analysis/IVDescriptors.cpp
index 8be5de3bf356f..efb6a5e97bc2b 100644
--- a/llvm/lib/Analysis/IVDescriptors.cpp
+++ b/llvm/lib/Analysis/IVDescriptors.cpp
@@ -1228,6 +1228,8 @@ unsigned RecurrenceDescriptor::getOpcode(RecurKind Kind) {
return Instruction::ICmp;
case RecurKind::FMax:
case RecurKind::FMin:
+ case RecurKind::FMaxNum:
+ case RecurKind::FMinNum:
case RecurKind::FMaximum:
case RecurKind::FMinimum:
case RecurKind::FMaximumNum:
diff --git a/llvm/unittests/Analysis/IVDescriptorsTest.cpp b/llvm/unittests/Analysis/IVDescriptorsTest.cpp
index 453800abf9cab..bc237296a36b1 100644
--- a/llvm/unittests/Analysis/IVDescriptorsTest.cpp
+++ b/llvm/unittests/Analysis/IVDescriptorsTest.cpp
@@ -259,3 +259,12 @@ for.end:
EXPECT_EQ(Kind, RecurKind::FMax);
});
}
+
+TEST(IVDescriptorsTest, GetOpcodeForFMinMaxNum) {
+ using RK = RecurKind;
+ // Ensure the "num" variants map to the same opcode as the plain ones.
+ EXPECT_EQ(RecurrenceDescriptor::getOpcode(RK::FMinNum),
+ RecurrenceDescriptor::getOpcode(RK::FMin));
+ EXPECT_EQ(RecurrenceDescriptor::getOpcode(RK::FMinNum),
+ RecurrenceDescriptor::getOpcode(RK::FMin));
+}
|
fhahn
left a comment
fhahn
left a comment
There was a problem hiding this comment.
This is very likely to be incorrect.
There's not really a corresponding opcode and treating them the same as FMax/FMin will be incorrect. It would probably be better to update your code not to query getOpcode for those recurrence kinds.
|
Thank you for the review and explanation. |
3 participants
Modify
RecurrenceDescriptor::getOpcodeto handleRecurKind::FMinNumandRecurKind::FMaxNumby returningInstruction::FCmp, matchingFMin/FMax.Without these cases, querying the opcode for minNum/maxNum reductions can hit
llvm_unreachablein code that inspects the recurrence kind. Also addedIVDescriptorsTest.GetOpcodeForFMinMaxNum(gtest) to verify the mapping.