Skip to content

[LoopRotate] Set loop back edge weight to not less than exit weight #86496

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 4 commits into from
Mar 28, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
10 changes: 10 additions & 0 deletions llvm/lib/Transforms/Utils/LoopRotationUtils.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -347,9 +347,19 @@ static void updateBranchWeights(BranchInst &PreHeaderBI, BranchInst &LoopBI,
// probabilities as if there are only 0-trip and 1-trip cases.
ExitWeight0 = OrigLoopExitWeight - OrigLoopBackedgeWeight;
}
} else {
// Theoretically, if the loop body must be executed at least once, the
// backedge count must be not less than exit count. However the branch
// weight collected by sampling-based PGO may be not very accurate due to
// sampling. Therefore this workaround is required here to avoid underflow
// of unsigned in following update of branch weight.
if (OrigLoopExitWeight > OrigLoopBackedgeWeight)
OrigLoopBackedgeWeight = OrigLoopExitWeight;
}
assert(OrigLoopExitWeight >= ExitWeight0 && "Bad branch weight");
ExitWeight1 = OrigLoopExitWeight - ExitWeight0;
EnterWeight = ExitWeight1;
assert(OrigLoopBackedgeWeight >= EnterWeight && "Bad branch weight");
LoopBackWeight = OrigLoopBackedgeWeight - EnterWeight;
} else if (OrigLoopExitWeight == 0) {
if (OrigLoopBackedgeWeight == 0) {
Expand Down
4 changes: 2 additions & 2 deletions llvm/test/Transforms/LoopRotate/update-branch-weights.ll
Original file line number Diff line number Diff line change
Expand Up @@ -240,7 +240,7 @@ loop_exit:

; BFI_AFTER-LABEL: block-frequency-info: func6_inaccurate_branch_weight
; BFI_AFTER: - entry: {{.*}} count = 1024
; BFI_AFTER: - loop_body: {{.*}} count = 4294967296
; BFI_AFTER: - loop_body: {{.*}} count = 1024
; BFI_AFTER: - loop_exit: {{.*}} count = 1024

; IR-LABEL: define void @func6_inaccurate_branch_weight(
Expand Down Expand Up @@ -292,4 +292,4 @@ loop_exit:
; IR: [[PROF_FUNC3_0]] = !{!"branch_weights", i32 0, i32 1}
; IR: [[PROF_FUNC4_0]] = !{!"branch_weights", i32 1, i32 0}
; IR: [[PROF_FUNC5_0]] = !{!"branch_weights", i32 0, i32 0}
; IR: [[PROF_FUNC6_0]] = !{!"branch_weights", i32 -1, i32 1024}
; IR: [[PROF_FUNC6_0]] = !{!"branch_weights", i32 0, i32 1024}