Add heuristics for irreducible loop metadata under PGO
Summary:
Add the following heuristics for irreducible loop metadata:
- When an irreducible loop header is missing the loop header weight metadata,
give it the minimum weight seen among other headers.
- Annotate indirectbr targets with the loop header weight metadata (as they are
likely to become irreducible loop headers after indirectbr tail duplication.)
These greatly improve the accuracy of the block frequency info of the Python
interpreter loop (eg. from ~3-16x off down to ~40-55% off) and the Python
performance (eg. unpack_sequence from ~50% slower to ~8% faster than GCC) due to
better register allocation under PGO.
Reviewers: davidxl
Reviewed By: davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D39980
llvm-svn: 318693
diff --git a/llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp b/llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp
index c92d483..47278e1 100644
--- a/llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp
+++ b/llvm/lib/Transforms/Instrumentation/PGOInstrumentation.cpp
@@ -1188,11 +1188,22 @@
}
}
+static bool isIndirectBrTarget(BasicBlock *BB) {
+ for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
+ if (isa<IndirectBrInst>((*PI)->getTerminator()))
+ return true;
+ }
+ return false;
+}
+
void PGOUseFunc::annotateIrrLoopHeaderWeights() {
DEBUG(dbgs() << "\nAnnotating irreducible loop header weights.\n");
// Find irr loop headers
for (auto &BB : F) {
- if (BFI->isIrrLoopHeader(&BB)) {
+ // As a heuristic also annotate indrectbr targets as they have a high chance
+ // to become an irreducible loop header after the indirectbr tail
+ // duplication.
+ if (BFI->isIrrLoopHeader(&BB) || isIndirectBrTarget(&BB)) {
TerminatorInst *TI = BB.getTerminator();
const UseBBInfo &BBCountInfo = getBBInfo(&BB);
setIrrLoopHeaderMetadata(M, TI, BBCountInfo.CountValue);