Despite previous effort D148569: [PseudoProbe] Clean up dwarf discriminator and avoid duplicating factor. to avoid screwing up existing disrminator field, I'm still seeing some call probes getting a non-zero discriminator eventually in non-FS mode. It could be related to callsite merge. While they are investigated I'm disabling discriminator emission for non-FS mode. This avoids breaking the compatiblity with older tools like llvm-profgen and bolt.
Details
Details
Diff Detail
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Comment Actions
thanks for the fix, lgtm.
llvm/lib/CodeGen/AsmPrinter/PseudoProbePrinter.cpp | ||
---|---|---|
50 | should we put in an assert to make sure we catch lurking non-zero discriminators? i.e. when FS is off, probe should always have zero discriminator? |
llvm/lib/CodeGen/AsmPrinter/PseudoProbePrinter.cpp | ||
---|---|---|
50 | Good point. Actually there is already an assert on IR level https://github.com/llvm/llvm-project/blob/b7d9322b4963e620dfd12246816e6f7b2da5fd88/llvm/lib/IR/PseudoProbe.cpp#L67, let me also add an assert here for MIR. |
should we put in an assert to make sure we catch lurking non-zero discriminators? i.e. when FS is off, probe should always have zero discriminator?