Another example would be where the algorithm stops merging functions (with call edges) because the temporal locality degrades too much, but this can be done in a separate patch.

I think D158674 handled this. So abandoning this revision.

Moved to https://github.com/llvm/llvm-project/pull/65506

ping

Fix test.

In D107534#4622487, @nickdesaulniers wrote:

Does this work with -mllvm -trap-unreachable?

Thanks.

In D158429#4605882, @jhenderson wrote:

Looks good (should it have had a unit test?). Not sure why this was landed without review though?

• Add check for stats.
• Change the option name to -gc-empty-basic-blocks.
• Rebase

fix comment.

clang-format.

Rename the test.

Apply final changes.

Add the cloning profile format to the function comment.

nit.

clang-format.

fixes.

nit.

Add BasicBlockSections.cpp changes.

Remove .gitignore from the patch.

Remove unused variable.

Remove redundant check, refactor, rebase.

clang-format.

Omit empty basic blocks which have their address taken, and EH pad blocks.

Remove unwanted change.

Thanks for the updated results. They look more consistent now.

Thanks for providing the evaluation results. This shows nice improvements when huge pages are not used. My intuition is that using huge pages reduces the opportunity from function reordering.

I am also interested in getting more details about your evaluation. Currently, LLD uses CallChainClustering for FDO (https://github.com/llvm-mirror/lld/blob/master/ELF/CallGraphSort.cpp). I wonder how much perf. improvement we get if we hook this into LLD.

• rebase.

• Address reviewer comments.

• Extract filename via DICompileUnits. Match on filename attribute only.

Thank you for the review.

Update the comment.

@jhenderson Another friendly ping.

• Add error handling for invalid encoding.
• Add llvm-readobj test.

• clang-format.

@jhenderson Thanks for reviewing. Please let me know if I have addressed your comments.