nickdesaulniers added a comment. In D20401#3824713 <https://reviews.llvm.org/D20401#3824713>, @sammccall wrote:
> Thanks Nick for the info! No kernel experience here, so if you have any > particular suggestions about how to measure the workload you care about it'd > be much appreciated (e.g. are particular files that are slow enough to > measure in isolation, or is it better to do a full build) @sammccall I wrote up instructions for how to profile a Linux kernel build with LLVM. https://github.com/ClangBuiltLinux/profiling/tree/main/perf A build on a 72+ threaded workstation should only take ~1 minute. Can you please give it a shot and let me know off-thread if you encounter any issues? One thing I found about that workflow: just this week I upgraded to a zen2-based threadripper workstation. It appears that zen2 has issues using per-thread LBR. https://www.spinics.net/lists/linux-perf-users/msg23103.html (There's follow up responses that I don't see yet in the archive, but it looks like there's pending Linux kernel patches to get that working.) https://lore.kernel.org/lkml/166155216401.401.5809694678609694438.tip-bot2@tip-bot2/ https://lore.kernel.org/lkml/20220829113347.295-1-ravi.bango...@amd.com/ So you might want to ensure you're running those instructions on an intel box, for now. I'm also happy to hop on a virtual call with you and @hokein anytime. Repository: rL LLVM CHANGES SINCE LAST ACTION https://reviews.llvm.org/D20401/new/ https://reviews.llvm.org/D20401 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
