[clang] [llvm] Disable pre-link LTO vectorisation and loop unroll for at O2 and O3 optimisation levels (PR #195886)

Luke Lau via cfe-commits cfe-commits at lists.llvm.org
Tue Oct 27 09:50:33 PDT 2026


lukel97 wrote:

> All that said, I think overall the pipeline is in a pretty good shape with geomean being neutral or improved, even if individual benchmarks can regress, it's a good starting point especially considering that it has not enjoyed the years of optimisations and tuning for specific phase order that the old pipeline had.
> 
> Does anyone disagree with committing these patches?

No objections on my side, the cactus_r change is recent and not a performance regression from a previous release. So I don't think it's a blocker.

Before this lands it would be good to get a final run of SPEC CPU 2026 numbers on AArch64 so it can be included in the PR description. Thanks for pushing this through

https://github.com/llvm/llvm-project/pull/195886


More information about the cfe-commits mailing list