[llvm] [RISCV] Add late optimization pass for RISC-V to optimize branch instructions (PR #131728)

Craig Topper via llvm-commits llvm-commits at lists.llvm.org
Thu Mar 20 12:55:52 PDT 2025


topperc wrote:

> > Sorry for the delay, I was seeing some strange numbers, so I ran everything twice: https://lnt.lukelau.me/db_default/v4/nts/338?show_delta=yes&show_previous=yes&show_stddev=yes&show_mad=yes&show_all=yes&show_all_samples=yes&show_sample_counts=yes&show_small_diff=yes&num_comparison_runs=0&test_filter=&test_min_value_filter=&aggregation_fn=min&MW_confidence_lv=0.05&compare_to=337&submit=Update
> > 
> > * I wouldn't trust the result of gcc_r, there is a high variance there
> > * The regression on xalancbmk_r is strange, I'll have to take a closer look at it
> 
> @mikhailramalho so what I'm hearing is that we should count this as a pretty nice gain in xz? I know xalancbmk usually has high variance too, but this looks quite large.

Does https://github.com/llvm/llvm-project/pull/117060 also improve xz?

https://github.com/llvm/llvm-project/pull/131728


More information about the llvm-commits mailing list