[llvm] [AArch64] Neoverse V2 FeatureDisableLatencySchedHeuristic (PR #140897)

[Igor Kirillov via llvm-commits]

igogo-x86 wrote:

Hi, @sjoerdmeijer! There’s a bit of a story here: PR #136329 caused about a 1.8% performance regression in 525.x264 on Neoverse-V2 concentrated in `x264_pixel_satd_16x16` and `x264_pixel_satd_8x8`. I put a patch to address that (PR #146694), but in the meantime, your patch seems to have made codegen a bit less efficient in those functions.

With my current patch, we’re only recovering around 0.3–0.4% of the lost performance, while reverting your patch on top gives an additional 1.4–1.5% improvement. I’m not sure if this is something we can address easily - the scheduler is tricky territory - but I just wanted to bring it to your attention in case you have any thoughts.

https://github.com/llvm/llvm-project/pull/140897