[PATCH] D159050: [PowerPC] set min jump table entries to 64

[ChenZheng via Phabricator via llvm-commits]

shchenz added a comment.

> I don't have access to any relevant PPC hardware to benchmark, but 64 seems excessively high. I mean, I can imagine it benefiting certain micro-benchmarks if it helps branch prediction in some hot code. But it's high enough that it likely also hurts other benchmarks, and maybe increases codesize.

Thanks for taking a look. I agree with you that setting this value to 64 has the possibility to hurt some applications. 64 is an experiment result based on benchmarks we(IBM Power team) care about currently. This value improves some benchmarks and does not have obvious negative impact for other ones. So we decide to change the default value. codesize is a good point, we can find out a new value for code size purpose if codesize becomes a concern for us later.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D159050/new/

https://reviews.llvm.org/D159050