[llvm] [LoopVectorizer] Prune VFs based on plan register pressure (PR #132190)

[Sam Tebbs via llvm-commits]

SamTebbs33 wrote:

> Performance has restored back. Thank you! I also see a significant performance improvement on zen4 - just wondering if this is expected (I did not verify which commit caused this, but it is in the same timeframe as the follow-up fix here).

Thanks for checking, I hoped that the regression fix would solve your issue. If a higher VF has lead to the performance improvement on zen4 then it could be thanks to this.

https://github.com/llvm/llvm-project/pull/132190