[llvm] [LoopVectorize] Improve Vectorization of Low Trip Count Loops (PR #195823)

[Jack Styles via llvm-commits]

Stylie777 wrote:

> Is this true? Looking at tc5_v4i32_reject_minsize, it gets vectorized, followed by scalar loop AFAICT

Apologies I missed it had done that, I've just pushed an update (df65dc1b1610cda5b6404c7eb296ed94ecaae1f7) which will consider both minsize/optsize on the function and `-Os` or `-Oz`.

https://github.com/llvm/llvm-project/pull/195823