pedroclobo wrote:

This PR stack (#237783, #237784, #237785, #237786) is the first to introduce 
bytes in IR that didn't have them before. Code from frontends that don't emit 
byte types can now reach optimization passes and the backends with byte values.

I ran benchmarks from the Phoronix Test Suite against this PR and found no 
significant regressions in compile time, run time, peak memory usage, or binary 
size. One of the benchmarks crashed in the loop vectorizer (see #237767).

The compile time tracker didn't show any meaningful regressions either.


https://github.com/llvm/llvm-project/pull/237786
_______________________________________________
llvm-branch-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits

Reply via email to