AaronBallman wrote:

> Actually, I don't think we should use function multi-versioning here. x86-64 
> distributions that are aware of the performance should enable 
> `-march=x86-64-v2` (roughly anything from Intel Nehalem 2008+ or AMD 
> Bulldozer/Jaguar 2011+), which implies SSE4.2. Then the SSE4.2 code will be 
> static.

Ideally, we would use multi-versioning here because that gets the best 
performance without making people performance tune the compiler when they 
build. This improves throughput for our pre- and post-commit CI pipeline, debug 
builds our developers are using, etc.

That said, I think it's reasonable to expect distributions and downstreams to 
do that performance tuning themselves if that's the only viable way forward at 
the moment.

https://github.com/llvm/llvm-project/pull/180631
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to