How do I use the MLGO RegAlloc model in my llvm build?

[Manish Kausik H]

Hey Everyone!

I wanted to experiment using the MLGO RegAlloc Model in my llvm build, but was unable to find a pre-trained model for it within llvm or the mlgo github repo. Do we have pre-trained models for RegAlloc, or the only way is to train them ourselves? 

Also, do we have any resources on how to use MLGOPerf?

Any help would be appreciated!

Regards,

Manish Kausik H

[Mircea Trofin]

https://github.com/google/ml-compiler-opt/releases for pretrained.

Mileage may vary for retraining regalloc, because the reward signal isn't great, hence the (for example) the gsoc project for latency prediction.

Also, could you please use discourse.llvm.org (with #mlgo tag) for discussions, this group is mostly for simplifying meeting calendar invites.

Thanks!

--
You received this message because you are subscribed to the Google Groups "llvm-mlgo" group.
To unsubscribe from this group and stop receiving emails from it, send an email to llvm-mlgo+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/llvm-mlgo/e4583dcc-4850-4de1-949b-c7073e500998n%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.