https://github.com/erichkeane commented:
I'm on the fence as defined now. I think this is absolutely the right direction (an attribute on a `cir.func` that tells us additional information for optimization purposes), but I think the data it is currently holding isn't particularly useful. Also, the 'instance-method' not telling us what the parent object type is, is a touch awkward/perhaps limiting as well? On the other hand, I very much expect this attribute to evolve over time, and perhaps this is a good starting point. I AM a little uncomfortable with it as a starting point at the moment because I suspect that the opts you have next are not necessarily sound because of the information in it, but I am a little OK discussing that in the next patches? https://github.com/llvm/llvm-project/pull/207261 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
