nikic wrote:

Right, but I think we're still a good bit away from treating ptrtoint as a 
side-effecting instruction (we need to migrate to the byte type to prevent 
implicit insertion of ptrtoint). So I think for now we need to ensure that the 
ptrtoint does not get dropped as a result of this transform.

This is similar to SCEV where it's also technically correct to model ptrtoint 
as ptrtoaddr, but we can't actually do it because it will end up dropping the 
ptrtoint and lose the provenance escape.

https://github.com/llvm/llvm-project/pull/210729
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to