jhuber6 wrote: > I've been battling figuring out when I have a "target ID" and when it's the > actual CPU name, none of which are useful because I just want the > OffloadArch, which not all targets even have definitions for.
The target-id itself was a huge mistake, I would've rejected it if it were proposed today. > IMO everywhere that has an OffloadKind argument must have an accompanying > bound arch; they shouldn't be separable. I have another patch that turns this > into a better behaved struct. I didn't remember that we used the OffloadKind in these methods, that also seems a bit strange as I would've figured that would be implicit in the JobAction, but we don't have access to that here? https://github.com/llvm/llvm-project/pull/196504 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
