kasuga-fj wrote:
> do you think is there a way this can be salvaged and made correct? In other
> words, my question here is if it is worth leaving a TODO in comments
> explaining what the opportunity is here.
I've thought about something similar once, but I think it's difficult to make
it sound. For some dependence tests, what nested addrecs like `{{c,+,b},+,a}`
(i.e., `[a*i + b*j + c]`) implies is unclear, and probably have the same
problem as
https://github.com/llvm/llvm-project/pull/176367#issuecomment-3849431540. That
is, to correct this, we may need to do something similar to what we need to fix
the MIV tests. If that's the case, I think it's better to leave nested addrecs
to the MIV tests rather than trying to handle them by the RDIV tests.
https://github.com/llvm/llvm-project/pull/179654
_______________________________________________
llvm-branch-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
