arsenm added a comment. In D79744#2165929 <https://reviews.llvm.org/D79744#2165929>, @rjmccall wrote:
> Arguably we should add this attribute to all indirect arguments. I can > understand not wanting to update all the test cases, but you could probably > avoid adding a new IndirectByRef kind of ABIArgInfo by treating kernels > specially in ConstructAttributeList. > > Or, sorry, I forget — is this semantically necessary because the argument is > to constant memory and the callee has to copy it to form the mutable local? > If so, I think (1) the above statement about theoretically using `byref` on > all arguments still applies and (2) we do need a new ABIArgInfo kind, but we > should name it something like IndirectAliased. Yes, it's semantically needed to insert the copy from constant memory. I originally interpreted a copy as necessary if the indirect addrspace did not match the stack address space, which is a sort of roundabout way of achieving the same thing CHANGES SINCE LAST ACTION https://reviews.llvm.org/D79744/new/ https://reviews.llvm.org/D79744 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
