[PATCH] D74644: TTI: Add addrspace parameters to memcpy lowering functions
Matt Arsenault via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Wed Mar 4 09:33:04 PST 2020
arsenm added a comment.
In D74644#1904579 <https://reviews.llvm.org/D74644#1904579>, @kbarton wrote:
> To me the grouping that you have here makes the most sense, but I don't have a strong opinion either way.
>
> Is the idea with this that you can override the types used for the mempy based on either the source or destination address space? If not, why is the address space needed (out of curiosity)?
Yes. Any query that involves memory that doesn't have an address space ends up being a problem eventually. For AMDGPU we would prefer to use wider accesses for certain address spaces
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D74644/new/
https://reviews.llvm.org/D74644
More information about the llvm-commits
mailing list