[PATCH] D93715: AMDGPU: Don't fold AGPR copy pairs that need a temp VGPR
Matt Arsenault via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Mon Jan 4 10:13:20 PST 2021
arsenm added a comment.
In D93715#2477411 <https://reviews.llvm.org/D93715#2477411>, @kerbowa wrote:
> I'm working on eliminating some AGPR->VGPR->AGPR copies which commonly cause problems in loops where the VGPR is considered live across iterations by the register allocator even if the VGPR copy should be sunk. Why do we need this temp VGPR?
Because you can't directly copy between AGPRs (or directly from SGPR to AGPR)
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D93715/new/
https://reviews.llvm.org/D93715
More information about the llvm-commits
mailing list