[PATCH] D117562: [AMDGPU] Sink immediate VGPR defs if high RP
Valery Pykhtin via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Wed Jan 26 03:51:29 PST 2022
vpykhtin added a comment.
In D117562#3271947 <https://reviews.llvm.org/D117562#3271947>, @foad wrote:
>
> As for RA not sinking defs back into the loop, I guess this is because RA does not know about occupancy? So here's a radical idea: after (pre-RA) scheduling, when we know what occupancy we //want// to achieve, why not make that occupancy a hard requirement, marking all other vgprs reserved? Then RA would have to stick to the budget, spilling if necessary. And yes spilling can be really bad, but just failing to meet your occupancy target can also be really bad.
I think we don't know the occupancy we want to achieve unless there is a user requirement. In general its hard to say what is more beneficial: to have larger occupancy or less spills
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D117562/new/
https://reviews.llvm.org/D117562
More information about the llvm-commits
mailing list