[PATCH] D119475: [AMDGPU] Add scheduler pass to rematerialize trivial defs

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec added a comment.

In D119475#3315986 <https://reviews.llvm.org/D119475#3315986>, @arsenm wrote:

> I don't understand why the scheduler should be doing rematerialization. The rematerialization should have been done by the coalescer if it would reduce pressure

Not even a coalescer, it shall be done in RA, but RA only does it when it has to spill and knows nothing about occupancy. Scheduler is a closest place where we have all RP information and can reasonably intervene. In a long run we need to consider teaching RA about occupancy.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D119475/new/

https://reviews.llvm.org/D119475