[all-commits] [llvm/llvm-project] 64de40: [CodeGen][AMDGPU] Prepare rematerializer for multi...

[Lucas Ramirez via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 64de40e5cbfba35eb4118254326889dabb04bb6c
      https://github.com/llvm/llvm-project/commit/64de40e5cbfba35eb4118254326889dabb04bb6c
  Author: Lucas Ramirez <11032120+lucas-rami at users.noreply.github.com>
  Date:   2026-07-14 (Tue, 14 Jul 2026)

  Changed paths:
    M llvm/include/llvm/CodeGen/Rematerializer.h
    M llvm/lib/CodeGen/Rematerializer.cpp
    M llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp
    M llvm/unittests/CodeGen/RematerializerTest.cpp

  Log Message:
  -----------
  [CodeGen][AMDGPU] Prepare rematerializer for multi-def remat support (NFC) (#197579)

This makes some NFCs to the rematerializer before adding initial support
for rematerializing registers with multiple defs. The main change is
that, in the representation of (un)rematerializable register
dependencies, we drop references to machine operand indices which lose
meaning in the multi-def case.

Other minor changes listed below.

- Removal of `DefRegion` argument to `Rematerializer::recreateReg`.
Registers are always re-created in their original region so there is no
need to set their region again.
- Removal of `InsertPos` unused argument to
`Rematerializer::postRematerialization`.
- Refactor of how AMDGPU's scheduler checks whether a given register is
rematerializable.



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications