[PATCH] D73485: [AMDGPU] Simplify DS and SM cases in getMemOperandsWithOffset

[Jay Foad via Phabricator via llvm-commits]

foad marked an inline comment as done.
foad added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/SIInstrInfo.cpp:276-277
+      if (!BaseOp) {
+        // DS_CONSUME/DS_APPEND use M0 for the base address, but there is no
+        // operand that we can push to represent that here.
         return false;
----------------
arsenm wrote:
> Couldn't this report the implicit m0 read?
Good point, but I've left it as a TODO because I'm not sure how to test it. I'm not sure if clustering would ever apply to DS_CONSUME/DS_APPEND or any other instructions that use M0 like this.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D73485/new/

https://reviews.llvm.org/D73485