[PATCH] D91308: AMDGPU: Split large offsets when selecting global saddr mode

[Jay Foad via Phabricator via llvm-commits]

foad added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp:1847
       ImmOffset = COffsetVal;
+    } else if (!LHS->isDivergent()) {
+      SDLoc SL(N);
----------------
I would add `&& COffsetVal >= 0`. I don't think there's any point trying this for large negative offsets, because it will always fail at the `isUint<32>` check below. That will simplify the code because you don't need to bother with the "signed division by a power of two" thing.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D91308/new/

https://reviews.llvm.org/D91308