[PATCH] D103230: [AMDGPU] Disable NSA for BVH instructions when appropriate

[Carl Ritson via Phabricator via llvm-commits]

critson marked 4 inline comments as done.
critson added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp:4812
+    // Build a single vector containing all the operands so far prepared.
+    const unsigned LaneCount = NumVAddrs <= 8 ? 8 : 16;
+
----------------
foad wrote:
> Why do we have to round up to 8 or 16?
I cannot remember why I did this for GlobalIsel.
It certainly works without, perhaps it didn't when I first wrote this.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D103230/new/

https://reviews.llvm.org/D103230