[llvm] [AMDGPU][True16] Select _t16 D16 loads/stores for FLAT/GLOBAL/SCRATCH (PR #210191)

Jay Foad via llvm-commits llvm-commits at lists.llvm.org
Fri Jul 17 01:52:51 PDT 2026


================
@@ -263,6 +263,12 @@ void AMDGPUDAGToDAGISelLegacy::getAnalysisUsage(AnalysisUsage &AU) const {
 
 bool AMDGPUDAGToDAGISel::matchLoadD16FromBuildVector(SDNode *N) const {
   assert(Subtarget->d16PreservesUnusedBits());
+  // In true16 mode the two 16-bit halves of a packed vector live in
----------------
jayfoad wrote:

Instead of this you could return early from PreprocessISelDAG like we do for !d16PreservesUnusedBits.

https://github.com/llvm/llvm-project/pull/210191


More information about the llvm-commits mailing list