[PATCH] D48586: [AMDGPU] Early expansion of 32 bit udiv/urem

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp:492-494
+  Value *MUL64_HILO = Builder.CreateBitCast(MUL64, VectorType::get(I32Ty, 2));
+  return std::make_pair(Builder.CreateExtractElement(MUL64_HILO, Zero),
+                        Builder.CreateExtractElement(MUL64_HILO, One));
----------------
While this is what we do in the DAG, this isn't really the canonical IR way to do this. Truncate, and shift + truncate should probably be used instead here


https://reviews.llvm.org/D48586