[PATCH] D118351: [AMDGPU][GlobalISel] Code quality: don't expand G_BUILD_VECTOR_TRUNC if not neccessary

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp:2612
+      // the left by 16.
+      if (LoOpc == TargetOpcode::G_IMPLICIT_DEF) {
+        if (isLshrBy16(*HiMI))
----------------
I think we should have combined out any implicit def inputs into something else. Trying to optimize as part of a lowering expansion is generally a last resort strategy


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D118351/new/

https://reviews.llvm.org/D118351