[PATCH] D54042: [AMDGPU] Extend the SI Load/Store optimizer to combine more things.

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added a comment.

I think these cases should mostly be handled by an IR pass to merge load/store intrinsics, and we need to fix handling of 3 element vectors in SelectionDAG



================
Comment at: lib/Target/AMDGPU/SILoadStoreOptimizer.cpp:112
+
+const DenseMap<unsigned, OpcodeInfo> OpcodeInfoMap{
+    {AMDGPU::S_LOAD_DWORD_IMM, {S_LOAD_IMM, 1, SBASE}},
----------------
No global static initializer


Repository:
  rL LLVM

https://reviews.llvm.org/D54042