[PATCH] D125279: [AMDGPU] Add llvm.amdgc.global.load.lds intrinsic

Mon May 16 14:39:13 PDT 2022

arsenm accepted this revision.
arsenm added a comment.
This revision is now accepted and ready to land.

Typo amdgc. in description

================
Comment at: llvm/include/llvm/IR/IntrinsicsAMDGPU.td:1789
+   llvm_i32_ty,                        // Data byte size: 1/2/4
+   llvm_i32_ty,                        // imm offset
+   llvm_i32_ty],                       // auxiliary data (imm, cachepolicy (bit 0 = glc/sc0,
----------------
Does this really need the offset parameter? It's not a buffer intrinsic with the funny different swizzling behavior between scalar and vector, so can't we just match this from the base pointer?

================
Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:1314-1315
+    Info.memVT = EVT::getIntegerVT(CI.getContext(), Width * 8);
+    Info.flags |= MachineMemOperand::MOLoad | MachineMemOperand::MOStore |
+                  MachineMemOperand::MOVolatile;
+    return true;
----------------
I guess this API just isn't strong enough to handle 2 MMOs

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D125279/new/

https://reviews.llvm.org/D125279