[PATCH] D25802: [AMDGPU] Check if type transforms to i16 (VI+) when getting AMDGPUISD::FFBH_U32

[Matt Arsenault via llvm-commits]

arsenm added inline comments.


================
Comment at: lib/Target/AMDGPU/AMDGPUISelLowering.cpp:2461
+  if (LegalVT != MVT::i32 &&
+        (Subtarget->getGeneration() >= AMDGPUSubtarget::VOLCANIC_ISLANDS &&
+         LegalVT != MVT::i16))
----------------
Subtarget->hasI16


================
Comment at: lib/Target/AMDGPU/AMDGPUISelLowering.cpp:2468
 
-  SDValue FFBH = DAG.getNode(AMDGPUISD::FFBH_U32, SL, MVT::i32, Op);
+  SDValue FFBH_U32 = DAG.getNode(AMDGPUISD::FFBH_U32, DL, MVT::i32, Op);
   if (VT != MVT::i32)
----------------
Variable name probably shouldn't match the node type name


================
Comment at: lib/Target/AMDGPU/AMDGPUISelLowering.h:33
+  /// compare.
+  SDValue getFFBH_U32(SelectionDAG &DAG, SDValue Op, SDLoc DL) const;
+
----------------
I think SDLoc is supposed to be passed by const reference now


https://reviews.llvm.org/D25802