[PATCH] D52416: Allow FP types for atomicrmw xchg

[JF Bastien via Phabricator via llvm-commits]

jfb added a comment.

What does this look like with half, fp80, fp128?
Does this not change anything on other architectures?



================
Comment at: lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp:2171
+  EVT VT = N->getValueType(0);
+  assert(VT == MVT::f16 && "unhandled case");
+
----------------
I don't get why only FP16 is accessible here.


================
Comment at: lib/Target/AMDGPU/AMDGPUISelLowering.cpp:290
 
-
   setOperationAction(ISD::Constant, MVT::i32, Legal);
----------------
?


https://reviews.llvm.org/D52416