[PATCH] D38315: [ARM] Armv8.2-A FP16 code generation (part 1/2)

[Oliver Stannard via Phabricator via llvm-commits]

olista01 added a comment.

Ok, I agree with the idea of committing this as a starting point and developing it gradually. Just a few nits left.



================
Comment at: lib/Target/ARM/ARMCallingConv.td:191
+
+  CCIfType<[f16], CCBitConvertToType<i16>>,
+
----------------
There are still some changes in this file, are they required for this patch?


================
Comment at: lib/Target/ARM/ARMISelLowering.cpp:4939
+static SDValue ExpandBITCAST(SDNode *N, SelectionDAG &DAG,
+                             const ARMSubtarget *Subtarget) {
   const TargetLowering &TLI = DAG.getTargetLoweringInfo();
----------------
The new argument is unused, should the change below be guarded using this?


================
Comment at: lib/Target/ARM/ARMInstrVFP.td:712
 def VCVTBDH : ADuI<0b11101, 0b11, 0b0011, 0b01, 0,
+                   //(outs HPR:$Sd), (ins DPR:$Dm),
                    (outs SPR:$Sd), (ins DPR:$Dm),
----------------
Commented-out code


https://reviews.llvm.org/D38315