[llvm] [LoongArch] Pass 'half' in the lower 16 bits of an f32 value when F extension is enabled (PR #109368)

[via llvm-commits]

================
@@ -1354,6 +1358,40 @@ SDValue LoongArchTargetLowering::lowerVECTOR_SHUFFLE(SDValue Op,
   return SDValue();
 }
 
+SDValue LoongArchTargetLowering::lowerFP_TO_FP16(SDValue Op,
+                                                 SelectionDAG &DAG) const {
+  // Custom lower to ensure the libcall return is passed in an FPR on hard
+  // float ABIs.
+  SDLoc DL(Op);
+  MakeLibCallOptions CallOptions;
+  SDValue Op0 = Op.getOperand(0);
+  SDValue Chain = SDValue();
+  RTLIB::Libcall LC = RTLIB::getFPROUND(Op0.getValueType(), MVT::f16);
+  SDValue Res;
+  std::tie(Res, Chain) =
+      makeLibCall(DAG, LC, MVT::f32, Op0, CallOptions, DL, Chain);
+  if (Subtarget.is64Bit())
+    return DAG.getNode(LoongArchISD::MOVFR2GR_S_LA64, DL, MVT::i64, Res);
+  return DAG.getBitcast(MVT::i32, Res);
+}
+
+SDValue LoongArchTargetLowering::lowerFP16_TO_FP(SDValue Op,
+                                                 SelectionDAG &DAG) const {
+  // Custom lower to ensure the libcall argument is passed in an FPR on hard
+  // float ABIs.
----------------
heiher wrote:

We're not trying to special case the ABI, but rather ensuring compliance with the ABI rules for floating-point operations. Specifically, the argument for `f32 __gnu_h2f_ieee(f16)` needs to be passed via the FPR, as per the floating-point ABI, rather than the GPR. Custom lowering ensures that the argument is correctly passed through the FPR in cases where the default behavior doesn't align with this requirement. Thanks.

https://github.com/llvm/llvm-project/pull/109368