rename utility function

Author: broxigarchen

llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp

@@ -3578,11 +3578,12 @@ SDValue AMDGPUTargetLowering::LowerFP_TO_FP16(SDValue Op, SelectionDAG &DAG) con
     return SDValue();
   }
 
-  return LowerF64ToF16(N0, DL, DAG);
+  return LowerF64ToF16Safe(N0, DL, DAG);
 }
 
-SDValue AMDGPUTargetLowering::LowerF64ToF16(SDValue Src, const SDLoc &DL,
-                                            SelectionDAG &DAG) const {
+// return node in i32
+SDValue AMDGPUTargetLowering::LowerF64ToF16Safe(SDValue Src, const SDLoc &DL,
+                                                SelectionDAG &DAG) const {
   assert(Src.getSimpleValueType() == MVT::f64);
 
   // f64 -> f16 conversion using round-to-nearest-even rounding mode.

llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h

@@ -97,7 +97,8 @@ class AMDGPUTargetLowering : public TargetLowering {
   SDValue LowerFP_TO_FP16(SDValue Op, SelectionDAG &DAG) const;
   SDValue LowerFP_TO_INT(SDValue Op, SelectionDAG &DAG) const;
 
-  SDValue LowerF64ToF16(SDValue Src, const SDLoc &DL, SelectionDAG &DAG) const;
+  SDValue LowerF64ToF16Safe(SDValue Src, const SDLoc &DL,
+                            SelectionDAG &DAG) const;
 
   SDValue LowerSIGN_EXTEND_INREG(SDValue Op, SelectionDAG &DAG) const;
 

llvm/lib/Target/AMDGPU/SIISelLowering.cpp

@@ -6909,7 +6909,7 @@ SDValue SITargetLowering::lowerFP_ROUND(SDValue Op, SelectionDAG &DAG) const {
       SDValue Src32 = DAG.getNode(ISD::FP_ROUND, DL, MVT::f32, Src, Flags);
       return DAG.getNode(ISD::FP_ROUND, DL, MVT::f16, Src32, Flags);
     }
-    SDValue FpToFp16 = LowerF64ToF16(Src, DL, DAG);
+    SDValue FpToFp16 = LowerF64ToF16Safe(Src, DL, DAG);
     SDValue Trunc = DAG.getNode(ISD::TRUNCATE, DL, MVT::i16, FpToFp16);
     return DAG.getNode(ISD::BITCAST, DL, MVT::f16, Trunc);
   }

llvm/test/CodeGen/AMDGPU/fptrunc.f16.ll

@@ -756,9 +756,7 @@ define amdgpu_kernel void @fptrunc_v2f64_to_v2f16(
 ; GFX950-SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; GFX950-SDAG-NEXT:    v_cvt_f32_f64_e32 v2, v[2:3]
 ; GFX950-SDAG-NEXT:    v_cvt_f32_f64_e32 v0, v[0:1]
-; GFX950-SDAG-NEXT:    v_cvt_f16_f32_e32 v1, v2
-; GFX950-SDAG-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX950-SDAG-NEXT:    v_lshl_or_b32 v0, v1, 16, v0
+; GFX950-SDAG-NEXT:    v_cvt_pk_f16_f32 v0, v0, v2
 ; GFX950-SDAG-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; GFX950-SDAG-NEXT:    s_endpgm
 ;