update patch

Author: broxigarchen

llvm/lib/Target/AMDGPU/SIISelLowering.cpp

@@ -6899,9 +6899,20 @@ SDValue SITargetLowering::lowerFP_ROUND(SDValue Op, SelectionDAG &DAG) const {
     if (Op.getOpcode() != ISD::FP_ROUND)
       return Op;
 
-    SDValue FpToFp16 = DAG.getNode(ISD::FP_TO_FP16, DL, MVT::i32, Src);
-    SDValue Trunc = DAG.getNode(ISD::TRUNCATE, DL, MVT::i16, FpToFp16);
-    return DAG.getNode(ISD::BITCAST, DL, MVT::f16, Trunc);
+    if (Subtarget->has16BitInsts()) {
+      if (getTargetMachine().Options.UnsafeFPMath) {
+        SDValue Flags = Op.getOperand(1);
+        SDValue Src32 = DAG.getNode(ISD::FP_ROUND, DL, MVT::f32, Src, Flags);
+        return DAG.getNode(ISD::FP_ROUND, DL, MVT::f16, Src32, Flags);
+      } else {
+        SDValue FpToFp16 = LowerF64ToF16(Src, MVT::i16, DL, DAG);
+        return DAG.getNode(ISD::BITCAST, DL, MVT::f16, FpToFp16);
+      }
+    } else {
+      SDValue FpToFp16 = DAG.getNode(ISD::FP_TO_FP16, DL, MVT::i32, Src);
+      SDValue Trunc = DAG.getNode(ISD::TRUNCATE, DL, MVT::i16, FpToFp16);
+      return DAG.getNode(ISD::BITCAST, DL, MVT::f16, Trunc);
+    }
   }
 
   assert(DstVT.getScalarType() == MVT::bf16 &&

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

@@ -720,7 +720,7 @@ define amdgpu_kernel void @fptrunc_v2f64_to_v2f16(
 ; GFX9-SDAG-NEXT:    v_cvt_f32_f64_e32 v0, v[0:1]
 ; GFX9-SDAG-NEXT:    v_cvt_f16_f32_e32 v1, v2
 ; GFX9-SDAG-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX9-SDAG-NEXT:    v_lshl_or_b32 v0, v1, 16, v0
+; GFX9-SDAG-NEXT:    v_pack_b32_f16 v0, v0, v1
 ; GFX9-SDAG-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; GFX9-SDAG-NEXT:    s_endpgm
 ;
@@ -814,14 +814,13 @@ define amdgpu_kernel void @fptrunc_v2f64_to_v2f16(
 ; GFX11-SDAG-FAKE16-NEXT:    buffer_load_b128 v[0:3], off, s[8:11], 0
 ; GFX11-SDAG-FAKE16-NEXT:    s_mov_b32 s5, s1
 ; GFX11-SDAG-FAKE16-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-SDAG-FAKE16-NEXT:    v_cvt_f32_f64_e32 v2, v[2:3]
 ; GFX11-SDAG-FAKE16-NEXT:    v_cvt_f32_f64_e32 v0, v[0:1]
-; GFX11-SDAG-FAKE16-NEXT:    v_cvt_f32_f64_e32 v1, v[2:3]
 ; GFX11-SDAG-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-FAKE16-NEXT:    v_cvt_f16_f32_e32 v1, v2
 ; GFX11-SDAG-FAKE16-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX11-SDAG-FAKE16-NEXT:    v_cvt_f16_f32_e32 v1, v1
-; GFX11-SDAG-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-SDAG-FAKE16-NEXT:    v_and_b32_e32 v0, 0xffff, v0
-; GFX11-SDAG-FAKE16-NEXT:    v_lshl_or_b32 v0, v1, 16, v0
+; GFX11-SDAG-FAKE16-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-FAKE16-NEXT:    v_pack_b32_f16 v0, v0, v1
 ; GFX11-SDAG-FAKE16-NEXT:    buffer_store_b32 v0, off, s[4:7], 0
 ; GFX11-SDAG-FAKE16-NEXT:    s_endpgm
 ;