[LoongArch] Supports FP_TO_SINT operation for fp16 (#118303)

Fixes #118301

Author: heiher

llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp

@@ -1669,16 +1669,19 @@ SDValue LoongArchTargetLowering::lowerFP_TO_SINT(SDValue Op,
                                                  SelectionDAG &DAG) const {
 
   SDLoc DL(Op);
+  SDValue Op0 = Op.getOperand(0);
+
+  if (Op0.getValueType() == MVT::f16)
+    Op0 = DAG.getNode(ISD::FP_EXTEND, DL, MVT::f32, Op0);
 
   if (Op.getValueSizeInBits() > 32 && Subtarget.hasBasicF() &&
       !Subtarget.hasBasicD()) {
-    SDValue Dst =
-        DAG.getNode(LoongArchISD::FTINT, DL, MVT::f32, Op.getOperand(0));
+    SDValue Dst = DAG.getNode(LoongArchISD::FTINT, DL, MVT::f32, Op0);
     return DAG.getNode(LoongArchISD::MOVFR2GR_S_LA64, DL, MVT::i64, Dst);
   }
 
   EVT FPTy = EVT::getFloatingPointVT(Op.getValueSizeInBits());
-  SDValue Trunc = DAG.getNode(LoongArchISD::FTINT, DL, FPTy, Op.getOperand(0));
+  SDValue Trunc = DAG.getNode(LoongArchISD::FTINT, DL, FPTy, Op0);
   return DAG.getNode(ISD::BITCAST, DL, Op.getValueType(), Trunc);
 }
 
@@ -2872,6 +2875,8 @@ void LoongArchTargetLowering::ReplaceNodeResults(
     EVT FVT = EVT::getFloatingPointVT(N->getValueSizeInBits(0));
     if (getTypeAction(*DAG.getContext(), Src.getValueType()) !=
         TargetLowering::TypeSoftenFloat) {
+      if (Src.getValueType() == MVT::f16)
+        Src = DAG.getNode(ISD::FP_EXTEND, DL, MVT::f32, Src);
       SDValue Dst = DAG.getNode(LoongArchISD::FTINT, DL, FVT, Src);
       Results.push_back(DAG.getNode(ISD::BITCAST, DL, VT, Dst));
       return;

llvm/test/CodeGen/LoongArch/fp16-promote.ll

@@ -378,3 +378,84 @@ define half @freeze_half_poison(half %maybe.poison) nounwind {
   %t1 = fadd half %y1, %y1
   ret half %t1
 }
+
+define signext i32 @test_half_to_s32(half %a) nounwind {
+; LA32-LABEL: test_half_to_s32:
+; LA32:       # %bb.0: # %entry
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    bl %plt(__gnu_h2f_ieee)
+; LA32-NEXT:    ftintrz.w.s $fa0, $fa0
+; LA32-NEXT:    movfr2gr.s $a0, $fa0
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: test_half_to_s32:
+; LA64:       # %bb.0: # %entry
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    bl %plt(__gnu_h2f_ieee)
+; LA64-NEXT:    ftintrz.w.s $fa0, $fa0
+; LA64-NEXT:    movfr2gr.s $a0, $fa0
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+entry:
+  %conv = fptosi half %a to i32
+  ret i32 %conv
+}
+
+define zeroext i32 @test_half_to_s32_u32(half %a) nounwind {
+; LA32-LABEL: test_half_to_s32_u32:
+; LA32:       # %bb.0: # %entry
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    bl %plt(__gnu_h2f_ieee)
+; LA32-NEXT:    ftintrz.w.s $fa0, $fa0
+; LA32-NEXT:    movfr2gr.s $a0, $fa0
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: test_half_to_s32_u32:
+; LA64:       # %bb.0: # %entry
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    bl %plt(__gnu_h2f_ieee)
+; LA64-NEXT:    ftintrz.w.s $fa0, $fa0
+; LA64-NEXT:    movfr2gr.s $a0, $fa0
+; LA64-NEXT:    bstrpick.d $a0, $a0, 31, 0
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+entry:
+  %conv = fptosi half %a to i32
+  ret i32 %conv
+}
+
+define i64 @test_half_to_i64(half %a) nounwind {
+; LA32-LABEL: test_half_to_i64:
+; LA32:       # %bb.0: # %entry
+; LA32-NEXT:    addi.w $sp, $sp, -16
+; LA32-NEXT:    st.w $ra, $sp, 12 # 4-byte Folded Spill
+; LA32-NEXT:    bl %plt(__gnu_h2f_ieee)
+; LA32-NEXT:    bl %plt(__fixsfdi)
+; LA32-NEXT:    ld.w $ra, $sp, 12 # 4-byte Folded Reload
+; LA32-NEXT:    addi.w $sp, $sp, 16
+; LA32-NEXT:    ret
+;
+; LA64-LABEL: test_half_to_i64:
+; LA64:       # %bb.0: # %entry
+; LA64-NEXT:    addi.d $sp, $sp, -16
+; LA64-NEXT:    st.d $ra, $sp, 8 # 8-byte Folded Spill
+; LA64-NEXT:    bl %plt(__gnu_h2f_ieee)
+; LA64-NEXT:    ftintrz.l.s $fa0, $fa0
+; LA64-NEXT:    movfr2gr.d $a0, $fa0
+; LA64-NEXT:    ld.d $ra, $sp, 8 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 16
+; LA64-NEXT:    ret
+entry:
+  %conv = fptosi half %a to i64
+  ret i64 %conv
+}