R600: Custom lower f64 frint for pre-CI

[arsenm]

Ping

[arsenm]

r211182

lib/Target/R600/AMDGPUISelLowering.h

   /// \returns The resulting chain.
   SDValue LowerUDIVREM(SDValue Op, SelectionDAG &DAG) const;
   SDValue LowerFREM(SDValue Op, SelectionDAG &DAG) const;
+  SDValue LowerFRINT(SDValue Op, SelectionDAG &DAG) const;
   SDValue LowerUINT_TO_FP(SDValue Op, SelectionDAG &DAG) const;
 
 protected:

lib/Target/R600/AMDGPUISelLowering.cpp

   setOperationAction(ISD::FREM, MVT::f32, Custom);
   setOperationAction(ISD::FREM, MVT::f64, Custom);
 
+  if (Subtarget->getGeneration() < AMDGPUSubtarget::SEA_ISLANDS) {
+    setOperationAction(ISD::FRINT, MVT::f64, Custom);
+  }
+
   if (!Subtarget->hasBFI()) {
     // fcopysign can be done in a single instruction with BFI.
     setOperationAction(ISD::FCOPYSIGN, MVT::f32, Expand);

   case ISD::INTRINSIC_WO_CHAIN: return LowerINTRINSIC_WO_CHAIN(Op, DAG);
   case ISD::UDIVREM: return LowerUDIVREM(Op, DAG);
   case ISD::FREM: return LowerFREM(Op, DAG);
+  case ISD::FRINT: return LowerFRINT(Op, DAG);
   case ISD::UINT_TO_FP: return LowerUINT_TO_FP(Op, DAG);
   }
   return Op;

   return DAG.getNode(ISD::FSUB, SL, VT, X, Mul);
 }
 
+SDValue AMDGPUTargetLowering::LowerFRINT(SDValue Op, SelectionDAG &DAG) const {
+  SDLoc SL(Op);
+  SDValue Src = Op.getOperand(0);
+
+  assert(Op.getValueType() == MVT::f64);
+
+  SDValue C1 = DAG.getConstantFP(0x1.0p+52, MVT::f64);
+  SDValue CopySign = DAG.getNode(ISD::FCOPYSIGN, SL, MVT::f64, C1, Src);
+
+  SDValue Tmp1 = DAG.getNode(ISD::FADD, SL, MVT::f64, Src, CopySign);
+  SDValue Tmp2 = DAG.getNode(ISD::FSUB, SL, MVT::f64, Tmp1, CopySign);
+
+  SDValue Fabs = DAG.getNode(ISD::FABS, SL, MVT::f64, Src);
+  SDValue C2 = DAG.getConstantFP(0x1.fffffffffffffp+51, MVT::f64);
+
+  EVT SetCCVT = getSetCCResultType(*DAG.getContext(), MVT::f64);
+  SDValue Cond = DAG.getSetCC(SL, SetCCVT, Fabs, C2, ISD::SETOGT);
+
+  return DAG.getSelect(SL, MVT::f64, Cond, Src, Tmp2);
+}
+
 SDValue AMDGPUTargetLowering::LowerUINT_TO_FP(SDValue Op,
                                                SelectionDAG &DAG) const {
   SDValue S0 = Op.getOperand(0);

test/CodeGen/R600/llvm.rint.f64.ll

 ; RUN: llc -march=r600 -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
 
-; FUNC-LABEL: @f64
+; FUNC-LABEL: @rint_f64
 ; CI: V_RNDNE_F64_e32
-define void @f64(double addrspace(1)* %out, double %in) {
+
+; SI-DAG: V_ADD_F64
+; SI-DAG: V_ADD_F64
+; SI-DAG V_CMP_GT_F64_e64
+; SI: V_CNDMASK_B32
+; SI: V_CNDMASK_B32
+; SI: S_ENDPGM
+define void @rint_f64(double addrspace(1)* %out, double %in) {
 entry:
   %0 = call double @llvm.rint.f64(double %in)
   store double %0, double addrspace(1)* %out
   ret void
 }
 
-; FUNC-LABEL: @v2f64
+; FUNC-LABEL: @rint_v2f64
 ; CI: V_RNDNE_F64_e32
 ; CI: V_RNDNE_F64_e32
-define void @v2f64(<2 x double> addrspace(1)* %out, <2 x double> %in) {
+define void @rint_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %in) {
 entry:
   %0 = call <2 x double> @llvm.rint.v2f64(<2 x double> %in)
   store <2 x double> %0, <2 x double> addrspace(1)* %out
   ret void
 }
 
-; FUNC-LABEL: @v4f64
+; FUNC-LABEL: @rint_v4f64
 ; CI: V_RNDNE_F64_e32
 ; CI: V_RNDNE_F64_e32
 ; CI: V_RNDNE_F64_e32
 ; CI: V_RNDNE_F64_e32
-define void @v4f64(<4 x double> addrspace(1)* %out, <4 x double> %in) {
+define void @rint_v4f64(<4 x double> addrspace(1)* %out, <4 x double> %in) {
 entry:
   %0 = call <4 x double> @llvm.rint.v4f64(<4 x double> %in)
   store <4 x double> %0, <4 x double> addrspace(1)* %out