mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2025-02-21 06:30:16 +00:00
R600: Move MIN/MAX matching from LowerOperation() to PerformDAGCombine()
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@208429 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
parent
bfaccd494f
commit
87b983680c
@ -267,6 +267,7 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(TargetMachine &TM) :
|
|||||||
}
|
}
|
||||||
|
|
||||||
setTargetDAGCombine(ISD::MUL);
|
setTargetDAGCombine(ISD::MUL);
|
||||||
|
setTargetDAGCombine(ISD::SELECT_CC);
|
||||||
}
|
}
|
||||||
|
|
||||||
//===----------------------------------------------------------------------===//
|
//===----------------------------------------------------------------------===//
|
||||||
@ -748,16 +749,16 @@ SDValue AMDGPUTargetLowering::LowerIntrinsicLRP(SDValue Op,
|
|||||||
}
|
}
|
||||||
|
|
||||||
/// \brief Generate Min/Max node
|
/// \brief Generate Min/Max node
|
||||||
SDValue AMDGPUTargetLowering::LowerMinMax(SDValue Op,
|
SDValue AMDGPUTargetLowering::CombineMinMax(SDNode *N,
|
||||||
SelectionDAG &DAG) const {
|
SelectionDAG &DAG) const {
|
||||||
SDLoc DL(Op);
|
SDLoc DL(N);
|
||||||
EVT VT = Op.getValueType();
|
EVT VT = N->getValueType(0);
|
||||||
|
|
||||||
SDValue LHS = Op.getOperand(0);
|
SDValue LHS = N->getOperand(0);
|
||||||
SDValue RHS = Op.getOperand(1);
|
SDValue RHS = N->getOperand(1);
|
||||||
SDValue True = Op.getOperand(2);
|
SDValue True = N->getOperand(2);
|
||||||
SDValue False = Op.getOperand(3);
|
SDValue False = N->getOperand(3);
|
||||||
SDValue CC = Op.getOperand(4);
|
SDValue CC = N->getOperand(4);
|
||||||
|
|
||||||
if (VT != MVT::f32 ||
|
if (VT != MVT::f32 ||
|
||||||
!((LHS == True && RHS == False) || (LHS == False && RHS == True))) {
|
!((LHS == True && RHS == False) || (LHS == False && RHS == True))) {
|
||||||
@ -804,7 +805,7 @@ SDValue AMDGPUTargetLowering::LowerMinMax(SDValue Op,
|
|||||||
case ISD::SETCC_INVALID:
|
case ISD::SETCC_INVALID:
|
||||||
llvm_unreachable("Invalid setcc condcode!");
|
llvm_unreachable("Invalid setcc condcode!");
|
||||||
}
|
}
|
||||||
return Op;
|
return SDValue();
|
||||||
}
|
}
|
||||||
|
|
||||||
SDValue AMDGPUTargetLowering::SplitVectorLoad(const SDValue &Op,
|
SDValue AMDGPUTargetLowering::SplitVectorLoad(const SDValue &Op,
|
||||||
@ -1283,6 +1284,9 @@ SDValue AMDGPUTargetLowering::PerformDAGCombine(SDNode *N,
|
|||||||
simplifyI24(N1, DCI);
|
simplifyI24(N1, DCI);
|
||||||
return SDValue();
|
return SDValue();
|
||||||
}
|
}
|
||||||
|
case ISD::SELECT_CC: {
|
||||||
|
return CombineMinMax(N, DAG);
|
||||||
|
}
|
||||||
}
|
}
|
||||||
return SDValue();
|
return SDValue();
|
||||||
}
|
}
|
||||||
|
@ -107,7 +107,7 @@ public:
|
|||||||
|
|
||||||
SDValue LowerIntrinsicIABS(SDValue Op, SelectionDAG &DAG) const;
|
SDValue LowerIntrinsicIABS(SDValue Op, SelectionDAG &DAG) const;
|
||||||
SDValue LowerIntrinsicLRP(SDValue Op, SelectionDAG &DAG) const;
|
SDValue LowerIntrinsicLRP(SDValue Op, SelectionDAG &DAG) const;
|
||||||
SDValue LowerMinMax(SDValue Op, SelectionDAG &DAG) const;
|
SDValue CombineMinMax(SDNode *N, SelectionDAG &DAG) const;
|
||||||
const char* getTargetNodeName(unsigned Opcode) const override;
|
const char* getTargetNodeName(unsigned Opcode) const override;
|
||||||
|
|
||||||
virtual SDNode *PostISelFolding(MachineSDNode *N,
|
virtual SDNode *PostISelFolding(MachineSDNode *N,
|
||||||
|
@ -986,13 +986,6 @@ SDValue R600TargetLowering::LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const
|
|||||||
return DAG.getNode(ISD::BITCAST, DL, VT, SelectNode);
|
return DAG.getNode(ISD::BITCAST, DL, VT, SelectNode);
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
||||||
// Possible Min/Max pattern
|
|
||||||
SDValue MinMax = LowerMinMax(Op, DAG);
|
|
||||||
if (MinMax.getNode()) {
|
|
||||||
return MinMax;
|
|
||||||
}
|
|
||||||
|
|
||||||
// If we make it this for it means we have no native instructions to handle
|
// If we make it this for it means we have no native instructions to handle
|
||||||
// this SELECT_CC, so we must lower it.
|
// this SELECT_CC, so we must lower it.
|
||||||
SDValue HWTrue, HWFalse;
|
SDValue HWTrue, HWFalse;
|
||||||
@ -1672,6 +1665,11 @@ SDValue R600TargetLowering::PerformDAGCombine(SDNode *N,
|
|||||||
}
|
}
|
||||||
|
|
||||||
case ISD::SELECT_CC: {
|
case ISD::SELECT_CC: {
|
||||||
|
// Try common optimizations
|
||||||
|
SDValue Ret = AMDGPUTargetLowering::PerformDAGCombine(N, DCI);
|
||||||
|
if (Ret.getNode())
|
||||||
|
return Ret;
|
||||||
|
|
||||||
// fold selectcc (selectcc x, y, a, b, cc), b, a, b, seteq ->
|
// fold selectcc (selectcc x, y, a, b, cc), b, a, b, seteq ->
|
||||||
// selectcc x, y, a, b, inv(cc)
|
// selectcc x, y, a, b, inv(cc)
|
||||||
//
|
//
|
||||||
|
@ -902,12 +902,6 @@ SDValue SITargetLowering::LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const {
|
|||||||
EVT VT = Op.getValueType();
|
EVT VT = Op.getValueType();
|
||||||
SDLoc DL(Op);
|
SDLoc DL(Op);
|
||||||
|
|
||||||
// Possible Min/Max pattern
|
|
||||||
SDValue MinMax = LowerMinMax(Op, DAG);
|
|
||||||
if (MinMax.getNode()) {
|
|
||||||
return MinMax;
|
|
||||||
}
|
|
||||||
|
|
||||||
SDValue Cond = DAG.getNode(ISD::SETCC, DL, MVT::i1, LHS, RHS, CC);
|
SDValue Cond = DAG.getNode(ISD::SETCC, DL, MVT::i1, LHS, RHS, CC);
|
||||||
return DAG.getNode(ISD::SELECT, DL, VT, Cond, True, False);
|
return DAG.getNode(ISD::SELECT, DL, VT, Cond, True, False);
|
||||||
}
|
}
|
||||||
|
@ -1,7 +1,7 @@
|
|||||||
; RUN: llc < %s -march=r600 | FileCheck %s
|
; RUN: llc < %s -march=r600 | FileCheck %s
|
||||||
|
|
||||||
;CHECK: DOT4 * T{{[0-9]\.W}} (MASKED)
|
;CHECK: DOT4 * T{{[0-9]\.W}} (MASKED)
|
||||||
;CHECK: MAX T{{[0-9].[XYZW]}}, 0.0, PV.X
|
;CHECK: MAX T{{[0-9].[XYZW]}}, PV.X, 0.0
|
||||||
|
|
||||||
define void @main(<4 x float> inreg %reg0, <4 x float> inreg %reg1, <4 x float> inreg %reg2, <4 x float> inreg %reg3, <4 x float> inreg %reg4, <4 x float> inreg %reg5, <4 x float> inreg %reg6, <4 x float> inreg %reg7) #0 {
|
define void @main(<4 x float> inreg %reg0, <4 x float> inreg %reg1, <4 x float> inreg %reg2, <4 x float> inreg %reg3, <4 x float> inreg %reg4, <4 x float> inreg %reg5, <4 x float> inreg %reg6, <4 x float> inreg %reg7) #0 {
|
||||||
main_body:
|
main_body:
|
||||||
|
Loading…
x
Reference in New Issue
Block a user