RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2024-12-26 05:56:51 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

R600: Reenable llvm.R600.load.input/interp.input for compatibility

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194484 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Vincent Lejeune 2013-11-12 16:26:47 +00:00
parent 9e2838e29b
commit 837dd95d6c
2 changed files with 47 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

45
lib/Target/R600/R600ISelLowering.cpp
View File

@ -554,6 +554,51 @@ SDValue R600TargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG) const
SDLoc DL(Op); SDLoc DL(Op);
switch(IntrinsicID) { switch(IntrinsicID) {
default: return AMDGPUTargetLowering::LowerOperation(Op, DAG); default: return AMDGPUTargetLowering::LowerOperation(Op, DAG);
case AMDGPUIntrinsic::R600_load_input: {
int64_t RegIndex = cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue();
unsigned Reg = AMDGPU::R600_TReg32RegClass.getRegister(RegIndex);
MachineFunction &MF = DAG.getMachineFunction();
MachineRegisterInfo &MRI = MF.getRegInfo();
MRI.addLiveIn(Reg);
return DAG.getCopyFromReg(DAG.getEntryNode(),
SDLoc(DAG.getEntryNode()), Reg, VT);
}
case AMDGPUIntrinsic::R600_interp_input: {
int slot = cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue();
int ijb = cast<ConstantSDNode>(Op.getOperand(2))->getSExtValue();
MachineSDNode *interp;
if (ijb < 0) {
const MachineFunction &MF = DAG.getMachineFunction();
const R600InstrInfo *TII =
static_cast<const R600InstrInfo*>(MF.getTarget().getInstrInfo());
interp = DAG.getMachineNode(AMDGPU::INTERP_VEC_LOAD, DL,
MVT::v4f32, DAG.getTargetConstant(slot / 4 , MVT::i32));
return DAG.getTargetExtractSubreg(
TII->getRegisterInfo().getSubRegFromChannel(slot % 4),
DL, MVT::f32, SDValue(interp, 0));
}
MachineFunction &MF = DAG.getMachineFunction();
MachineRegisterInfo &MRI = MF.getRegInfo();
unsigned RegisterI = AMDGPU::R600_TReg32RegClass.getRegister(2 * ijb);
unsigned RegisterJ = AMDGPU::R600_TReg32RegClass.getRegister(2 * ijb + 1);
MRI.addLiveIn(RegisterI);
MRI.addLiveIn(RegisterJ);
SDValue RegisterINode = DAG.getCopyFromReg(DAG.getEntryNode(),
SDLoc(DAG.getEntryNode()), RegisterI, MVT::f32);
SDValue RegisterJNode = DAG.getCopyFromReg(DAG.getEntryNode(),
SDLoc(DAG.getEntryNode()), RegisterJ, MVT::f32);
if (slot % 4 < 2)
interp = DAG.getMachineNode(AMDGPU::INTERP_PAIR_XY, DL,
MVT::f32, MVT::f32, DAG.getTargetConstant(slot / 4 , MVT::i32),
RegisterJNode, RegisterINode);
else
interp = DAG.getMachineNode(AMDGPU::INTERP_PAIR_ZW, DL,
MVT::f32, MVT::f32, DAG.getTargetConstant(slot / 4 , MVT::i32),
RegisterJNode, RegisterINode);
return SDValue(interp, slot % 2);
}
case AMDGPUIntrinsic::R600_interp_xy: case AMDGPUIntrinsic::R600_interp_xy:
case AMDGPUIntrinsic::R600_interp_zw: { case AMDGPUIntrinsic::R600_interp_zw: {
int slot = cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue(); int slot = cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue();

2
lib/Target/R600/R600Intrinsics.td
View File

@ -39,6 +39,8 @@ let TargetPrefix = "R600", isTarget = 1 in {
llvm_i32_ty // coord_type_w llvm_i32_ty // coord_type_w
], [IntrNoMem]>; ], [IntrNoMem]>;
def int_R600_load_input :
Intrinsic<[llvm_float_ty], [llvm_i32_ty], [IntrNoMem]>;
def int_R600_interp_input : def int_R600_interp_input :
Intrinsic<[llvm_float_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>; Intrinsic<[llvm_float_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
def int_R600_interp_const : def int_R600_interp_const :