setOperationAction(ISD::SIGN_EXTEND, MVT::i64, Custom);
setOperationAction(ISD::ZERO_EXTEND, MVT::i64, Custom);
- setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i1, Custom);
+ setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i1, Legal);
setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v2i1, Custom);
setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v4i1, Custom);
- setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i8, Custom);
+ setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i8, Legal);
setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v2i8, Custom);
setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v4i8, Custom);
- setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i16, Custom);
+ setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i16, Legal);
setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v2i16, Custom);
setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v4i16, Custom);
bool SITargetLowering::allowsUnalignedMemoryAccesses(EVT VT,
unsigned AddrSpace,
bool *IsFast) const {
+ if (IsFast)
+ *IsFast = false;
+
// XXX: This depends on the address space and also we may want to revist
// the alignment values we specify in the DataLayout.
+
+ // TODO: I think v3i32 should allow unaligned accesses on CI with DS_READ_B96,
+ // which isn't a simple VT.
if (!VT.isSimple() || VT == MVT::Other)
return false;
+
+ // XXX - CI changes say "Support for unaligned memory accesses" but I don't
+ // see what for specifically. The wording everywhere else seems to be the
+ // same.
+
+ // 3.6.4 - Operations using pairs of VGPRs (for example: double-floats) have
+ // no alignment restrictions.
+ if (AddrSpace == AMDGPUAS::PRIVATE_ADDRESS) {
+ // Using any pair of GPRs should be the same as any other pair.
+ if (IsFast)
+ *IsFast = true;
+ return VT.bitsGE(MVT::i64);
+ }
+
+ // XXX - The only mention I see of this in the ISA manual is for LDS direct
+ // reads the "byte address and must be dword aligned". Is it also true for the
+ // normal loads and stores?
+ if (AddrSpace == AMDGPUAS::LOCAL_ADDRESS)
+ return false;
+
+ // 8.1.6 - For Dword or larger reads or writes, the two LSBs of the
+ // byte-address are ignored, thus forcing Dword alignment.
+ if (IsFast)
+ *IsFast = true;
return VT.bitsGT(MVT::i32);
}
if (I->getOpcode() == Opcode)
return *I;
}
- return 0;
+ return nullptr;
}
/// This transforms the control flow intrinsics to get the branch destination as
SDNode *Intr = BRCOND.getOperand(1).getNode();
SDValue Target = BRCOND.getOperand(2);
- SDNode *BR = 0;
+ SDNode *BR = nullptr;
if (Intr->getOpcode() == ISD::SETCC) {
// As long as we negate the condition everything is fine
SDValue Arg0 = N->getOperand(0);
SDValue Arg1 = N->getOperand(1);
SDValue CC = N->getOperand(2);
- ConstantSDNode * C = NULL;
+ ConstantSDNode * C = nullptr;
ISD::CondCode CCOp = dyn_cast<CondCodeSDNode>(CC)->get();
// i1 setcc (sext(i1), 0, setne) -> i1 setcc(i1, 0, setne)
MachineSDNode *Mov = dyn_cast<MachineSDNode>(Operand);
const SIInstrInfo *TII =
static_cast<const SIInstrInfo*>(getTargetMachine().getInstrInfo());
- if (Mov == 0 || !TII->isMov(Mov->getMachineOpcode()))
+ if (!Mov || !TII->isMov(Mov->getMachineOpcode()))
return false;
const SDValue &Op = Mov->getOperand(0);
}
return TRI.getPhysRegClass(Reg);
}
- default: return NULL;
+ default: return nullptr;
}
}
const MCInstrDesc &Desc = TII->get(Op->getMachineOpcode());
// Commuted opcode if available
int OpcodeRev = Desc->isCommutable() ? TII->commuteOpcode(Opcode) : -1;
- const MCInstrDesc *DescRev = OpcodeRev == -1 ? 0 : &TII->get(OpcodeRev);
+ const MCInstrDesc *DescRev = OpcodeRev == -1 ? nullptr : &TII->get(OpcodeRev);
assert(!DescRev || DescRev->getNumDefs() == NumDefs);
assert(!DescRev || DescRev->getNumOperands() == NumOps);
// e64 version if available, -1 otherwise
int OpcodeE64 = AMDGPU::getVOPe64(Opcode);
- const MCInstrDesc *DescE64 = OpcodeE64 == -1 ? 0 : &TII->get(OpcodeE64);
+ const MCInstrDesc *DescE64 = OpcodeE64 == -1 ? nullptr : &TII->get(OpcodeE64);
assert(!DescE64 || DescE64->getNumDefs() == NumDefs);
assert(!DescE64 || DescE64->getNumOperands() == (NumOps + 4));
std::swap(Ops[0], Ops[1]);
Desc = DescRev;
- DescRev = 0;
+ DescRev = nullptr;
continue;
}
}
Immediate = -1;
Promote2e64 = true;
Desc = DescE64;
- DescE64 = 0;
+ DescE64 = nullptr;
}
}
}