let PrintMethod = "printNEONModImmOperand";
let ParserMatchClass = nImmSplatI32AsmOperand;
}
+def nImmSplatNotI16AsmOperand : AsmOperandClass { let Name = "NEONi16splatNot"; }
+def nImmSplatNotI16 : Operand<i32> {
+ let ParserMatchClass = nImmSplatNotI16AsmOperand;
+}
+def nImmSplatNotI32AsmOperand : AsmOperandClass { let Name = "NEONi32splatNot"; }
+def nImmSplatNotI32 : Operand<i32> {
+ let ParserMatchClass = nImmSplatNotI32AsmOperand;
+}
def nImmVMOVI32AsmOperand : AsmOperandClass { let Name = "NEONi32vmov"; }
def nImmVMOVI32 : Operand<i32> {
let PrintMethod = "printNEONModImmOperand";
let ParserMatchClass = nImmVMOVI32AsmOperand;
}
+
+def nImmVMOVI16AsmOperandByteReplicate :
+ AsmOperandClass {
+ let Name = "NEONi16vmovByteReplicate";
+ let PredicateMethod = "isNEONi16ByteReplicate";
+ let RenderMethod = "addNEONvmovByteReplicateOperands";
+}
+def nImmVMOVI32AsmOperandByteReplicate :
+ AsmOperandClass {
+ let Name = "NEONi32vmovByteReplicate";
+ let PredicateMethod = "isNEONi32ByteReplicate";
+ let RenderMethod = "addNEONvmovByteReplicateOperands";
+}
+def nImmVMVNI16AsmOperandByteReplicate :
+ AsmOperandClass {
+ let Name = "NEONi16invByteReplicate";
+ let PredicateMethod = "isNEONi16ByteReplicate";
+ let RenderMethod = "addNEONinvByteReplicateOperands";
+}
+def nImmVMVNI32AsmOperandByteReplicate :
+ AsmOperandClass {
+ let Name = "NEONi32invByteReplicate";
+ let PredicateMethod = "isNEONi32ByteReplicate";
+ let RenderMethod = "addNEONinvByteReplicateOperands";
+}
+
+def nImmVMOVI16ByteReplicate : Operand<i32> {
+ let PrintMethod = "printNEONModImmOperand";
+ let ParserMatchClass = nImmVMOVI16AsmOperandByteReplicate;
+}
+def nImmVMOVI32ByteReplicate : Operand<i32> {
+ let PrintMethod = "printNEONModImmOperand";
+ let ParserMatchClass = nImmVMOVI32AsmOperandByteReplicate;
+}
+def nImmVMVNI16ByteReplicate : Operand<i32> {
+ let PrintMethod = "printNEONModImmOperand";
+ let ParserMatchClass = nImmVMVNI16AsmOperandByteReplicate;
+}
+def nImmVMVNI32ByteReplicate : Operand<i32> {
+ let PrintMethod = "printNEONModImmOperand";
+ let ParserMatchClass = nImmVMVNI32AsmOperandByteReplicate;
+}
+
def nImmVMOVI32NegAsmOperand : AsmOperandClass { let Name = "NEONi32vmovNeg"; }
def nImmVMOVI32Neg : Operand<i32> {
let PrintMethod = "printNEONModImmOperand";
let MIOperandInfo = (ops DPR:$Vd, i32imm:$idx);
}
+def dword_alignedload : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+ return cast<LoadSDNode>(N)->getAlignment() >= 8;
+}]>;
+def dword_alignedstore : PatFrag<(ops node:$val, node:$ptr),
+ (store node:$val, node:$ptr), [{
+ return cast<StoreSDNode>(N)->getAlignment() >= 8;
+}]>;
+def word_alignedload : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+ return cast<LoadSDNode>(N)->getAlignment() == 4;
+}]>;
+def word_alignedstore : PatFrag<(ops node:$val, node:$ptr),
+ (store node:$val, node:$ptr), [{
+ return cast<StoreSDNode>(N)->getAlignment() == 4;
+}]>;
+def hword_alignedload : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+ return cast<LoadSDNode>(N)->getAlignment() == 2;
+}]>;
+def hword_alignedstore : PatFrag<(ops node:$val, node:$ptr),
+ (store node:$val, node:$ptr), [{
+ return cast<StoreSDNode>(N)->getAlignment() == 2;
+}]>;
+def byte_alignedload : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+ return cast<LoadSDNode>(N)->getAlignment() == 1;
+}]>;
+def byte_alignedstore : PatFrag<(ops node:$val, node:$ptr),
+ (store node:$val, node:$ptr), [{
+ return cast<StoreSDNode>(N)->getAlignment() == 1;
+}]>;
+def non_word_alignedload : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+ return cast<LoadSDNode>(N)->getAlignment() < 4;
+}]>;
+def non_word_alignedstore : PatFrag<(ops node:$val, node:$ptr),
+ (store node:$val, node:$ptr), [{
+ return cast<StoreSDNode>(N)->getAlignment() < 4;
+}]>;
//===----------------------------------------------------------------------===//
// NEON-specific DAG Nodes.
def NEONvshl : SDNode<"ARMISD::VSHL", SDTARMVSH>;
def NEONvshrs : SDNode<"ARMISD::VSHRs", SDTARMVSH>;
def NEONvshru : SDNode<"ARMISD::VSHRu", SDTARMVSH>;
-def NEONvshlls : SDNode<"ARMISD::VSHLLs", SDTARMVSHX>;
-def NEONvshllu : SDNode<"ARMISD::VSHLLu", SDTARMVSHX>;
-def NEONvshlli : SDNode<"ARMISD::VSHLLi", SDTARMVSHX>;
def NEONvshrn : SDNode<"ARMISD::VSHRN", SDTARMVSHX>;
def NEONvrshrs : SDNode<"ARMISD::VRSHRs", SDTARMVSH>;
def NEONvmulls : SDNode<"ARMISD::VMULLs", SDTARMVMULL>;
def NEONvmullu : SDNode<"ARMISD::VMULLu", SDTARMVMULL>;
-def SDTARMFMAX : SDTypeProfile<1, 2, [SDTCisVT<0, f32>, SDTCisSameAs<0, 1>,
- SDTCisSameAs<0, 2>]>;
-def NEONfmax : SDNode<"ARMISD::FMAX", SDTARMFMAX>;
-def NEONfmin : SDNode<"ARMISD::FMIN", SDTARMFMAX>;
-
def NEONimmAllZerosV: PatLeaf<(NEONvmovImm (i32 timm)), [{
ConstantSDNode *ConstVal = cast<ConstantSDNode>(N->getOperand(0));
unsigned EltBits = 0;
(ins addrmode6:$addr, am6offset:$offset, QQQQPR:$src), itin,
"$addr.addr = $wb, $src = $dst">;
-let mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1 in {
+let mayLoad = 1, hasSideEffects = 0, hasExtraDefRegAllocReq = 1 in {
// VLD1 : Vector Load (multiple single elements)
-class VLD1D<bits<4> op7_4, string Dt>
+class VLD1D<bits<4> op7_4, string Dt, Operand AddrMode>
: NLdSt<0,0b10,0b0111,op7_4, (outs VecListOneD:$Vd),
- (ins addrmode6:$Rn), IIC_VLD1,
+ (ins AddrMode:$Rn), IIC_VLD1,
"vld1", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVLDInstruction";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
-class VLD1Q<bits<4> op7_4, string Dt>
+class VLD1Q<bits<4> op7_4, string Dt, Operand AddrMode>
: NLdSt<0,0b10,0b1010,op7_4, (outs VecListDPair:$Vd),
- (ins addrmode6:$Rn), IIC_VLD1x2,
+ (ins AddrMode:$Rn), IIC_VLD1x2,
"vld1", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
-def VLD1d8 : VLD1D<{0,0,0,?}, "8">;
-def VLD1d16 : VLD1D<{0,1,0,?}, "16">;
-def VLD1d32 : VLD1D<{1,0,0,?}, "32">;
-def VLD1d64 : VLD1D<{1,1,0,?}, "64">;
+def VLD1d8 : VLD1D<{0,0,0,?}, "8", addrmode6align64>;
+def VLD1d16 : VLD1D<{0,1,0,?}, "16", addrmode6align64>;
+def VLD1d32 : VLD1D<{1,0,0,?}, "32", addrmode6align64>;
+def VLD1d64 : VLD1D<{1,1,0,?}, "64", addrmode6align64>;
-def VLD1q8 : VLD1Q<{0,0,?,?}, "8">;
-def VLD1q16 : VLD1Q<{0,1,?,?}, "16">;
-def VLD1q32 : VLD1Q<{1,0,?,?}, "32">;
-def VLD1q64 : VLD1Q<{1,1,?,?}, "64">;
+def VLD1q8 : VLD1Q<{0,0,?,?}, "8", addrmode6align64or128>;
+def VLD1q16 : VLD1Q<{0,1,?,?}, "16", addrmode6align64or128>;
+def VLD1q32 : VLD1Q<{1,0,?,?}, "32", addrmode6align64or128>;
+def VLD1q64 : VLD1Q<{1,1,?,?}, "64", addrmode6align64or128>;
// ...with address register writeback:
-multiclass VLD1DWB<bits<4> op7_4, string Dt> {
+multiclass VLD1DWB<bits<4> op7_4, string Dt, Operand AddrMode> {
def _fixed : NLdSt<0,0b10, 0b0111,op7_4, (outs VecListOneD:$Vd, GPR:$wb),
- (ins addrmode6:$Rn), IIC_VLD1u,
+ (ins AddrMode:$Rn), IIC_VLD1u,
"vld1", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVLDInstruction";
- let AsmMatchConverter = "cvtVLDwbFixed";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
def _register : NLdSt<0,0b10,0b0111,op7_4, (outs VecListOneD:$Vd, GPR:$wb),
- (ins addrmode6:$Rn, rGPR:$Rm), IIC_VLD1u,
+ (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1u,
"vld1", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVLDInstruction";
- let AsmMatchConverter = "cvtVLDwbRegister";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
}
-multiclass VLD1QWB<bits<4> op7_4, string Dt> {
+multiclass VLD1QWB<bits<4> op7_4, string Dt, Operand AddrMode> {
def _fixed : NLdSt<0,0b10,0b1010,op7_4, (outs VecListDPair:$Vd, GPR:$wb),
- (ins addrmode6:$Rn), IIC_VLD1x2u,
+ (ins AddrMode:$Rn), IIC_VLD1x2u,
"vld1", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
- let AsmMatchConverter = "cvtVLDwbFixed";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
def _register : NLdSt<0,0b10,0b1010,op7_4, (outs VecListDPair:$Vd, GPR:$wb),
- (ins addrmode6:$Rn, rGPR:$Rm), IIC_VLD1x2u,
+ (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1x2u,
"vld1", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
- let AsmMatchConverter = "cvtVLDwbRegister";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
}
-defm VLD1d8wb : VLD1DWB<{0,0,0,?}, "8">;
-defm VLD1d16wb : VLD1DWB<{0,1,0,?}, "16">;
-defm VLD1d32wb : VLD1DWB<{1,0,0,?}, "32">;
-defm VLD1d64wb : VLD1DWB<{1,1,0,?}, "64">;
-defm VLD1q8wb : VLD1QWB<{0,0,?,?}, "8">;
-defm VLD1q16wb : VLD1QWB<{0,1,?,?}, "16">;
-defm VLD1q32wb : VLD1QWB<{1,0,?,?}, "32">;
-defm VLD1q64wb : VLD1QWB<{1,1,?,?}, "64">;
+defm VLD1d8wb : VLD1DWB<{0,0,0,?}, "8", addrmode6align64>;
+defm VLD1d16wb : VLD1DWB<{0,1,0,?}, "16", addrmode6align64>;
+defm VLD1d32wb : VLD1DWB<{1,0,0,?}, "32", addrmode6align64>;
+defm VLD1d64wb : VLD1DWB<{1,1,0,?}, "64", addrmode6align64>;
+defm VLD1q8wb : VLD1QWB<{0,0,?,?}, "8", addrmode6align64or128>;
+defm VLD1q16wb : VLD1QWB<{0,1,?,?}, "16", addrmode6align64or128>;
+defm VLD1q32wb : VLD1QWB<{1,0,?,?}, "32", addrmode6align64or128>;
+defm VLD1q64wb : VLD1QWB<{1,1,?,?}, "64", addrmode6align64or128>;
// ...with 3 registers
-class VLD1D3<bits<4> op7_4, string Dt>
+class VLD1D3<bits<4> op7_4, string Dt, Operand AddrMode>
: NLdSt<0,0b10,0b0110,op7_4, (outs VecListThreeD:$Vd),
- (ins addrmode6:$Rn), IIC_VLD1x3, "vld1", Dt,
+ (ins AddrMode:$Rn), IIC_VLD1x3, "vld1", Dt,
"$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVLDInstruction";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
-multiclass VLD1D3WB<bits<4> op7_4, string Dt> {
+multiclass VLD1D3WB<bits<4> op7_4, string Dt, Operand AddrMode> {
def _fixed : NLdSt<0,0b10,0b0110, op7_4, (outs VecListThreeD:$Vd, GPR:$wb),
- (ins addrmode6:$Rn), IIC_VLD1x2u,
+ (ins AddrMode:$Rn), IIC_VLD1x2u,
"vld1", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVLDInstruction";
- let AsmMatchConverter = "cvtVLDwbFixed";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
def _register : NLdSt<0,0b10,0b0110,op7_4, (outs VecListThreeD:$Vd, GPR:$wb),
- (ins addrmode6:$Rn, rGPR:$Rm), IIC_VLD1x2u,
+ (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1x2u,
"vld1", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVLDInstruction";
- let AsmMatchConverter = "cvtVLDwbRegister";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
}
-def VLD1d8T : VLD1D3<{0,0,0,?}, "8">;
-def VLD1d16T : VLD1D3<{0,1,0,?}, "16">;
-def VLD1d32T : VLD1D3<{1,0,0,?}, "32">;
-def VLD1d64T : VLD1D3<{1,1,0,?}, "64">;
+def VLD1d8T : VLD1D3<{0,0,0,?}, "8", addrmode6align64>;
+def VLD1d16T : VLD1D3<{0,1,0,?}, "16", addrmode6align64>;
+def VLD1d32T : VLD1D3<{1,0,0,?}, "32", addrmode6align64>;
+def VLD1d64T : VLD1D3<{1,1,0,?}, "64", addrmode6align64>;
-defm VLD1d8Twb : VLD1D3WB<{0,0,0,?}, "8">;
-defm VLD1d16Twb : VLD1D3WB<{0,1,0,?}, "16">;
-defm VLD1d32Twb : VLD1D3WB<{1,0,0,?}, "32">;
-defm VLD1d64Twb : VLD1D3WB<{1,1,0,?}, "64">;
+defm VLD1d8Twb : VLD1D3WB<{0,0,0,?}, "8", addrmode6align64>;
+defm VLD1d16Twb : VLD1D3WB<{0,1,0,?}, "16", addrmode6align64>;
+defm VLD1d32Twb : VLD1D3WB<{1,0,0,?}, "32", addrmode6align64>;
+defm VLD1d64Twb : VLD1D3WB<{1,1,0,?}, "64", addrmode6align64>;
def VLD1d64TPseudo : VLDQQPseudo<IIC_VLD1x3>;
+def VLD1d64TPseudoWB_fixed : VLDQQWBfixedPseudo<IIC_VLD1x3>;
+def VLD1d64TPseudoWB_register : VLDQQWBregisterPseudo<IIC_VLD1x3>;
// ...with 4 registers
-class VLD1D4<bits<4> op7_4, string Dt>
+class VLD1D4<bits<4> op7_4, string Dt, Operand AddrMode>
: NLdSt<0, 0b10, 0b0010, op7_4, (outs VecListFourD:$Vd),
- (ins addrmode6:$Rn), IIC_VLD1x4, "vld1", Dt,
+ (ins AddrMode:$Rn), IIC_VLD1x4, "vld1", Dt,
"$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
-multiclass VLD1D4WB<bits<4> op7_4, string Dt> {
+multiclass VLD1D4WB<bits<4> op7_4, string Dt, Operand AddrMode> {
def _fixed : NLdSt<0,0b10,0b0010, op7_4, (outs VecListFourD:$Vd, GPR:$wb),
- (ins addrmode6:$Rn), IIC_VLD1x2u,
+ (ins AddrMode:$Rn), IIC_VLD1x2u,
"vld1", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
- let AsmMatchConverter = "cvtVLDwbFixed";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
def _register : NLdSt<0,0b10,0b0010,op7_4, (outs VecListFourD:$Vd, GPR:$wb),
- (ins addrmode6:$Rn, rGPR:$Rm), IIC_VLD1x2u,
+ (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1x2u,
"vld1", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
- let AsmMatchConverter = "cvtVLDwbRegister";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
}
-def VLD1d8Q : VLD1D4<{0,0,?,?}, "8">;
-def VLD1d16Q : VLD1D4<{0,1,?,?}, "16">;
-def VLD1d32Q : VLD1D4<{1,0,?,?}, "32">;
-def VLD1d64Q : VLD1D4<{1,1,?,?}, "64">;
+def VLD1d8Q : VLD1D4<{0,0,?,?}, "8", addrmode6align64or128or256>;
+def VLD1d16Q : VLD1D4<{0,1,?,?}, "16", addrmode6align64or128or256>;
+def VLD1d32Q : VLD1D4<{1,0,?,?}, "32", addrmode6align64or128or256>;
+def VLD1d64Q : VLD1D4<{1,1,?,?}, "64", addrmode6align64or128or256>;
-defm VLD1d8Qwb : VLD1D4WB<{0,0,?,?}, "8">;
-defm VLD1d16Qwb : VLD1D4WB<{0,1,?,?}, "16">;
-defm VLD1d32Qwb : VLD1D4WB<{1,0,?,?}, "32">;
-defm VLD1d64Qwb : VLD1D4WB<{1,1,?,?}, "64">;
+defm VLD1d8Qwb : VLD1D4WB<{0,0,?,?}, "8", addrmode6align64or128or256>;
+defm VLD1d16Qwb : VLD1D4WB<{0,1,?,?}, "16", addrmode6align64or128or256>;
+defm VLD1d32Qwb : VLD1D4WB<{1,0,?,?}, "32", addrmode6align64or128or256>;
+defm VLD1d64Qwb : VLD1D4WB<{1,1,?,?}, "64", addrmode6align64or128or256>;
def VLD1d64QPseudo : VLDQQPseudo<IIC_VLD1x4>;
+def VLD1d64QPseudoWB_fixed : VLDQQWBfixedPseudo<IIC_VLD1x4>;
+def VLD1d64QPseudoWB_register : VLDQQWBregisterPseudo<IIC_VLD1x4>;
// VLD2 : Vector Load (multiple 2-element structures)
class VLD2<bits<4> op11_8, bits<4> op7_4, string Dt, RegisterOperand VdTy,
- InstrItinClass itin>
+ InstrItinClass itin, Operand AddrMode>
: NLdSt<0, 0b10, op11_8, op7_4, (outs VdTy:$Vd),
- (ins addrmode6:$Rn), itin,
+ (ins AddrMode:$Rn), itin,
"vld2", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
+ let DecoderMethod = "DecodeVLDST2Instruction";
}
-def VLD2d8 : VLD2<0b1000, {0,0,?,?}, "8", VecListDPair, IIC_VLD2>;
-def VLD2d16 : VLD2<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VLD2>;
-def VLD2d32 : VLD2<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VLD2>;
+def VLD2d8 : VLD2<0b1000, {0,0,?,?}, "8", VecListDPair, IIC_VLD2,
+ addrmode6align64or128>;
+def VLD2d16 : VLD2<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VLD2,
+ addrmode6align64or128>;
+def VLD2d32 : VLD2<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VLD2,
+ addrmode6align64or128>;
-def VLD2q8 : VLD2<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VLD2x2>;
-def VLD2q16 : VLD2<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VLD2x2>;
-def VLD2q32 : VLD2<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VLD2x2>;
+def VLD2q8 : VLD2<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VLD2x2,
+ addrmode6align64or128or256>;
+def VLD2q16 : VLD2<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VLD2x2,
+ addrmode6align64or128or256>;
+def VLD2q32 : VLD2<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VLD2x2,
+ addrmode6align64or128or256>;
def VLD2q8Pseudo : VLDQQPseudo<IIC_VLD2x2>;
def VLD2q16Pseudo : VLDQQPseudo<IIC_VLD2x2>;
// ...with address register writeback:
multiclass VLD2WB<bits<4> op11_8, bits<4> op7_4, string Dt,
- RegisterOperand VdTy, InstrItinClass itin> {
+ RegisterOperand VdTy, InstrItinClass itin, Operand AddrMode> {
def _fixed : NLdSt<0, 0b10, op11_8, op7_4, (outs VdTy:$Vd, GPR:$wb),
- (ins addrmode6:$Rn), itin,
+ (ins AddrMode:$Rn), itin,
"vld2", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
- let AsmMatchConverter = "cvtVLDwbFixed";
+ let DecoderMethod = "DecodeVLDST2Instruction";
}
def _register : NLdSt<0, 0b10, op11_8, op7_4, (outs VdTy:$Vd, GPR:$wb),
- (ins addrmode6:$Rn, rGPR:$Rm), itin,
+ (ins AddrMode:$Rn, rGPR:$Rm), itin,
"vld2", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
- let AsmMatchConverter = "cvtVLDwbRegister";
+ let DecoderMethod = "DecodeVLDST2Instruction";
}
}
-defm VLD2d8wb : VLD2WB<0b1000, {0,0,?,?}, "8", VecListDPair, IIC_VLD2u>;
-defm VLD2d16wb : VLD2WB<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VLD2u>;
-defm VLD2d32wb : VLD2WB<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VLD2u>;
+defm VLD2d8wb : VLD2WB<0b1000, {0,0,?,?}, "8", VecListDPair, IIC_VLD2u,
+ addrmode6align64or128>;
+defm VLD2d16wb : VLD2WB<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VLD2u,
+ addrmode6align64or128>;
+defm VLD2d32wb : VLD2WB<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VLD2u,
+ addrmode6align64or128>;
-defm VLD2q8wb : VLD2WB<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VLD2x2u>;
-defm VLD2q16wb : VLD2WB<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VLD2x2u>;
-defm VLD2q32wb : VLD2WB<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VLD2x2u>;
+defm VLD2q8wb : VLD2WB<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VLD2x2u,
+ addrmode6align64or128or256>;
+defm VLD2q16wb : VLD2WB<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VLD2x2u,
+ addrmode6align64or128or256>;
+defm VLD2q32wb : VLD2WB<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VLD2x2u,
+ addrmode6align64or128or256>;
def VLD2q8PseudoWB_fixed : VLDQQWBfixedPseudo<IIC_VLD2x2u>;
def VLD2q16PseudoWB_fixed : VLDQQWBfixedPseudo<IIC_VLD2x2u>;
def VLD2q32PseudoWB_register : VLDQQWBregisterPseudo<IIC_VLD2x2u>;
// ...with double-spaced registers
-def VLD2b8 : VLD2<0b1001, {0,0,?,?}, "8", VecListDPairSpaced, IIC_VLD2>;
-def VLD2b16 : VLD2<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VLD2>;
-def VLD2b32 : VLD2<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VLD2>;
-defm VLD2b8wb : VLD2WB<0b1001, {0,0,?,?}, "8", VecListDPairSpaced, IIC_VLD2u>;
-defm VLD2b16wb : VLD2WB<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VLD2u>;
-defm VLD2b32wb : VLD2WB<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VLD2u>;
+def VLD2b8 : VLD2<0b1001, {0,0,?,?}, "8", VecListDPairSpaced, IIC_VLD2,
+ addrmode6align64or128>;
+def VLD2b16 : VLD2<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VLD2,
+ addrmode6align64or128>;
+def VLD2b32 : VLD2<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VLD2,
+ addrmode6align64or128>;
+defm VLD2b8wb : VLD2WB<0b1001, {0,0,?,?}, "8", VecListDPairSpaced, IIC_VLD2u,
+ addrmode6align64or128>;
+defm VLD2b16wb : VLD2WB<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VLD2u,
+ addrmode6align64or128>;
+defm VLD2b32wb : VLD2WB<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VLD2u,
+ addrmode6align64or128>;
// VLD3 : Vector Load (multiple 3-element structures)
class VLD3D<bits<4> op11_8, bits<4> op7_4, string Dt>
"vld3", Dt, "\\{$Vd, $dst2, $dst3\\}, $Rn", "", []> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVLDInstruction";
+ let DecoderMethod = "DecodeVLDST3Instruction";
}
def VLD3d8 : VLD3D<0b0100, {0,0,0,?}, "8">;
"vld3", Dt, "\\{$Vd, $dst2, $dst3\\}, $Rn$Rm",
"$Rn.addr = $wb", []> {
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVLDInstruction";
+ let DecoderMethod = "DecodeVLDST3Instruction";
}
def VLD3d8_UPD : VLD3DWB<0b0100, {0,0,0,?}, "8">;
"vld4", Dt, "\\{$Vd, $dst2, $dst3, $dst4\\}, $Rn", "", []> {
let Rm = 0b1111;
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
+ let DecoderMethod = "DecodeVLDST4Instruction";
}
def VLD4d8 : VLD4D<0b0000, {0,0,?,?}, "8">;
"vld4", Dt, "\\{$Vd, $dst2, $dst3, $dst4\\}, $Rn$Rm",
"$Rn.addr = $wb", []> {
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
+ let DecoderMethod = "DecodeVLDST4Instruction";
}
def VLD4d8_UPD : VLD4DWB<0b0000, {0,0,?,?}, "8">;
def VLD4q16oddPseudo_UPD : VLDQQQQWBPseudo<IIC_VLD4u>;
def VLD4q32oddPseudo_UPD : VLDQQQQWBPseudo<IIC_VLD4u>;
-} // mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1
+} // mayLoad = 1, hasSideEffects = 0, hasExtraDefRegAllocReq = 1
// Classes for VLD*LN pseudo-instructions with multi-register operands.
// These are expanded to real instructions after register allocation.
(f32 (load addrmode6:$addr)), imm:$lane),
(VLD1LNq32Pseudo addrmode6:$addr, QPR:$src, imm:$lane)>;
-let mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1 in {
+let mayLoad = 1, hasSideEffects = 0, hasExtraDefRegAllocReq = 1 in {
// ...with address register writeback:
class VLD1LNWB<bits<4> op11_8, bits<4> op7_4, string Dt>
def VLD4LNq16Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD4lnu>;
def VLD4LNq32Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD4lnu>;
-} // mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1
+} // mayLoad = 1, hasSideEffects = 0, hasExtraDefRegAllocReq = 1
// VLD1DUP : Vector Load (single element to all lanes)
-class VLD1DUP<bits<4> op7_4, string Dt, ValueType Ty, PatFrag LoadOp>
+class VLD1DUP<bits<4> op7_4, string Dt, ValueType Ty, PatFrag LoadOp,
+ Operand AddrMode>
: NLdSt<1, 0b10, 0b1100, op7_4, (outs VecListOneDAllLanes:$Vd),
- (ins addrmode6dup:$Rn),
+ (ins AddrMode:$Rn),
IIC_VLD1dup, "vld1", Dt, "$Vd, $Rn", "",
[(set VecListOneDAllLanes:$Vd,
- (Ty (NEONvdup (i32 (LoadOp addrmode6dup:$Rn)))))]> {
+ (Ty (NEONvdup (i32 (LoadOp AddrMode:$Rn)))))]> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD1DupInstruction";
}
-def VLD1DUPd8 : VLD1DUP<{0,0,0,?}, "8", v8i8, extloadi8>;
-def VLD1DUPd16 : VLD1DUP<{0,1,0,?}, "16", v4i16, extloadi16>;
-def VLD1DUPd32 : VLD1DUP<{1,0,0,?}, "32", v2i32, load>;
+def VLD1DUPd8 : VLD1DUP<{0,0,0,?}, "8", v8i8, extloadi8,
+ addrmode6dupalignNone>;
+def VLD1DUPd16 : VLD1DUP<{0,1,0,?}, "16", v4i16, extloadi16,
+ addrmode6dupalign16>;
+def VLD1DUPd32 : VLD1DUP<{1,0,0,?}, "32", v2i32, load,
+ addrmode6dupalign32>;
def : Pat<(v2f32 (NEONvdup (f32 (load addrmode6dup:$addr)))),
(VLD1DUPd32 addrmode6:$addr)>;
-class VLD1QDUP<bits<4> op7_4, string Dt, ValueType Ty, PatFrag LoadOp>
+class VLD1QDUP<bits<4> op7_4, string Dt, ValueType Ty, PatFrag LoadOp,
+ Operand AddrMode>
: NLdSt<1, 0b10, 0b1100, op7_4, (outs VecListDPairAllLanes:$Vd),
- (ins addrmode6dup:$Rn), IIC_VLD1dup,
+ (ins AddrMode:$Rn), IIC_VLD1dup,
"vld1", Dt, "$Vd, $Rn", "",
[(set VecListDPairAllLanes:$Vd,
- (Ty (NEONvdup (i32 (LoadOp addrmode6dup:$Rn)))))]> {
+ (Ty (NEONvdup (i32 (LoadOp AddrMode:$Rn)))))]> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD1DupInstruction";
}
-def VLD1DUPq8 : VLD1QDUP<{0,0,1,0}, "8", v16i8, extloadi8>;
-def VLD1DUPq16 : VLD1QDUP<{0,1,1,?}, "16", v8i16, extloadi16>;
-def VLD1DUPq32 : VLD1QDUP<{1,0,1,?}, "32", v4i32, load>;
+def VLD1DUPq8 : VLD1QDUP<{0,0,1,0}, "8", v16i8, extloadi8,
+ addrmode6dupalignNone>;
+def VLD1DUPq16 : VLD1QDUP<{0,1,1,?}, "16", v8i16, extloadi16,
+ addrmode6dupalign16>;
+def VLD1DUPq32 : VLD1QDUP<{1,0,1,?}, "32", v4i32, load,
+ addrmode6dupalign32>;
def : Pat<(v4f32 (NEONvdup (f32 (load addrmode6dup:$addr)))),
(VLD1DUPq32 addrmode6:$addr)>;
-let mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1 in {
+let mayLoad = 1, hasSideEffects = 0, hasExtraDefRegAllocReq = 1 in {
// ...with address register writeback:
-multiclass VLD1DUPWB<bits<4> op7_4, string Dt> {
+multiclass VLD1DUPWB<bits<4> op7_4, string Dt, Operand AddrMode> {
def _fixed : NLdSt<1, 0b10, 0b1100, op7_4,
(outs VecListOneDAllLanes:$Vd, GPR:$wb),
- (ins addrmode6dup:$Rn), IIC_VLD1dupu,
+ (ins AddrMode:$Rn), IIC_VLD1dupu,
"vld1", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD1DupInstruction";
- let AsmMatchConverter = "cvtVLDwbFixed";
}
def _register : NLdSt<1, 0b10, 0b1100, op7_4,
(outs VecListOneDAllLanes:$Vd, GPR:$wb),
- (ins addrmode6dup:$Rn, rGPR:$Rm), IIC_VLD1dupu,
+ (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1dupu,
"vld1", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD1DupInstruction";
- let AsmMatchConverter = "cvtVLDwbRegister";
}
}
-multiclass VLD1QDUPWB<bits<4> op7_4, string Dt> {
+multiclass VLD1QDUPWB<bits<4> op7_4, string Dt, Operand AddrMode> {
def _fixed : NLdSt<1, 0b10, 0b1100, op7_4,
(outs VecListDPairAllLanes:$Vd, GPR:$wb),
- (ins addrmode6dup:$Rn), IIC_VLD1dupu,
+ (ins AddrMode:$Rn), IIC_VLD1dupu,
"vld1", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD1DupInstruction";
- let AsmMatchConverter = "cvtVLDwbFixed";
}
def _register : NLdSt<1, 0b10, 0b1100, op7_4,
(outs VecListDPairAllLanes:$Vd, GPR:$wb),
- (ins addrmode6dup:$Rn, rGPR:$Rm), IIC_VLD1dupu,
+ (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD1dupu,
"vld1", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD1DupInstruction";
- let AsmMatchConverter = "cvtVLDwbRegister";
}
}
-defm VLD1DUPd8wb : VLD1DUPWB<{0,0,0,0}, "8">;
-defm VLD1DUPd16wb : VLD1DUPWB<{0,1,0,?}, "16">;
-defm VLD1DUPd32wb : VLD1DUPWB<{1,0,0,?}, "32">;
+defm VLD1DUPd8wb : VLD1DUPWB<{0,0,0,0}, "8", addrmode6dupalignNone>;
+defm VLD1DUPd16wb : VLD1DUPWB<{0,1,0,?}, "16", addrmode6dupalign16>;
+defm VLD1DUPd32wb : VLD1DUPWB<{1,0,0,?}, "32", addrmode6dupalign32>;
-defm VLD1DUPq8wb : VLD1QDUPWB<{0,0,1,0}, "8">;
-defm VLD1DUPq16wb : VLD1QDUPWB<{0,1,1,?}, "16">;
-defm VLD1DUPq32wb : VLD1QDUPWB<{1,0,1,?}, "32">;
+defm VLD1DUPq8wb : VLD1QDUPWB<{0,0,1,0}, "8", addrmode6dupalignNone>;
+defm VLD1DUPq16wb : VLD1QDUPWB<{0,1,1,?}, "16", addrmode6dupalign16>;
+defm VLD1DUPq32wb : VLD1QDUPWB<{1,0,1,?}, "32", addrmode6dupalign32>;
// VLD2DUP : Vector Load (single 2-element structure to all lanes)
-class VLD2DUP<bits<4> op7_4, string Dt, RegisterOperand VdTy>
+class VLD2DUP<bits<4> op7_4, string Dt, RegisterOperand VdTy, Operand AddrMode>
: NLdSt<1, 0b10, 0b1101, op7_4, (outs VdTy:$Vd),
- (ins addrmode6dup:$Rn), IIC_VLD2dup,
+ (ins AddrMode:$Rn), IIC_VLD2dup,
"vld2", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD2DupInstruction";
}
-def VLD2DUPd8 : VLD2DUP<{0,0,0,?}, "8", VecListDPairAllLanes>;
-def VLD2DUPd16 : VLD2DUP<{0,1,0,?}, "16", VecListDPairAllLanes>;
-def VLD2DUPd32 : VLD2DUP<{1,0,0,?}, "32", VecListDPairAllLanes>;
+def VLD2DUPd8 : VLD2DUP<{0,0,0,?}, "8", VecListDPairAllLanes,
+ addrmode6dupalign16>;
+def VLD2DUPd16 : VLD2DUP<{0,1,0,?}, "16", VecListDPairAllLanes,
+ addrmode6dupalign32>;
+def VLD2DUPd32 : VLD2DUP<{1,0,0,?}, "32", VecListDPairAllLanes,
+ addrmode6dupalign64>;
+// HACK this one, VLD2DUPd8x2 must be changed at the same time with VLD2b8 or
+// "vld2.8 {d0[], d2[]}, [r4:32]" will become "vld2.8 {d0, d2}, [r4:32]".
// ...with double-spaced registers
-def VLD2DUPd8x2 : VLD2DUP<{0,0,1,?}, "8", VecListDPairSpacedAllLanes>;
-def VLD2DUPd16x2 : VLD2DUP<{0,1,1,?}, "16", VecListDPairSpacedAllLanes>;
-def VLD2DUPd32x2 : VLD2DUP<{1,0,1,?}, "32", VecListDPairSpacedAllLanes>;
+def VLD2DUPd8x2 : VLD2DUP<{0,0,1,?}, "8", VecListDPairSpacedAllLanes,
+ addrmode6dupalign16>;
+def VLD2DUPd16x2 : VLD2DUP<{0,1,1,?}, "16", VecListDPairSpacedAllLanes,
+ addrmode6dupalign32>;
+def VLD2DUPd32x2 : VLD2DUP<{1,0,1,?}, "32", VecListDPairSpacedAllLanes,
+ addrmode6dupalign64>;
// ...with address register writeback:
-multiclass VLD2DUPWB<bits<4> op7_4, string Dt, RegisterOperand VdTy> {
+multiclass VLD2DUPWB<bits<4> op7_4, string Dt, RegisterOperand VdTy,
+ Operand AddrMode> {
def _fixed : NLdSt<1, 0b10, 0b1101, op7_4,
(outs VdTy:$Vd, GPR:$wb),
- (ins addrmode6dup:$Rn), IIC_VLD2dupu,
+ (ins AddrMode:$Rn), IIC_VLD2dupu,
"vld2", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD2DupInstruction";
- let AsmMatchConverter = "cvtVLDwbFixed";
}
def _register : NLdSt<1, 0b10, 0b1101, op7_4,
(outs VdTy:$Vd, GPR:$wb),
- (ins addrmode6dup:$Rn, rGPR:$Rm), IIC_VLD2dupu,
+ (ins AddrMode:$Rn, rGPR:$Rm), IIC_VLD2dupu,
"vld2", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD2DupInstruction";
- let AsmMatchConverter = "cvtVLDwbRegister";
}
}
-defm VLD2DUPd8wb : VLD2DUPWB<{0,0,0,0}, "8", VecListDPairAllLanes>;
-defm VLD2DUPd16wb : VLD2DUPWB<{0,1,0,?}, "16", VecListDPairAllLanes>;
-defm VLD2DUPd32wb : VLD2DUPWB<{1,0,0,?}, "32", VecListDPairAllLanes>;
+defm VLD2DUPd8wb : VLD2DUPWB<{0,0,0,0}, "8", VecListDPairAllLanes,
+ addrmode6dupalign16>;
+defm VLD2DUPd16wb : VLD2DUPWB<{0,1,0,?}, "16", VecListDPairAllLanes,
+ addrmode6dupalign32>;
+defm VLD2DUPd32wb : VLD2DUPWB<{1,0,0,?}, "32", VecListDPairAllLanes,
+ addrmode6dupalign64>;
-defm VLD2DUPd8x2wb : VLD2DUPWB<{0,0,1,0}, "8", VecListDPairSpacedAllLanes>;
-defm VLD2DUPd16x2wb : VLD2DUPWB<{0,1,1,?}, "16", VecListDPairSpacedAllLanes>;
-defm VLD2DUPd32x2wb : VLD2DUPWB<{1,0,1,?}, "32", VecListDPairSpacedAllLanes>;
+defm VLD2DUPd8x2wb : VLD2DUPWB<{0,0,1,0}, "8", VecListDPairSpacedAllLanes,
+ addrmode6dupalign16>;
+defm VLD2DUPd16x2wb : VLD2DUPWB<{0,1,1,?}, "16", VecListDPairSpacedAllLanes,
+ addrmode6dupalign32>;
+defm VLD2DUPd32x2wb : VLD2DUPWB<{1,0,1,?}, "32", VecListDPairSpacedAllLanes,
+ addrmode6dupalign64>;
// VLD3DUP : Vector Load (single 3-element structure to all lanes)
class VLD3DUP<bits<4> op7_4, string Dt>
def VLD3DUPq32 : VLD3DUP<{1,0,1,?}, "32">;
// ...with address register writeback:
-class VLD3DUPWB<bits<4> op7_4, string Dt>
+class VLD3DUPWB<bits<4> op7_4, string Dt, Operand AddrMode>
: NLdSt<1, 0b10, 0b1110, op7_4, (outs DPR:$Vd, DPR:$dst2, DPR:$dst3, GPR:$wb),
- (ins addrmode6dup:$Rn, am6offset:$Rm), IIC_VLD3dupu,
+ (ins AddrMode:$Rn, am6offset:$Rm), IIC_VLD3dupu,
"vld3", Dt, "\\{$Vd[], $dst2[], $dst3[]\\}, $Rn$Rm",
"$Rn.addr = $wb", []> {
let Inst{4} = 0;
let DecoderMethod = "DecodeVLD3DupInstruction";
}
-def VLD3DUPd8_UPD : VLD3DUPWB<{0,0,0,0}, "8">;
-def VLD3DUPd16_UPD : VLD3DUPWB<{0,1,0,?}, "16">;
-def VLD3DUPd32_UPD : VLD3DUPWB<{1,0,0,?}, "32">;
+def VLD3DUPd8_UPD : VLD3DUPWB<{0,0,0,0}, "8", addrmode6dupalign64>;
+def VLD3DUPd16_UPD : VLD3DUPWB<{0,1,0,?}, "16", addrmode6dupalign64>;
+def VLD3DUPd32_UPD : VLD3DUPWB<{1,0,0,?}, "32", addrmode6dupalign64>;
-def VLD3DUPq8_UPD : VLD3DUPWB<{0,0,1,0}, "8">;
-def VLD3DUPq16_UPD : VLD3DUPWB<{0,1,1,?}, "16">;
-def VLD3DUPq32_UPD : VLD3DUPWB<{1,0,1,?}, "32">;
+def VLD3DUPq8_UPD : VLD3DUPWB<{0,0,1,0}, "8", addrmode6dupalign64>;
+def VLD3DUPq16_UPD : VLD3DUPWB<{0,1,1,?}, "16", addrmode6dupalign64>;
+def VLD3DUPq32_UPD : VLD3DUPWB<{1,0,1,?}, "32", addrmode6dupalign64>;
def VLD3DUPd8Pseudo_UPD : VLDQQWBPseudo<IIC_VLD3dupu>;
def VLD3DUPd16Pseudo_UPD : VLDQQWBPseudo<IIC_VLD3dupu>;
def VLD4DUPd16Pseudo_UPD : VLDQQWBPseudo<IIC_VLD4dupu>;
def VLD4DUPd32Pseudo_UPD : VLDQQWBPseudo<IIC_VLD4dupu>;
-} // mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1
+} // mayLoad = 1, hasSideEffects = 0, hasExtraDefRegAllocReq = 1
-let mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1 in {
+let mayStore = 1, hasSideEffects = 0, hasExtraSrcRegAllocReq = 1 in {
// Classes for VST* pseudo-instructions with multi-register operands.
// These are expanded to real instructions after register allocation.
"$addr.addr = $wb">;
// VST1 : Vector Store (multiple single elements)
-class VST1D<bits<4> op7_4, string Dt>
- : NLdSt<0,0b00,0b0111,op7_4, (outs), (ins addrmode6:$Rn, VecListOneD:$Vd),
+class VST1D<bits<4> op7_4, string Dt, Operand AddrMode>
+ : NLdSt<0,0b00,0b0111,op7_4, (outs), (ins AddrMode:$Rn, VecListOneD:$Vd),
IIC_VST1, "vst1", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVSTInstruction";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
-class VST1Q<bits<4> op7_4, string Dt>
- : NLdSt<0,0b00,0b1010,op7_4, (outs), (ins addrmode6:$Rn, VecListDPair:$Vd),
+class VST1Q<bits<4> op7_4, string Dt, Operand AddrMode>
+ : NLdSt<0,0b00,0b1010,op7_4, (outs), (ins AddrMode:$Rn, VecListDPair:$Vd),
IIC_VST1x2, "vst1", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
-def VST1d8 : VST1D<{0,0,0,?}, "8">;
-def VST1d16 : VST1D<{0,1,0,?}, "16">;
-def VST1d32 : VST1D<{1,0,0,?}, "32">;
-def VST1d64 : VST1D<{1,1,0,?}, "64">;
+def VST1d8 : VST1D<{0,0,0,?}, "8", addrmode6align64>;
+def VST1d16 : VST1D<{0,1,0,?}, "16", addrmode6align64>;
+def VST1d32 : VST1D<{1,0,0,?}, "32", addrmode6align64>;
+def VST1d64 : VST1D<{1,1,0,?}, "64", addrmode6align64>;
-def VST1q8 : VST1Q<{0,0,?,?}, "8">;
-def VST1q16 : VST1Q<{0,1,?,?}, "16">;
-def VST1q32 : VST1Q<{1,0,?,?}, "32">;
-def VST1q64 : VST1Q<{1,1,?,?}, "64">;
+def VST1q8 : VST1Q<{0,0,?,?}, "8", addrmode6align64or128>;
+def VST1q16 : VST1Q<{0,1,?,?}, "16", addrmode6align64or128>;
+def VST1q32 : VST1Q<{1,0,?,?}, "32", addrmode6align64or128>;
+def VST1q64 : VST1Q<{1,1,?,?}, "64", addrmode6align64or128>;
// ...with address register writeback:
-multiclass VST1DWB<bits<4> op7_4, string Dt> {
+multiclass VST1DWB<bits<4> op7_4, string Dt, Operand AddrMode> {
def _fixed : NLdSt<0,0b00, 0b0111,op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, VecListOneD:$Vd), IIC_VLD1u,
+ (ins AddrMode:$Rn, VecListOneD:$Vd), IIC_VLD1u,
"vst1", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbFixed";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
def _register : NLdSt<0,0b00,0b0111,op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, rGPR:$Rm, VecListOneD:$Vd),
+ (ins AddrMode:$Rn, rGPR:$Rm, VecListOneD:$Vd),
IIC_VLD1u,
"vst1", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbRegister";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
}
-multiclass VST1QWB<bits<4> op7_4, string Dt> {
+multiclass VST1QWB<bits<4> op7_4, string Dt, Operand AddrMode> {
def _fixed : NLdSt<0,0b00,0b1010,op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, VecListDPair:$Vd), IIC_VLD1x2u,
+ (ins AddrMode:$Rn, VecListDPair:$Vd), IIC_VLD1x2u,
"vst1", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbFixed";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
def _register : NLdSt<0,0b00,0b1010,op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, rGPR:$Rm, VecListDPair:$Vd),
+ (ins AddrMode:$Rn, rGPR:$Rm, VecListDPair:$Vd),
IIC_VLD1x2u,
"vst1", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbRegister";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
}
-defm VST1d8wb : VST1DWB<{0,0,0,?}, "8">;
-defm VST1d16wb : VST1DWB<{0,1,0,?}, "16">;
-defm VST1d32wb : VST1DWB<{1,0,0,?}, "32">;
-defm VST1d64wb : VST1DWB<{1,1,0,?}, "64">;
+defm VST1d8wb : VST1DWB<{0,0,0,?}, "8", addrmode6align64>;
+defm VST1d16wb : VST1DWB<{0,1,0,?}, "16", addrmode6align64>;
+defm VST1d32wb : VST1DWB<{1,0,0,?}, "32", addrmode6align64>;
+defm VST1d64wb : VST1DWB<{1,1,0,?}, "64", addrmode6align64>;
-defm VST1q8wb : VST1QWB<{0,0,?,?}, "8">;
-defm VST1q16wb : VST1QWB<{0,1,?,?}, "16">;
-defm VST1q32wb : VST1QWB<{1,0,?,?}, "32">;
-defm VST1q64wb : VST1QWB<{1,1,?,?}, "64">;
+defm VST1q8wb : VST1QWB<{0,0,?,?}, "8", addrmode6align64or128>;
+defm VST1q16wb : VST1QWB<{0,1,?,?}, "16", addrmode6align64or128>;
+defm VST1q32wb : VST1QWB<{1,0,?,?}, "32", addrmode6align64or128>;
+defm VST1q64wb : VST1QWB<{1,1,?,?}, "64", addrmode6align64or128>;
// ...with 3 registers
-class VST1D3<bits<4> op7_4, string Dt>
+class VST1D3<bits<4> op7_4, string Dt, Operand AddrMode>
: NLdSt<0, 0b00, 0b0110, op7_4, (outs),
- (ins addrmode6:$Rn, VecListThreeD:$Vd),
+ (ins AddrMode:$Rn, VecListThreeD:$Vd),
IIC_VST1x3, "vst1", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVSTInstruction";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
-multiclass VST1D3WB<bits<4> op7_4, string Dt> {
+multiclass VST1D3WB<bits<4> op7_4, string Dt, Operand AddrMode> {
def _fixed : NLdSt<0,0b00,0b0110,op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, VecListThreeD:$Vd), IIC_VLD1x3u,
+ (ins AddrMode:$Rn, VecListThreeD:$Vd), IIC_VLD1x3u,
"vst1", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbFixed";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
def _register : NLdSt<0,0b00,0b0110,op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, rGPR:$Rm, VecListThreeD:$Vd),
+ (ins AddrMode:$Rn, rGPR:$Rm, VecListThreeD:$Vd),
IIC_VLD1x3u,
"vst1", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbRegister";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
}
-def VST1d8T : VST1D3<{0,0,0,?}, "8">;
-def VST1d16T : VST1D3<{0,1,0,?}, "16">;
-def VST1d32T : VST1D3<{1,0,0,?}, "32">;
-def VST1d64T : VST1D3<{1,1,0,?}, "64">;
+def VST1d8T : VST1D3<{0,0,0,?}, "8", addrmode6align64>;
+def VST1d16T : VST1D3<{0,1,0,?}, "16", addrmode6align64>;
+def VST1d32T : VST1D3<{1,0,0,?}, "32", addrmode6align64>;
+def VST1d64T : VST1D3<{1,1,0,?}, "64", addrmode6align64>;
-defm VST1d8Twb : VST1D3WB<{0,0,0,?}, "8">;
-defm VST1d16Twb : VST1D3WB<{0,1,0,?}, "16">;
-defm VST1d32Twb : VST1D3WB<{1,0,0,?}, "32">;
-defm VST1d64Twb : VST1D3WB<{1,1,0,?}, "64">;
+defm VST1d8Twb : VST1D3WB<{0,0,0,?}, "8", addrmode6align64>;
+defm VST1d16Twb : VST1D3WB<{0,1,0,?}, "16", addrmode6align64>;
+defm VST1d32Twb : VST1D3WB<{1,0,0,?}, "32", addrmode6align64>;
+defm VST1d64Twb : VST1D3WB<{1,1,0,?}, "64", addrmode6align64>;
def VST1d64TPseudo : VSTQQPseudo<IIC_VST1x3>;
-def VST1d64TPseudoWB_fixed : VSTQQWBPseudo<IIC_VST1x3u>;
+def VST1d64TPseudoWB_fixed : VSTQQWBfixedPseudo<IIC_VST1x3u>;
def VST1d64TPseudoWB_register : VSTQQWBPseudo<IIC_VST1x3u>;
// ...with 4 registers
-class VST1D4<bits<4> op7_4, string Dt>
+class VST1D4<bits<4> op7_4, string Dt, Operand AddrMode>
: NLdSt<0, 0b00, 0b0010, op7_4, (outs),
- (ins addrmode6:$Rn, VecListFourD:$Vd),
+ (ins AddrMode:$Rn, VecListFourD:$Vd),
IIC_VST1x4, "vst1", Dt, "$Vd, $Rn", "",
[]> {
let Rm = 0b1111;
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
-multiclass VST1D4WB<bits<4> op7_4, string Dt> {
+multiclass VST1D4WB<bits<4> op7_4, string Dt, Operand AddrMode> {
def _fixed : NLdSt<0,0b00,0b0010,op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, VecListFourD:$Vd), IIC_VLD1x4u,
+ (ins AddrMode:$Rn, VecListFourD:$Vd), IIC_VLD1x4u,
"vst1", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbFixed";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
def _register : NLdSt<0,0b00,0b0010,op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, rGPR:$Rm, VecListFourD:$Vd),
+ (ins AddrMode:$Rn, rGPR:$Rm, VecListFourD:$Vd),
IIC_VLD1x4u,
"vst1", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbRegister";
+ let DecoderMethod = "DecodeVLDST1Instruction";
}
}
-def VST1d8Q : VST1D4<{0,0,?,?}, "8">;
-def VST1d16Q : VST1D4<{0,1,?,?}, "16">;
-def VST1d32Q : VST1D4<{1,0,?,?}, "32">;
-def VST1d64Q : VST1D4<{1,1,?,?}, "64">;
+def VST1d8Q : VST1D4<{0,0,?,?}, "8", addrmode6align64or128or256>;
+def VST1d16Q : VST1D4<{0,1,?,?}, "16", addrmode6align64or128or256>;
+def VST1d32Q : VST1D4<{1,0,?,?}, "32", addrmode6align64or128or256>;
+def VST1d64Q : VST1D4<{1,1,?,?}, "64", addrmode6align64or128or256>;
-defm VST1d8Qwb : VST1D4WB<{0,0,?,?}, "8">;
-defm VST1d16Qwb : VST1D4WB<{0,1,?,?}, "16">;
-defm VST1d32Qwb : VST1D4WB<{1,0,?,?}, "32">;
-defm VST1d64Qwb : VST1D4WB<{1,1,?,?}, "64">;
+defm VST1d8Qwb : VST1D4WB<{0,0,?,?}, "8", addrmode6align64or128or256>;
+defm VST1d16Qwb : VST1D4WB<{0,1,?,?}, "16", addrmode6align64or128or256>;
+defm VST1d32Qwb : VST1D4WB<{1,0,?,?}, "32", addrmode6align64or128or256>;
+defm VST1d64Qwb : VST1D4WB<{1,1,?,?}, "64", addrmode6align64or128or256>;
def VST1d64QPseudo : VSTQQPseudo<IIC_VST1x4>;
-def VST1d64QPseudoWB_fixed : VSTQQWBPseudo<IIC_VST1x4u>;
+def VST1d64QPseudoWB_fixed : VSTQQWBfixedPseudo<IIC_VST1x4u>;
def VST1d64QPseudoWB_register : VSTQQWBPseudo<IIC_VST1x4u>;
// VST2 : Vector Store (multiple 2-element structures)
class VST2<bits<4> op11_8, bits<4> op7_4, string Dt, RegisterOperand VdTy,
- InstrItinClass itin>
- : NLdSt<0, 0b00, op11_8, op7_4, (outs), (ins addrmode6:$Rn, VdTy:$Vd),
+ InstrItinClass itin, Operand AddrMode>
+ : NLdSt<0, 0b00, op11_8, op7_4, (outs), (ins AddrMode:$Rn, VdTy:$Vd),
itin, "vst2", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
+ let DecoderMethod = "DecodeVLDST2Instruction";
}
-def VST2d8 : VST2<0b1000, {0,0,?,?}, "8", VecListDPair, IIC_VST2>;
-def VST2d16 : VST2<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VST2>;
-def VST2d32 : VST2<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VST2>;
+def VST2d8 : VST2<0b1000, {0,0,?,?}, "8", VecListDPair, IIC_VST2,
+ addrmode6align64or128>;
+def VST2d16 : VST2<0b1000, {0,1,?,?}, "16", VecListDPair, IIC_VST2,
+ addrmode6align64or128>;
+def VST2d32 : VST2<0b1000, {1,0,?,?}, "32", VecListDPair, IIC_VST2,
+ addrmode6align64or128>;
-def VST2q8 : VST2<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VST2x2>;
-def VST2q16 : VST2<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VST2x2>;
-def VST2q32 : VST2<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VST2x2>;
+def VST2q8 : VST2<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VST2x2,
+ addrmode6align64or128or256>;
+def VST2q16 : VST2<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VST2x2,
+ addrmode6align64or128or256>;
+def VST2q32 : VST2<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VST2x2,
+ addrmode6align64or128or256>;
def VST2q8Pseudo : VSTQQPseudo<IIC_VST2x2>;
def VST2q16Pseudo : VSTQQPseudo<IIC_VST2x2>;
// ...with address register writeback:
multiclass VST2DWB<bits<4> op11_8, bits<4> op7_4, string Dt,
- RegisterOperand VdTy> {
+ RegisterOperand VdTy, Operand AddrMode> {
def _fixed : NLdSt<0, 0b00, op11_8, op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, VdTy:$Vd), IIC_VLD1u,
+ (ins AddrMode:$Rn, VdTy:$Vd), IIC_VLD1u,
"vst2", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbFixed";
+ let DecoderMethod = "DecodeVLDST2Instruction";
}
def _register : NLdSt<0, 0b00, op11_8, op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, rGPR:$Rm, VdTy:$Vd), IIC_VLD1u,
+ (ins AddrMode:$Rn, rGPR:$Rm, VdTy:$Vd), IIC_VLD1u,
"vst2", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbRegister";
+ let DecoderMethod = "DecodeVLDST2Instruction";
}
}
-multiclass VST2QWB<bits<4> op7_4, string Dt> {
+multiclass VST2QWB<bits<4> op7_4, string Dt, Operand AddrMode> {
def _fixed : NLdSt<0, 0b00, 0b0011, op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, VecListFourD:$Vd), IIC_VLD1u,
+ (ins AddrMode:$Rn, VecListFourD:$Vd), IIC_VLD1u,
"vst2", Dt, "$Vd, $Rn!",
"$Rn.addr = $wb", []> {
let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbFixed";
+ let DecoderMethod = "DecodeVLDST2Instruction";
}
def _register : NLdSt<0, 0b00, 0b0011, op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, rGPR:$Rm, VecListFourD:$Vd),
+ (ins AddrMode:$Rn, rGPR:$Rm, VecListFourD:$Vd),
IIC_VLD1u,
"vst2", Dt, "$Vd, $Rn, $Rm",
"$Rn.addr = $wb", []> {
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
- let AsmMatchConverter = "cvtVSTwbRegister";
+ let DecoderMethod = "DecodeVLDST2Instruction";
}
}
-defm VST2d8wb : VST2DWB<0b1000, {0,0,?,?}, "8", VecListDPair>;
-defm VST2d16wb : VST2DWB<0b1000, {0,1,?,?}, "16", VecListDPair>;
-defm VST2d32wb : VST2DWB<0b1000, {1,0,?,?}, "32", VecListDPair>;
+defm VST2d8wb : VST2DWB<0b1000, {0,0,?,?}, "8", VecListDPair,
+ addrmode6align64or128>;
+defm VST2d16wb : VST2DWB<0b1000, {0,1,?,?}, "16", VecListDPair,
+ addrmode6align64or128>;
+defm VST2d32wb : VST2DWB<0b1000, {1,0,?,?}, "32", VecListDPair,
+ addrmode6align64or128>;
-defm VST2q8wb : VST2QWB<{0,0,?,?}, "8">;
-defm VST2q16wb : VST2QWB<{0,1,?,?}, "16">;
-defm VST2q32wb : VST2QWB<{1,0,?,?}, "32">;
+defm VST2q8wb : VST2QWB<{0,0,?,?}, "8", addrmode6align64or128or256>;
+defm VST2q16wb : VST2QWB<{0,1,?,?}, "16", addrmode6align64or128or256>;
+defm VST2q32wb : VST2QWB<{1,0,?,?}, "32", addrmode6align64or128or256>;
def VST2q8PseudoWB_fixed : VSTQQWBfixedPseudo<IIC_VST2x2u>;
def VST2q16PseudoWB_fixed : VSTQQWBfixedPseudo<IIC_VST2x2u>;
def VST2q32PseudoWB_register : VSTQQWBregisterPseudo<IIC_VST2x2u>;
// ...with double-spaced registers
-def VST2b8 : VST2<0b1001, {0,0,?,?}, "8", VecListDPairSpaced, IIC_VST2>;
-def VST2b16 : VST2<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VST2>;
-def VST2b32 : VST2<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VST2>;
-defm VST2b8wb : VST2DWB<0b1001, {0,0,?,?}, "8", VecListDPairSpaced>;
-defm VST2b16wb : VST2DWB<0b1001, {0,1,?,?}, "16", VecListDPairSpaced>;
-defm VST2b32wb : VST2DWB<0b1001, {1,0,?,?}, "32", VecListDPairSpaced>;
+def VST2b8 : VST2<0b1001, {0,0,?,?}, "8", VecListDPairSpaced, IIC_VST2,
+ addrmode6align64or128>;
+def VST2b16 : VST2<0b1001, {0,1,?,?}, "16", VecListDPairSpaced, IIC_VST2,
+ addrmode6align64or128>;
+def VST2b32 : VST2<0b1001, {1,0,?,?}, "32", VecListDPairSpaced, IIC_VST2,
+ addrmode6align64or128>;
+defm VST2b8wb : VST2DWB<0b1001, {0,0,?,?}, "8", VecListDPairSpaced,
+ addrmode6align64or128>;
+defm VST2b16wb : VST2DWB<0b1001, {0,1,?,?}, "16", VecListDPairSpaced,
+ addrmode6align64or128>;
+defm VST2b32wb : VST2DWB<0b1001, {1,0,?,?}, "32", VecListDPairSpaced,
+ addrmode6align64or128>;
// VST3 : Vector Store (multiple 3-element structures)
class VST3D<bits<4> op11_8, bits<4> op7_4, string Dt>
"vst3", Dt, "\\{$Vd, $src2, $src3\\}, $Rn", "", []> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVSTInstruction";
+ let DecoderMethod = "DecodeVLDST3Instruction";
}
def VST3d8 : VST3D<0b0100, {0,0,0,?}, "8">;
"vst3", Dt, "\\{$Vd, $src2, $src3\\}, $Rn$Rm",
"$Rn.addr = $wb", []> {
let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVSTInstruction";
+ let DecoderMethod = "DecodeVLDST3Instruction";
}
def VST3d8_UPD : VST3DWB<0b0100, {0,0,0,?}, "8">;
"", []> {
let Rm = 0b1111;
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
+ let DecoderMethod = "DecodeVLDST4Instruction";
}
def VST4d8 : VST4D<0b0000, {0,0,?,?}, "8">;
"vst4", Dt, "\\{$Vd, $src2, $src3, $src4\\}, $Rn$Rm",
"$Rn.addr = $wb", []> {
let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
+ let DecoderMethod = "DecodeVLDST4Instruction";
}
def VST4d8_UPD : VST4DWB<0b0000, {0,0,?,?}, "8">;
def VST4q16oddPseudo_UPD : VSTQQQQWBPseudo<IIC_VST4u>;
def VST4q32oddPseudo_UPD : VSTQQQQWBPseudo<IIC_VST4u>;
-} // mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1
+} // mayStore = 1, hasSideEffects = 0, hasExtraSrcRegAllocReq = 1
// Classes for VST*LN pseudo-instructions with multi-register operands.
// These are expanded to real instructions after register allocation.
def VST1LNd16 : VST1LN<0b0100, {?,?,0,?}, "16", v4i16, truncstorei16,
NEONvgetlaneu, addrmode6> {
let Inst{7-6} = lane{1-0};
- let Inst{4} = Rn{5};
+ let Inst{4} = Rn{4};
}
def VST1LNd32 : VST1LN<0b1000, {?,0,?,?}, "32", v2i32, store, extractelt,
def VST1LNd16_UPD : VST1LNWB<0b0100, {?,?,0,?}, "16", v4i16, post_truncsti16,
NEONvgetlaneu, addrmode6> {
let Inst{7-6} = lane{1-0};
- let Inst{4} = Rn{5};
+ let Inst{4} = Rn{4};
}
def VST1LNd32_UPD : VST1LNWB<0b1000, {?,0,?,?}, "32", v2i32, post_store,
extractelt, addrmode6oneL32> {
def VST1LNq16Pseudo_UPD : VST1QLNWBPseudo<v8i16, post_truncsti16,NEONvgetlaneu>;
def VST1LNq32Pseudo_UPD : VST1QLNWBPseudo<v4i32, post_store, extractelt>;
-let mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1 in {
+let mayStore = 1, hasSideEffects = 0, hasExtraSrcRegAllocReq = 1 in {
// VST2LN : Vector Store (single 2-element structure from one lane)
class VST2LN<bits<4> op11_8, bits<4> op7_4, string Dt>
def VST4LNq16Pseudo_UPD : VSTQQQQLNWBPseudo<IIC_VST4lnu>;
def VST4LNq32Pseudo_UPD : VSTQQQQLNWBPseudo<IIC_VST4lnu>;
-} // mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1
-
+} // mayStore = 1, hasSideEffects = 0, hasExtraSrcRegAllocReq = 1
+
+// Use vld1/vst1 for unaligned f64 load / store
+def : Pat<(f64 (hword_alignedload addrmode6:$addr)),
+ (VLD1d16 addrmode6:$addr)>, Requires<[IsLE]>;
+def : Pat<(hword_alignedstore (f64 DPR:$value), addrmode6:$addr),
+ (VST1d16 addrmode6:$addr, DPR:$value)>, Requires<[IsLE]>;
+def : Pat<(f64 (byte_alignedload addrmode6:$addr)),
+ (VLD1d8 addrmode6:$addr)>, Requires<[IsLE]>;
+def : Pat<(byte_alignedstore (f64 DPR:$value), addrmode6:$addr),
+ (VST1d8 addrmode6:$addr, DPR:$value)>, Requires<[IsLE]>;
+def : Pat<(f64 (non_word_alignedload addrmode6:$addr)),
+ (VLD1d64 addrmode6:$addr)>, Requires<[IsBE]>;
+def : Pat<(non_word_alignedstore (f64 DPR:$value), addrmode6:$addr),
+ (VST1d64 addrmode6:$addr, DPR:$value)>, Requires<[IsBE]>;
+
+// Use vld1/vst1 for Q and QQ. Also use them for unaligned v2f64
+// load / store if it's legal.
+def : Pat<(v2f64 (dword_alignedload addrmode6:$addr)),
+ (VLD1q64 addrmode6:$addr)>;
+def : Pat<(dword_alignedstore (v2f64 QPR:$value), addrmode6:$addr),
+ (VST1q64 addrmode6:$addr, QPR:$value)>;
+def : Pat<(v2f64 (word_alignedload addrmode6:$addr)),
+ (VLD1q32 addrmode6:$addr)>, Requires<[IsLE]>;
+def : Pat<(word_alignedstore (v2f64 QPR:$value), addrmode6:$addr),
+ (VST1q32 addrmode6:$addr, QPR:$value)>, Requires<[IsLE]>;
+def : Pat<(v2f64 (hword_alignedload addrmode6:$addr)),
+ (VLD1q16 addrmode6:$addr)>, Requires<[IsLE]>;
+def : Pat<(hword_alignedstore (v2f64 QPR:$value), addrmode6:$addr),
+ (VST1q16 addrmode6:$addr, QPR:$value)>, Requires<[IsLE]>;
+def : Pat<(v2f64 (byte_alignedload addrmode6:$addr)),
+ (VLD1q8 addrmode6:$addr)>, Requires<[IsLE]>;
+def : Pat<(byte_alignedstore (v2f64 QPR:$value), addrmode6:$addr),
+ (VST1q8 addrmode6:$addr, QPR:$value)>, Requires<[IsLE]>;
//===----------------------------------------------------------------------===//
// NEON pattern fragments
// Extract D sub-registers of Q registers.
def DSubReg_i8_reg : SDNodeXForm<imm, [{
assert(ARM::dsub_7 == ARM::dsub_0+7 && "Unexpected subreg numbering");
- return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue()/8, MVT::i32);
+ return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue()/8, SDLoc(N),
+ MVT::i32);
}]>;
def DSubReg_i16_reg : SDNodeXForm<imm, [{
assert(ARM::dsub_7 == ARM::dsub_0+7 && "Unexpected subreg numbering");
- return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue()/4, MVT::i32);
+ return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue()/4, SDLoc(N),
+ MVT::i32);
}]>;
def DSubReg_i32_reg : SDNodeXForm<imm, [{
assert(ARM::dsub_7 == ARM::dsub_0+7 && "Unexpected subreg numbering");
- return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue()/2, MVT::i32);
+ return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue()/2, SDLoc(N),
+ MVT::i32);
}]>;
def DSubReg_f64_reg : SDNodeXForm<imm, [{
assert(ARM::dsub_7 == ARM::dsub_0+7 && "Unexpected subreg numbering");
- return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue(), MVT::i32);
+ return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue(), SDLoc(N),
+ MVT::i32);
}]>;
// Extract S sub-registers of Q/D registers.
def SSubReg_f32_reg : SDNodeXForm<imm, [{
assert(ARM::ssub_3 == ARM::ssub_0+3 && "Unexpected subreg numbering");
- return CurDAG->getTargetConstant(ARM::ssub_0 + N->getZExtValue(), MVT::i32);
+ return CurDAG->getTargetConstant(ARM::ssub_0 + N->getZExtValue(), SDLoc(N),
+ MVT::i32);
}]>;
// Translate lane numbers from Q registers to D subregs.
def SubReg_i8_lane : SDNodeXForm<imm, [{
- return CurDAG->getTargetConstant(N->getZExtValue() & 7, MVT::i32);
+ return CurDAG->getTargetConstant(N->getZExtValue() & 7, SDLoc(N), MVT::i32);
}]>;
def SubReg_i16_lane : SDNodeXForm<imm, [{
- return CurDAG->getTargetConstant(N->getZExtValue() & 3, MVT::i32);
+ return CurDAG->getTargetConstant(N->getZExtValue() & 3, SDLoc(N), MVT::i32);
}]>;
def SubReg_i32_lane : SDNodeXForm<imm, [{
- return CurDAG->getTargetConstant(N->getZExtValue() & 1, MVT::i32);
+ return CurDAG->getTargetConstant(N->getZExtValue() & 1, SDLoc(N), MVT::i32);
}]>;
//===----------------------------------------------------------------------===//
(ins QPR:$Vm), itin, OpcodeStr, Dt, "$Vd, $Vm", "",
[(set QPR:$Vd, (ResTy (IntOp (OpTy QPR:$Vm))))]>;
+// Same as above, but not predicated.
+class N2VDIntnp<bits<2> op17_16, bits<3> op10_8, bit op7,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType ResTy, ValueType OpTy, SDPatternOperator IntOp>
+ : N2Vnp<0b10, op17_16, op10_8, op7, 0, (outs DPR:$Vd), (ins DPR:$Vm),
+ itin, OpcodeStr, Dt,
+ [(set DPR:$Vd, (ResTy (IntOp (OpTy DPR:$Vm))))]>;
+
+class N2VQIntnp<bits<2> op17_16, bits<3> op10_8, bit op7,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType ResTy, ValueType OpTy, SDPatternOperator IntOp>
+ : N2Vnp<0b10, op17_16, op10_8, op7, 1, (outs QPR:$Vd), (ins QPR:$Vm),
+ itin, OpcodeStr, Dt,
+ [(set QPR:$Vd, (ResTy (IntOp (OpTy QPR:$Vm))))]>;
+
+// Similar to NV2VQIntnp with some more encoding bits exposed (crypto).
+class N2VQIntXnp<bits<2> op19_18, bits<2> op17_16, bits<3> op10_8, bit op6,
+ bit op7, InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType ResTy, ValueType OpTy, SDPatternOperator IntOp>
+ : N2Vnp<op19_18, op17_16, op10_8, op7, op6, (outs QPR:$Vd), (ins QPR:$Vm),
+ itin, OpcodeStr, Dt,
+ [(set QPR:$Vd, (ResTy (IntOp (OpTy QPR:$Vm))))]>;
+
+// Same as N2VQIntXnp but with Vd as a src register.
+class N2VQIntX2np<bits<2> op19_18, bits<2> op17_16, bits<3> op10_8, bit op6,
+ bit op7, InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType ResTy, ValueType OpTy, SDPatternOperator IntOp>
+ : N2Vnp<op19_18, op17_16, op10_8, op7, op6,
+ (outs QPR:$Vd), (ins QPR:$src, QPR:$Vm),
+ itin, OpcodeStr, Dt,
+ [(set QPR:$Vd, (ResTy (IntOp (OpTy QPR:$src), (OpTy QPR:$Vm))))]> {
+ let Constraints = "$src = $Vd";
+}
+
// Narrow 2-register operations.
class N2VN<bits<2> op24_23, bits<2> op21_20, bits<2> op19_18,
bits<2> op17_16, bits<5> op11_7, bit op6, bit op4,
let TwoOperandAliasConstraint = "$Vn = $Vd";
let isCommutable = Commutable;
}
+
+class N3VDIntnp<bits<5> op27_23, bits<2> op21_20, bits<4> op11_8, bit op6,
+ bit op4, Format f, InstrItinClass itin, string OpcodeStr,
+ string Dt, ValueType ResTy, ValueType OpTy,
+ SDPatternOperator IntOp, bit Commutable>
+ : N3Vnp<op27_23, op21_20, op11_8, op6, op4,
+ (outs DPR:$Vd), (ins DPR:$Vn, DPR:$Vm), N3RegFrm, itin, OpcodeStr, Dt,
+ [(set DPR:$Vd, (ResTy (IntOp (OpTy DPR:$Vn), (OpTy DPR:$Vm))))]>;
+
class N3VDIntSL<bits<2> op21_20, bits<4> op11_8, InstrItinClass itin,
string OpcodeStr, string Dt, ValueType Ty, SDPatternOperator IntOp>
: N3VLane32<0, 1, op21_20, op11_8, 1, 0,
imm:$lane)))))]> {
let isCommutable = 0;
}
+
class N3VDIntSL16<bits<2> op21_20, bits<4> op11_8, InstrItinClass itin,
string OpcodeStr, string Dt, ValueType Ty, SDPatternOperator IntOp>
: N3VLane16<0, 1, op21_20, op11_8, 1, 0,
let TwoOperandAliasConstraint = "$Vn = $Vd";
let isCommutable = Commutable;
}
+
+class N3VQIntnp<bits<5> op27_23, bits<2> op21_20, bits<4> op11_8, bit op6,
+ bit op4, Format f, InstrItinClass itin, string OpcodeStr,
+ string Dt, ValueType ResTy, ValueType OpTy,
+ SDPatternOperator IntOp, bit Commutable>
+ : N3Vnp<op27_23, op21_20, op11_8, op6, op4,
+ (outs QPR:$Vd), (ins QPR:$Vn, QPR:$Vm), f, itin, OpcodeStr, Dt,
+ [(set QPR:$Vd, (ResTy (IntOp (OpTy QPR:$Vn), (OpTy QPR:$Vm))))]>;
+
+// Same as N3VQIntnp but with Vd as a src register.
+class N3VQInt3np<bits<5> op27_23, bits<2> op21_20, bits<4> op11_8, bit op6,
+ bit op4, Format f, InstrItinClass itin, string OpcodeStr,
+ string Dt, ValueType ResTy, ValueType OpTy,
+ SDPatternOperator IntOp, bit Commutable>
+ : N3Vnp<op27_23, op21_20, op11_8, op6, op4,
+ (outs QPR:$Vd), (ins QPR:$src, QPR:$Vn, QPR:$Vm),
+ f, itin, OpcodeStr, Dt,
+ [(set QPR:$Vd, (ResTy (IntOp (OpTy QPR:$src), (OpTy QPR:$Vn),
+ (OpTy QPR:$Vm))))]> {
+ let Constraints = "$src = $Vd";
+}
+
class N3VQIntSL<bits<2> op21_20, bits<4> op11_8, InstrItinClass itin,
string OpcodeStr, string Dt,
ValueType ResTy, ValueType OpTy, SDPatternOperator IntOp>
imm:$lane)))))))]>;
class N3VDMulOpSL16<bits<2> op21_20, bits<4> op11_8, InstrItinClass itin,
string OpcodeStr, string Dt,
- ValueType Ty, SDNode MulOp, SDNode ShOp>
+ ValueType Ty, SDPatternOperator MulOp, SDPatternOperator ShOp>
: N3VLane16<0, 1, op21_20, op11_8, 1, 0,
(outs DPR:$Vd),
(ins DPR:$src1, DPR:$Vn, DPR_8:$Vm, VectorIndex16:$lane),
class N3VQMulOpSL16<bits<2> op21_20, bits<4> op11_8, InstrItinClass itin,
string OpcodeStr, string Dt,
ValueType ResTy, ValueType OpTy,
- SDNode MulOp, SDNode ShOp>
+ SDPatternOperator MulOp, SDPatternOperator ShOp>
: N3VLane16<1, 1, op21_20, op11_8, 1, 0,
(outs QPR:$Vd),
(ins QPR:$src1, QPR:$Vn, DPR_8:$Vm, VectorIndex16:$lane),
[(set QPR:$Vd, (TyQ (OpNode (TyD DPR:$Vn), (TyD DPR:$Vm))))]> {
let isCommutable = Commutable;
}
+
class N3VLSL<bit op24, bits<2> op21_20, bits<4> op11_8,
InstrItinClass itin, string OpcodeStr, string Dt,
ValueType TyQ, ValueType TyD, SDNode OpNode>
[(set QPR:$Vd, (TyQ (IntOp (TyD DPR:$Vn), (TyD DPR:$Vm))))]> {
let isCommutable = Commutable;
}
+
+// Same as above, but not predicated.
+class N3VLIntnp<bits<5> op27_23, bits<2> op21_20, bits<4> op11_8, bit op6,
+ bit op4, InstrItinClass itin, string OpcodeStr,
+ string Dt, ValueType ResTy, ValueType OpTy,
+ SDPatternOperator IntOp, bit Commutable>
+ : N3Vnp<op27_23, op21_20, op11_8, op6, op4,
+ (outs QPR:$Vd), (ins DPR:$Vn, DPR:$Vm), N3RegFrm, itin, OpcodeStr, Dt,
+ [(set QPR:$Vd, (ResTy (IntOp (OpTy DPR:$Vn), (OpTy DPR:$Vm))))]>;
+
class N3VLIntSL<bit op24, bits<2> op21_20, bits<4> op11_8, InstrItinClass itin,
string OpcodeStr, string Dt,
ValueType ResTy, ValueType OpTy, SDPatternOperator IntOp>
// Long shift by immediate.
class N2VLSh<bit op24, bit op23, bits<4> op11_8, bit op7, bit op6, bit op4,
string OpcodeStr, string Dt,
- ValueType ResTy, ValueType OpTy, Operand ImmTy, SDNode OpNode>
+ ValueType ResTy, ValueType OpTy, Operand ImmTy,
+ SDPatternOperator OpNode>
: N2VImm<op24, op23, op11_8, op7, op6, op4,
(outs QPR:$Vd), (ins DPR:$Vm, ImmTy:$SIMM), N2RegVShLFrm,
IIC_VSHLiD, OpcodeStr, Dt, "$Vd, $Vm, $SIMM", "",
- [(set QPR:$Vd, (ResTy (OpNode (OpTy DPR:$Vm),
- (i32 imm:$SIMM))))]>;
+ [(set QPR:$Vd, (ResTy (OpNode (OpTy DPR:$Vm), ImmTy:$SIMM)))]>;
// Narrow shift by immediate.
class N2VNSh<bit op24, bit op23, bits<4> op11_8, bit op7, bit op6, bit op4,
InstrItinClass itin, string OpcodeStr, string Dt,
- ValueType ResTy, ValueType OpTy, Operand ImmTy, SDNode OpNode>
+ ValueType ResTy, ValueType OpTy, Operand ImmTy,
+ SDPatternOperator OpNode>
: N2VImm<op24, op23, op11_8, op7, op6, op4,
(outs DPR:$Vd), (ins QPR:$Vm, ImmTy:$SIMM), N2RegVShRFrm, itin,
OpcodeStr, Dt, "$Vd, $Vm, $SIMM", "",
[(set DPR:$Vd, (ResTy (OpNode (OpTy QPR:$Vm),
- (i32 imm:$SIMM))))]>;
+ (i32 ImmTy:$SIMM))))]>;
// Shift right by immediate and accumulate,
// both double- and quad-register.
multiclass N3VMulOpSL_HS<bits<4> op11_8,
InstrItinClass itinD16, InstrItinClass itinD32,
InstrItinClass itinQ16, InstrItinClass itinQ32,
- string OpcodeStr, string Dt, SDNode ShOp> {
+ string OpcodeStr, string Dt, SDPatternOperator ShOp> {
def v4i16 : N3VDMulOpSL16<0b01, op11_8, itinD16,
OpcodeStr, !strconcat(Dt, "16"), v4i16, mul, ShOp>;
def v2i32 : N3VDMulOpSL<0b10, op11_8, itinD32,
}
// Neon 3-argument intrinsics,
-// element sizes of 8, 16 and 32 bits:
-multiclass N3VInt3_QHS<bit op24, bit op23, bits<4> op11_8, bit op4,
- InstrItinClass itinD, InstrItinClass itinQ,
+// element sizes of 16 and 32 bits:
+multiclass N3VInt3_HS<bit op24, bit op23, bits<4> op11_8, bit op4,
+ InstrItinClass itinD16, InstrItinClass itinD32,
+ InstrItinClass itinQ16, InstrItinClass itinQ32,
string OpcodeStr, string Dt, SDPatternOperator IntOp> {
// 64-bit vector types.
- def v8i8 : N3VDInt3<op24, op23, 0b00, op11_8, op4, itinD,
- OpcodeStr, !strconcat(Dt, "8"), v8i8, v8i8, IntOp>;
- def v4i16 : N3VDInt3<op24, op23, 0b01, op11_8, op4, itinD,
+ def v4i16 : N3VDInt3<op24, op23, 0b01, op11_8, op4, itinD16,
OpcodeStr, !strconcat(Dt, "16"), v4i16, v4i16, IntOp>;
- def v2i32 : N3VDInt3<op24, op23, 0b10, op11_8, op4, itinD,
+ def v2i32 : N3VDInt3<op24, op23, 0b10, op11_8, op4, itinD32,
OpcodeStr, !strconcat(Dt, "32"), v2i32, v2i32, IntOp>;
// 128-bit vector types.
- def v16i8 : N3VQInt3<op24, op23, 0b00, op11_8, op4, itinQ,
- OpcodeStr, !strconcat(Dt, "8"), v16i8, v16i8, IntOp>;
- def v8i16 : N3VQInt3<op24, op23, 0b01, op11_8, op4, itinQ,
+ def v8i16 : N3VQInt3<op24, op23, 0b01, op11_8, op4, itinQ16,
OpcodeStr, !strconcat(Dt, "16"), v8i16, v8i16, IntOp>;
- def v4i32 : N3VQInt3<op24, op23, 0b10, op11_8, op4, itinQ,
+ def v4i32 : N3VQInt3<op24, op23, 0b10, op11_8, op4, itinQ32,
OpcodeStr, !strconcat(Dt, "32"), v4i32, v4i32, IntOp>;
}
+// element sizes of 8, 16 and 32 bits:
+multiclass N3VInt3_QHS<bit op24, bit op23, bits<4> op11_8, bit op4,
+ InstrItinClass itinD16, InstrItinClass itinD32,
+ InstrItinClass itinQ16, InstrItinClass itinQ32,
+ string OpcodeStr, string Dt, SDPatternOperator IntOp>
+ :N3VInt3_HS <op24, op23, op11_8, op4, itinD16, itinD32,
+ itinQ16, itinQ32, OpcodeStr, Dt, IntOp>{
+ // 64-bit vector types.
+ def v8i8 : N3VDInt3<op24, op23, 0b00, op11_8, op4, itinD16,
+ OpcodeStr, !strconcat(Dt, "8"), v8i8, v8i8, IntOp>;
+ // 128-bit vector types.
+ def v16i8 : N3VQInt3<op24, op23, 0b00, op11_8, op4, itinQ16,
+ OpcodeStr, !strconcat(Dt, "8"), v16i8, v16i8, IntOp>;
+}
// Neon Long Multiply-Op vector operations,
// element sizes of 8, 16 and 32 bits:
// Neon Shift Long operations,
// element sizes of 8, 16, 32 bits:
multiclass N2VLSh_QHS<bit op24, bit op23, bits<4> op11_8, bit op7, bit op6,
- bit op4, string OpcodeStr, string Dt, SDNode OpNode> {
+ bit op4, string OpcodeStr, string Dt,
+ SDPatternOperator OpNode> {
def v8i16 : N2VLSh<op24, op23, op11_8, op7, op6, op4,
OpcodeStr, !strconcat(Dt, "8"), v8i16, v8i8, imm1_7, OpNode> {
let Inst{21-19} = 0b001; // imm6 = 001xxx
// element sizes of 16, 32, 64 bits:
multiclass N2VNSh_HSD<bit op24, bit op23, bits<4> op11_8, bit op7, bit op6,
bit op4, InstrItinClass itin, string OpcodeStr, string Dt,
- SDNode OpNode> {
+ SDPatternOperator OpNode> {
def v8i8 : N2VNSh<op24, op23, op11_8, op7, op6, op4, itin,
OpcodeStr, !strconcat(Dt, "16"),
v8i8, v8i16, shr_imm8, OpNode> {
IIC_VBINi4D, IIC_VBINi4D, IIC_VBINi4Q, IIC_VBINi4Q,
"vqadd", "u", int_arm_neon_vqaddu, 1>;
// VADDHN : Vector Add and Narrow Returning High Half (D = Q + Q)
-defm VADDHN : N3VNInt_HSD<0,1,0b0100,0, "vaddhn", "i",
- int_arm_neon_vaddhn, 1>;
+defm VADDHN : N3VNInt_HSD<0,1,0b0100,0, "vaddhn", "i", null_frag, 1>;
// VRADDHN : Vector Rounding Add and Narrow Returning High Half (D = Q + Q)
defm VRADDHN : N3VNInt_HSD<1,1,0b0100,0, "vraddhn", "i",
int_arm_neon_vraddhn, 1>;
+def : Pat<(v8i8 (trunc (NEONvshru (add (v8i16 QPR:$Vn), QPR:$Vm), 8))),
+ (VADDHNv8i8 QPR:$Vn, QPR:$Vm)>;
+def : Pat<(v4i16 (trunc (NEONvshru (add (v4i32 QPR:$Vn), QPR:$Vm), 16))),
+ (VADDHNv4i16 QPR:$Vn, QPR:$Vm)>;
+def : Pat<(v2i32 (trunc (NEONvshru (add (v2i64 QPR:$Vn), QPR:$Vm), 32))),
+ (VADDHNv2i32 QPR:$Vn, QPR:$Vm)>;
+
// Vector Multiply Operations.
// VMUL : Vector Multiply (integer, polynomial and floating-point)
(DSubReg_i32_reg imm:$lane))),
(SubReg_i32_lane imm:$lane)))>;
+
+def : Pat<(v2f32 (fmul DPR:$Rn, (NEONvdup (f32 SPR:$Rm)))),
+ (VMULslfd DPR:$Rn,
+ (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)), SPR:$Rm, ssub_0),
+ (i32 0))>;
+def : Pat<(v4f32 (fmul QPR:$Rn, (NEONvdup (f32 SPR:$Rm)))),
+ (VMULslfq QPR:$Rn,
+ (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)), SPR:$Rm, ssub_0),
+ (i32 0))>;
+
+
// VQDMULH : Vector Saturating Doubling Multiply Returning High Half
defm VQDMULH : N3VInt_HS<0, 0, 0b1011, 0, N3RegFrm, IIC_VMULi16D, IIC_VMULi32D,
IIC_VMULi16Q, IIC_VMULi32Q,
(SubReg_i32_lane imm:$lane)))>;
// VMULL : Vector Multiply Long (integer and polynomial) (Q = D * D)
-defm VMULLs : N3VL_QHS<0,1,0b1100,0, IIC_VMULi16D, IIC_VMULi32D,
- "vmull", "s", NEONvmulls, 1>;
-defm VMULLu : N3VL_QHS<1,1,0b1100,0, IIC_VMULi16D, IIC_VMULi32D,
- "vmull", "u", NEONvmullu, 1>;
-def VMULLp : N3VLInt<0, 1, 0b00, 0b1110, 0, IIC_VMULi16D, "vmull", "p8",
- v8i16, v8i8, int_arm_neon_vmullp, 1>;
+let PostEncoderMethod = "NEONThumb2DataIPostEncoder",
+ DecoderNamespace = "NEONData" in {
+ defm VMULLs : N3VL_QHS<0,1,0b1100,0, IIC_VMULi16D, IIC_VMULi32D,
+ "vmull", "s", NEONvmulls, 1>;
+ defm VMULLu : N3VL_QHS<1,1,0b1100,0, IIC_VMULi16D, IIC_VMULi32D,
+ "vmull", "u", NEONvmullu, 1>;
+ def VMULLp8 : N3VLInt<0, 1, 0b00, 0b1110, 0, IIC_VMULi16D, "vmull", "p8",
+ v8i16, v8i8, int_arm_neon_vmullp, 1>;
+ def VMULLp64 : N3VLIntnp<0b00101, 0b10, 0b1110, 0, 0, NoItinerary,
+ "vmull", "p64", v2i64, v1i64, int_arm_neon_vmullp, 1>,
+ Requires<[HasV8, HasCrypto]>;
+}
defm VMULLsls : N3VLSL_HS<0, 0b1010, IIC_VMULi16D, "vmull", "s", NEONvmulls>;
defm VMULLslu : N3VLSL_HS<1, 0b1010, IIC_VMULi16D, "vmull", "u", NEONvmullu>;
defm VMLALsls : N3VLMulOpSL_HS<0, 0b0010, "vmlal", "s", NEONvmulls, add>;
defm VMLALslu : N3VLMulOpSL_HS<1, 0b0010, "vmlal", "u", NEONvmullu, add>;
+let Predicates = [HasNEON, HasV8_1a] in {
+ // v8.1a Neon Rounding Double Multiply-Op vector operations,
+ // VQRDMLAH : Vector Saturating Rounding Doubling Multiply Accumulate Long
+ // (Q += D * D)
+ defm VQRDMLAH : N3VInt3_HS<1, 0, 0b1011, 1, IIC_VMACi16D, IIC_VMACi32D,
+ IIC_VMACi16Q, IIC_VMACi32Q, "vqrdmlah", "s",
+ null_frag>;
+ def : Pat<(v4i16 (int_arm_neon_vqadds
+ (v4i16 DPR:$src1),
+ (v4i16 (int_arm_neon_vqrdmulh (v4i16 DPR:$Vn),
+ (v4i16 DPR:$Vm))))),
+ (v4i16 (VQRDMLAHv4i16 DPR:$src1, DPR:$Vn, DPR:$Vm))>;
+ def : Pat<(v2i32 (int_arm_neon_vqadds
+ (v2i32 DPR:$src1),
+ (v2i32 (int_arm_neon_vqrdmulh (v2i32 DPR:$Vn),
+ (v2i32 DPR:$Vm))))),
+ (v2i32 (VQRDMLAHv2i32 DPR:$src1, DPR:$Vn, DPR:$Vm))>;
+ def : Pat<(v8i16 (int_arm_neon_vqadds
+ (v8i16 QPR:$src1),
+ (v8i16 (int_arm_neon_vqrdmulh (v8i16 QPR:$Vn),
+ (v8i16 QPR:$Vm))))),
+ (v8i16 (VQRDMLAHv8i16 QPR:$src1, QPR:$Vn, QPR:$Vm))>;
+ def : Pat<(v4i32 (int_arm_neon_vqadds
+ (v4i32 QPR:$src1),
+ (v4i32 (int_arm_neon_vqrdmulh (v4i32 QPR:$Vn),
+ (v4i32 QPR:$Vm))))),
+ (v4i32 (VQRDMLAHv4i32 QPR:$src1, QPR:$Vn, QPR:$Vm))>;
+
+ defm VQRDMLAHsl : N3VMulOpSL_HS<0b1110, IIC_VMACi16D, IIC_VMACi32D,
+ IIC_VMACi16Q, IIC_VMACi32Q, "vqrdmlah", "s",
+ null_frag>;
+ def : Pat<(v4i16 (int_arm_neon_vqadds
+ (v4i16 DPR:$src1),
+ (v4i16 (int_arm_neon_vqrdmulh
+ (v4i16 DPR:$Vn),
+ (v4i16 (NEONvduplane (v4i16 DPR_8:$Vm),
+ imm:$lane)))))),
+ (v4i16 (VQRDMLAHslv4i16 DPR:$src1, DPR:$Vn, DPR_8:$Vm,
+ imm:$lane))>;
+ def : Pat<(v2i32 (int_arm_neon_vqadds
+ (v2i32 DPR:$src1),
+ (v2i32 (int_arm_neon_vqrdmulh
+ (v2i32 DPR:$Vn),
+ (v2i32 (NEONvduplane (v2i32 DPR_VFP2:$Vm),
+ imm:$lane)))))),
+ (v2i32 (VQRDMLAHslv2i32 DPR:$src1, DPR:$Vn, DPR_VFP2:$Vm,
+ imm:$lane))>;
+ def : Pat<(v8i16 (int_arm_neon_vqadds
+ (v8i16 QPR:$src1),
+ (v8i16 (int_arm_neon_vqrdmulh
+ (v8i16 QPR:$src2),
+ (v8i16 (NEONvduplane (v8i16 QPR:$src3),
+ imm:$lane)))))),
+ (v8i16 (VQRDMLAHslv8i16 (v8i16 QPR:$src1),
+ (v8i16 QPR:$src2),
+ (v4i16 (EXTRACT_SUBREG
+ QPR:$src3,
+ (DSubReg_i16_reg imm:$lane))),
+ (SubReg_i16_lane imm:$lane)))>;
+ def : Pat<(v4i32 (int_arm_neon_vqadds
+ (v4i32 QPR:$src1),
+ (v4i32 (int_arm_neon_vqrdmulh
+ (v4i32 QPR:$src2),
+ (v4i32 (NEONvduplane (v4i32 QPR:$src3),
+ imm:$lane)))))),
+ (v4i32 (VQRDMLAHslv4i32 (v4i32 QPR:$src1),
+ (v4i32 QPR:$src2),
+ (v2i32 (EXTRACT_SUBREG
+ QPR:$src3,
+ (DSubReg_i32_reg imm:$lane))),
+ (SubReg_i32_lane imm:$lane)))>;
+
+ // VQRDMLSH : Vector Saturating Rounding Doubling Multiply Subtract Long
+ // (Q -= D * D)
+ defm VQRDMLSH : N3VInt3_HS<1, 0, 0b1100, 1, IIC_VMACi16D, IIC_VMACi32D,
+ IIC_VMACi16Q, IIC_VMACi32Q, "vqrdmlsh", "s",
+ null_frag>;
+ def : Pat<(v4i16 (int_arm_neon_vqsubs
+ (v4i16 DPR:$src1),
+ (v4i16 (int_arm_neon_vqrdmulh (v4i16 DPR:$Vn),
+ (v4i16 DPR:$Vm))))),
+ (v4i16 (VQRDMLSHv4i16 DPR:$src1, DPR:$Vn, DPR:$Vm))>;
+ def : Pat<(v2i32 (int_arm_neon_vqsubs
+ (v2i32 DPR:$src1),
+ (v2i32 (int_arm_neon_vqrdmulh (v2i32 DPR:$Vn),
+ (v2i32 DPR:$Vm))))),
+ (v2i32 (VQRDMLSHv2i32 DPR:$src1, DPR:$Vn, DPR:$Vm))>;
+ def : Pat<(v8i16 (int_arm_neon_vqsubs
+ (v8i16 QPR:$src1),
+ (v8i16 (int_arm_neon_vqrdmulh (v8i16 QPR:$Vn),
+ (v8i16 QPR:$Vm))))),
+ (v8i16 (VQRDMLSHv8i16 QPR:$src1, QPR:$Vn, QPR:$Vm))>;
+ def : Pat<(v4i32 (int_arm_neon_vqsubs
+ (v4i32 QPR:$src1),
+ (v4i32 (int_arm_neon_vqrdmulh (v4i32 QPR:$Vn),
+ (v4i32 QPR:$Vm))))),
+ (v4i32 (VQRDMLSHv4i32 QPR:$src1, QPR:$Vn, QPR:$Vm))>;
+
+ defm VQRDMLSHsl : N3VMulOpSL_HS<0b1111, IIC_VMACi16D, IIC_VMACi32D,
+ IIC_VMACi16Q, IIC_VMACi32Q, "vqrdmlsh", "s",
+ null_frag>;
+ def : Pat<(v4i16 (int_arm_neon_vqsubs
+ (v4i16 DPR:$src1),
+ (v4i16 (int_arm_neon_vqrdmulh
+ (v4i16 DPR:$Vn),
+ (v4i16 (NEONvduplane (v4i16 DPR_8:$Vm),
+ imm:$lane)))))),
+ (v4i16 (VQRDMLSHslv4i16 DPR:$src1, DPR:$Vn, DPR_8:$Vm, imm:$lane))>;
+ def : Pat<(v2i32 (int_arm_neon_vqsubs
+ (v2i32 DPR:$src1),
+ (v2i32 (int_arm_neon_vqrdmulh
+ (v2i32 DPR:$Vn),
+ (v2i32 (NEONvduplane (v2i32 DPR_VFP2:$Vm),
+ imm:$lane)))))),
+ (v2i32 (VQRDMLSHslv2i32 DPR:$src1, DPR:$Vn, DPR_VFP2:$Vm,
+ imm:$lane))>;
+ def : Pat<(v8i16 (int_arm_neon_vqsubs
+ (v8i16 QPR:$src1),
+ (v8i16 (int_arm_neon_vqrdmulh
+ (v8i16 QPR:$src2),
+ (v8i16 (NEONvduplane (v8i16 QPR:$src3),
+ imm:$lane)))))),
+ (v8i16 (VQRDMLSHslv8i16 (v8i16 QPR:$src1),
+ (v8i16 QPR:$src2),
+ (v4i16 (EXTRACT_SUBREG
+ QPR:$src3,
+ (DSubReg_i16_reg imm:$lane))),
+ (SubReg_i16_lane imm:$lane)))>;
+ def : Pat<(v4i32 (int_arm_neon_vqsubs
+ (v4i32 QPR:$src1),
+ (v4i32 (int_arm_neon_vqrdmulh
+ (v4i32 QPR:$src2),
+ (v4i32 (NEONvduplane (v4i32 QPR:$src3),
+ imm:$lane)))))),
+ (v4i32 (VQRDMLSHslv4i32 (v4i32 QPR:$src1),
+ (v4i32 QPR:$src2),
+ (v2i32 (EXTRACT_SUBREG
+ QPR:$src3,
+ (DSubReg_i32_reg imm:$lane))),
+ (SubReg_i32_lane imm:$lane)))>;
+}
// VQDMLAL : Vector Saturating Doubling Multiply Accumulate Long (Q += D * D)
defm VQDMLAL : N3VLInt3_HS<0, 1, 0b1001, 0, IIC_VMACi16D, IIC_VMACi32D,
- "vqdmlal", "s", int_arm_neon_vqdmlal>;
-defm VQDMLALsl: N3VLInt3SL_HS<0, 0b0011, "vqdmlal", "s", int_arm_neon_vqdmlal>;
+ "vqdmlal", "s", null_frag>;
+defm VQDMLALsl: N3VLInt3SL_HS<0, 0b0011, "vqdmlal", "s", null_frag>;
+
+def : Pat<(v4i32 (int_arm_neon_vqadds (v4i32 QPR:$src1),
+ (v4i32 (int_arm_neon_vqdmull (v4i16 DPR:$Vn),
+ (v4i16 DPR:$Vm))))),
+ (VQDMLALv4i32 QPR:$src1, DPR:$Vn, DPR:$Vm)>;
+def : Pat<(v2i64 (int_arm_neon_vqadds (v2i64 QPR:$src1),
+ (v2i64 (int_arm_neon_vqdmull (v2i32 DPR:$Vn),
+ (v2i32 DPR:$Vm))))),
+ (VQDMLALv2i64 QPR:$src1, DPR:$Vn, DPR:$Vm)>;
+def : Pat<(v4i32 (int_arm_neon_vqadds (v4i32 QPR:$src1),
+ (v4i32 (int_arm_neon_vqdmull (v4i16 DPR:$Vn),
+ (v4i16 (NEONvduplane (v4i16 DPR_8:$Vm),
+ imm:$lane)))))),
+ (VQDMLALslv4i16 QPR:$src1, DPR:$Vn, DPR_8:$Vm, imm:$lane)>;
+def : Pat<(v2i64 (int_arm_neon_vqadds (v2i64 QPR:$src1),
+ (v2i64 (int_arm_neon_vqdmull (v2i32 DPR:$Vn),
+ (v2i32 (NEONvduplane (v2i32 DPR_VFP2:$Vm),
+ imm:$lane)))))),
+ (VQDMLALslv2i32 QPR:$src1, DPR:$Vn, DPR_VFP2:$Vm, imm:$lane)>;
// VMLS : Vector Multiply Subtract (integer and floating-point)
defm VMLS : N3VMulOp_QHS<1, 0, 0b1001, 0, IIC_VMACi16D, IIC_VMACi32D,
// VQDMLSL : Vector Saturating Doubling Multiply Subtract Long (Q -= D * D)
defm VQDMLSL : N3VLInt3_HS<0, 1, 0b1011, 0, IIC_VMACi16D, IIC_VMACi32D,
- "vqdmlsl", "s", int_arm_neon_vqdmlsl>;
-defm VQDMLSLsl: N3VLInt3SL_HS<0, 0b111, "vqdmlsl", "s", int_arm_neon_vqdmlsl>;
+ "vqdmlsl", "s", null_frag>;
+defm VQDMLSLsl: N3VLInt3SL_HS<0, 0b0111, "vqdmlsl", "s", null_frag>;
+
+def : Pat<(v4i32 (int_arm_neon_vqsubs (v4i32 QPR:$src1),
+ (v4i32 (int_arm_neon_vqdmull (v4i16 DPR:$Vn),
+ (v4i16 DPR:$Vm))))),
+ (VQDMLSLv4i32 QPR:$src1, DPR:$Vn, DPR:$Vm)>;
+def : Pat<(v2i64 (int_arm_neon_vqsubs (v2i64 QPR:$src1),
+ (v2i64 (int_arm_neon_vqdmull (v2i32 DPR:$Vn),
+ (v2i32 DPR:$Vm))))),
+ (VQDMLSLv2i64 QPR:$src1, DPR:$Vn, DPR:$Vm)>;
+def : Pat<(v4i32 (int_arm_neon_vqsubs (v4i32 QPR:$src1),
+ (v4i32 (int_arm_neon_vqdmull (v4i16 DPR:$Vn),
+ (v4i16 (NEONvduplane (v4i16 DPR_8:$Vm),
+ imm:$lane)))))),
+ (VQDMLSLslv4i16 QPR:$src1, DPR:$Vn, DPR_8:$Vm, imm:$lane)>;
+def : Pat<(v2i64 (int_arm_neon_vqsubs (v2i64 QPR:$src1),
+ (v2i64 (int_arm_neon_vqdmull (v2i32 DPR:$Vn),
+ (v2i32 (NEONvduplane (v2i32 DPR_VFP2:$Vm),
+ imm:$lane)))))),
+ (VQDMLSLslv2i32 QPR:$src1, DPR:$Vn, DPR_VFP2:$Vm, imm:$lane)>;
// Fused Vector Multiply-Accumulate and Fused Multiply-Subtract Operations.
def VFMAfd : N3VDMulOp<0, 0, 0b00, 0b1100, 1, IIC_VFMACD, "vfma", "f32",
v2f32, fmul_su, fadd_mlx>,
- Requires<[HasVFP4,UseFusedMAC]>;
+ Requires<[HasNEON,HasVFP4,UseFusedMAC]>;
def VFMAfq : N3VQMulOp<0, 0, 0b00, 0b1100, 1, IIC_VFMACQ, "vfma", "f32",
v4f32, fmul_su, fadd_mlx>,
- Requires<[HasVFP4,UseFusedMAC]>;
+ Requires<[HasNEON,HasVFP4,UseFusedMAC]>;
// Fused Vector Multiply Subtract (floating-point)
def VFMSfd : N3VDMulOp<0, 0, 0b10, 0b1100, 1, IIC_VFMACD, "vfms", "f32",
v2f32, fmul_su, fsub_mlx>,
- Requires<[HasVFP4,UseFusedMAC]>;
+ Requires<[HasNEON,HasVFP4,UseFusedMAC]>;
def VFMSfq : N3VQMulOp<0, 0, 0b10, 0b1100, 1, IIC_VFMACQ, "vfms", "f32",
v4f32, fmul_su, fsub_mlx>,
- Requires<[HasVFP4,UseFusedMAC]>;
+ Requires<[HasNEON,HasVFP4,UseFusedMAC]>;
// Match @llvm.fma.* intrinsics
def : Pat<(v2f32 (fma DPR:$Vn, DPR:$Vm, DPR:$src1)),
IIC_VSUBi4D, IIC_VSUBi4D, IIC_VSUBi4Q, IIC_VSUBi4Q,
"vqsub", "u", int_arm_neon_vqsubu, 0>;
// VSUBHN : Vector Subtract and Narrow Returning High Half (D = Q - Q)
-defm VSUBHN : N3VNInt_HSD<0,1,0b0110,0, "vsubhn", "i",
- int_arm_neon_vsubhn, 0>;
+defm VSUBHN : N3VNInt_HSD<0,1,0b0110,0, "vsubhn", "i", null_frag, 0>;
// VRSUBHN : Vector Rounding Subtract and Narrow Returning High Half (D=Q-Q)
defm VRSUBHN : N3VNInt_HSD<1,1,0b0110,0, "vrsubhn", "i",
int_arm_neon_vrsubhn, 0>;
+def : Pat<(v8i8 (trunc (NEONvshru (sub (v8i16 QPR:$Vn), QPR:$Vm), 8))),
+ (VSUBHNv8i8 QPR:$Vn, QPR:$Vm)>;
+def : Pat<(v4i16 (trunc (NEONvshru (sub (v4i32 QPR:$Vn), QPR:$Vm), 16))),
+ (VSUBHNv4i16 QPR:$Vn, QPR:$Vm)>;
+def : Pat<(v2i32 (trunc (NEONvshru (sub (v2i64 QPR:$Vn), QPR:$Vm), 32))),
+ (VSUBHNv2i32 QPR:$Vn, QPR:$Vm)>;
+
// Vector Comparisons.
// VCEQ : Vector Compare Equal
def VCEQfq : N3VQ<0,0,0b00,0b1110,0, IIC_VBINQ, "vceq", "f32", v4i32, v4f32,
NEONvceq, 1>;
+let TwoOperandAliasConstraint = "$Vm = $Vd" in
defm VCEQz : N2V_QHS_cmp<0b11, 0b11, 0b01, 0b00010, 0, "vceq", "i",
"$Vd, $Vm, #0", NEONvceqz>;
def VCGEfq : N3VQ<1,0,0b00,0b1110,0, IIC_VBINQ, "vcge", "f32", v4i32, v4f32,
NEONvcge, 0>;
+let TwoOperandAliasConstraint = "$Vm = $Vd" in {
defm VCGEz : N2V_QHS_cmp<0b11, 0b11, 0b01, 0b00001, 0, "vcge", "s",
"$Vd, $Vm, #0", NEONvcgez>;
defm VCLEz : N2V_QHS_cmp<0b11, 0b11, 0b01, 0b00011, 0, "vcle", "s",
"$Vd, $Vm, #0", NEONvclez>;
+}
// VCGT : Vector Compare Greater Than
defm VCGTs : N3V_QHS<0, 0, 0b0011, 0, IIC_VSUBi4D, IIC_VSUBi4D, IIC_VSUBi4Q,
def VCGTfq : N3VQ<1,0,0b10,0b1110,0, IIC_VBINQ, "vcgt", "f32", v4i32, v4f32,
NEONvcgt, 0>;
+let TwoOperandAliasConstraint = "$Vm = $Vd" in {
defm VCGTz : N2V_QHS_cmp<0b11, 0b11, 0b01, 0b00000, 0, "vcgt", "s",
"$Vd, $Vm, #0", NEONvcgtz>;
defm VCLTz : N2V_QHS_cmp<0b11, 0b11, 0b01, 0b00100, 0, "vclt", "s",
"$Vd, $Vm, #0", NEONvcltz>;
+}
// VACGE : Vector Absolute Compare Greater Than or Equal (aka VCAGE)
def VACGEd : N3VDInt<1, 0, 0b00, 0b1110, 1, N3RegFrm, IIC_VBIND, "vacge",
- "f32", v2i32, v2f32, int_arm_neon_vacged, 0>;
+ "f32", v2i32, v2f32, int_arm_neon_vacge, 0>;
def VACGEq : N3VQInt<1, 0, 0b00, 0b1110, 1, N3RegFrm, IIC_VBINQ, "vacge",
- "f32", v4i32, v4f32, int_arm_neon_vacgeq, 0>;
+ "f32", v4i32, v4f32, int_arm_neon_vacge, 0>;
// VACGT : Vector Absolute Compare Greater Than (aka VCAGT)
def VACGTd : N3VDInt<1, 0, 0b10, 0b1110, 1, N3RegFrm, IIC_VBIND, "vacgt",
- "f32", v2i32, v2f32, int_arm_neon_vacgtd, 0>;
+ "f32", v2i32, v2f32, int_arm_neon_vacgt, 0>;
def VACGTq : N3VQInt<1, 0, 0b10, 0b1110, 1, N3RegFrm, IIC_VBINQ, "vacgt",
- "f32", v4i32, v4f32, int_arm_neon_vacgtq, 0>;
+ "f32", v4i32, v4f32, int_arm_neon_vacgt, 0>;
// VTST : Vector Test Bits
defm VTST : N3V_QHS<0, 0, 0b1000, 1, IIC_VBINi4D, IIC_VBINi4D, IIC_VBINi4Q,
IIC_VBINi4Q, "vtst", "", NEONvtst, 1>;
+def: NEONInstAlias<"vaclt${p}.f32 $Vd, $Vn, $Vm",
+ (VACGTd DPR:$Vd, DPR:$Vm, DPR:$Vn, pred:$p)>;
+def: NEONInstAlias<"vaclt${p}.f32 $Vd, $Vn, $Vm",
+ (VACGTq QPR:$Vd, QPR:$Vm, QPR:$Vn, pred:$p)>;
+def: NEONInstAlias<"vacle${p}.f32 $Vd, $Vn, $Vm",
+ (VACGEd DPR:$Vd, DPR:$Vm, DPR:$Vn, pred:$p)>;
+def: NEONInstAlias<"vacle${p}.f32 $Vd, $Vn, $Vm",
+ (VACGEq QPR:$Vd, QPR:$Vm, QPR:$Vn, pred:$p)>;
+
+def: NEONInstAlias<"vaclt${p}.f32 $Vd, $Vm",
+ (VACGTd DPR:$Vd, DPR:$Vm, DPR:$Vd, pred:$p)>;
+def: NEONInstAlias<"vaclt${p}.f32 $Vd, $Vm",
+ (VACGTq QPR:$Vd, QPR:$Vm, QPR:$Vd, pred:$p)>;
+def: NEONInstAlias<"vacle${p}.f32 $Vd, $Vm",
+ (VACGEd DPR:$Vd, DPR:$Vm, DPR:$Vd, pred:$p)>;
+def: NEONInstAlias<"vacle${p}.f32 $Vd, $Vm",
+ (VACGEq QPR:$Vd, QPR:$Vm, QPR:$Vd, pred:$p)>;
+
// Vector Bitwise Operations.
def vnotd : PatFrag<(ops node:$in),
"vbsl", "$Vd, $Vn, $Vm", "$src1 = $Vd",
[(set DPR:$Vd,
(v2i32 (NEONvbsl DPR:$src1, DPR:$Vn, DPR:$Vm)))]>;
+def : Pat<(v8i8 (int_arm_neon_vbsl (v8i8 DPR:$src1),
+ (v8i8 DPR:$Vn), (v8i8 DPR:$Vm))),
+ (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
+ Requires<[HasNEON]>;
+def : Pat<(v4i16 (int_arm_neon_vbsl (v4i16 DPR:$src1),
+ (v4i16 DPR:$Vn), (v4i16 DPR:$Vm))),
+ (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
+ Requires<[HasNEON]>;
+def : Pat<(v2i32 (int_arm_neon_vbsl (v2i32 DPR:$src1),
+ (v2i32 DPR:$Vn), (v2i32 DPR:$Vm))),
+ (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
+ Requires<[HasNEON]>;
+def : Pat<(v2f32 (int_arm_neon_vbsl (v2f32 DPR:$src1),
+ (v2f32 DPR:$Vn), (v2f32 DPR:$Vm))),
+ (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
+ Requires<[HasNEON]>;
+def : Pat<(v1i64 (int_arm_neon_vbsl (v1i64 DPR:$src1),
+ (v1i64 DPR:$Vn), (v1i64 DPR:$Vm))),
+ (VBSLd DPR:$src1, DPR:$Vn, DPR:$Vm)>,
+ Requires<[HasNEON]>;
def : Pat<(v2i32 (or (and DPR:$Vn, DPR:$Vd),
(and DPR:$Vm, (vnotd DPR:$Vd)))),
- (VBSLd DPR:$Vd, DPR:$Vn, DPR:$Vm)>;
+ (VBSLd DPR:$Vd, DPR:$Vn, DPR:$Vm)>,
+ Requires<[HasNEON]>;
+
+def : Pat<(v1i64 (or (and DPR:$Vn, DPR:$Vd),
+ (and DPR:$Vm, (vnotd DPR:$Vd)))),
+ (VBSLd DPR:$Vd, DPR:$Vn, DPR:$Vm)>,
+ Requires<[HasNEON]>;
def VBSLq : N3VX<1, 0, 0b01, 0b0001, 1, 1, (outs QPR:$Vd),
(ins QPR:$src1, QPR:$Vn, QPR:$Vm),
[(set QPR:$Vd,
(v4i32 (NEONvbsl QPR:$src1, QPR:$Vn, QPR:$Vm)))]>;
+def : Pat<(v16i8 (int_arm_neon_vbsl (v16i8 QPR:$src1),
+ (v16i8 QPR:$Vn), (v16i8 QPR:$Vm))),
+ (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
+ Requires<[HasNEON]>;
+def : Pat<(v8i16 (int_arm_neon_vbsl (v8i16 QPR:$src1),
+ (v8i16 QPR:$Vn), (v8i16 QPR:$Vm))),
+ (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
+ Requires<[HasNEON]>;
+def : Pat<(v4i32 (int_arm_neon_vbsl (v4i32 QPR:$src1),
+ (v4i32 QPR:$Vn), (v4i32 QPR:$Vm))),
+ (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
+ Requires<[HasNEON]>;
+def : Pat<(v4f32 (int_arm_neon_vbsl (v4f32 QPR:$src1),
+ (v4f32 QPR:$Vn), (v4f32 QPR:$Vm))),
+ (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
+ Requires<[HasNEON]>;
+def : Pat<(v2i64 (int_arm_neon_vbsl (v2i64 QPR:$src1),
+ (v2i64 QPR:$Vn), (v2i64 QPR:$Vm))),
+ (VBSLq QPR:$src1, QPR:$Vn, QPR:$Vm)>,
+ Requires<[HasNEON]>;
+
def : Pat<(v4i32 (or (and QPR:$Vn, QPR:$Vd),
(and QPR:$Vm, (vnotq QPR:$Vd)))),
- (VBSLq QPR:$Vd, QPR:$Vn, QPR:$Vm)>;
+ (VBSLq QPR:$Vd, QPR:$Vn, QPR:$Vm)>,
+ Requires<[HasNEON]>;
+def : Pat<(v2i64 (or (and QPR:$Vn, QPR:$Vd),
+ (and QPR:$Vm, (vnotq QPR:$Vd)))),
+ (VBSLq QPR:$Vd, QPR:$Vn, QPR:$Vm)>,
+ Requires<[HasNEON]>;
// VBIF : Vector Bitwise Insert if False
// like VBSL but with: "vbif $dst, $src3, $src1", "$src2 = $dst",
defm VABDLu : N3VLIntExt_QHS<1,1,0b0111,0, IIC_VSUBi4Q,
"vabdl", "u", int_arm_neon_vabdu, zext, 1>;
+def abd_shr :
+ PatFrag<(ops node:$in1, node:$in2, node:$shift),
+ (NEONvshrs (sub (zext node:$in1),
+ (zext node:$in2)), (i32 $shift))>;
+
+def : Pat<(xor (v4i32 (bitconvert (v8i16 (abd_shr (v8i8 DPR:$opA), (v8i8 DPR:$opB), 15)))),
+ (v4i32 (bitconvert (v8i16 (add (sub (zext (v8i8 DPR:$opA)),
+ (zext (v8i8 DPR:$opB))),
+ (v8i16 (abd_shr (v8i8 DPR:$opA), (v8i8 DPR:$opB), 15))))))),
+ (VABDLuv8i16 DPR:$opA, DPR:$opB)>;
+
+def : Pat<(xor (v4i32 (abd_shr (v4i16 DPR:$opA), (v4i16 DPR:$opB), 31)),
+ (v4i32 (add (sub (zext (v4i16 DPR:$opA)),
+ (zext (v4i16 DPR:$opB))),
+ (abd_shr (v4i16 DPR:$opA), (v4i16 DPR:$opB), 31)))),
+ (VABDLuv4i32 DPR:$opA, DPR:$opB)>;
+
+def : Pat<(xor (v4i32 (bitconvert (v2i64 (abd_shr (v2i32 DPR:$opA), (v2i32 DPR:$opB), 63)))),
+ (v4i32 (bitconvert (v2i64 (add (sub (zext (v2i32 DPR:$opA)),
+ (zext (v2i32 DPR:$opB))),
+ (abd_shr (v2i32 DPR:$opA), (v2i32 DPR:$opB), 63)))))),
+ (VABDLuv2i64 DPR:$opA, DPR:$opB)>;
+
// VABA : Vector Absolute Difference and Accumulate
defm VABAs : N3VIntOp_QHS<0,0,0b0111,1, IIC_VABAD, IIC_VABAQ,
"vaba", "s", int_arm_neon_vabds, add>;
// VMAX : Vector Maximum
defm VMAXs : N3VInt_QHS<0, 0, 0b0110, 0, N3RegFrm,
IIC_VSUBi4D, IIC_VSUBi4D, IIC_VSUBi4Q, IIC_VSUBi4Q,
- "vmax", "s", int_arm_neon_vmaxs, 1>;
+ "vmax", "s", smax, 1>;
defm VMAXu : N3VInt_QHS<1, 0, 0b0110, 0, N3RegFrm,
IIC_VSUBi4D, IIC_VSUBi4D, IIC_VSUBi4Q, IIC_VSUBi4Q,
- "vmax", "u", int_arm_neon_vmaxu, 1>;
+ "vmax", "u", umax, 1>;
def VMAXfd : N3VDInt<0, 0, 0b00, 0b1111, 0, N3RegFrm, IIC_VBIND,
"vmax", "f32",
- v2f32, v2f32, int_arm_neon_vmaxs, 1>;
+ v2f32, v2f32, fmaxnan, 1>;
def VMAXfq : N3VQInt<0, 0, 0b00, 0b1111, 0, N3RegFrm, IIC_VBINQ,
"vmax", "f32",
- v4f32, v4f32, int_arm_neon_vmaxs, 1>;
+ v4f32, v4f32, fmaxnan, 1>;
+
+// VMAXNM
+let PostEncoderMethod = "NEONThumb2V8PostEncoder", DecoderNamespace = "v8NEON" in {
+ def VMAXNMND : N3VDIntnp<0b00110, 0b00, 0b1111, 0, 1,
+ N3RegFrm, NoItinerary, "vmaxnm", "f32",
+ v2f32, v2f32, fmaxnum, 1>,
+ Requires<[HasV8, HasNEON]>;
+ def VMAXNMNQ : N3VQIntnp<0b00110, 0b00, 0b1111, 1, 1,
+ N3RegFrm, NoItinerary, "vmaxnm", "f32",
+ v4f32, v4f32, fmaxnum, 1>,
+ Requires<[HasV8, HasNEON]>;
+}
// VMIN : Vector Minimum
defm VMINs : N3VInt_QHS<0, 0, 0b0110, 1, N3RegFrm,
IIC_VSUBi4D, IIC_VSUBi4D, IIC_VSUBi4Q, IIC_VSUBi4Q,
- "vmin", "s", int_arm_neon_vmins, 1>;
+ "vmin", "s", smin, 1>;
defm VMINu : N3VInt_QHS<1, 0, 0b0110, 1, N3RegFrm,
IIC_VSUBi4D, IIC_VSUBi4D, IIC_VSUBi4Q, IIC_VSUBi4Q,
- "vmin", "u", int_arm_neon_vminu, 1>;
+ "vmin", "u", umin, 1>;
def VMINfd : N3VDInt<0, 0, 0b10, 0b1111, 0, N3RegFrm, IIC_VBIND,
"vmin", "f32",
- v2f32, v2f32, int_arm_neon_vmins, 1>;
+ v2f32, v2f32, fminnan, 1>;
def VMINfq : N3VQInt<0, 0, 0b10, 0b1111, 0, N3RegFrm, IIC_VBINQ,
"vmin", "f32",
- v4f32, v4f32, int_arm_neon_vmins, 1>;
+ v4f32, v4f32, fminnan, 1>;
+
+// VMINNM
+let PostEncoderMethod = "NEONThumb2V8PostEncoder", DecoderNamespace = "v8NEON" in {
+ def VMINNMND : N3VDIntnp<0b00110, 0b10, 0b1111, 0, 1,
+ N3RegFrm, NoItinerary, "vminnm", "f32",
+ v2f32, v2f32, fminnum, 1>,
+ Requires<[HasV8, HasNEON]>;
+ def VMINNMNQ : N3VQIntnp<0b00110, 0b10, 0b1111, 1, 1,
+ N3RegFrm, NoItinerary, "vminnm", "f32",
+ v4f32, v4f32, fminnum, 1>,
+ Requires<[HasV8, HasNEON]>;
+}
// Vector Pairwise Operations.
NEONvshru>;
// VSHLL : Vector Shift Left Long
-defm VSHLLs : N2VLSh_QHS<0, 1, 0b1010, 0, 0, 1, "vshll", "s", NEONvshlls>;
-defm VSHLLu : N2VLSh_QHS<1, 1, 0b1010, 0, 0, 1, "vshll", "u", NEONvshllu>;
+defm VSHLLs : N2VLSh_QHS<0, 1, 0b1010, 0, 0, 1, "vshll", "s",
+ PatFrag<(ops node:$LHS, node:$RHS), (NEONvshl (sext node:$LHS), node:$RHS)>>;
+defm VSHLLu : N2VLSh_QHS<1, 1, 0b1010, 0, 0, 1, "vshll", "u",
+ PatFrag<(ops node:$LHS, node:$RHS), (NEONvshl (zext node:$LHS), node:$RHS)>>;
// VSHLL : Vector Shift Left Long (with maximum shift count)
class N2VLShMax<bit op24, bit op23, bits<6> op21_16, bits<4> op11_8, bit op7,
bit op6, bit op4, string OpcodeStr, string Dt, ValueType ResTy,
- ValueType OpTy, Operand ImmTy, SDNode OpNode>
+ ValueType OpTy, Operand ImmTy>
: N2VLSh<op24, op23, op11_8, op7, op6, op4, OpcodeStr, Dt,
- ResTy, OpTy, ImmTy, OpNode> {
+ ResTy, OpTy, ImmTy, null_frag> {
let Inst{21-16} = op21_16;
let DecoderMethod = "DecodeVSHLMaxInstruction";
}
def VSHLLi8 : N2VLShMax<1, 1, 0b110010, 0b0011, 0, 0, 0, "vshll", "i8",
- v8i16, v8i8, imm8, NEONvshlli>;
+ v8i16, v8i8, imm8>;
def VSHLLi16 : N2VLShMax<1, 1, 0b110110, 0b0011, 0, 0, 0, "vshll", "i16",
- v4i32, v4i16, imm16, NEONvshlli>;
+ v4i32, v4i16, imm16>;
def VSHLLi32 : N2VLShMax<1, 1, 0b111010, 0b0011, 0, 0, 0, "vshll", "i32",
- v2i64, v2i32, imm32, NEONvshlli>;
+ v2i64, v2i32, imm32>;
+
+def : Pat<(v8i16 (NEONvshl (zext (v8i8 DPR:$Rn)), (i32 8))),
+ (VSHLLi8 DPR:$Rn, 8)>;
+def : Pat<(v4i32 (NEONvshl (zext (v4i16 DPR:$Rn)), (i32 16))),
+ (VSHLLi16 DPR:$Rn, 16)>;
+def : Pat<(v2i64 (NEONvshl (zext (v2i32 DPR:$Rn)), (i32 32))),
+ (VSHLLi32 DPR:$Rn, 32)>;
+def : Pat<(v8i16 (NEONvshl (sext (v8i8 DPR:$Rn)), (i32 8))),
+ (VSHLLi8 DPR:$Rn, 8)>;
+def : Pat<(v4i32 (NEONvshl (sext (v4i16 DPR:$Rn)), (i32 16))),
+ (VSHLLi16 DPR:$Rn, 16)>;
+def : Pat<(v2i64 (NEONvshl (sext (v2i32 DPR:$Rn)), (i32 32))),
+ (VSHLLi32 DPR:$Rn, 32)>;
// VSHRN : Vector Shift Right and Narrow
defm VSHRN : N2VNSh_HSD<0,1,0b1000,0,0,1, IIC_VSHLiD, "vshrn", "i",
- NEONvshrn>;
+ PatFrag<(ops node:$Rn, node:$amt),
+ (trunc (NEONvshrs node:$Rn, node:$amt))>>;
+
+def : Pat<(v8i8 (trunc (NEONvshru (v8i16 QPR:$Vn), shr_imm8:$amt))),
+ (VSHRNv8i8 QPR:$Vn, shr_imm8:$amt)>;
+def : Pat<(v4i16 (trunc (NEONvshru (v4i32 QPR:$Vn), shr_imm16:$amt))),
+ (VSHRNv4i16 QPR:$Vn, shr_imm16:$amt)>;
+def : Pat<(v2i32 (trunc (NEONvshru (v2i64 QPR:$Vn), shr_imm32:$amt))),
+ (VSHRNv2i32 QPR:$Vn, shr_imm32:$amt)>;
// VRSHL : Vector Rounding Shift
defm VRSHLs : N3VInt_QHSDSh<0, 0, 0b0101, 0, N3RegVShFrm,
defm VABS : N2VInt_QHS<0b11, 0b11, 0b01, 0b00110, 0,
IIC_VUNAiD, IIC_VUNAiQ, "vabs", "s",
int_arm_neon_vabs>;
-def VABSfd : N2VDInt<0b11, 0b11, 0b10, 0b01, 0b01110, 0,
- IIC_VUNAD, "vabs", "f32",
- v2f32, v2f32, int_arm_neon_vabs>;
-def VABSfq : N2VQInt<0b11, 0b11, 0b10, 0b01, 0b01110, 0,
- IIC_VUNAQ, "vabs", "f32",
- v4f32, v4f32, int_arm_neon_vabs>;
+def VABSfd : N2VD<0b11, 0b11, 0b10, 0b01, 0b01110, 0,
+ "vabs", "f32",
+ v2f32, v2f32, fabs>;
+def VABSfq : N2VQ<0b11, 0b11, 0b10, 0b01, 0b01110, 0,
+ "vabs", "f32",
+ v4f32, v4f32, fabs>;
+
+def : Pat<(xor (v2i32 (bitconvert (v8i8 (NEONvshrs DPR:$src, (i32 7))))),
+ (v2i32 (bitconvert (v8i8 (add DPR:$src,
+ (NEONvshrs DPR:$src, (i32 7))))))),
+ (VABSv8i8 DPR:$src)>;
+def : Pat<(xor (v2i32 (bitconvert (v4i16 (NEONvshrs DPR:$src, (i32 15))))),
+ (v2i32 (bitconvert (v4i16 (add DPR:$src,
+ (NEONvshrs DPR:$src, (i32 15))))))),
+ (VABSv4i16 DPR:$src)>;
+def : Pat<(xor (v2i32 (NEONvshrs DPR:$src, (i32 31))),
+ (v2i32 (add DPR:$src, (NEONvshrs DPR:$src, (i32 31))))),
+ (VABSv2i32 DPR:$src)>;
+def : Pat<(xor (v4i32 (bitconvert (v16i8 (NEONvshrs QPR:$src, (i32 7))))),
+ (v4i32 (bitconvert (v16i8 (add QPR:$src,
+ (NEONvshrs QPR:$src, (i32 7))))))),
+ (VABSv16i8 QPR:$src)>;
+def : Pat<(xor (v4i32 (bitconvert (v8i16 (NEONvshrs QPR:$src, (i32 15))))),
+ (v4i32 (bitconvert (v8i16 (add QPR:$src,
+ (NEONvshrs QPR:$src, (i32 15))))))),
+ (VABSv8i16 QPR:$src)>;
+def : Pat<(xor (v4i32 (NEONvshrs QPR:$src, (i32 31))),
+ (v4i32 (add QPR:$src, (NEONvshrs QPR:$src, (i32 31))))),
+ (VABSv4i32 QPR:$src)>;
// VQABS : Vector Saturating Absolute Value
defm VQABS : N2VInt_QHS<0b11, 0b11, 0b00, 0b01110, 0,
// VCLZ : Vector Count Leading Zeros
defm VCLZ : N2VInt_QHS<0b11, 0b11, 0b00, 0b01001, 0,
IIC_VCNTiD, IIC_VCNTiQ, "vclz", "i",
- int_arm_neon_vclz>;
+ ctlz>;
// VCNT : Vector Count One Bits
def VCNTd : N2VDInt<0b11, 0b11, 0b00, 0b00, 0b01010, 0,
IIC_VCNTiD, "vcnt", "8",
- v8i8, v8i8, int_arm_neon_vcnt>;
+ v8i8, v8i8, ctpop>;
def VCNTq : N2VQInt<0b11, 0b11, 0b00, 0b00, 0b01010, 0,
IIC_VCNTiQ, "vcnt", "8",
- v16i8, v16i8, int_arm_neon_vcnt>;
+ v16i8, v16i8, ctpop>;
// Vector Swap
def VSWPd : N2VX<0b11, 0b11, 0b00, 0b10, 0b00000, 0, 0,
// Vector Move Operations.
// VMOV : Vector Move (Register)
-def : InstAlias<"vmov${p} $Vd, $Vm",
- (VORRd DPR:$Vd, DPR:$Vm, DPR:$Vm, pred:$p)>;
-def : InstAlias<"vmov${p} $Vd, $Vm",
- (VORRq QPR:$Vd, QPR:$Vm, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmov${p} $Vd, $Vm",
+ (VORRd DPR:$Vd, DPR:$Vm, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmov${p} $Vd, $Vm",
+ (VORRq QPR:$Vd, QPR:$Vm, QPR:$Vm, pred:$p)>;
// VMOV : Vector Move (Immediate)
[(set QPR:$Vd, (v4f32 (NEONvmovFPImm timm:$SIMM)))]>;
} // isReMaterializable
+// Add support for bytes replication feature, so it could be GAS compatible.
+// E.g. instructions below:
+// "vmov.i32 d0, 0xffffffff"
+// "vmov.i32 d0, 0xabababab"
+// "vmov.i16 d0, 0xabab"
+// are incorrect, but we could deal with such cases.
+// For last two instructions, for example, it should emit:
+// "vmov.i8 d0, 0xab"
+def : NEONInstAlias<"vmov${p}.i16 $Vd, $Vm",
+ (VMOVv8i8 DPR:$Vd, nImmVMOVI16ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmov${p}.i32 $Vd, $Vm",
+ (VMOVv8i8 DPR:$Vd, nImmVMOVI32ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmov${p}.i16 $Vd, $Vm",
+ (VMOVv16i8 QPR:$Vd, nImmVMOVI16ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmov${p}.i32 $Vd, $Vm",
+ (VMOVv16i8 QPR:$Vd, nImmVMOVI32ByteReplicate:$Vm, pred:$p)>;
+
+// Also add same support for VMVN instructions. So instruction:
+// "vmvn.i32 d0, 0xabababab"
+// actually means:
+// "vmov.i8 d0, 0x54"
+def : NEONInstAlias<"vmvn${p}.i16 $Vd, $Vm",
+ (VMOVv8i8 DPR:$Vd, nImmVMVNI16ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmvn${p}.i32 $Vd, $Vm",
+ (VMOVv8i8 DPR:$Vd, nImmVMVNI32ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmvn${p}.i16 $Vd, $Vm",
+ (VMOVv16i8 QPR:$Vd, nImmVMVNI16ByteReplicate:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmvn${p}.i32 $Vd, $Vm",
+ (VMOVv16i8 QPR:$Vd, nImmVMVNI32ByteReplicate:$Vm, pred:$p)>;
+
+// On some CPUs the two instructions "vmov.i32 dD, #0" and "vmov.i32 qD, #0"
+// require zero cycles to execute so they should be used wherever possible for
+// setting a register to zero.
+
+// Even without these pseudo-insts we would probably end up with the correct
+// instruction, but we could not mark the general ones with "isAsCheapAsAMove"
+// since they are sometimes rather expensive (in general).
+
+let AddedComplexity = 50, isAsCheapAsAMove = 1, isReMaterializable = 1 in {
+ def VMOVD0 : ARMPseudoExpand<(outs DPR:$Vd), (ins), 4, IIC_VMOVImm,
+ [(set DPR:$Vd, (v2i32 NEONimmAllZerosV))],
+ (VMOVv2i32 DPR:$Vd, 0, (ops 14, zero_reg))>,
+ Requires<[HasZCZ]>;
+ def VMOVQ0 : ARMPseudoExpand<(outs QPR:$Vd), (ins), 4, IIC_VMOVImm,
+ [(set QPR:$Vd, (v4i32 NEONimmAllZerosV))],
+ (VMOVv4i32 QPR:$Vd, 0, (ops 14, zero_reg))>,
+ Requires<[HasZCZ]>;
+}
+
// VMOV : Vector Get Lane (move scalar to ARM core register)
def VGETLNs8 : NVGetLane<{1,1,1,0,0,1,?,1}, 0b1011, {?,?},
(outs GPR:$R), (ins DPR:$V, VectorIndex32:$lane),
IIC_VMOVSI, "vmov", "32", "$R, $V$lane",
[(set GPR:$R, (extractelt (v2i32 DPR:$V),
- imm:$lane))]> {
+ imm:$lane))]>,
+ Requires<[HasVFP2, HasFastVGETLNi32]> {
let Inst{21} = lane{0};
}
// def VGETLNf32: see FMRDH and FMRDL in ARMInstrVFP.td
def : Pat<(extractelt (v4i32 QPR:$src), imm:$lane),
(VGETLNi32 (v2i32 (EXTRACT_SUBREG QPR:$src,
(DSubReg_i32_reg imm:$lane))),
- (SubReg_i32_lane imm:$lane))>;
+ (SubReg_i32_lane imm:$lane))>,
+ Requires<[HasNEON, HasFastVGETLNi32]>;
+def : Pat<(extractelt (v2i32 DPR:$src), imm:$lane),
+ (COPY_TO_REGCLASS
+ (i32 (EXTRACT_SUBREG DPR:$src, (SSubReg_f32_reg imm:$lane))), GPR)>,
+ Requires<[HasNEON, HasSlowVGETLNi32]>;
+def : Pat<(extractelt (v4i32 QPR:$src), imm:$lane),
+ (COPY_TO_REGCLASS
+ (i32 (EXTRACT_SUBREG QPR:$src, (SSubReg_f32_reg imm:$lane))), GPR)>,
+ Requires<[HasNEON, HasSlowVGETLNi32]>;
def : Pat<(extractelt (v2f32 DPR:$src1), imm:$src2),
(EXTRACT_SUBREG (v2f32 (COPY_TO_REGCLASS (v2f32 DPR:$src1),DPR_VFP2)),
(SSubReg_f32_reg imm:$src2))>;
(ins DPR:$src1, GPR:$R, VectorIndex32:$lane),
IIC_VMOVISL, "vmov", "32", "$V$lane, $R",
[(set DPR:$V, (insertelt (v2i32 DPR:$src1),
- GPR:$R, imm:$lane))]> {
+ GPR:$R, imm:$lane))]>,
+ Requires<[HasVFP2]> {
let Inst{21} = lane{0};
+ // This instruction is equivalent as
+ // $V = INSERT_SUBREG $src1, $R, translateImmToSubIdx($imm)
+ let isInsertSubreg = 1;
}
}
def : Pat<(vector_insert (v16i8 QPR:$src1), GPR:$src2, imm:$lane),
def VDUP8d : VDUPD<0b11101100, 0b00, "8", v8i8>;
def VDUP16d : VDUPD<0b11101000, 0b01, "16", v4i16>;
-def VDUP32d : VDUPD<0b11101000, 0b00, "32", v2i32>;
+def VDUP32d : VDUPD<0b11101000, 0b00, "32", v2i32>,
+ Requires<[HasNEON, HasFastVDUP32]>;
def VDUP8q : VDUPQ<0b11101110, 0b00, "8", v16i8>;
def VDUP16q : VDUPQ<0b11101010, 0b01, "16", v8i16>;
def VDUP32q : VDUPQ<0b11101010, 0b00, "32", v4i32>;
-def : Pat<(v2f32 (NEONvdup (f32 (bitconvert GPR:$R)))), (VDUP32d GPR:$R)>;
+// NEONvdup patterns for uarchs with fast VDUP.32.
+def : Pat<(v2f32 (NEONvdup (f32 (bitconvert GPR:$R)))), (VDUP32d GPR:$R)>,
+ Requires<[HasNEON,HasFastVDUP32]>;
def : Pat<(v4f32 (NEONvdup (f32 (bitconvert GPR:$R)))), (VDUP32q GPR:$R)>;
+// NEONvdup patterns for uarchs with slow VDUP.32 - use VMOVDRR instead.
+def : Pat<(v2i32 (NEONvdup (i32 GPR:$R))), (VMOVDRR GPR:$R, GPR:$R)>,
+ Requires<[HasNEON,HasSlowVDUP32]>;
+def : Pat<(v2f32 (NEONvdup (f32 (bitconvert GPR:$R)))), (VMOVDRR GPR:$R, GPR:$R)>,
+ Requires<[HasNEON,HasSlowVDUP32]>;
+
// VDUP : Vector Duplicate Lane (from scalar to all elements)
class VDUPLND<bits<4> op19_16, string OpcodeStr, string Dt,
(DSubReg_i32_reg imm:$lane))),
(SubReg_i32_lane imm:$lane)))>;
-def VDUPfdf : PseudoNeonI<(outs DPR:$dst), (ins SPR:$src), IIC_VMOVD, "",
- [(set DPR:$dst, (v2f32 (NEONvdup (f32 SPR:$src))))]>;
-def VDUPfqf : PseudoNeonI<(outs QPR:$dst), (ins SPR:$src), IIC_VMOVD, "",
- [(set QPR:$dst, (v4f32 (NEONvdup (f32 SPR:$src))))]>;
+def : Pat<(v2f32 (NEONvdup (f32 SPR:$src))),
+ (v2f32 (VDUPLN32d (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
+ SPR:$src, ssub_0), (i32 0)))>;
+def : Pat<(v4f32 (NEONvdup (f32 SPR:$src))),
+ (v4f32 (VDUPLN32q (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
+ SPR:$src, ssub_0), (i32 0)))>;
// VMOVN : Vector Narrowing Move
defm VMOVN : N2VN_HSD<0b11,0b11,0b10,0b00100,0,0, IIC_VMOVN,
def VCVTu2fq : N2VQ<0b11, 0b11, 0b10, 0b11, 0b01101, 0, "vcvt", "f32.u32",
v4f32, v4i32, uint_to_fp>;
+// VCVT{A, N, P, M}
+multiclass VCVT_FPI<string op, bits<3> op10_8, SDPatternOperator IntS,
+ SDPatternOperator IntU> {
+ let PostEncoderMethod = "NEONThumb2V8PostEncoder", DecoderNamespace = "v8NEON" in {
+ def SD : N2VDIntnp<0b11, op10_8, 0, NoItinerary, !strconcat("vcvt", op),
+ "s32.f32", v2i32, v2f32, IntS>, Requires<[HasV8, HasNEON]>;
+ def SQ : N2VQIntnp<0b11, op10_8, 0, NoItinerary, !strconcat("vcvt", op),
+ "s32.f32", v4i32, v4f32, IntS>, Requires<[HasV8, HasNEON]>;
+ def UD : N2VDIntnp<0b11, op10_8, 1, NoItinerary, !strconcat("vcvt", op),
+ "u32.f32", v2i32, v2f32, IntU>, Requires<[HasV8, HasNEON]>;
+ def UQ : N2VQIntnp<0b11, op10_8, 1, NoItinerary, !strconcat("vcvt", op),
+ "u32.f32", v4i32, v4f32, IntU>, Requires<[HasV8, HasNEON]>;
+ }
+}
+
+defm VCVTAN : VCVT_FPI<"a", 0b000, int_arm_neon_vcvtas, int_arm_neon_vcvtau>;
+defm VCVTNN : VCVT_FPI<"n", 0b001, int_arm_neon_vcvtns, int_arm_neon_vcvtnu>;
+defm VCVTPN : VCVT_FPI<"p", 0b010, int_arm_neon_vcvtps, int_arm_neon_vcvtpu>;
+defm VCVTMN : VCVT_FPI<"m", 0b011, int_arm_neon_vcvtms, int_arm_neon_vcvtmu>;
+
// VCVT : Vector Convert Between Floating-Point and Fixed-Point.
let DecoderMethod = "DecodeVCVTD" in {
def VCVTf2xsd : N2VCvtD<0, 1, 0b1111, 0, 1, "vcvt", "s32.f32",
v4f32, v4i32, int_arm_neon_vcvtfxu2fp>;
}
+def : NEONInstAlias<"vcvt${p}.s32.f32 $Dd, $Dm, #0",
+ (VCVTf2sd DPR:$Dd, DPR:$Dm, pred:$p)>;
+def : NEONInstAlias<"vcvt${p}.u32.f32 $Dd, $Dm, #0",
+ (VCVTf2ud DPR:$Dd, DPR:$Dm, pred:$p)>;
+def : NEONInstAlias<"vcvt${p}.f32.s32 $Dd, $Dm, #0",
+ (VCVTs2fd DPR:$Dd, DPR:$Dm, pred:$p)>;
+def : NEONInstAlias<"vcvt${p}.f32.u32 $Dd, $Dm, #0",
+ (VCVTu2fd DPR:$Dd, DPR:$Dm, pred:$p)>;
+
+def : NEONInstAlias<"vcvt${p}.s32.f32 $Qd, $Qm, #0",
+ (VCVTf2sq QPR:$Qd, QPR:$Qm, pred:$p)>;
+def : NEONInstAlias<"vcvt${p}.u32.f32 $Qd, $Qm, #0",
+ (VCVTf2uq QPR:$Qd, QPR:$Qm, pred:$p)>;
+def : NEONInstAlias<"vcvt${p}.f32.s32 $Qd, $Qm, #0",
+ (VCVTs2fq QPR:$Qd, QPR:$Qm, pred:$p)>;
+def : NEONInstAlias<"vcvt${p}.f32.u32 $Qd, $Qm, #0",
+ (VCVTu2fq QPR:$Qd, QPR:$Qm, pred:$p)>;
+
+
// VCVT : Vector Convert Between Half-Precision and Single-Precision.
def VCVTf2h : N2VNInt<0b11, 0b11, 0b01, 0b10, 0b01100, 0, 0,
IIC_VUNAQ, "vcvt", "f16.f32",
IIC_VEXTD, OpcodeStr, Dt, "$Vd, $Vn, $Vm, $index", "",
[(set DPR:$Vd, (Ty (NEONvext (Ty DPR:$Vn),
(Ty DPR:$Vm), imm:$index)))]> {
- bits<4> index;
- let Inst{11-8} = index{3-0};
+ bits<3> index;
+ let Inst{11} = 0b0;
+ let Inst{10-8} = index{2-0};
}
class VEXTq<string OpcodeStr, string Dt, ValueType Ty, Operand immTy>
}
def VEXTd8 : VEXTd<"vext", "8", v8i8, imm0_7> {
- let Inst{11-8} = index{3-0};
+ let Inst{10-8} = index{2-0};
}
def VEXTd16 : VEXTd<"vext", "16", v4i16, imm0_3> {
- let Inst{11-9} = index{2-0};
+ let Inst{10-9} = index{1-0};
let Inst{8} = 0b0;
}
def VEXTd32 : VEXTd<"vext", "32", v2i32, imm0_1> {
- let Inst{11-10} = index{1-0};
+ let Inst{10} = index{0};
let Inst{9-8} = 0b00;
}
def : Pat<(v2f32 (NEONvext (v2f32 DPR:$Vn),
IIC_VTBX4, "$orig = $dst", []>;
} // DecoderMethod = "DecodeTBLInstruction"
+// VRINT : Vector Rounding
+multiclass VRINT_FPI<string op, bits<3> op9_7, SDPatternOperator Int> {
+ let PostEncoderMethod = "NEONThumb2V8PostEncoder", DecoderNamespace = "v8NEON" in {
+ def D : N2VDIntnp<0b10, 0b100, 0, NoItinerary,
+ !strconcat("vrint", op), "f32",
+ v2f32, v2f32, Int>, Requires<[HasV8, HasNEON]> {
+ let Inst{9-7} = op9_7;
+ }
+ def Q : N2VQIntnp<0b10, 0b100, 0, NoItinerary,
+ !strconcat("vrint", op), "f32",
+ v4f32, v4f32, Int>, Requires<[HasV8, HasNEON]> {
+ let Inst{9-7} = op9_7;
+ }
+ }
+
+ def : NEONInstAlias<!strconcat("vrint", op, ".f32.f32\t$Dd, $Dm"),
+ (!cast<Instruction>(NAME#"D") DPR:$Dd, DPR:$Dm)>;
+ def : NEONInstAlias<!strconcat("vrint", op, ".f32.f32\t$Qd, $Qm"),
+ (!cast<Instruction>(NAME#"Q") QPR:$Qd, QPR:$Qm)>;
+}
+
+defm VRINTNN : VRINT_FPI<"n", 0b000, int_arm_neon_vrintn>;
+defm VRINTXN : VRINT_FPI<"x", 0b001, int_arm_neon_vrintx>;
+defm VRINTAN : VRINT_FPI<"a", 0b010, int_arm_neon_vrinta>;
+defm VRINTZN : VRINT_FPI<"z", 0b011, int_arm_neon_vrintz>;
+defm VRINTMN : VRINT_FPI<"m", 0b101, int_arm_neon_vrintm>;
+defm VRINTPN : VRINT_FPI<"p", 0b111, int_arm_neon_vrintp>;
+
+// Cryptography instructions
+let PostEncoderMethod = "NEONThumb2DataIPostEncoder",
+ DecoderNamespace = "v8Crypto", hasSideEffects = 0 in {
+ class AES<string op, bit op7, bit op6, SDPatternOperator Int>
+ : N2VQIntXnp<0b00, 0b00, 0b011, op6, op7, NoItinerary,
+ !strconcat("aes", op), "8", v16i8, v16i8, Int>,
+ Requires<[HasV8, HasCrypto]>;
+ class AES2Op<string op, bit op7, bit op6, SDPatternOperator Int>
+ : N2VQIntX2np<0b00, 0b00, 0b011, op6, op7, NoItinerary,
+ !strconcat("aes", op), "8", v16i8, v16i8, Int>,
+ Requires<[HasV8, HasCrypto]>;
+ class N2SHA<string op, bits<2> op17_16, bits<3> op10_8, bit op7, bit op6,
+ SDPatternOperator Int>
+ : N2VQIntXnp<0b10, op17_16, op10_8, op6, op7, NoItinerary,
+ !strconcat("sha", op), "32", v4i32, v4i32, Int>,
+ Requires<[HasV8, HasCrypto]>;
+ class N2SHA2Op<string op, bits<2> op17_16, bits<3> op10_8, bit op7, bit op6,
+ SDPatternOperator Int>
+ : N2VQIntX2np<0b10, op17_16, op10_8, op6, op7, NoItinerary,
+ !strconcat("sha", op), "32", v4i32, v4i32, Int>,
+ Requires<[HasV8, HasCrypto]>;
+ class N3SHA3Op<string op, bits<5> op27_23, bits<2> op21_20, SDPatternOperator Int>
+ : N3VQInt3np<op27_23, op21_20, 0b1100, 1, 0, N3RegFrm, NoItinerary,
+ !strconcat("sha", op), "32", v4i32, v4i32, Int, 0>,
+ Requires<[HasV8, HasCrypto]>;
+}
+
+def AESD : AES2Op<"d", 0, 1, int_arm_neon_aesd>;
+def AESE : AES2Op<"e", 0, 0, int_arm_neon_aese>;
+def AESIMC : AES<"imc", 1, 1, int_arm_neon_aesimc>;
+def AESMC : AES<"mc", 1, 0, int_arm_neon_aesmc>;
+
+def SHA1H : N2SHA<"1h", 0b01, 0b010, 1, 1, null_frag>;
+def SHA1SU1 : N2SHA2Op<"1su1", 0b10, 0b011, 1, 0, int_arm_neon_sha1su1>;
+def SHA256SU0 : N2SHA2Op<"256su0", 0b10, 0b011, 1, 1, int_arm_neon_sha256su0>;
+def SHA1C : N3SHA3Op<"1c", 0b00100, 0b00, null_frag>;
+def SHA1M : N3SHA3Op<"1m", 0b00100, 0b10, null_frag>;
+def SHA1P : N3SHA3Op<"1p", 0b00100, 0b01, null_frag>;
+def SHA1SU0 : N3SHA3Op<"1su0", 0b00100, 0b11, int_arm_neon_sha1su0>;
+def SHA256H : N3SHA3Op<"256h", 0b00110, 0b00, int_arm_neon_sha256h>;
+def SHA256H2 : N3SHA3Op<"256h2", 0b00110, 0b01, int_arm_neon_sha256h2>;
+def SHA256SU1 : N3SHA3Op<"256su1", 0b00110, 0b10, int_arm_neon_sha256su1>;
+
+def : Pat<(i32 (int_arm_neon_sha1h i32:$Rn)),
+ (COPY_TO_REGCLASS (f32 (EXTRACT_SUBREG
+ (SHA1H (SUBREG_TO_REG (i64 0),
+ (f32 (COPY_TO_REGCLASS i32:$Rn, SPR)),
+ ssub_0)),
+ ssub_0)), GPR)>;
+
+def : Pat<(v4i32 (int_arm_neon_sha1c v4i32:$hash_abcd, i32:$hash_e, v4i32:$wk)),
+ (SHA1C v4i32:$hash_abcd,
+ (SUBREG_TO_REG (i64 0),
+ (f32 (COPY_TO_REGCLASS i32:$hash_e, SPR)),
+ ssub_0),
+ v4i32:$wk)>;
+
+def : Pat<(v4i32 (int_arm_neon_sha1m v4i32:$hash_abcd, i32:$hash_e, v4i32:$wk)),
+ (SHA1M v4i32:$hash_abcd,
+ (SUBREG_TO_REG (i64 0),
+ (f32 (COPY_TO_REGCLASS i32:$hash_e, SPR)),
+ ssub_0),
+ v4i32:$wk)>;
+
+def : Pat<(v4i32 (int_arm_neon_sha1p v4i32:$hash_abcd, i32:$hash_e, v4i32:$wk)),
+ (SHA1P v4i32:$hash_abcd,
+ (SUBREG_TO_REG (i64 0),
+ (f32 (COPY_TO_REGCLASS i32:$hash_e, SPR)),
+ ssub_0),
+ v4i32:$wk)>;
+
//===----------------------------------------------------------------------===//
// NEON instructions for single-precision FP math
//===----------------------------------------------------------------------===//
(v2f32 (COPY_TO_REGCLASS (v2f32 (IMPLICIT_DEF)), DPR_VFP2)),
SPR:$b, ssub_0)), DPR_VFP2)), ssub_0)>;
+class NVCVTIFPat<SDNode OpNode, NeonI Inst>
+ : NEONFPPat<(f32 (OpNode GPR:$a)),
+ (f32 (EXTRACT_SUBREG
+ (v2f32 (Inst
+ (INSERT_SUBREG
+ (v2f32 (IMPLICIT_DEF)),
+ (i32 (COPY_TO_REGCLASS GPR:$a, SPR)), ssub_0))),
+ ssub_0))>;
+class NVCVTFIPat<SDNode OpNode, NeonI Inst>
+ : NEONFPPat<(i32 (OpNode SPR:$a)),
+ (i32 (EXTRACT_SUBREG
+ (v2f32 (Inst (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
+ SPR:$a, ssub_0))),
+ ssub_0))>;
+
def : N3VSPat<fadd, VADDfd>;
def : N3VSPat<fsub, VSUBfd>;
def : N3VSPat<fmul, VMULfd>;
Requires<[HasVFP4, UseNEONForFP, UseFusedMAC]>;
def : N2VSPat<fabs, VABSfd>;
def : N2VSPat<fneg, VNEGfd>;
-def : N3VSPat<NEONfmax, VMAXfd>;
-def : N3VSPat<NEONfmin, VMINfd>;
-def : N2VSPat<arm_ftosi, VCVTf2sd>;
-def : N2VSPat<arm_ftoui, VCVTf2ud>;
-def : N2VSPat<arm_sitof, VCVTs2fd>;
-def : N2VSPat<arm_uitof, VCVTu2fd>;
+def : N3VSPat<fmaxnan, VMAXfd>, Requires<[HasNEON]>;
+def : N3VSPat<fminnan, VMINfd>, Requires<[HasNEON]>;
+def : NVCVTFIPat<fp_to_sint, VCVTf2sd>;
+def : NVCVTFIPat<fp_to_uint, VCVTf2ud>;
+def : NVCVTIFPat<sint_to_fp, VCVTs2fd>;
+def : NVCVTIFPat<uint_to_fp, VCVTu2fd>;
+
+// NEON doesn't have any f64 conversions, so provide patterns to make
+// sure the VFP conversions match when extracting from a vector.
+def : VFPPat<(f64 (sint_to_fp (extractelt (v2i32 DPR:$src), imm:$lane))),
+ (VSITOD (EXTRACT_SUBREG DPR:$src, (SSubReg_f32_reg imm:$lane)))>;
+def : VFPPat<(f64 (sint_to_fp (extractelt (v4i32 QPR:$src), imm:$lane))),
+ (VSITOD (EXTRACT_SUBREG QPR:$src, (SSubReg_f32_reg imm:$lane)))>;
+def : VFPPat<(f64 (uint_to_fp (extractelt (v2i32 DPR:$src), imm:$lane))),
+ (VUITOD (EXTRACT_SUBREG DPR:$src, (SSubReg_f32_reg imm:$lane)))>;
+def : VFPPat<(f64 (uint_to_fp (extractelt (v4i32 QPR:$src), imm:$lane))),
+ (VUITOD (EXTRACT_SUBREG QPR:$src, (SSubReg_f32_reg imm:$lane)))>;
+
+
+// Prefer VMOVDRR for i32 -> f32 bitcasts, it can write all DPR registers.
+def : Pat<(f32 (bitconvert GPR:$a)),
+ (EXTRACT_SUBREG (VMOVDRR GPR:$a, GPR:$a), ssub_0)>,
+ Requires<[HasNEON, DontUseVMOVSR]>;
//===----------------------------------------------------------------------===//
// Non-Instruction Patterns
//===----------------------------------------------------------------------===//
// bit_convert
-def : Pat<(v1i64 (bitconvert (v2i32 DPR:$src))), (v1i64 DPR:$src)>;
-def : Pat<(v1i64 (bitconvert (v4i16 DPR:$src))), (v1i64 DPR:$src)>;
-def : Pat<(v1i64 (bitconvert (v8i8 DPR:$src))), (v1i64 DPR:$src)>;
+let Predicates = [IsLE] in {
+ def : Pat<(v1i64 (bitconvert (v2i32 DPR:$src))), (v1i64 DPR:$src)>;
+ def : Pat<(v1i64 (bitconvert (v4i16 DPR:$src))), (v1i64 DPR:$src)>;
+ def : Pat<(v1i64 (bitconvert (v8i8 DPR:$src))), (v1i64 DPR:$src)>;
+}
def : Pat<(v1i64 (bitconvert (f64 DPR:$src))), (v1i64 DPR:$src)>;
-def : Pat<(v1i64 (bitconvert (v2f32 DPR:$src))), (v1i64 DPR:$src)>;
-def : Pat<(v2i32 (bitconvert (v1i64 DPR:$src))), (v2i32 DPR:$src)>;
-def : Pat<(v2i32 (bitconvert (v4i16 DPR:$src))), (v2i32 DPR:$src)>;
-def : Pat<(v2i32 (bitconvert (v8i8 DPR:$src))), (v2i32 DPR:$src)>;
-def : Pat<(v2i32 (bitconvert (f64 DPR:$src))), (v2i32 DPR:$src)>;
+let Predicates = [IsLE] in {
+ def : Pat<(v1i64 (bitconvert (v2f32 DPR:$src))), (v1i64 DPR:$src)>;
+ def : Pat<(v2i32 (bitconvert (v1i64 DPR:$src))), (v2i32 DPR:$src)>;
+ def : Pat<(v2i32 (bitconvert (v4i16 DPR:$src))), (v2i32 DPR:$src)>;
+ def : Pat<(v2i32 (bitconvert (v8i8 DPR:$src))), (v2i32 DPR:$src)>;
+ def : Pat<(v2i32 (bitconvert (f64 DPR:$src))), (v2i32 DPR:$src)>;
+}
def : Pat<(v2i32 (bitconvert (v2f32 DPR:$src))), (v2i32 DPR:$src)>;
-def : Pat<(v4i16 (bitconvert (v1i64 DPR:$src))), (v4i16 DPR:$src)>;
-def : Pat<(v4i16 (bitconvert (v2i32 DPR:$src))), (v4i16 DPR:$src)>;
-def : Pat<(v4i16 (bitconvert (v8i8 DPR:$src))), (v4i16 DPR:$src)>;
-def : Pat<(v4i16 (bitconvert (f64 DPR:$src))), (v4i16 DPR:$src)>;
-def : Pat<(v4i16 (bitconvert (v2f32 DPR:$src))), (v4i16 DPR:$src)>;
-def : Pat<(v8i8 (bitconvert (v1i64 DPR:$src))), (v8i8 DPR:$src)>;
-def : Pat<(v8i8 (bitconvert (v2i32 DPR:$src))), (v8i8 DPR:$src)>;
-def : Pat<(v8i8 (bitconvert (v4i16 DPR:$src))), (v8i8 DPR:$src)>;
-def : Pat<(v8i8 (bitconvert (f64 DPR:$src))), (v8i8 DPR:$src)>;
-def : Pat<(v8i8 (bitconvert (v2f32 DPR:$src))), (v8i8 DPR:$src)>;
+let Predicates = [IsLE] in {
+ def : Pat<(v4i16 (bitconvert (v1i64 DPR:$src))), (v4i16 DPR:$src)>;
+ def : Pat<(v4i16 (bitconvert (v2i32 DPR:$src))), (v4i16 DPR:$src)>;
+ def : Pat<(v4i16 (bitconvert (v8i8 DPR:$src))), (v4i16 DPR:$src)>;
+ def : Pat<(v4i16 (bitconvert (f64 DPR:$src))), (v4i16 DPR:$src)>;
+ def : Pat<(v4i16 (bitconvert (v2f32 DPR:$src))), (v4i16 DPR:$src)>;
+ def : Pat<(v8i8 (bitconvert (v1i64 DPR:$src))), (v8i8 DPR:$src)>;
+ def : Pat<(v8i8 (bitconvert (v2i32 DPR:$src))), (v8i8 DPR:$src)>;
+ def : Pat<(v8i8 (bitconvert (v4i16 DPR:$src))), (v8i8 DPR:$src)>;
+ def : Pat<(v8i8 (bitconvert (f64 DPR:$src))), (v8i8 DPR:$src)>;
+ def : Pat<(v8i8 (bitconvert (v2f32 DPR:$src))), (v8i8 DPR:$src)>;
+}
def : Pat<(f64 (bitconvert (v1i64 DPR:$src))), (f64 DPR:$src)>;
-def : Pat<(f64 (bitconvert (v2i32 DPR:$src))), (f64 DPR:$src)>;
-def : Pat<(f64 (bitconvert (v4i16 DPR:$src))), (f64 DPR:$src)>;
-def : Pat<(f64 (bitconvert (v8i8 DPR:$src))), (f64 DPR:$src)>;
-def : Pat<(f64 (bitconvert (v2f32 DPR:$src))), (f64 DPR:$src)>;
-def : Pat<(v2f32 (bitconvert (f64 DPR:$src))), (v2f32 DPR:$src)>;
-def : Pat<(v2f32 (bitconvert (v1i64 DPR:$src))), (v2f32 DPR:$src)>;
+let Predicates = [IsLE] in {
+ def : Pat<(f64 (bitconvert (v2i32 DPR:$src))), (f64 DPR:$src)>;
+ def : Pat<(f64 (bitconvert (v4i16 DPR:$src))), (f64 DPR:$src)>;
+ def : Pat<(f64 (bitconvert (v8i8 DPR:$src))), (f64 DPR:$src)>;
+ def : Pat<(f64 (bitconvert (v2f32 DPR:$src))), (f64 DPR:$src)>;
+ def : Pat<(v2f32 (bitconvert (f64 DPR:$src))), (v2f32 DPR:$src)>;
+ def : Pat<(v2f32 (bitconvert (v1i64 DPR:$src))), (v2f32 DPR:$src)>;
+}
def : Pat<(v2f32 (bitconvert (v2i32 DPR:$src))), (v2f32 DPR:$src)>;
-def : Pat<(v2f32 (bitconvert (v4i16 DPR:$src))), (v2f32 DPR:$src)>;
-def : Pat<(v2f32 (bitconvert (v8i8 DPR:$src))), (v2f32 DPR:$src)>;
+let Predicates = [IsLE] in {
+ def : Pat<(v2f32 (bitconvert (v4i16 DPR:$src))), (v2f32 DPR:$src)>;
+ def : Pat<(v2f32 (bitconvert (v8i8 DPR:$src))), (v2f32 DPR:$src)>;
+}
-def : Pat<(v2i64 (bitconvert (v4i32 QPR:$src))), (v2i64 QPR:$src)>;
-def : Pat<(v2i64 (bitconvert (v8i16 QPR:$src))), (v2i64 QPR:$src)>;
-def : Pat<(v2i64 (bitconvert (v16i8 QPR:$src))), (v2i64 QPR:$src)>;
+let Predicates = [IsLE] in {
+ def : Pat<(v2i64 (bitconvert (v4i32 QPR:$src))), (v2i64 QPR:$src)>;
+ def : Pat<(v2i64 (bitconvert (v8i16 QPR:$src))), (v2i64 QPR:$src)>;
+ def : Pat<(v2i64 (bitconvert (v16i8 QPR:$src))), (v2i64 QPR:$src)>;
+}
def : Pat<(v2i64 (bitconvert (v2f64 QPR:$src))), (v2i64 QPR:$src)>;
-def : Pat<(v2i64 (bitconvert (v4f32 QPR:$src))), (v2i64 QPR:$src)>;
-def : Pat<(v4i32 (bitconvert (v2i64 QPR:$src))), (v4i32 QPR:$src)>;
-def : Pat<(v4i32 (bitconvert (v8i16 QPR:$src))), (v4i32 QPR:$src)>;
-def : Pat<(v4i32 (bitconvert (v16i8 QPR:$src))), (v4i32 QPR:$src)>;
-def : Pat<(v4i32 (bitconvert (v2f64 QPR:$src))), (v4i32 QPR:$src)>;
+let Predicates = [IsLE] in {
+ def : Pat<(v2i64 (bitconvert (v4f32 QPR:$src))), (v2i64 QPR:$src)>;
+ def : Pat<(v4i32 (bitconvert (v2i64 QPR:$src))), (v4i32 QPR:$src)>;
+ def : Pat<(v4i32 (bitconvert (v8i16 QPR:$src))), (v4i32 QPR:$src)>;
+ def : Pat<(v4i32 (bitconvert (v16i8 QPR:$src))), (v4i32 QPR:$src)>;
+ def : Pat<(v4i32 (bitconvert (v2f64 QPR:$src))), (v4i32 QPR:$src)>;
+}
def : Pat<(v4i32 (bitconvert (v4f32 QPR:$src))), (v4i32 QPR:$src)>;
-def : Pat<(v8i16 (bitconvert (v2i64 QPR:$src))), (v8i16 QPR:$src)>;
-def : Pat<(v8i16 (bitconvert (v4i32 QPR:$src))), (v8i16 QPR:$src)>;
-def : Pat<(v8i16 (bitconvert (v16i8 QPR:$src))), (v8i16 QPR:$src)>;
-def : Pat<(v8i16 (bitconvert (v2f64 QPR:$src))), (v8i16 QPR:$src)>;
-def : Pat<(v8i16 (bitconvert (v4f32 QPR:$src))), (v8i16 QPR:$src)>;
-def : Pat<(v16i8 (bitconvert (v2i64 QPR:$src))), (v16i8 QPR:$src)>;
-def : Pat<(v16i8 (bitconvert (v4i32 QPR:$src))), (v16i8 QPR:$src)>;
-def : Pat<(v16i8 (bitconvert (v8i16 QPR:$src))), (v16i8 QPR:$src)>;
-def : Pat<(v16i8 (bitconvert (v2f64 QPR:$src))), (v16i8 QPR:$src)>;
-def : Pat<(v16i8 (bitconvert (v4f32 QPR:$src))), (v16i8 QPR:$src)>;
-def : Pat<(v4f32 (bitconvert (v2i64 QPR:$src))), (v4f32 QPR:$src)>;
+let Predicates = [IsLE] in {
+ def : Pat<(v8i16 (bitconvert (v2i64 QPR:$src))), (v8i16 QPR:$src)>;
+ def : Pat<(v8i16 (bitconvert (v4i32 QPR:$src))), (v8i16 QPR:$src)>;
+ def : Pat<(v8i16 (bitconvert (v16i8 QPR:$src))), (v8i16 QPR:$src)>;
+ def : Pat<(v8i16 (bitconvert (v2f64 QPR:$src))), (v8i16 QPR:$src)>;
+ def : Pat<(v8i16 (bitconvert (v4f32 QPR:$src))), (v8i16 QPR:$src)>;
+ def : Pat<(v16i8 (bitconvert (v2i64 QPR:$src))), (v16i8 QPR:$src)>;
+ def : Pat<(v16i8 (bitconvert (v4i32 QPR:$src))), (v16i8 QPR:$src)>;
+ def : Pat<(v16i8 (bitconvert (v8i16 QPR:$src))), (v16i8 QPR:$src)>;
+ def : Pat<(v16i8 (bitconvert (v2f64 QPR:$src))), (v16i8 QPR:$src)>;
+ def : Pat<(v16i8 (bitconvert (v4f32 QPR:$src))), (v16i8 QPR:$src)>;
+ def : Pat<(v4f32 (bitconvert (v2i64 QPR:$src))), (v4f32 QPR:$src)>;
+}
def : Pat<(v4f32 (bitconvert (v4i32 QPR:$src))), (v4f32 QPR:$src)>;
-def : Pat<(v4f32 (bitconvert (v8i16 QPR:$src))), (v4f32 QPR:$src)>;
-def : Pat<(v4f32 (bitconvert (v16i8 QPR:$src))), (v4f32 QPR:$src)>;
-def : Pat<(v4f32 (bitconvert (v2f64 QPR:$src))), (v4f32 QPR:$src)>;
+let Predicates = [IsLE] in {
+ def : Pat<(v4f32 (bitconvert (v8i16 QPR:$src))), (v4f32 QPR:$src)>;
+ def : Pat<(v4f32 (bitconvert (v16i8 QPR:$src))), (v4f32 QPR:$src)>;
+ def : Pat<(v4f32 (bitconvert (v2f64 QPR:$src))), (v4f32 QPR:$src)>;
+}
def : Pat<(v2f64 (bitconvert (v2i64 QPR:$src))), (v2f64 QPR:$src)>;
-def : Pat<(v2f64 (bitconvert (v4i32 QPR:$src))), (v2f64 QPR:$src)>;
-def : Pat<(v2f64 (bitconvert (v8i16 QPR:$src))), (v2f64 QPR:$src)>;
-def : Pat<(v2f64 (bitconvert (v16i8 QPR:$src))), (v2f64 QPR:$src)>;
-def : Pat<(v2f64 (bitconvert (v4f32 QPR:$src))), (v2f64 QPR:$src)>;
+let Predicates = [IsLE] in {
+ def : Pat<(v2f64 (bitconvert (v4i32 QPR:$src))), (v2f64 QPR:$src)>;
+ def : Pat<(v2f64 (bitconvert (v8i16 QPR:$src))), (v2f64 QPR:$src)>;
+ def : Pat<(v2f64 (bitconvert (v16i8 QPR:$src))), (v2f64 QPR:$src)>;
+ def : Pat<(v2f64 (bitconvert (v4f32 QPR:$src))), (v2f64 QPR:$src)>;
+}
+
+let Predicates = [IsBE] in {
+ // 64 bit conversions
+ def : Pat<(v1i64 (bitconvert (v2i32 DPR:$src))), (VREV64d32 DPR:$src)>;
+ def : Pat<(v1i64 (bitconvert (v4i16 DPR:$src))), (VREV64d16 DPR:$src)>;
+ def : Pat<(v1i64 (bitconvert (v8i8 DPR:$src))), (VREV64d8 DPR:$src)>;
+ def : Pat<(v1i64 (bitconvert (v2f32 DPR:$src))), (VREV64d32 DPR:$src)>;
+ def : Pat<(v2i32 (bitconvert (v1i64 DPR:$src))), (VREV64d32 DPR:$src)>;
+ def : Pat<(v2i32 (bitconvert (v4i16 DPR:$src))), (VREV32d16 DPR:$src)>;
+ def : Pat<(v2i32 (bitconvert (v8i8 DPR:$src))), (VREV32d8 DPR:$src)>;
+ def : Pat<(v2i32 (bitconvert (f64 DPR:$src))), (VREV64d32 DPR:$src)>;
+ def : Pat<(v4i16 (bitconvert (v1i64 DPR:$src))), (VREV64d16 DPR:$src)>;
+ def : Pat<(v4i16 (bitconvert (v2i32 DPR:$src))), (VREV32d16 DPR:$src)>;
+ def : Pat<(v4i16 (bitconvert (v8i8 DPR:$src))), (VREV16d8 DPR:$src)>;
+ def : Pat<(v4i16 (bitconvert (f64 DPR:$src))), (VREV64d16 DPR:$src)>;
+ def : Pat<(v4i16 (bitconvert (v2f32 DPR:$src))), (VREV32d16 DPR:$src)>;
+ def : Pat<(v8i8 (bitconvert (v1i64 DPR:$src))), (VREV64d8 DPR:$src)>;
+ def : Pat<(v8i8 (bitconvert (v2i32 DPR:$src))), (VREV32d8 DPR:$src)>;
+ def : Pat<(v8i8 (bitconvert (v4i16 DPR:$src))), (VREV16d8 DPR:$src)>;
+ def : Pat<(v8i8 (bitconvert (f64 DPR:$src))), (VREV64d8 DPR:$src)>;
+ def : Pat<(v8i8 (bitconvert (v2f32 DPR:$src))), (VREV32d8 DPR:$src)>;
+ def : Pat<(f64 (bitconvert (v2i32 DPR:$src))), (VREV64d32 DPR:$src)>;
+ def : Pat<(f64 (bitconvert (v4i16 DPR:$src))), (VREV64d16 DPR:$src)>;
+ def : Pat<(f64 (bitconvert (v8i8 DPR:$src))), (VREV64d8 DPR:$src)>;
+ def : Pat<(f64 (bitconvert (v2f32 DPR:$src))), (VREV64d32 DPR:$src)>;
+ def : Pat<(v2f32 (bitconvert (f64 DPR:$src))), (VREV64d32 DPR:$src)>;
+ def : Pat<(v2f32 (bitconvert (v1i64 DPR:$src))), (VREV64d32 DPR:$src)>;
+ def : Pat<(v2f32 (bitconvert (v4i16 DPR:$src))), (VREV32d16 DPR:$src)>;
+ def : Pat<(v2f32 (bitconvert (v8i8 DPR:$src))), (VREV32d8 DPR:$src)>;
+
+ // 128 bit conversions
+ def : Pat<(v2i64 (bitconvert (v4i32 QPR:$src))), (VREV64q32 QPR:$src)>;
+ def : Pat<(v2i64 (bitconvert (v8i16 QPR:$src))), (VREV64q16 QPR:$src)>;
+ def : Pat<(v2i64 (bitconvert (v16i8 QPR:$src))), (VREV64q8 QPR:$src)>;
+ def : Pat<(v2i64 (bitconvert (v4f32 QPR:$src))), (VREV64q32 QPR:$src)>;
+ def : Pat<(v4i32 (bitconvert (v2i64 QPR:$src))), (VREV64q32 QPR:$src)>;
+ def : Pat<(v4i32 (bitconvert (v8i16 QPR:$src))), (VREV32q16 QPR:$src)>;
+ def : Pat<(v4i32 (bitconvert (v16i8 QPR:$src))), (VREV32q8 QPR:$src)>;
+ def : Pat<(v4i32 (bitconvert (v2f64 QPR:$src))), (VREV64q32 QPR:$src)>;
+ def : Pat<(v8i16 (bitconvert (v2i64 QPR:$src))), (VREV64q16 QPR:$src)>;
+ def : Pat<(v8i16 (bitconvert (v4i32 QPR:$src))), (VREV32q16 QPR:$src)>;
+ def : Pat<(v8i16 (bitconvert (v16i8 QPR:$src))), (VREV16q8 QPR:$src)>;
+ def : Pat<(v8i16 (bitconvert (v2f64 QPR:$src))), (VREV64q16 QPR:$src)>;
+ def : Pat<(v8i16 (bitconvert (v4f32 QPR:$src))), (VREV32q16 QPR:$src)>;
+ def : Pat<(v16i8 (bitconvert (v2i64 QPR:$src))), (VREV64q8 QPR:$src)>;
+ def : Pat<(v16i8 (bitconvert (v4i32 QPR:$src))), (VREV32q8 QPR:$src)>;
+ def : Pat<(v16i8 (bitconvert (v8i16 QPR:$src))), (VREV16q8 QPR:$src)>;
+ def : Pat<(v16i8 (bitconvert (v2f64 QPR:$src))), (VREV64q8 QPR:$src)>;
+ def : Pat<(v16i8 (bitconvert (v4f32 QPR:$src))), (VREV32q8 QPR:$src)>;
+ def : Pat<(v4f32 (bitconvert (v2i64 QPR:$src))), (VREV64q32 QPR:$src)>;
+ def : Pat<(v4f32 (bitconvert (v8i16 QPR:$src))), (VREV32q16 QPR:$src)>;
+ def : Pat<(v4f32 (bitconvert (v16i8 QPR:$src))), (VREV32q8 QPR:$src)>;
+ def : Pat<(v4f32 (bitconvert (v2f64 QPR:$src))), (VREV64q32 QPR:$src)>;
+ def : Pat<(v2f64 (bitconvert (v4i32 QPR:$src))), (VREV64q32 QPR:$src)>;
+ def : Pat<(v2f64 (bitconvert (v8i16 QPR:$src))), (VREV64q16 QPR:$src)>;
+ def : Pat<(v2f64 (bitconvert (v16i8 QPR:$src))), (VREV64q8 QPR:$src)>;
+ def : Pat<(v2f64 (bitconvert (v4f32 QPR:$src))), (VREV64q32 QPR:$src)>;
+}
+
+// Fold extracting an element out of a v2i32 into a vfp register.
+def : Pat<(f32 (bitconvert (i32 (extractelt (v2i32 DPR:$src), imm:$lane)))),
+ (f32 (EXTRACT_SUBREG DPR:$src, (SSubReg_f32_reg imm:$lane)))>;
// Vector lengthening move with load, matching extending loads.
// extload, zextload and sextload for a standard lengthening load. Example:
-// Lengthen_Single<"8", "i16", "i8"> =
-// Pat<(v8i16 (extloadvi8 addrmode6oneL32:$addr))
-// (VMOVLuv8i16 (VLD1LNd32 addrmode6oneL32:$addr,
-// (f64 (IMPLICIT_DEF)), (i32 0)))>;
+// Lengthen_Single<"8", "i16", "8"> =
+// Pat<(v8i16 (extloadvi8 addrmode6:$addr))
+// (VMOVLuv8i16 (VLD1d8 addrmode6:$addr,
+// (f64 (IMPLICIT_DEF)), (i32 0)))>;
multiclass Lengthen_Single<string DestLanes, string DestTy, string SrcTy> {
+ let AddedComplexity = 10 in {
def _Any : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
- (!cast<PatFrag>("extloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (!cast<PatFrag>("extloadvi" # SrcTy) addrmode6:$addr)),
(!cast<Instruction>("VMOVLuv" # DestLanes # DestTy)
- (VLD1LNd32 addrmode6oneL32:$addr,
- (f64 (IMPLICIT_DEF)), (i32 0)))>;
+ (!cast<Instruction>("VLD1d" # SrcTy) addrmode6:$addr))>;
+
def _Z : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
- (!cast<PatFrag>("zextloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (!cast<PatFrag>("zextloadvi" # SrcTy) addrmode6:$addr)),
(!cast<Instruction>("VMOVLuv" # DestLanes # DestTy)
- (VLD1LNd32 addrmode6oneL32:$addr,
- (f64 (IMPLICIT_DEF)), (i32 0)))>;
+ (!cast<Instruction>("VLD1d" # SrcTy) addrmode6:$addr))>;
+
def _S : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
- (!cast<PatFrag>("sextloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (!cast<PatFrag>("sextloadvi" # SrcTy) addrmode6:$addr)),
(!cast<Instruction>("VMOVLsv" # DestLanes # DestTy)
- (VLD1LNd32 addrmode6oneL32:$addr,
- (f64 (IMPLICIT_DEF)), (i32 0)))>;
+ (!cast<Instruction>("VLD1d" # SrcTy) addrmode6:$addr))>;
+ }
}
// extload, zextload and sextload for a lengthening load which only uses
// half the lanes available. Example:
// Lengthen_HalfSingle<"4", "i16", "8", "i16", "i8"> =
// Pat<(v4i16 (extloadvi8 addrmode6oneL32:$addr)),
-// (EXTRACT_SUBREG (VMOVLuv8i16 (VLD1LNd32 addrmode6oneL32:$addr,
+// (EXTRACT_SUBREG (VMOVLuv8i16 (VLD1LNd32 addrmode6oneL32:$addr,
// (f64 (IMPLICIT_DEF)), (i32 0))),
// dsub_0)>;
multiclass Lengthen_HalfSingle<string DestLanes, string DestTy, string SrcTy,
dsub_0)>;
}
+// The following class definition is basically a copy of the
+// Lengthen_HalfSingle definition above, however with an additional parameter
+// "RevLanes" to select the correct VREV32dXX instruction. This is to convert
+// data loaded by VLD1LN into proper vector format in big endian mode.
+multiclass Lengthen_HalfSingle_Big_Endian<string DestLanes, string DestTy, string SrcTy,
+ string InsnLanes, string InsnTy, string RevLanes> {
+ def _Any : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+ (!cast<PatFrag>("extloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # InsnLanes # InsnTy)
+ (!cast<Instruction>("VREV32d" # RevLanes)
+ (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0)))),
+ dsub_0)>;
+ def _Z : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+ (!cast<PatFrag>("zextloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # InsnLanes # InsnTy)
+ (!cast<Instruction>("VREV32d" # RevLanes)
+ (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0)))),
+ dsub_0)>;
+ def _S : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+ (!cast<PatFrag>("sextloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # InsnLanes # InsnTy)
+ (!cast<Instruction>("VREV32d" # RevLanes)
+ (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0)))),
+ dsub_0)>;
+}
+
// extload, zextload and sextload for a lengthening load followed by another
// lengthening load, to quadruple the initial length.
//
-// Lengthen_Double<"4", "i32", "i8", "8", "i16", "4", "i32", qsub_0> =
-// Pat<(v4i32 (extloadvi8 addrmode5:$addr))
+// Lengthen_Double<"4", "i32", "i8", "8", "i16", "4", "i32"> =
+// Pat<(v4i32 (extloadvi8 addrmode6oneL32:$addr))
// (EXTRACT_SUBREG (VMOVLuv4i32
// (EXTRACT_SUBREG (VMOVLuv8i16 (VLD1LNd32 addrmode6oneL32:$addr,
// (f64 (IMPLICIT_DEF)),
(!cast<PatFrag>("extloadv" # SrcTy) addrmode6oneL32:$addr)),
(!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
(EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
- (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
+ (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
dsub_0))>;
def _Z : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
(!cast<PatFrag>("zextloadv" # SrcTy) addrmode6oneL32:$addr)),
(!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
(EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
- (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
+ (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
dsub_0))>;
def _S : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
(!cast<PatFrag>("sextloadv" # SrcTy) addrmode6oneL32:$addr)),
(!cast<Instruction>("VMOVLsv" # Insn2Lanes # Insn2Ty)
(EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # Insn1Lanes # Insn1Ty)
- (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
+ (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
+ dsub_0))>;
+}
+
+// The following class definition is basically a copy of the
+// Lengthen_Double definition above, however with an additional parameter
+// "RevLanes" to select the correct VREV32dXX instruction. This is to convert
+// data loaded by VLD1LN into proper vector format in big endian mode.
+multiclass Lengthen_Double_Big_Endian<string DestLanes, string DestTy, string SrcTy,
+ string Insn1Lanes, string Insn1Ty, string Insn2Lanes,
+ string Insn2Ty, string RevLanes> {
+ def _Any : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+ (!cast<PatFrag>("extloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
+ (!cast<Instruction>("VREV32d" # RevLanes)
+ (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0)))),
+ dsub_0))>;
+ def _Z : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+ (!cast<PatFrag>("zextloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
+ (!cast<Instruction>("VREV32d" # RevLanes)
+ (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0)))),
+ dsub_0))>;
+ def _S : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+ (!cast<PatFrag>("sextloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (!cast<Instruction>("VMOVLsv" # Insn2Lanes # Insn2Ty)
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # Insn1Lanes # Insn1Ty)
+ (!cast<Instruction>("VREV32d" # RevLanes)
+ (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0)))),
dsub_0))>;
}
// requiring half the available lanes (a 64-bit outcome instead of a 128-bit).
//
// Lengthen_HalfDouble<"2", "i32", "i8", "8", "i16", "4", "i32"> =
-// Pat<(v4i32 (extloadvi8 addrmode5:$addr))
+// Pat<(v2i32 (extloadvi8 addrmode6:$addr))
// (EXTRACT_SUBREG (VMOVLuv4i32
-// (EXTRACT_SUBREG (VMOVLuv8i16 (VLD1LNd32 addrmode6oneL32:$addr,
+// (EXTRACT_SUBREG (VMOVLuv8i16 (VLD1LNd16 addrmode6:$addr,
// (f64 (IMPLICIT_DEF)), (i32 0))),
// dsub_0)),
// dsub_0)>;
string Insn1Lanes, string Insn1Ty, string Insn2Lanes,
string Insn2Ty> {
def _Any : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
- (!cast<PatFrag>("extloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (!cast<PatFrag>("extloadv" # SrcTy) addrmode6:$addr)),
(EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
(EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
- (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
+ (VLD1LNd16 addrmode6:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
dsub_0)),
dsub_0)>;
def _Z : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
- (!cast<PatFrag>("zextloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (!cast<PatFrag>("zextloadv" # SrcTy) addrmode6:$addr)),
(EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
(EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
- (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
+ (VLD1LNd16 addrmode6:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
dsub_0)),
dsub_0)>;
def _S : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
- (!cast<PatFrag>("sextloadv" # SrcTy) addrmode6oneL32:$addr)),
+ (!cast<PatFrag>("sextloadv" # SrcTy) addrmode6:$addr)),
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # Insn2Lanes # Insn2Ty)
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # Insn1Lanes # Insn1Ty)
+ (VLD1LNd16 addrmode6:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
+ dsub_0)),
+ dsub_0)>;
+}
+
+// The following class definition is basically a copy of the
+// Lengthen_HalfDouble definition above, however with an additional VREV16d8
+// instruction to convert data loaded by VLD1LN into proper vector format
+// in big endian mode.
+multiclass Lengthen_HalfDouble_Big_Endian<string DestLanes, string DestTy, string SrcTy,
+ string Insn1Lanes, string Insn1Ty, string Insn2Lanes,
+ string Insn2Ty> {
+ def _Any : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+ (!cast<PatFrag>("extloadv" # SrcTy) addrmode6:$addr)),
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
+ (!cast<Instruction>("VREV16d8")
+ (VLD1LNd16 addrmode6:$addr, (f64 (IMPLICIT_DEF)), (i32 0)))),
+ dsub_0)),
+ dsub_0)>;
+ def _Z : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+ (!cast<PatFrag>("zextloadv" # SrcTy) addrmode6:$addr)),
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn2Lanes # Insn2Ty)
+ (EXTRACT_SUBREG (!cast<Instruction>("VMOVLuv" # Insn1Lanes # Insn1Ty)
+ (!cast<Instruction>("VREV16d8")
+ (VLD1LNd16 addrmode6:$addr, (f64 (IMPLICIT_DEF)), (i32 0)))),
+ dsub_0)),
+ dsub_0)>;
+ def _S : Pat<(!cast<ValueType>("v" # DestLanes # DestTy)
+ (!cast<PatFrag>("sextloadv" # SrcTy) addrmode6:$addr)),
(EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # Insn2Lanes # Insn2Ty)
(EXTRACT_SUBREG (!cast<Instruction>("VMOVLsv" # Insn1Lanes # Insn1Ty)
- (VLD1LNd32 addrmode6oneL32:$addr, (f64 (IMPLICIT_DEF)), (i32 0))),
+ (!cast<Instruction>("VREV16d8")
+ (VLD1LNd16 addrmode6:$addr, (f64 (IMPLICIT_DEF)), (i32 0)))),
dsub_0)),
dsub_0)>;
}
-defm : Lengthen_Single<"8", "i16", "i8">; // v8i8 -> v8i16
-defm : Lengthen_Single<"4", "i32", "i16">; // v4i16 -> v4i32
-defm : Lengthen_Single<"2", "i64", "i32">; // v2i32 -> v2i64
+defm : Lengthen_Single<"8", "i16", "8">; // v8i8 -> v8i16
+defm : Lengthen_Single<"4", "i32", "16">; // v4i16 -> v4i32
+defm : Lengthen_Single<"2", "i64", "32">; // v2i32 -> v2i64
+
+let Predicates = [IsLE] in {
+ defm : Lengthen_HalfSingle<"4", "i16", "i8", "8", "i16">; // v4i8 -> v4i16
+ defm : Lengthen_HalfSingle<"2", "i32", "i16", "4", "i32">; // v2i16 -> v2i32
+
+ // Double lengthening - v4i8 -> v4i16 -> v4i32
+ defm : Lengthen_Double<"4", "i32", "i8", "8", "i16", "4", "i32">;
+ // v2i8 -> v2i16 -> v2i32
+ defm : Lengthen_HalfDouble<"2", "i32", "i8", "8", "i16", "4", "i32">;
+ // v2i16 -> v2i32 -> v2i64
+ defm : Lengthen_Double<"2", "i64", "i16", "4", "i32", "2", "i64">;
+}
-defm : Lengthen_HalfSingle<"4", "i16", "i8", "8", "i16">; // v4i8 -> v4i16
-defm : Lengthen_HalfSingle<"2", "i16", "i8", "8", "i16">; // v2i8 -> v2i16
-defm : Lengthen_HalfSingle<"2", "i32", "i16", "4", "i32">; // v2i16 -> v2i32
+let Predicates = [IsBE] in {
+ defm : Lengthen_HalfSingle_Big_Endian<"4", "i16", "i8", "8", "i16", "8">; // v4i8 -> v4i16
+ defm : Lengthen_HalfSingle_Big_Endian<"2", "i32", "i16", "4", "i32", "16">; // v2i16 -> v2i32
-// Double lengthening - v4i8 -> v4i16 -> v4i32
-defm : Lengthen_Double<"4", "i32", "i8", "8", "i16", "4", "i32">;
-// v2i8 -> v2i16 -> v2i32
-defm : Lengthen_HalfDouble<"2", "i32", "i8", "8", "i16", "4", "i32">;
-// v2i16 -> v2i32 -> v2i64
-defm : Lengthen_Double<"2", "i64", "i16", "4", "i32", "2", "i64">;
+ // Double lengthening - v4i8 -> v4i16 -> v4i32
+ defm : Lengthen_Double_Big_Endian<"4", "i32", "i8", "8", "i16", "4", "i32", "8">;
+ // v2i8 -> v2i16 -> v2i32
+ defm : Lengthen_HalfDouble_Big_Endian<"2", "i32", "i8", "8", "i16", "4", "i32">;
+ // v2i16 -> v2i32 -> v2i64
+ defm : Lengthen_Double_Big_Endian<"2", "i64", "i16", "4", "i32", "2", "i64", "16">;
+}
// Triple lengthening - v2i8 -> v2i16 -> v2i32 -> v2i64
-def : Pat<(v2i64 (extloadvi8 addrmode6oneL32:$addr)),
- (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
- (VLD1LNd32 addrmode6oneL32:$addr,
- (f64 (IMPLICIT_DEF)), (i32 0))), dsub_0)), dsub_0))>;
-def : Pat<(v2i64 (zextloadvi8 addrmode6oneL32:$addr)),
- (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
- (VLD1LNd32 addrmode6oneL32:$addr,
- (f64 (IMPLICIT_DEF)), (i32 0))), dsub_0)), dsub_0))>;
-def : Pat<(v2i64 (sextloadvi8 addrmode6oneL32:$addr)),
- (VMOVLsv2i64 (EXTRACT_SUBREG (VMOVLsv4i32 (EXTRACT_SUBREG (VMOVLsv8i16
- (VLD1LNd32 addrmode6oneL32:$addr,
- (f64 (IMPLICIT_DEF)), (i32 0))), dsub_0)), dsub_0))>;
+let Predicates = [IsLE] in {
+ def : Pat<(v2i64 (extloadvi8 addrmode6:$addr)),
+ (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
+ (VLD1LNd16 addrmode6:$addr,
+ (f64 (IMPLICIT_DEF)), (i32 0))), dsub_0)), dsub_0))>;
+ def : Pat<(v2i64 (zextloadvi8 addrmode6:$addr)),
+ (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
+ (VLD1LNd16 addrmode6:$addr,
+ (f64 (IMPLICIT_DEF)), (i32 0))), dsub_0)), dsub_0))>;
+ def : Pat<(v2i64 (sextloadvi8 addrmode6:$addr)),
+ (VMOVLsv2i64 (EXTRACT_SUBREG (VMOVLsv4i32 (EXTRACT_SUBREG (VMOVLsv8i16
+ (VLD1LNd16 addrmode6:$addr,
+ (f64 (IMPLICIT_DEF)), (i32 0))), dsub_0)), dsub_0))>;
+}
+// The following patterns are basically a copy of the patterns above,
+// however with an additional VREV16d instruction to convert data
+// loaded by VLD1LN into proper vector format in big endian mode.
+let Predicates = [IsBE] in {
+ def : Pat<(v2i64 (extloadvi8 addrmode6:$addr)),
+ (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
+ (!cast<Instruction>("VREV16d8")
+ (VLD1LNd16 addrmode6:$addr,
+ (f64 (IMPLICIT_DEF)), (i32 0)))), dsub_0)), dsub_0))>;
+ def : Pat<(v2i64 (zextloadvi8 addrmode6:$addr)),
+ (VMOVLuv2i64 (EXTRACT_SUBREG (VMOVLuv4i32 (EXTRACT_SUBREG (VMOVLuv8i16
+ (!cast<Instruction>("VREV16d8")
+ (VLD1LNd16 addrmode6:$addr,
+ (f64 (IMPLICIT_DEF)), (i32 0)))), dsub_0)), dsub_0))>;
+ def : Pat<(v2i64 (sextloadvi8 addrmode6:$addr)),
+ (VMOVLsv2i64 (EXTRACT_SUBREG (VMOVLsv4i32 (EXTRACT_SUBREG (VMOVLsv8i16
+ (!cast<Instruction>("VREV16d8")
+ (VLD1LNd16 addrmode6:$addr,
+ (f64 (IMPLICIT_DEF)), (i32 0)))), dsub_0)), dsub_0))>;
+}
//===----------------------------------------------------------------------===//
// Assembler aliases
(VORRd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
defm : NEONDTAnyInstAlias<"vorr${p}", "$Vdn, $Vm",
(VORRq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+// ... immediates
+def : NEONInstAlias<"vand${p}.i16 $Vd, $imm",
+ (VBICiv4i16 DPR:$Vd, nImmSplatNotI16:$imm, pred:$p)>;
+def : NEONInstAlias<"vand${p}.i32 $Vd, $imm",
+ (VBICiv2i32 DPR:$Vd, nImmSplatNotI32:$imm, pred:$p)>;
+def : NEONInstAlias<"vand${p}.i16 $Vd, $imm",
+ (VBICiv8i16 QPR:$Vd, nImmSplatNotI16:$imm, pred:$p)>;
+def : NEONInstAlias<"vand${p}.i32 $Vd, $imm",
+ (VBICiv4i32 QPR:$Vd, nImmSplatNotI32:$imm, pred:$p)>;
+
// VLD1 single-lane pseudo-instructions. These need special handling for
// the lane index that an InstAlias can't handle, so we use these instead.
def VLD1LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vld1${p}", ".8", "$list, $addr",
- (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD1LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vld1${p}", ".16", "$list, $addr",
- (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
+ pred:$p)>;
def VLD1LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vld1${p}", ".32", "$list, $addr",
- (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VLD1LNdWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld1${p}", ".8", "$list, $addr!",
- (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD1LNdWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld1${p}", ".16", "$list, $addr!",
- (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
+ pred:$p)>;
def VLD1LNdWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld1${p}", ".32", "$list, $addr!",
- (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VLD1LNdWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld1${p}", ".8", "$list, $addr, $Rm",
- (ins VecListOneDByteIndexed:$list, addrmode6:$addr,
+ (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
rGPR:$Rm, pred:$p)>;
def VLD1LNdWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld1${p}", ".16", "$list, $addr, $Rm",
- (ins VecListOneDHWordIndexed:$list, addrmode6:$addr,
+ (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
rGPR:$Rm, pred:$p)>;
def VLD1LNdWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld1${p}", ".32", "$list, $addr, $Rm",
- (ins VecListOneDWordIndexed:$list, addrmode6:$addr,
+ (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
rGPR:$Rm, pred:$p)>;
// VST1 single-lane pseudo-instructions. These need special handling for
// the lane index that an InstAlias can't handle, so we use these instead.
def VST1LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vst1${p}", ".8", "$list, $addr",
- (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST1LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vst1${p}", ".16", "$list, $addr",
- (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
+ pred:$p)>;
def VST1LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vst1${p}", ".32", "$list, $addr",
- (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VST1LNdWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst1${p}", ".8", "$list, $addr!",
- (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST1LNdWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst1${p}", ".16", "$list, $addr!",
- (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
+ pred:$p)>;
def VST1LNdWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst1${p}", ".32", "$list, $addr!",
- (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VST1LNdWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst1${p}", ".8", "$list, $addr, $Rm",
- (ins VecListOneDByteIndexed:$list, addrmode6:$addr,
+ (ins VecListOneDByteIndexed:$list, addrmode6alignNone:$addr,
rGPR:$Rm, pred:$p)>;
def VST1LNdWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst1${p}", ".16", "$list, $addr, $Rm",
- (ins VecListOneDHWordIndexed:$list, addrmode6:$addr,
+ (ins VecListOneDHWordIndexed:$list, addrmode6align16:$addr,
rGPR:$Rm, pred:$p)>;
def VST1LNdWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst1${p}", ".32", "$list, $addr, $Rm",
- (ins VecListOneDWordIndexed:$list, addrmode6:$addr,
+ (ins VecListOneDWordIndexed:$list, addrmode6align32:$addr,
rGPR:$Rm, pred:$p)>;
// VLD2 single-lane pseudo-instructions. These need special handling for
// the lane index that an InstAlias can't handle, so we use these instead.
def VLD2LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vld2${p}", ".8", "$list, $addr",
- (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
+ pred:$p)>;
def VLD2LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr",
- (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VLD2LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr",
- (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr, pred:$p)>;
def VLD2LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr",
- (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VLD2LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr",
- (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VLD2LNdWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld2${p}", ".8", "$list, $addr!",
- (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
+ pred:$p)>;
def VLD2LNdWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr!",
- (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VLD2LNdWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr!",
- (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VLD2LNqWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr!",
- (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VLD2LNqWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr!",
- (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VLD2LNdWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld2${p}", ".8", "$list, $addr, $Rm",
- (ins VecListTwoDByteIndexed:$list, addrmode6:$addr,
+ (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
rGPR:$Rm, pred:$p)>;
def VLD2LNdWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr, $Rm",
- (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr,
+ (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
rGPR:$Rm, pred:$p)>;
def VLD2LNdWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr, $Rm",
- (ins VecListTwoDWordIndexed:$list, addrmode6:$addr,
+ (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VLD2LNqWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld2${p}", ".16", "$list, $addr, $Rm",
- (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr,
+ (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
rGPR:$Rm, pred:$p)>;
def VLD2LNqWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld2${p}", ".32", "$list, $addr, $Rm",
- (ins VecListTwoQWordIndexed:$list, addrmode6:$addr,
+ (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
// VST2 single-lane pseudo-instructions. These need special handling for
// the lane index that an InstAlias can't handle, so we use these instead.
def VST2LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vst2${p}", ".8", "$list, $addr",
- (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
+ pred:$p)>;
def VST2LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vst2${p}", ".16", "$list, $addr",
- (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VST2LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr",
- (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VST2LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vst2${p}", ".16", "$list, $addr",
- (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VST2LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr",
- (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VST2LNdWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst2${p}", ".8", "$list, $addr!",
- (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
+ pred:$p)>;
def VST2LNdWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst2${p}", ".16", "$list, $addr!",
- (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VST2LNdWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr!",
- (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VST2LNqWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst2${p}", ".16", "$list, $addr!",
- (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VST2LNqWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr!",
- (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VST2LNdWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst2${p}", ".8", "$list, $addr, $Rm",
- (ins VecListTwoDByteIndexed:$list, addrmode6:$addr,
+ (ins VecListTwoDByteIndexed:$list, addrmode6align16:$addr,
rGPR:$Rm, pred:$p)>;
def VST2LNdWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst2${p}", ".16","$list, $addr, $Rm",
- (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr,
+ (ins VecListTwoDHWordIndexed:$list, addrmode6align32:$addr,
rGPR:$Rm, pred:$p)>;
def VST2LNdWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr, $Rm",
- (ins VecListTwoDWordIndexed:$list, addrmode6:$addr,
+ (ins VecListTwoDWordIndexed:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VST2LNqWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst2${p}", ".16","$list, $addr, $Rm",
- (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr,
+ (ins VecListTwoQHWordIndexed:$list, addrmode6align32:$addr,
rGPR:$Rm, pred:$p)>;
def VST2LNqWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst2${p}", ".32", "$list, $addr, $Rm",
- (ins VecListTwoQWordIndexed:$list, addrmode6:$addr,
+ (ins VecListTwoQWordIndexed:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
// VLD3 all-lanes pseudo-instructions. These need special handling for
// the lane index that an InstAlias can't handle, so we use these instead.
def VLD3DUPdAsm_8 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr",
- (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPdAsm_16: NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
- (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPdAsm_32: NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
- (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPqAsm_8 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr",
- (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPqAsm_16: NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
- (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPqAsm_32: NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
- (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPdWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr!",
- (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPdWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
- (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPdWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
- (ins VecListThreeDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPqWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr!",
- (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPqWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
- (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPqWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
- (ins VecListThreeQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
+ pred:$p)>;
def VLD3DUPdWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr, $Rm",
- (ins VecListThreeDAllLanes:$list, addrmode6:$addr,
+ (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3DUPdWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
- (ins VecListThreeDAllLanes:$list, addrmode6:$addr,
+ (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3DUPdWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
- (ins VecListThreeDAllLanes:$list, addrmode6:$addr,
+ (ins VecListThreeDAllLanes:$list, addrmode6dupalignNone:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3DUPqWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr, $Rm",
- (ins VecListThreeQAllLanes:$list, addrmode6:$addr,
+ (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3DUPqWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
- (ins VecListThreeQAllLanes:$list, addrmode6:$addr,
+ (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3DUPqWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
- (ins VecListThreeQAllLanes:$list, addrmode6:$addr,
+ (ins VecListThreeQAllLanes:$list, addrmode6dupalignNone:$addr,
rGPR:$Rm, pred:$p)>;
// VLD3 single-lane pseudo-instructions. These need special handling for
// the lane index that an InstAlias can't handle, so we use these instead.
def VLD3LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr",
- (ins VecListThreeDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD3LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
- (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDHWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD3LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
- (ins VecListThreeDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD3LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
- (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQHWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD3LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
- (ins VecListThreeQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD3LNdWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr!",
- (ins VecListThreeDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD3LNdWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
- (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDHWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD3LNdWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
- (ins VecListThreeDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD3LNqWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
- (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQHWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD3LNqWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
- (ins VecListThreeQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VLD3LNdWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr, $Rm",
- (ins VecListThreeDByteIndexed:$list, addrmode6:$addr,
+ (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3LNdWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
- (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr,
- rGPR:$Rm, pred:$p)>;
+ (ins VecListThreeDHWordIndexed:$list,
+ addrmode6alignNone:$addr, rGPR:$Rm, pred:$p)>;
def VLD3LNdWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
- (ins VecListThreeDWordIndexed:$list, addrmode6:$addr,
+ (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3LNqWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
- (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr,
- rGPR:$Rm, pred:$p)>;
+ (ins VecListThreeQHWordIndexed:$list,
+ addrmode6alignNone:$addr, rGPR:$Rm, pred:$p)>;
def VLD3LNqWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
- (ins VecListThreeQWordIndexed:$list, addrmode6:$addr,
+ (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
rGPR:$Rm, pred:$p)>;
// VLD3 multiple structure pseudo-instructions. These need special handling for
// the vector operands that the normal instructions don't yet model.
// FIXME: Remove these when the register classes and instructions are updated.
def VLD3dAsm_8 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3dAsm_16 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3dAsm_32 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3qAsm_8 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3qAsm_16 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3qAsm_32 : NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3dWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr!",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3dWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3dWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3qWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr!",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3qWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr!",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3qWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr!",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VLD3dWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr, $Rm",
- (ins VecListThreeD:$list, addrmode6:$addr,
+ (ins VecListThreeD:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3dWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
- (ins VecListThreeD:$list, addrmode6:$addr,
+ (ins VecListThreeD:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3dWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
- (ins VecListThreeD:$list, addrmode6:$addr,
+ (ins VecListThreeD:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3qWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".8", "$list, $addr, $Rm",
- (ins VecListThreeQ:$list, addrmode6:$addr,
+ (ins VecListThreeQ:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3qWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".16", "$list, $addr, $Rm",
- (ins VecListThreeQ:$list, addrmode6:$addr,
+ (ins VecListThreeQ:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VLD3qWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld3${p}", ".32", "$list, $addr, $Rm",
- (ins VecListThreeQ:$list, addrmode6:$addr,
+ (ins VecListThreeQ:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
// VST3 single-lane pseudo-instructions. These need special handling for
// the lane index that an InstAlias can't handle, so we use these instead.
def VST3LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr",
- (ins VecListThreeDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST3LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr",
- (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDHWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST3LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr",
- (ins VecListThreeDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST3LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr",
- (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQHWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST3LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr",
- (ins VecListThreeQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST3LNdWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr!",
- (ins VecListThreeDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST3LNdWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr!",
- (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDHWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST3LNdWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr!",
- (ins VecListThreeDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST3LNqWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr!",
- (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQHWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST3LNqWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr!",
- (ins VecListThreeQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
+ pred:$p)>;
def VST3LNdWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr, $Rm",
- (ins VecListThreeDByteIndexed:$list, addrmode6:$addr,
+ (ins VecListThreeDByteIndexed:$list, addrmode6alignNone:$addr,
rGPR:$Rm, pred:$p)>;
def VST3LNdWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr, $Rm",
- (ins VecListThreeDHWordIndexed:$list, addrmode6:$addr,
- rGPR:$Rm, pred:$p)>;
+ (ins VecListThreeDHWordIndexed:$list,
+ addrmode6alignNone:$addr, rGPR:$Rm, pred:$p)>;
def VST3LNdWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr, $Rm",
- (ins VecListThreeDWordIndexed:$list, addrmode6:$addr,
+ (ins VecListThreeDWordIndexed:$list, addrmode6alignNone:$addr,
rGPR:$Rm, pred:$p)>;
def VST3LNqWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr, $Rm",
- (ins VecListThreeQHWordIndexed:$list, addrmode6:$addr,
- rGPR:$Rm, pred:$p)>;
+ (ins VecListThreeQHWordIndexed:$list,
+ addrmode6alignNone:$addr, rGPR:$Rm, pred:$p)>;
def VST3LNqWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr, $Rm",
- (ins VecListThreeQWordIndexed:$list, addrmode6:$addr,
+ (ins VecListThreeQWordIndexed:$list, addrmode6alignNone:$addr,
rGPR:$Rm, pred:$p)>;
// the vector operands that the normal instructions don't yet model.
// FIXME: Remove these when the register classes and instructions are updated.
def VST3dAsm_8 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VST3dAsm_16 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VST3dAsm_32 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VST3qAsm_8 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VST3qAsm_16 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VST3qAsm_32 : NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VST3dWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr!",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VST3dWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr!",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VST3dWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr!",
- (ins VecListThreeD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeD:$list, addrmode6align64:$addr, pred:$p)>;
def VST3qWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr!",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VST3qWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr!",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VST3qWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr!",
- (ins VecListThreeQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListThreeQ:$list, addrmode6align64:$addr, pred:$p)>;
def VST3dWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr, $Rm",
- (ins VecListThreeD:$list, addrmode6:$addr,
+ (ins VecListThreeD:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VST3dWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr, $Rm",
- (ins VecListThreeD:$list, addrmode6:$addr,
+ (ins VecListThreeD:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VST3dWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr, $Rm",
- (ins VecListThreeD:$list, addrmode6:$addr,
+ (ins VecListThreeD:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VST3qWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".8", "$list, $addr, $Rm",
- (ins VecListThreeQ:$list, addrmode6:$addr,
+ (ins VecListThreeQ:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VST3qWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".16", "$list, $addr, $Rm",
- (ins VecListThreeQ:$list, addrmode6:$addr,
+ (ins VecListThreeQ:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VST3qWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst3${p}", ".32", "$list, $addr, $Rm",
- (ins VecListThreeQ:$list, addrmode6:$addr,
+ (ins VecListThreeQ:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
// VLD4 all-lanes pseudo-instructions. These need special handling for
// the lane index that an InstAlias can't handle, so we use these instead.
def VLD4DUPdAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
- (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDAllLanes:$list, addrmode6dupalign32:$addr,
+ pred:$p)>;
def VLD4DUPdAsm_16: NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
- (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDAllLanes:$list, addrmode6dupalign64:$addr,
+ pred:$p)>;
def VLD4DUPdAsm_32: NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
- (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDAllLanes:$list, addrmode6dupalign64or128:$addr,
+ pred:$p)>;
def VLD4DUPqAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
- (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQAllLanes:$list, addrmode6dupalign32:$addr,
+ pred:$p)>;
def VLD4DUPqAsm_16: NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
- (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQAllLanes:$list, addrmode6dupalign64:$addr,
+ pred:$p)>;
def VLD4DUPqAsm_32: NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
- (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQAllLanes:$list, addrmode6dupalign64or128:$addr,
+ pred:$p)>;
def VLD4DUPdWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
- (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDAllLanes:$list, addrmode6dupalign32:$addr,
+ pred:$p)>;
def VLD4DUPdWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
- (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDAllLanes:$list, addrmode6dupalign64:$addr,
+ pred:$p)>;
def VLD4DUPdWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
- (ins VecListFourDAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDAllLanes:$list, addrmode6dupalign64or128:$addr,
+ pred:$p)>;
def VLD4DUPqWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
- (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQAllLanes:$list, addrmode6dupalign32:$addr,
+ pred:$p)>;
def VLD4DUPqWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
- (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQAllLanes:$list, addrmode6dupalign64:$addr,
+ pred:$p)>;
def VLD4DUPqWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
- (ins VecListFourQAllLanes:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQAllLanes:$list, addrmode6dupalign64or128:$addr,
+ pred:$p)>;
def VLD4DUPdWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
- (ins VecListFourDAllLanes:$list, addrmode6:$addr,
+ (ins VecListFourDAllLanes:$list, addrmode6dupalign32:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4DUPdWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
- (ins VecListFourDAllLanes:$list, addrmode6:$addr,
+ (ins VecListFourDAllLanes:$list, addrmode6dupalign64:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4DUPdWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
- (ins VecListFourDAllLanes:$list, addrmode6:$addr,
- rGPR:$Rm, pred:$p)>;
+ (ins VecListFourDAllLanes:$list,
+ addrmode6dupalign64or128:$addr, rGPR:$Rm, pred:$p)>;
def VLD4DUPqWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
- (ins VecListFourQAllLanes:$list, addrmode6:$addr,
+ (ins VecListFourQAllLanes:$list, addrmode6dupalign32:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4DUPqWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
- (ins VecListFourQAllLanes:$list, addrmode6:$addr,
+ (ins VecListFourQAllLanes:$list, addrmode6dupalign64:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4DUPqWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
- (ins VecListFourQAllLanes:$list, addrmode6:$addr,
- rGPR:$Rm, pred:$p)>;
+ (ins VecListFourQAllLanes:$list,
+ addrmode6dupalign64or128:$addr, rGPR:$Rm, pred:$p)>;
// VLD4 single-lane pseudo-instructions. These need special handling for
// the lane index that an InstAlias can't handle, so we use these instead.
def VLD4LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
- (ins VecListFourDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VLD4LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
- (ins VecListFourDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VLD4LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
- (ins VecListFourDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDWordIndexed:$list, addrmode6align64or128:$addr,
+ pred:$p)>;
def VLD4LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
- (ins VecListFourQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VLD4LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
- (ins VecListFourQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQWordIndexed:$list, addrmode6align64or128:$addr,
+ pred:$p)>;
def VLD4LNdWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
- (ins VecListFourDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VLD4LNdWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
- (ins VecListFourDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VLD4LNdWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
- (ins VecListFourDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDWordIndexed:$list, addrmode6align64or128:$addr,
+ pred:$p)>;
def VLD4LNqWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
- (ins VecListFourQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VLD4LNqWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
- (ins VecListFourQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQWordIndexed:$list, addrmode6align64or128:$addr,
+ pred:$p)>;
def VLD4LNdWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
- (ins VecListFourDByteIndexed:$list, addrmode6:$addr,
+ (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4LNdWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
- (ins VecListFourDHWordIndexed:$list, addrmode6:$addr,
+ (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4LNdWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
- (ins VecListFourDWordIndexed:$list, addrmode6:$addr,
- rGPR:$Rm, pred:$p)>;
+ (ins VecListFourDWordIndexed:$list,
+ addrmode6align64or128:$addr, rGPR:$Rm, pred:$p)>;
def VLD4LNqWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
- (ins VecListFourQHWordIndexed:$list, addrmode6:$addr,
+ (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4LNqWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
- (ins VecListFourQWordIndexed:$list, addrmode6:$addr,
- rGPR:$Rm, pred:$p)>;
+ (ins VecListFourQWordIndexed:$list,
+ addrmode6align64or128:$addr, rGPR:$Rm, pred:$p)>;
// the vector operands that the normal instructions don't yet model.
// FIXME: Remove these when the register classes and instructions are updated.
def VLD4dAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4dAsm_16 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4dAsm_32 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4qAsm_8 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4qAsm_16 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4qAsm_32 : NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4dWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4dWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4dWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4qWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr!",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4qWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr!",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4qWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr!",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VLD4dWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
- (ins VecListFourD:$list, addrmode6:$addr,
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4dWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
- (ins VecListFourD:$list, addrmode6:$addr,
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4dWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
- (ins VecListFourD:$list, addrmode6:$addr,
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4qWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".8", "$list, $addr, $Rm",
- (ins VecListFourQ:$list, addrmode6:$addr,
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4qWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".16", "$list, $addr, $Rm",
- (ins VecListFourQ:$list, addrmode6:$addr,
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
def VLD4qWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vld4${p}", ".32", "$list, $addr, $Rm",
- (ins VecListFourQ:$list, addrmode6:$addr,
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
// VST4 single-lane pseudo-instructions. These need special handling for
// the lane index that an InstAlias can't handle, so we use these instead.
def VST4LNdAsm_8 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr",
- (ins VecListFourDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VST4LNdAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
- (ins VecListFourDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VST4LNdAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
- (ins VecListFourDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDWordIndexed:$list, addrmode6align64or128:$addr,
+ pred:$p)>;
def VST4LNqAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
- (ins VecListFourQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VST4LNqAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
- (ins VecListFourQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQWordIndexed:$list, addrmode6align64or128:$addr,
+ pred:$p)>;
def VST4LNdWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr!",
- (ins VecListFourDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
+ pred:$p)>;
def VST4LNdWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
- (ins VecListFourDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VST4LNdWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
- (ins VecListFourDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourDWordIndexed:$list, addrmode6align64or128:$addr,
+ pred:$p)>;
def VST4LNqWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
- (ins VecListFourQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
+ pred:$p)>;
def VST4LNqWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
- (ins VecListFourQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQWordIndexed:$list, addrmode6align64or128:$addr,
+ pred:$p)>;
def VST4LNdWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr, $Rm",
- (ins VecListFourDByteIndexed:$list, addrmode6:$addr,
+ (ins VecListFourDByteIndexed:$list, addrmode6align32:$addr,
rGPR:$Rm, pred:$p)>;
def VST4LNdWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
- (ins VecListFourDHWordIndexed:$list, addrmode6:$addr,
+ (ins VecListFourDHWordIndexed:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VST4LNdWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr, $Rm",
- (ins VecListFourDWordIndexed:$list, addrmode6:$addr,
- rGPR:$Rm, pred:$p)>;
+ (ins VecListFourDWordIndexed:$list,
+ addrmode6align64or128:$addr, rGPR:$Rm, pred:$p)>;
def VST4LNqWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
- (ins VecListFourQHWordIndexed:$list, addrmode6:$addr,
+ (ins VecListFourQHWordIndexed:$list, addrmode6align64:$addr,
rGPR:$Rm, pred:$p)>;
def VST4LNqWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr, $Rm",
- (ins VecListFourQWordIndexed:$list, addrmode6:$addr,
- rGPR:$Rm, pred:$p)>;
+ (ins VecListFourQWordIndexed:$list,
+ addrmode6align64or128:$addr, rGPR:$Rm, pred:$p)>;
// VST4 multiple structure pseudo-instructions. These need special handling for
// the vector operands that the normal instructions don't yet model.
// FIXME: Remove these when the register classes and instructions are updated.
def VST4dAsm_8 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4dAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4dAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4qAsm_8 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4qAsm_16 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4qAsm_32 : NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4dWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr!",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4dWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4dWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
- (ins VecListFourD:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4qWB_fixed_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr!",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4qWB_fixed_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr!",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4qWB_fixed_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr!",
- (ins VecListFourQ:$list, addrmode6:$addr, pred:$p)>;
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
+ pred:$p)>;
def VST4dWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr, $Rm",
- (ins VecListFourD:$list, addrmode6:$addr,
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
def VST4dWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
- (ins VecListFourD:$list, addrmode6:$addr,
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
def VST4dWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr, $Rm",
- (ins VecListFourD:$list, addrmode6:$addr,
+ (ins VecListFourD:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
def VST4qWB_register_Asm_8 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".8", "$list, $addr, $Rm",
- (ins VecListFourQ:$list, addrmode6:$addr,
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
def VST4qWB_register_Asm_16 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".16", "$list, $addr, $Rm",
- (ins VecListFourQ:$list, addrmode6:$addr,
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
def VST4qWB_register_Asm_32 :
NEONDataTypeAsmPseudoInst<"vst4${p}", ".32", "$list, $addr, $Rm",
- (ins VecListFourQ:$list, addrmode6:$addr,
+ (ins VecListFourQ:$list, addrmode6align64or128or256:$addr,
rGPR:$Rm, pred:$p)>;
-// VMOV takes an optional datatype suffix
+// VMOV/VMVN takes an optional datatype suffix
defm : NEONDTAnyInstAlias<"vmov${p}", "$Vd, $Vm",
(VORRd DPR:$Vd, DPR:$Vm, DPR:$Vm, pred:$p)>;
defm : NEONDTAnyInstAlias<"vmov${p}", "$Vd, $Vm",
(VORRq QPR:$Vd, QPR:$Vm, QPR:$Vm, pred:$p)>;
+defm : NEONDTAnyInstAlias<"vmvn${p}", "$Vd, $Vm",
+ (VMVNd DPR:$Vd, DPR:$Vm, pred:$p)>;
+defm : NEONDTAnyInstAlias<"vmvn${p}", "$Vd, $Vm",
+ (VMVNq QPR:$Vd, QPR:$Vm, pred:$p)>;
+
// VCLT (register) is an assembler alias for VCGT w/ the operands reversed.
// D-register versions.
def : NEONInstAlias<"vcle${p}.s8 $Dd, $Dn, $Dm",