let PrintMethod = "printNEONModImmOperand";
let ParserMatchClass = nImmVMOVI32AsmOperand;
}
+def nImmVMOVI32NegAsmOperand : AsmOperandClass { let Name = "NEONi32vmovNeg"; }
+def nImmVMOVI32Neg : Operand<i32> {
+ let PrintMethod = "printNEONModImmOperand";
+ let ParserMatchClass = nImmVMOVI32NegAsmOperand;
+}
+def nImmVMOVF32 : Operand<i32> {
+ let PrintMethod = "printFPImmOperand";
+ let ParserMatchClass = FPImmOperand;
+}
def nImmSplatI64AsmOperand : AsmOperandClass { let Name = "NEONi64splat"; }
def nImmSplatI64 : Operand<i32> {
let PrintMethod = "printNEONModImmOperand";
let MIOperandInfo = (ops i32imm);
}
+// Register list of one D register.
def VecListOneDAsmOperand : AsmOperandClass {
let Name = "VecListOneD";
let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListOperands";
}
def VecListOneD : RegisterOperand<DPR, "printVectorListOne"> {
let ParserMatchClass = VecListOneDAsmOperand;
def VecListTwoDAsmOperand : AsmOperandClass {
let Name = "VecListTwoD";
let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListOperands";
}
def VecListTwoD : RegisterOperand<DPR, "printVectorListTwo"> {
let ParserMatchClass = VecListTwoDAsmOperand;
def VecListThreeDAsmOperand : AsmOperandClass {
let Name = "VecListThreeD";
let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListOperands";
}
def VecListThreeD : RegisterOperand<DPR, "printVectorListThree"> {
let ParserMatchClass = VecListThreeDAsmOperand;
def VecListFourDAsmOperand : AsmOperandClass {
let Name = "VecListFourD";
let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListOperands";
}
def VecListFourD : RegisterOperand<DPR, "printVectorListFour"> {
let ParserMatchClass = VecListFourDAsmOperand;
def VecListTwoQAsmOperand : AsmOperandClass {
let Name = "VecListTwoQ";
let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListOperands";
}
-def VecListTwoQ : RegisterOperand<DPR, "printVectorListTwo"> {
+def VecListTwoQ : RegisterOperand<DPR, "printVectorListTwoSpaced"> {
let ParserMatchClass = VecListTwoQAsmOperand;
}
+// Register list of one D register, with "all lanes" subscripting.
+def VecListOneDAllLanesAsmOperand : AsmOperandClass {
+ let Name = "VecListOneDAllLanes";
+ let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListOperands";
+}
+def VecListOneDAllLanes : RegisterOperand<DPR, "printVectorListOneAllLanes"> {
+ let ParserMatchClass = VecListOneDAllLanesAsmOperand;
+}
+// Register list of two D registers, with "all lanes" subscripting.
+def VecListTwoDAllLanesAsmOperand : AsmOperandClass {
+ let Name = "VecListTwoDAllLanes";
+ let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListOperands";
+}
+def VecListTwoDAllLanes : RegisterOperand<DPR, "printVectorListTwoAllLanes"> {
+ let ParserMatchClass = VecListTwoDAllLanesAsmOperand;
+}
+// Register list of two D registers spaced by 2 (two sequential Q registers).
+def VecListTwoQAllLanesAsmOperand : AsmOperandClass {
+ let Name = "VecListTwoQAllLanes";
+ let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListOperands";
+}
+def VecListTwoQAllLanes : RegisterOperand<DPR,
+ "printVectorListTwoSpacedAllLanes"> {
+ let ParserMatchClass = VecListTwoQAllLanesAsmOperand;
+}
+
+// Register list of one D register, with byte lane subscripting.
+def VecListOneDByteIndexAsmOperand : AsmOperandClass {
+ let Name = "VecListOneDByteIndexed";
+ let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListIndexedOperands";
+}
+def VecListOneDByteIndexed : Operand<i32> {
+ let ParserMatchClass = VecListOneDByteIndexAsmOperand;
+ let MIOperandInfo = (ops DPR:$Vd, i32imm:$idx);
+}
+// ...with half-word lane subscripting.
+def VecListOneDHWordIndexAsmOperand : AsmOperandClass {
+ let Name = "VecListOneDHWordIndexed";
+ let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListIndexedOperands";
+}
+def VecListOneDHWordIndexed : Operand<i32> {
+ let ParserMatchClass = VecListOneDHWordIndexAsmOperand;
+ let MIOperandInfo = (ops DPR:$Vd, i32imm:$idx);
+}
+// ...with word lane subscripting.
+def VecListOneDWordIndexAsmOperand : AsmOperandClass {
+ let Name = "VecListOneDWordIndexed";
+ let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListIndexedOperands";
+}
+def VecListOneDWordIndexed : Operand<i32> {
+ let ParserMatchClass = VecListOneDWordIndexAsmOperand;
+ let MIOperandInfo = (ops DPR:$Vd, i32imm:$idx);
+}
+// Register list of two D registers with byte lane subscripting.
+def VecListTwoDByteIndexAsmOperand : AsmOperandClass {
+ let Name = "VecListTwoDByteIndexed";
+ let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListIndexedOperands";
+}
+def VecListTwoDByteIndexed : Operand<i32> {
+ let ParserMatchClass = VecListTwoDByteIndexAsmOperand;
+ let MIOperandInfo = (ops DPR:$Vd, i32imm:$idx);
+}
+// ...with half-word lane subscripting.
+def VecListTwoDHWordIndexAsmOperand : AsmOperandClass {
+ let Name = "VecListTwoDHWordIndexed";
+ let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListIndexedOperands";
+}
+def VecListTwoDHWordIndexed : Operand<i32> {
+ let ParserMatchClass = VecListTwoDHWordIndexAsmOperand;
+ let MIOperandInfo = (ops DPR:$Vd, i32imm:$idx);
+}
+// ...with word lane subscripting.
+def VecListTwoDWordIndexAsmOperand : AsmOperandClass {
+ let Name = "VecListTwoDWordIndexed";
+ let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListIndexedOperands";
+}
+def VecListTwoDWordIndexed : Operand<i32> {
+ let ParserMatchClass = VecListTwoDWordIndexAsmOperand;
+ let MIOperandInfo = (ops DPR:$Vd, i32imm:$idx);
+}
+// Register list of two Q registers with half-word lane subscripting.
+def VecListTwoQHWordIndexAsmOperand : AsmOperandClass {
+ let Name = "VecListTwoQHWordIndexed";
+ let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListIndexedOperands";
+}
+def VecListTwoQHWordIndexed : Operand<i32> {
+ let ParserMatchClass = VecListTwoQHWordIndexAsmOperand;
+ let MIOperandInfo = (ops DPR:$Vd, i32imm:$idx);
+}
+// ...with word lane subscripting.
+def VecListTwoQWordIndexAsmOperand : AsmOperandClass {
+ let Name = "VecListTwoQWordIndexed";
+ let ParserMethod = "parseVectorList";
+ let RenderMethod = "addVecListIndexedOperands";
+}
+def VecListTwoQWordIndexed : Operand<i32> {
+ let ParserMatchClass = VecListTwoQWordIndexAsmOperand;
+ let MIOperandInfo = (ops DPR:$Vd, i32imm:$idx);
+}
+
//===----------------------------------------------------------------------===//
// NEON-specific DAG Nodes.
//===----------------------------------------------------------------------===//
def SDTARMVMOVIMM : SDTypeProfile<1, 1, [SDTCisVec<0>, SDTCisVT<1, i32>]>;
def NEONvmovImm : SDNode<"ARMISD::VMOVIMM", SDTARMVMOVIMM>;
def NEONvmvnImm : SDNode<"ARMISD::VMVNIMM", SDTARMVMOVIMM>;
+def NEONvmovFPImm : SDNode<"ARMISD::VMOVFPIMM", SDTARMVMOVIMM>;
def SDTARMVORRIMM : SDTypeProfile<1, 2, [SDTCisVec<0>, SDTCisSameAs<0, 1>,
SDTCisVT<2, i32>]>;
: PseudoNLdSt<(outs QPR:$dst, GPR:$wb),
(ins addrmode6:$addr, rGPR:$offset), itin,
"$addr.addr = $wb">;
+
class VLDQQPseudo<InstrItinClass itin>
: PseudoNLdSt<(outs QQPR:$dst), (ins addrmode6:$addr), itin, "">;
class VLDQQWBPseudo<InstrItinClass itin>
: PseudoNLdSt<(outs QQPR:$dst, GPR:$wb),
(ins addrmode6:$addr, am6offset:$offset), itin,
"$addr.addr = $wb">;
+class VLDQQWBfixedPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs QQPR:$dst, GPR:$wb),
+ (ins addrmode6:$addr), itin,
+ "$addr.addr = $wb">;
+class VLDQQWBregisterPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs QQPR:$dst, GPR:$wb),
+ (ins addrmode6:$addr, rGPR:$offset), itin,
+ "$addr.addr = $wb">;
+
+
class VLDQQQQPseudo<InstrItinClass itin>
: PseudoNLdSt<(outs QQQQPR:$dst), (ins addrmode6:$addr, QQQQPR:$src),itin,
"$src = $dst">;
def VLD1d64QPseudo : VLDQQPseudo<IIC_VLD1x4>;
// VLD2 : Vector Load (multiple 2-element structures)
-class VLD2D<bits<4> op11_8, bits<4> op7_4, string Dt, RegisterOperand VdTy>
+class VLD2<bits<4> op11_8, bits<4> op7_4, string Dt, RegisterOperand VdTy,
+ InstrItinClass itin>
: NLdSt<0, 0b10, op11_8, op7_4, (outs VdTy:$Vd),
- (ins addrmode6:$Rn), IIC_VLD2,
- "vld2", Dt, "$Vd, $Rn", "", []> {
- let Rm = 0b1111;
- let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
-}
-class VLD2Q<bits<4> op7_4, string Dt, RegisterOperand VdTy>
- : NLdSt<0, 0b10, 0b0011, op7_4,
- (outs VdTy:$Vd),
- (ins addrmode6:$Rn), IIC_VLD2x2,
+ (ins addrmode6:$Rn), itin,
"vld2", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{5-4} = Rn{5-4};
let DecoderMethod = "DecodeVLDInstruction";
}
-def VLD2d8 : VLD2D<0b1000, {0,0,?,?}, "8", VecListTwoD>;
-def VLD2d16 : VLD2D<0b1000, {0,1,?,?}, "16", VecListTwoD>;
-def VLD2d32 : VLD2D<0b1000, {1,0,?,?}, "32", VecListTwoD>;
+def VLD2d8 : VLD2<0b1000, {0,0,?,?}, "8", VecListTwoD, IIC_VLD2>;
+def VLD2d16 : VLD2<0b1000, {0,1,?,?}, "16", VecListTwoD, IIC_VLD2>;
+def VLD2d32 : VLD2<0b1000, {1,0,?,?}, "32", VecListTwoD, IIC_VLD2>;
-def VLD2q8 : VLD2Q<{0,0,?,?}, "8", VecListFourD>;
-def VLD2q16 : VLD2Q<{0,1,?,?}, "16", VecListFourD>;
-def VLD2q32 : VLD2Q<{1,0,?,?}, "32", VecListFourD>;
+def VLD2q8 : VLD2<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VLD2x2>;
+def VLD2q16 : VLD2<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VLD2x2>;
+def VLD2q32 : VLD2<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VLD2x2>;
def VLD2d8Pseudo : VLDQPseudo<IIC_VLD2>;
def VLD2d16Pseudo : VLDQPseudo<IIC_VLD2>;
def VLD2q32Pseudo : VLDQQPseudo<IIC_VLD2x2>;
// ...with address register writeback:
-class VLD2DWB<bits<4> op11_8, bits<4> op7_4, string Dt, RegisterOperand VdTy>
- : NLdSt<0, 0b10, op11_8, op7_4, (outs VdTy:$Vd, GPR:$wb),
- (ins addrmode6:$Rn, am6offset:$Rm), IIC_VLD2u,
- "vld2", Dt, "$Vd, $Rn$Rm",
- "$Rn.addr = $wb", []> {
- let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
-}
-class VLD2QWB<bits<4> op7_4, string Dt, RegisterOperand VdTy>
- : NLdSt<0, 0b10, 0b0011, op7_4,
- (outs VdTy:$Vd, GPR:$wb),
- (ins addrmode6:$Rn, am6offset:$Rm), IIC_VLD2x2u,
- "vld2", Dt, "$Vd, $Rn$Rm",
- "$Rn.addr = $wb", []> {
- let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVLDInstruction";
+multiclass VLD2WB<bits<4> op11_8, bits<4> op7_4, string Dt,
+ RegisterOperand VdTy, InstrItinClass itin> {
+ def _fixed : NLdSt<0, 0b10, op11_8, op7_4, (outs VdTy:$Vd, GPR:$wb),
+ (ins addrmode6:$Rn), itin,
+ "vld2", Dt, "$Vd, $Rn!",
+ "$Rn.addr = $wb", []> {
+ let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
+ let Inst{5-4} = Rn{5-4};
+ let DecoderMethod = "DecodeVLDInstruction";
+ let AsmMatchConverter = "cvtVLDwbFixed";
+ }
+ def _register : NLdSt<0, 0b10, op11_8, op7_4, (outs VdTy:$Vd, GPR:$wb),
+ (ins addrmode6:$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";
+ }
}
-def VLD2d8_UPD : VLD2DWB<0b1000, {0,0,?,?}, "8", VecListTwoD>;
-def VLD2d16_UPD : VLD2DWB<0b1000, {0,1,?,?}, "16", VecListTwoD>;
-def VLD2d32_UPD : VLD2DWB<0b1000, {1,0,?,?}, "32", VecListTwoD>;
+defm VLD2d8wb : VLD2WB<0b1000, {0,0,?,?}, "8", VecListTwoD, IIC_VLD2u>;
+defm VLD2d16wb : VLD2WB<0b1000, {0,1,?,?}, "16", VecListTwoD, IIC_VLD2u>;
+defm VLD2d32wb : VLD2WB<0b1000, {1,0,?,?}, "32", VecListTwoD, IIC_VLD2u>;
-def VLD2q8_UPD : VLD2QWB<{0,0,?,?}, "8", VecListFourD>;
-def VLD2q16_UPD : VLD2QWB<{0,1,?,?}, "16", VecListFourD>;
-def VLD2q32_UPD : VLD2QWB<{1,0,?,?}, "32", VecListFourD>;
+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>;
-def VLD2d8Pseudo_UPD : VLDQWBPseudo<IIC_VLD2u>;
-def VLD2d16Pseudo_UPD : VLDQWBPseudo<IIC_VLD2u>;
-def VLD2d32Pseudo_UPD : VLDQWBPseudo<IIC_VLD2u>;
+def VLD2d8PseudoWB_fixed : VLDQWBfixedPseudo<IIC_VLD2u>;
+def VLD2d16PseudoWB_fixed : VLDQWBfixedPseudo<IIC_VLD2u>;
+def VLD2d32PseudoWB_fixed : VLDQWBfixedPseudo<IIC_VLD2u>;
+def VLD2d8PseudoWB_register : VLDQWBregisterPseudo<IIC_VLD2u>;
+def VLD2d16PseudoWB_register : VLDQWBregisterPseudo<IIC_VLD2u>;
+def VLD2d32PseudoWB_register : VLDQWBregisterPseudo<IIC_VLD2u>;
-def VLD2q8Pseudo_UPD : VLDQQWBPseudo<IIC_VLD2x2u>;
-def VLD2q16Pseudo_UPD : VLDQQWBPseudo<IIC_VLD2x2u>;
-def VLD2q32Pseudo_UPD : VLDQQWBPseudo<IIC_VLD2x2u>;
+def VLD2q8PseudoWB_fixed : VLDQQWBfixedPseudo<IIC_VLD2x2u>;
+def VLD2q16PseudoWB_fixed : VLDQQWBfixedPseudo<IIC_VLD2x2u>;
+def VLD2q32PseudoWB_fixed : VLDQQWBfixedPseudo<IIC_VLD2x2u>;
+def VLD2q8PseudoWB_register : VLDQQWBregisterPseudo<IIC_VLD2x2u>;
+def VLD2q16PseudoWB_register : VLDQQWBregisterPseudo<IIC_VLD2x2u>;
+def VLD2q32PseudoWB_register : VLDQQWBregisterPseudo<IIC_VLD2x2u>;
// ...with double-spaced registers
-def VLD2b8 : VLD2D<0b1001, {0,0,?,?}, "8", VecListTwoQ>;
-def VLD2b16 : VLD2D<0b1001, {0,1,?,?}, "16", VecListTwoQ>;
-def VLD2b32 : VLD2D<0b1001, {1,0,?,?}, "32", VecListTwoQ>;
-def VLD2b8_UPD : VLD2DWB<0b1001, {0,0,?,?}, "8", VecListTwoQ>;
-def VLD2b16_UPD : VLD2DWB<0b1001, {0,1,?,?}, "16", VecListTwoQ>;
-def VLD2b32_UPD : VLD2DWB<0b1001, {1,0,?,?}, "32", VecListTwoQ>;
+def VLD2b8 : VLD2<0b1001, {0,0,?,?}, "8", VecListTwoQ, IIC_VLD2>;
+def VLD2b16 : VLD2<0b1001, {0,1,?,?}, "16", VecListTwoQ, IIC_VLD2>;
+def VLD2b32 : VLD2<0b1001, {1,0,?,?}, "32", VecListTwoQ, IIC_VLD2>;
+defm VLD2b8wb : VLD2WB<0b1001, {0,0,?,?}, "8", VecListTwoQ, IIC_VLD2u>;
+defm VLD2b16wb : VLD2WB<0b1001, {0,1,?,?}, "16", VecListTwoQ, IIC_VLD2u>;
+defm VLD2b32wb : VLD2WB<0b1001, {1,0,?,?}, "32", VecListTwoQ, IIC_VLD2u>;
// VLD3 : Vector Load (multiple 3-element structures)
class VLD3D<bits<4> op11_8, bits<4> op7_4, string Dt>
}
def VLD1LNd16 : VLD1LN<0b0100, {?,?,0,?}, "16", v4i16, extloadi16> {
let Inst{7-6} = lane{1-0};
- let Inst{4} = Rn{4};
+ let Inst{5-4} = Rn{5-4};
}
def VLD1LNd32 : VLD1LN32<0b1000, {?,0,?,?}, "32", v2i32, load> {
let Inst{7} = lane{0};
- let Inst{5} = Rn{4};
- let Inst{4} = Rn{4};
+ let Inst{5-4} = Rn{5-4};
}
def VLD1LNq8Pseudo : VLD1QLNPseudo<v16i8, extloadi8>;
let Inst{7-6} = lane{1-0};
}
def VLD3LNd32_UPD : VLD3LNWB<0b1010, {?,0,0,0}, "32"> {
- let Inst{7} = lane{0};
+ let Inst{7} = lane{0};
}
def VLD3LNd8Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD3lnu>;
let Inst{7-6} = lane{1-0};
}
def VLD3LNq32_UPD : VLD3LNWB<0b1010, {?,1,0,0}, "32"> {
- let Inst{7} = lane{0};
+ let Inst{7} = lane{0};
}
def VLD3LNq16Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD3lnu>;
"\\{$Vd[$lane], $dst2[$lane], $dst3[$lane], $dst4[$lane]\\}, $Rn",
"$src1 = $Vd, $src2 = $dst2, $src3 = $dst3, $src4 = $dst4", []> {
let Rm = 0b1111;
- let Inst{4} = Rn{4};
+ let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD4LN";
}
let Inst{7-6} = lane{1-0};
}
def VLD4LNd32 : VLD4LN<0b1011, {?,0,?,?}, "32"> {
- let Inst{7} = lane{0};
+ let Inst{7} = lane{0};
let Inst{5} = Rn{5};
}
let Inst{7-6} = lane{1-0};
}
def VLD4LNq32 : VLD4LN<0b1011, {?,1,?,?}, "32"> {
- let Inst{7} = lane{0};
+ let Inst{7} = lane{0};
let Inst{5} = Rn{5};
}
let Inst{7-6} = lane{1-0};
}
def VLD4LNd32_UPD : VLD4LNWB<0b1011, {?,0,?,?}, "32"> {
- let Inst{7} = lane{0};
+ let Inst{7} = lane{0};
let Inst{5} = Rn{5};
}
let Inst{7-6} = lane{1-0};
}
def VLD4LNq32_UPD : VLD4LNWB<0b1011, {?,1,?,?}, "32"> {
- let Inst{7} = lane{0};
+ let Inst{7} = lane{0};
let Inst{5} = Rn{5};
}
// VLD1DUP : Vector Load (single element to all lanes)
class VLD1DUP<bits<4> op7_4, string Dt, ValueType Ty, PatFrag LoadOp>
- : NLdSt<1, 0b10, 0b1100, op7_4, (outs DPR:$Vd), (ins addrmode6dup:$Rn),
- IIC_VLD1dup, "vld1", Dt, "\\{$Vd[]\\}, $Rn", "",
- [(set DPR:$Vd, (Ty (NEONvdup (i32 (LoadOp addrmode6dup:$Rn)))))]> {
+ : NLdSt<1, 0b10, 0b1100, op7_4, (outs VecListOneDAllLanes:$Vd),
+ (ins addrmode6dup:$Rn),
+ IIC_VLD1dup, "vld1", Dt, "$Vd, $Rn", "",
+ [(set VecListOneDAllLanes:$Vd,
+ (Ty (NEONvdup (i32 (LoadOp addrmode6dup:$Rn)))))]> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD1DupInstruction";
let mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1 in {
class VLD1QDUP<bits<4> op7_4, string Dt>
- : NLdSt<1, 0b10, 0b1100, op7_4, (outs DPR:$Vd, DPR:$dst2),
+ : NLdSt<1, 0b10, 0b1100, op7_4, (outs VecListTwoDAllLanes:$Vd),
(ins addrmode6dup:$Rn), IIC_VLD1dup,
- "vld1", Dt, "\\{$Vd[], $dst2[]\\}, $Rn", "", []> {
+ "vld1", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD1DupInstruction";
def VLD1DUPq32 : VLD1QDUP<{1,0,1,?}, "32">;
// ...with address register writeback:
-class VLD1DUPWB<bits<4> op7_4, string Dt>
- : NLdSt<1, 0b10, 0b1100, op7_4, (outs DPR:$Vd, GPR:$wb),
- (ins addrmode6dup:$Rn, am6offset:$Rm), IIC_VLD1dupu,
- "vld1", Dt, "\\{$Vd[]\\}, $Rn$Rm", "$Rn.addr = $wb", []> {
- let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVLD1DupInstruction";
+multiclass VLD1DUPWB<bits<4> op7_4, string Dt> {
+ def _fixed : NLdSt<1, 0b10, 0b1100, op7_4,
+ (outs VecListOneDAllLanes:$Vd, GPR:$wb),
+ (ins addrmode6dup:$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:$Rm), IIC_VLD1dupu,
+ "vld1", Dt, "$Vd, $Rn, $Rm",
+ "$Rn.addr = $wb", []> {
+ let Inst{4} = Rn{4};
+ let DecoderMethod = "DecodeVLD1DupInstruction";
+ let AsmMatchConverter = "cvtVLDwbRegister";
+ }
}
-class VLD1QDUPWB<bits<4> op7_4, string Dt>
- : NLdSt<1, 0b10, 0b1100, op7_4, (outs DPR:$Vd, DPR:$dst2, GPR:$wb),
- (ins addrmode6dup:$Rn, am6offset:$Rm), IIC_VLD1dupu,
- "vld1", Dt, "\\{$Vd[], $dst2[]\\}, $Rn$Rm", "$Rn.addr = $wb", []> {
- let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVLD1DupInstruction";
+multiclass VLD1QDUPWB<bits<4> op7_4, string Dt> {
+ def _fixed : NLdSt<1, 0b10, 0b1100, op7_4,
+ (outs VecListTwoDAllLanes:$Vd, GPR:$wb),
+ (ins addrmode6dup:$Rn), IIC_VLD1dupu,
+ "vld1", Dt, "$Vd, $Rn!",
+ "$Rn.addr = $wb", []> {
+ let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
+ let Inst{4} = Rn{4};
+ let DecoderMethod = "DecodeVLD1DupInstruction";
+ let AsmMatchConverter = "cvtVLDwbFixed";
+ }
+ def _register : NLdSt<1, 0b10, 0b1100, op7_4,
+ (outs VecListTwoDAllLanes:$Vd, GPR:$wb),
+ (ins addrmode6dup:$Rn, rGPR:$Rm), IIC_VLD1dupu,
+ "vld1", Dt, "$Vd, $Rn, $Rm",
+ "$Rn.addr = $wb", []> {
+ let Inst{4} = Rn{4};
+ let DecoderMethod = "DecodeVLD1DupInstruction";
+ let AsmMatchConverter = "cvtVLDwbRegister";
+ }
}
-def VLD1DUPd8_UPD : VLD1DUPWB<{0,0,0,0}, "8">;
-def VLD1DUPd16_UPD : VLD1DUPWB<{0,1,0,?}, "16">;
-def VLD1DUPd32_UPD : VLD1DUPWB<{1,0,0,?}, "32">;
+defm VLD1DUPd8wb : VLD1DUPWB<{0,0,0,0}, "8">;
+defm VLD1DUPd16wb : VLD1DUPWB<{0,1,0,?}, "16">;
+defm VLD1DUPd32wb : VLD1DUPWB<{1,0,0,?}, "32">;
-def VLD1DUPq8_UPD : VLD1QDUPWB<{0,0,1,0}, "8">;
-def VLD1DUPq16_UPD : VLD1QDUPWB<{0,1,1,?}, "16">;
-def VLD1DUPq32_UPD : VLD1QDUPWB<{1,0,1,?}, "32">;
+defm VLD1DUPq8wb : VLD1QDUPWB<{0,0,1,0}, "8">;
+defm VLD1DUPq16wb : VLD1QDUPWB<{0,1,1,?}, "16">;
+defm VLD1DUPq32wb : VLD1QDUPWB<{1,0,1,?}, "32">;
-def VLD1DUPq8Pseudo_UPD : VLDQWBPseudo<IIC_VLD1dupu>;
-def VLD1DUPq16Pseudo_UPD : VLDQWBPseudo<IIC_VLD1dupu>;
-def VLD1DUPq32Pseudo_UPD : VLDQWBPseudo<IIC_VLD1dupu>;
+def VLD1DUPq8PseudoWB_fixed : VLDQWBfixedPseudo<IIC_VLD1dupu>;
+def VLD1DUPq16PseudoWB_fixed : VLDQWBfixedPseudo<IIC_VLD1dupu>;
+def VLD1DUPq32PseudoWB_fixed : VLDQWBfixedPseudo<IIC_VLD1dupu>;
+def VLD1DUPq8PseudoWB_register : VLDQWBregisterPseudo<IIC_VLD1dupu>;
+def VLD1DUPq16PseudoWB_register : VLDQWBregisterPseudo<IIC_VLD1dupu>;
+def VLD1DUPq32PseudoWB_register : VLDQWBregisterPseudo<IIC_VLD1dupu>;
// VLD2DUP : Vector Load (single 2-element structure to all lanes)
-class VLD2DUP<bits<4> op7_4, string Dt>
- : NLdSt<1, 0b10, 0b1101, op7_4, (outs DPR:$Vd, DPR:$dst2),
+class VLD2DUP<bits<4> op7_4, string Dt, RegisterOperand VdTy>
+ : NLdSt<1, 0b10, 0b1101, op7_4, (outs VdTy:$Vd),
(ins addrmode6dup:$Rn), IIC_VLD2dup,
- "vld2", Dt, "\\{$Vd[], $dst2[]\\}, $Rn", "", []> {
+ "vld2", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVLD2DupInstruction";
}
-def VLD2DUPd8 : VLD2DUP<{0,0,0,?}, "8">;
-def VLD2DUPd16 : VLD2DUP<{0,1,0,?}, "16">;
-def VLD2DUPd32 : VLD2DUP<{1,0,0,?}, "32">;
+def VLD2DUPd8 : VLD2DUP<{0,0,0,?}, "8", VecListTwoDAllLanes>;
+def VLD2DUPd16 : VLD2DUP<{0,1,0,?}, "16", VecListTwoDAllLanes>;
+def VLD2DUPd32 : VLD2DUP<{1,0,0,?}, "32", VecListTwoDAllLanes>;
def VLD2DUPd8Pseudo : VLDQPseudo<IIC_VLD2dup>;
def VLD2DUPd16Pseudo : VLDQPseudo<IIC_VLD2dup>;
def VLD2DUPd32Pseudo : VLDQPseudo<IIC_VLD2dup>;
// ...with double-spaced registers (not used for codegen):
-def VLD2DUPd8x2 : VLD2DUP<{0,0,1,?}, "8">;
-def VLD2DUPd16x2 : VLD2DUP<{0,1,1,?}, "16">;
-def VLD2DUPd32x2 : VLD2DUP<{1,0,1,?}, "32">;
+def VLD2DUPd8x2 : VLD2DUP<{0,0,1,?}, "8", VecListTwoQAllLanes>;
+def VLD2DUPd16x2 : VLD2DUP<{0,1,1,?}, "16", VecListTwoQAllLanes>;
+def VLD2DUPd32x2 : VLD2DUP<{1,0,1,?}, "32", VecListTwoQAllLanes>;
// ...with address register writeback:
-class VLD2DUPWB<bits<4> op7_4, string Dt>
- : NLdSt<1, 0b10, 0b1101, op7_4, (outs DPR:$Vd, DPR:$dst2, GPR:$wb),
- (ins addrmode6dup:$Rn, am6offset:$Rm), IIC_VLD2dupu,
- "vld2", Dt, "\\{$Vd[], $dst2[]\\}, $Rn$Rm", "$Rn.addr = $wb", []> {
- let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVLD2DupInstruction";
+multiclass VLD2DUPWB<bits<4> op7_4, string Dt, RegisterOperand VdTy> {
+ def _fixed : NLdSt<1, 0b10, 0b1101, op7_4,
+ (outs VdTy:$Vd, GPR:$wb),
+ (ins addrmode6dup:$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:$Rm), IIC_VLD2dupu,
+ "vld2", Dt, "$Vd, $Rn, $Rm",
+ "$Rn.addr = $wb", []> {
+ let Inst{4} = Rn{4};
+ let DecoderMethod = "DecodeVLD2DupInstruction";
+ let AsmMatchConverter = "cvtVLDwbRegister";
+ }
}
-def VLD2DUPd8_UPD : VLD2DUPWB<{0,0,0,0}, "8">;
-def VLD2DUPd16_UPD : VLD2DUPWB<{0,1,0,?}, "16">;
-def VLD2DUPd32_UPD : VLD2DUPWB<{1,0,0,?}, "32">;
+defm VLD2DUPd8wb : VLD2DUPWB<{0,0,0,0}, "8", VecListTwoDAllLanes>;
+defm VLD2DUPd16wb : VLD2DUPWB<{0,1,0,?}, "16", VecListTwoDAllLanes>;
+defm VLD2DUPd32wb : VLD2DUPWB<{1,0,0,?}, "32", VecListTwoDAllLanes>;
-def VLD2DUPd8x2_UPD : VLD2DUPWB<{0,0,1,0}, "8">;
-def VLD2DUPd16x2_UPD : VLD2DUPWB<{0,1,1,?}, "16">;
-def VLD2DUPd32x2_UPD : VLD2DUPWB<{1,0,1,?}, "32">;
+defm VLD2DUPd8x2wb : VLD2DUPWB<{0,0,1,0}, "8", VecListTwoQAllLanes>;
+defm VLD2DUPd16x2wb : VLD2DUPWB<{0,1,1,?}, "16", VecListTwoQAllLanes>;
+defm VLD2DUPd32x2wb : VLD2DUPWB<{1,0,1,?}, "32", VecListTwoQAllLanes>;
-def VLD2DUPd8Pseudo_UPD : VLDQWBPseudo<IIC_VLD2dupu>;
-def VLD2DUPd16Pseudo_UPD : VLDQWBPseudo<IIC_VLD2dupu>;
-def VLD2DUPd32Pseudo_UPD : VLDQWBPseudo<IIC_VLD2dupu>;
+def VLD2DUPd8PseudoWB_fixed : VLDQWBfixedPseudo <IIC_VLD2dupu>;
+def VLD2DUPd8PseudoWB_register : VLDQWBregisterPseudo<IIC_VLD2dupu>;
+def VLD2DUPd16PseudoWB_fixed : VLDQWBfixedPseudo <IIC_VLD2dupu>;
+def VLD2DUPd16PseudoWB_register : VLDQWBregisterPseudo<IIC_VLD2dupu>;
+def VLD2DUPd32PseudoWB_fixed : VLDQWBfixedPseudo <IIC_VLD2dupu>;
+def VLD2DUPd32PseudoWB_register : VLDQWBregisterPseudo<IIC_VLD2dupu>;
// VLD3DUP : Vector Load (single 3-element structure to all lanes)
class VLD3DUP<bits<4> op7_4, string Dt>
// ...with 3 registers
class VST1D3<bits<4> op7_4, string Dt>
: NLdSt<0, 0b00, 0b0110, op7_4, (outs),
- (ins addrmode6:$Rn, DPR:$Vd, DPR:$src2, DPR:$src3),
- IIC_VST1x3, "vst1", Dt, "\\{$Vd, $src2, $src3\\}, $Rn", "", []> {
+ (ins addrmode6:$Rn, VecListThreeD:$Vd),
+ IIC_VST1x3, "vst1", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVSTInstruction";
}
-class VST1D3WB<bits<4> op7_4, string Dt>
- : NLdSt<0, 0b00, 0b0110, op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, am6offset:$Rm,
- DPR:$Vd, DPR:$src2, DPR:$src3),
- IIC_VST1x3u, "vst1", Dt, "\\{$Vd, $src2, $src3\\}, $Rn$Rm",
- "$Rn.addr = $wb", []> {
- let Inst{4} = Rn{4};
- let DecoderMethod = "DecodeVSTInstruction";
+multiclass VST1D3WB<bits<4> op7_4, string Dt> {
+ def _fixed : NLdSt<0,0b00,0b0110,op7_4, (outs GPR:$wb),
+ (ins addrmode6:$Rn, VecListThreeD:$Vd), IIC_VLD1x3u,
+ "vst1", Dt, "$Vd, $Rn!",
+ "$Rn.addr = $wb", []> {
+ let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
+ let Inst{5-4} = Rn{5-4};
+ let DecoderMethod = "DecodeVSTInstruction";
+ let AsmMatchConverter = "cvtVSTwbFixed";
+ }
+ def _register : NLdSt<0,0b00,0b0110,op7_4, (outs GPR:$wb),
+ (ins addrmode6:$Rn, rGPR:$Rm, VecListThreeD:$Vd),
+ IIC_VLD1x3u,
+ "vst1", Dt, "$Vd, $Rn, $Rm",
+ "$Rn.addr = $wb", []> {
+ let Inst{5-4} = Rn{5-4};
+ let DecoderMethod = "DecodeVSTInstruction";
+ let AsmMatchConverter = "cvtVSTwbRegister";
+ }
}
-def VST1d8T : VST1D3<{0,0,0,?}, "8">;
-def VST1d16T : VST1D3<{0,1,0,?}, "16">;
-def VST1d32T : VST1D3<{1,0,0,?}, "32">;
-def VST1d64T : VST1D3<{1,1,0,?}, "64">;
+def VST1d8T : VST1D3<{0,0,0,?}, "8">;
+def VST1d16T : VST1D3<{0,1,0,?}, "16">;
+def VST1d32T : VST1D3<{1,0,0,?}, "32">;
+def VST1d64T : VST1D3<{1,1,0,?}, "64">;
-def VST1d8T_UPD : VST1D3WB<{0,0,0,?}, "8">;
-def VST1d16T_UPD : VST1D3WB<{0,1,0,?}, "16">;
-def VST1d32T_UPD : VST1D3WB<{1,0,0,?}, "32">;
-def VST1d64T_UPD : VST1D3WB<{1,1,0,?}, "64">;
+defm VST1d8Twb : VST1D3WB<{0,0,0,?}, "8">;
+defm VST1d16Twb : VST1D3WB<{0,1,0,?}, "16">;
+defm VST1d32Twb : VST1D3WB<{1,0,0,?}, "32">;
+defm VST1d64Twb : VST1D3WB<{1,1,0,?}, "64">;
-def VST1d64TPseudo : VSTQQPseudo<IIC_VST1x3>;
-def VST1d64TPseudo_UPD : VSTQQWBPseudo<IIC_VST1x3u>;
+def VST1d64TPseudo : VSTQQPseudo<IIC_VST1x3>;
+def VST1d64TPseudoWB_fixed : VSTQQWBPseudo<IIC_VST1x3u>;
+def VST1d64TPseudoWB_register : VSTQQWBPseudo<IIC_VST1x3u>;
// ...with 4 registers
class VST1D4<bits<4> op7_4, string Dt>
: NLdSt<0, 0b00, 0b0010, op7_4, (outs),
- (ins addrmode6:$Rn, DPR:$Vd, DPR:$src2, DPR:$src3, DPR:$src4),
- IIC_VST1x4, "vst1", Dt, "\\{$Vd, $src2, $src3, $src4\\}, $Rn", "",
+ (ins addrmode6:$Rn, VecListFourD:$Vd),
+ IIC_VST1x4, "vst1", Dt, "$Vd, $Rn", "",
[]> {
let Rm = 0b1111;
let Inst{5-4} = Rn{5-4};
let DecoderMethod = "DecodeVSTInstruction";
}
-class VST1D4WB<bits<4> op7_4, string Dt>
- : NLdSt<0, 0b00, 0b0010, op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, am6offset:$Rm,
- DPR:$Vd, DPR:$src2, DPR:$src3, DPR:$src4), IIC_VST1x4u,
- "vst1", Dt, "\\{$Vd, $src2, $src3, $src4\\}, $Rn$Rm",
- "$Rn.addr = $wb", []> {
- let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
+multiclass VST1D4WB<bits<4> op7_4, string Dt> {
+ def _fixed : NLdSt<0,0b00,0b0010,op7_4, (outs GPR:$wb),
+ (ins addrmode6:$Rn, VecListFourD:$Vd), IIC_VLD1x4u,
+ "vst1", Dt, "$Vd, $Rn!",
+ "$Rn.addr = $wb", []> {
+ let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
+ let Inst{5-4} = Rn{5-4};
+ let DecoderMethod = "DecodeVSTInstruction";
+ let AsmMatchConverter = "cvtVSTwbFixed";
+ }
+ def _register : NLdSt<0,0b00,0b0010,op7_4, (outs GPR:$wb),
+ (ins addrmode6:$Rn, rGPR:$Rm, VecListFourD:$Vd),
+ IIC_VLD1x4u,
+ "vst1", Dt, "$Vd, $Rn, $Rm",
+ "$Rn.addr = $wb", []> {
+ let Inst{5-4} = Rn{5-4};
+ let DecoderMethod = "DecodeVSTInstruction";
+ let AsmMatchConverter = "cvtVSTwbRegister";
+ }
}
-def VST1d8Q : VST1D4<{0,0,?,?}, "8">;
-def VST1d16Q : VST1D4<{0,1,?,?}, "16">;
-def VST1d32Q : VST1D4<{1,0,?,?}, "32">;
-def VST1d64Q : VST1D4<{1,1,?,?}, "64">;
+def VST1d8Q : VST1D4<{0,0,?,?}, "8">;
+def VST1d16Q : VST1D4<{0,1,?,?}, "16">;
+def VST1d32Q : VST1D4<{1,0,?,?}, "32">;
+def VST1d64Q : VST1D4<{1,1,?,?}, "64">;
-def VST1d8Q_UPD : VST1D4WB<{0,0,?,?}, "8">;
-def VST1d16Q_UPD : VST1D4WB<{0,1,?,?}, "16">;
-def VST1d32Q_UPD : VST1D4WB<{1,0,?,?}, "32">;
-def VST1d64Q_UPD : VST1D4WB<{1,1,?,?}, "64">;
+defm VST1d8Qwb : VST1D4WB<{0,0,?,?}, "8">;
+defm VST1d16Qwb : VST1D4WB<{0,1,?,?}, "16">;
+defm VST1d32Qwb : VST1D4WB<{1,0,?,?}, "32">;
+defm VST1d64Qwb : VST1D4WB<{1,1,?,?}, "64">;
-def VST1d64QPseudo : VSTQQPseudo<IIC_VST1x4>;
-def VST1d64QPseudo_UPD : VSTQQWBPseudo<IIC_VST1x4u>;
+def VST1d64QPseudo : VSTQQPseudo<IIC_VST1x4>;
+def VST1d64QPseudoWB_fixed : VSTQQWBPseudo<IIC_VST1x4u>;
+def VST1d64QPseudoWB_register : VSTQQWBPseudo<IIC_VST1x4u>;
// VST2 : Vector Store (multiple 2-element structures)
-class VST2D<bits<4> op11_8, bits<4> op7_4, string Dt>
- : NLdSt<0, 0b00, op11_8, op7_4, (outs),
- (ins addrmode6:$Rn, DPR:$Vd, DPR:$src2),
- IIC_VST2, "vst2", Dt, "\\{$Vd, $src2\\}, $Rn", "", []> {
- let Rm = 0b1111;
- let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
-}
-class VST2Q<bits<4> op7_4, string Dt>
- : NLdSt<0, 0b00, 0b0011, op7_4, (outs),
- (ins addrmode6:$Rn, DPR:$Vd, DPR:$src2, DPR:$src3, DPR:$src4),
- IIC_VST2x2, "vst2", Dt, "\\{$Vd, $src2, $src3, $src4\\}, $Rn",
- "", []> {
+class VST2<bits<4> op11_8, bits<4> op7_4, string Dt, RegisterOperand VdTy,
+ InstrItinClass itin>
+ : NLdSt<0, 0b00, op11_8, op7_4, (outs), (ins addrmode6:$Rn, VdTy:$Vd),
+ itin, "vst2", Dt, "$Vd, $Rn", "", []> {
let Rm = 0b1111;
let Inst{5-4} = Rn{5-4};
let DecoderMethod = "DecodeVSTInstruction";
}
-def VST2d8 : VST2D<0b1000, {0,0,?,?}, "8">;
-def VST2d16 : VST2D<0b1000, {0,1,?,?}, "16">;
-def VST2d32 : VST2D<0b1000, {1,0,?,?}, "32">;
+def VST2d8 : VST2<0b1000, {0,0,?,?}, "8", VecListTwoD, IIC_VST2>;
+def VST2d16 : VST2<0b1000, {0,1,?,?}, "16", VecListTwoD, IIC_VST2>;
+def VST2d32 : VST2<0b1000, {1,0,?,?}, "32", VecListTwoD, IIC_VST2>;
-def VST2q8 : VST2Q<{0,0,?,?}, "8">;
-def VST2q16 : VST2Q<{0,1,?,?}, "16">;
-def VST2q32 : VST2Q<{1,0,?,?}, "32">;
+def VST2q8 : VST2<0b0011, {0,0,?,?}, "8", VecListFourD, IIC_VST2x2>;
+def VST2q16 : VST2<0b0011, {0,1,?,?}, "16", VecListFourD, IIC_VST2x2>;
+def VST2q32 : VST2<0b0011, {1,0,?,?}, "32", VecListFourD, IIC_VST2x2>;
def VST2d8Pseudo : VSTQPseudo<IIC_VST2>;
def VST2d16Pseudo : VSTQPseudo<IIC_VST2>;
def VST2q32Pseudo : VSTQQPseudo<IIC_VST2x2>;
// ...with address register writeback:
-class VST2DWB<bits<4> op11_8, bits<4> op7_4, string Dt>
- : NLdSt<0, 0b00, op11_8, op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, am6offset:$Rm, DPR:$Vd, DPR:$src2),
- IIC_VST2u, "vst2", Dt, "\\{$Vd, $src2\\}, $Rn$Rm",
- "$Rn.addr = $wb", []> {
- let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
+multiclass VST2DWB<bits<4> op11_8, bits<4> op7_4, string Dt,
+ RegisterOperand VdTy> {
+ def _fixed : NLdSt<0, 0b00, op11_8, op7_4, (outs GPR:$wb),
+ (ins addrmode6:$Rn, VdTy:$Vd), IIC_VLD1u,
+ "vst2", Dt, "$Vd, $Rn!",
+ "$Rn.addr = $wb", []> {
+ let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
+ let Inst{5-4} = Rn{5-4};
+ let DecoderMethod = "DecodeVSTInstruction";
+ let AsmMatchConverter = "cvtVSTwbFixed";
+ }
+ def _register : NLdSt<0, 0b00, op11_8, op7_4, (outs GPR:$wb),
+ (ins addrmode6:$Rn, rGPR:$Rm, VdTy:$Vd), IIC_VLD1u,
+ "vst2", Dt, "$Vd, $Rn, $Rm",
+ "$Rn.addr = $wb", []> {
+ let Inst{5-4} = Rn{5-4};
+ let DecoderMethod = "DecodeVSTInstruction";
+ let AsmMatchConverter = "cvtVSTwbRegister";
+ }
}
-class VST2QWB<bits<4> op7_4, string Dt>
- : NLdSt<0, 0b00, 0b0011, op7_4, (outs GPR:$wb),
- (ins addrmode6:$Rn, am6offset:$Rm,
- DPR:$Vd, DPR:$src2, DPR:$src3, DPR:$src4), IIC_VST2x2u,
- "vst2", Dt, "\\{$Vd, $src2, $src3, $src4\\}, $Rn$Rm",
- "$Rn.addr = $wb", []> {
- let Inst{5-4} = Rn{5-4};
- let DecoderMethod = "DecodeVSTInstruction";
+multiclass VST2QWB<bits<4> op7_4, string Dt> {
+ def _fixed : NLdSt<0, 0b00, 0b0011, op7_4, (outs GPR:$wb),
+ (ins addrmode6:$Rn, VecListFourD:$Vd), IIC_VLD1u,
+ "vst2", Dt, "$Vd, $Rn!",
+ "$Rn.addr = $wb", []> {
+ let Rm = 0b1101; // NLdSt will assign to the right encoding bits.
+ let Inst{5-4} = Rn{5-4};
+ let DecoderMethod = "DecodeVSTInstruction";
+ let AsmMatchConverter = "cvtVSTwbFixed";
+ }
+ def _register : NLdSt<0, 0b00, 0b0011, op7_4, (outs GPR:$wb),
+ (ins addrmode6:$Rn, rGPR:$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";
+ }
}
-def VST2d8_UPD : VST2DWB<0b1000, {0,0,?,?}, "8">;
-def VST2d16_UPD : VST2DWB<0b1000, {0,1,?,?}, "16">;
-def VST2d32_UPD : VST2DWB<0b1000, {1,0,?,?}, "32">;
+defm VST2d8wb : VST2DWB<0b1000, {0,0,?,?}, "8", VecListTwoD>;
+defm VST2d16wb : VST2DWB<0b1000, {0,1,?,?}, "16", VecListTwoD>;
+defm VST2d32wb : VST2DWB<0b1000, {1,0,?,?}, "32", VecListTwoD>;
-def VST2q8_UPD : VST2QWB<{0,0,?,?}, "8">;
-def VST2q16_UPD : VST2QWB<{0,1,?,?}, "16">;
-def VST2q32_UPD : VST2QWB<{1,0,?,?}, "32">;
+defm VST2q8wb : VST2QWB<{0,0,?,?}, "8">;
+defm VST2q16wb : VST2QWB<{0,1,?,?}, "16">;
+defm VST2q32wb : VST2QWB<{1,0,?,?}, "32">;
-def VST2d8Pseudo_UPD : VSTQWBPseudo<IIC_VST2u>;
-def VST2d16Pseudo_UPD : VSTQWBPseudo<IIC_VST2u>;
-def VST2d32Pseudo_UPD : VSTQWBPseudo<IIC_VST2u>;
+def VST2d8PseudoWB_fixed : VSTQWBPseudo<IIC_VST2u>;
+def VST2d16PseudoWB_fixed : VSTQWBPseudo<IIC_VST2u>;
+def VST2d32PseudoWB_fixed : VSTQWBPseudo<IIC_VST2u>;
+def VST2d8PseudoWB_register : VSTQWBPseudo<IIC_VST2u>;
+def VST2d16PseudoWB_register : VSTQWBPseudo<IIC_VST2u>;
+def VST2d32PseudoWB_register : VSTQWBPseudo<IIC_VST2u>;
-def VST2q8Pseudo_UPD : VSTQQWBPseudo<IIC_VST2x2u>;
-def VST2q16Pseudo_UPD : VSTQQWBPseudo<IIC_VST2x2u>;
-def VST2q32Pseudo_UPD : VSTQQWBPseudo<IIC_VST2x2u>;
+def VST2q8PseudoWB_fixed : VSTQQWBPseudo<IIC_VST2x2u>;
+def VST2q16PseudoWB_fixed : VSTQQWBPseudo<IIC_VST2x2u>;
+def VST2q32PseudoWB_fixed : VSTQQWBPseudo<IIC_VST2x2u>;
+def VST2q8PseudoWB_register : VSTQQWBPseudo<IIC_VST2x2u>;
+def VST2q16PseudoWB_register : VSTQQWBPseudo<IIC_VST2x2u>;
+def VST2q32PseudoWB_register : VSTQQWBPseudo<IIC_VST2x2u>;
// ...with double-spaced registers
-def VST2b8 : VST2D<0b1001, {0,0,?,?}, "8">;
-def VST2b16 : VST2D<0b1001, {0,1,?,?}, "16">;
-def VST2b32 : VST2D<0b1001, {1,0,?,?}, "32">;
-def VST2b8_UPD : VST2DWB<0b1001, {0,0,?,?}, "8">;
-def VST2b16_UPD : VST2DWB<0b1001, {0,1,?,?}, "16">;
-def VST2b32_UPD : VST2DWB<0b1001, {1,0,?,?}, "32">;
+def VST2b8 : VST2<0b1001, {0,0,?,?}, "8", VecListTwoQ, IIC_VST2>;
+def VST2b16 : VST2<0b1001, {0,1,?,?}, "16", VecListTwoQ, IIC_VST2>;
+def VST2b32 : VST2<0b1001, {1,0,?,?}, "32", VecListTwoQ, IIC_VST2>;
+defm VST2b8wb : VST2DWB<0b1001, {0,0,?,?}, "8", VecListTwoQ>;
+defm VST2b16wb : VST2DWB<0b1001, {0,1,?,?}, "16", VecListTwoQ>;
+defm VST2b32wb : VST2DWB<0b1001, {1,0,?,?}, "32", VecListTwoQ>;
// VST3 : Vector Store (multiple 3-element structures)
class VST3D<bits<4> op11_8, bits<4> op7_4, string Dt>
// ...with address register writeback:
class VST2LNWB<bits<4> op11_8, bits<4> op7_4, string Dt>
: NLdStLn<1, 0b00, op11_8, op7_4, (outs GPR:$wb),
- (ins addrmode6:$addr, am6offset:$offset,
- DPR:$src1, DPR:$src2, nohash_imm:$lane), IIC_VST2lnu, "vst2", Dt,
- "\\{$src1[$lane], $src2[$lane]\\}, $addr$offset",
- "$addr.addr = $wb", []> {
+ (ins addrmode6:$Rn, am6offset:$Rm,
+ DPR:$Vd, DPR:$src2, nohash_imm:$lane), IIC_VST2lnu, "vst2", Dt,
+ "\\{$Vd[$lane], $src2[$lane]\\}, $Rn$Rm",
+ "$Rn.addr = $wb", []> {
let Inst{4} = Rn{4};
let DecoderMethod = "DecodeVST2LN";
}
// Long shift by immediate.
class N2VLSh<bit op24, bit op23, bits<4> op11_8, bit op7, bit op6, bit op4,
string OpcodeStr, string Dt,
- ValueType ResTy, ValueType OpTy, SDNode OpNode>
+ ValueType ResTy, ValueType OpTy, Operand ImmTy, SDNode OpNode>
: N2VImm<op24, op23, op11_8, op7, op6, op4,
- (outs QPR:$Vd), (ins DPR:$Vm, i32imm:$SIMM), N2RegVShLFrm,
+ (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))))]>;
v4i32, v4i32, OpNode, Commutable>;
}
-multiclass N3VSL_HS<bits<4> op11_8, string OpcodeStr, string Dt, SDNode ShOp> {
- def v4i16 : N3VDSL16<0b01, op11_8, OpcodeStr, !strconcat(Dt, "16"),
- v4i16, ShOp>;
- def v2i32 : N3VDSL<0b10, op11_8, IIC_VMULi32D, OpcodeStr, !strconcat(Dt,"32"),
- v2i32, ShOp>;
- def v8i16 : N3VQSL16<0b01, op11_8, OpcodeStr, !strconcat(Dt, "16"),
- v8i16, v4i16, ShOp>;
- def v4i32 : N3VQSL<0b10, op11_8, IIC_VMULi32Q, OpcodeStr, !strconcat(Dt,"32"),
+multiclass N3VSL_HS<bits<4> op11_8, string OpcodeStr, SDNode ShOp> {
+ def v4i16 : N3VDSL16<0b01, op11_8, OpcodeStr, "i16", v4i16, ShOp>;
+ def v2i32 : N3VDSL<0b10, op11_8, IIC_VMULi32D, OpcodeStr, "i32", v2i32, ShOp>;
+ def v8i16 : N3VQSL16<0b01, op11_8, OpcodeStr, "i16", v8i16, v4i16, ShOp>;
+ def v4i32 : N3VQSL<0b10, op11_8, IIC_VMULi32Q, OpcodeStr, "i32",
v4i32, v2i32, ShOp>;
}
multiclass N2VLSh_QHS<bit op24, bit op23, bits<4> op11_8, bit op7, bit op6,
bit op4, string OpcodeStr, string Dt, SDNode OpNode> {
def v8i16 : N2VLSh<op24, op23, op11_8, op7, op6, op4,
- OpcodeStr, !strconcat(Dt, "8"), v8i16, v8i8, OpNode> {
+ OpcodeStr, !strconcat(Dt, "8"), v8i16, v8i8, imm1_7, OpNode> {
let Inst{21-19} = 0b001; // imm6 = 001xxx
}
def v4i32 : N2VLSh<op24, op23, op11_8, op7, op6, op4,
- OpcodeStr, !strconcat(Dt, "16"), v4i32, v4i16, OpNode> {
+ OpcodeStr, !strconcat(Dt, "16"), v4i32, v4i16, imm1_15, OpNode> {
let Inst{21-20} = 0b01; // imm6 = 01xxxx
}
def v2i64 : N2VLSh<op24, op23, op11_8, op7, op6, op4,
- OpcodeStr, !strconcat(Dt, "32"), v2i64, v2i32, OpNode> {
+ OpcodeStr, !strconcat(Dt, "32"), v2i64, v2i32, imm1_31, OpNode> {
let Inst{21} = 0b1; // imm6 = 1xxxxx
}
}
v2f32, v2f32, fmul, 1>;
def VMULfq : N3VQ<1, 0, 0b00, 0b1101, 1, IIC_VFMULQ, "vmul", "f32",
v4f32, v4f32, fmul, 1>;
-defm VMULsl : N3VSL_HS<0b1000, "vmul", "i", mul>;
+defm VMULsl : N3VSL_HS<0b1000, "vmul", mul>;
def VMULslfd : N3VDSL<0b10, 0b1001, IIC_VBIND, "vmul", "f32", v2f32, fmul>;
def VMULslfq : N3VQSL<0b10, 0b1001, IIC_VBINQ, "vmul", "f32", v4f32,
v2f32, fmul>;
// 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, SDNode OpNode>
+ ValueType OpTy, Operand ImmTy, SDNode OpNode>
: N2VLSh<op24, op23, op11_8, op7, op6, op4, OpcodeStr, Dt,
- ResTy, OpTy, OpNode> {
+ ResTy, OpTy, ImmTy, OpNode> {
let Inst{21-16} = op21_16;
let DecoderMethod = "DecodeVSHLMaxInstruction";
}
def VSHLLi8 : N2VLShMax<1, 1, 0b110010, 0b0011, 0, 0, 0, "vshll", "i8",
- v8i16, v8i8, NEONvshlli>;
+ v8i16, v8i8, imm8, NEONvshlli>;
def VSHLLi16 : N2VLShMax<1, 1, 0b110110, 0b0011, 0, 0, 0, "vshll", "i16",
- v4i32, v4i16, NEONvshlli>;
+ v4i32, v4i16, imm16, NEONvshlli>;
def VSHLLi32 : N2VLShMax<1, 1, 0b111010, 0b0011, 0, 0, 0, "vshll", "i32",
- v2i64, v2i32, NEONvshlli>;
+ v2i64, v2i32, imm32, NEONvshlli>;
// VSHRN : Vector Shift Right and Narrow
defm VSHRN : N2VNSh_HSD<0,1,0b1000,0,0,1, IIC_VSHLiD, "vshrn", "i",
(ins nImmSplatI64:$SIMM), IIC_VMOVImm,
"vmov", "i64", "$Vd, $SIMM", "",
[(set QPR:$Vd, (v2i64 (NEONvmovImm timm:$SIMM)))]>;
+
+def VMOVv2f32 : N1ModImm<1, 0b000, 0b1111, 0, 0, 0, 1, (outs DPR:$Vd),
+ (ins nImmVMOVF32:$SIMM), IIC_VMOVImm,
+ "vmov", "f32", "$Vd, $SIMM", "",
+ [(set DPR:$Vd, (v2f32 (NEONvmovFPImm timm:$SIMM)))]>;
+def VMOVv4f32 : N1ModImm<1, 0b000, 0b1111, 0, 1, 0, 1, (outs QPR:$Vd),
+ (ins nImmVMOVF32:$SIMM), IIC_VMOVImm,
+ "vmov", "f32", "$Vd, $SIMM", "",
+ [(set QPR:$Vd, (v4f32 (NEONvmovFPImm timm:$SIMM)))]>;
} // isReMaterializable
// VMOV : Vector Get Lane (move scalar to ARM core register)
v4f32, v4i32, uint_to_fp>;
// VCVT : Vector Convert Between Floating-Point and Fixed-Point.
+let DecoderMethod = "DecodeVCVTD" in {
def VCVTf2xsd : N2VCvtD<0, 1, 0b1111, 0, 1, "vcvt", "s32.f32",
v2i32, v2f32, int_arm_neon_vcvtfp2fxs>;
def VCVTf2xud : N2VCvtD<1, 1, 0b1111, 0, 1, "vcvt", "u32.f32",
v2f32, v2i32, int_arm_neon_vcvtfxs2fp>;
def VCVTxu2fd : N2VCvtD<1, 1, 0b1110, 0, 1, "vcvt", "f32.u32",
v2f32, v2i32, int_arm_neon_vcvtfxu2fp>;
+}
+let DecoderMethod = "DecodeVCVTQ" in {
def VCVTf2xsq : N2VCvtQ<0, 1, 0b1111, 0, 1, "vcvt", "s32.f32",
v4i32, v4f32, int_arm_neon_vcvtfp2fxs>;
def VCVTf2xuq : N2VCvtQ<1, 1, 0b1111, 0, 1, "vcvt", "u32.f32",
v4f32, v4i32, int_arm_neon_vcvtfxs2fp>;
def VCVTxu2fq : N2VCvtQ<1, 1, 0b1110, 0, 1, "vcvt", "f32.u32",
v4f32, v4i32, int_arm_neon_vcvtfxu2fp>;
+}
// VCVT : Vector Convert Between Half-Precision and Single-Precision.
def VCVTf2h : N2VNInt<0b11, 0b11, 0b01, 0b10, 0b01100, 0, 0,
// VEXT : Vector Extract
-class VEXTd<string OpcodeStr, string Dt, ValueType Ty>
+class VEXTd<string OpcodeStr, string Dt, ValueType Ty, Operand immTy>
: N3V<0,1,0b11,{?,?,?,?},0,0, (outs DPR:$Vd),
- (ins DPR:$Vn, DPR:$Vm, i32imm:$index), NVExtFrm,
+ (ins DPR:$Vn, DPR:$Vm, immTy:$index), NVExtFrm,
IIC_VEXTD, OpcodeStr, Dt, "$Vd, $Vn, $Vm, $index", "",
[(set DPR:$Vd, (Ty (NEONvext (Ty DPR:$Vn),
- (Ty DPR:$Vm), imm:$index)))]> {
+ (Ty DPR:$Vm), imm:$index)))]> {
bits<4> index;
let Inst{11-8} = index{3-0};
}
-class VEXTq<string OpcodeStr, string Dt, ValueType Ty>
+class VEXTq<string OpcodeStr, string Dt, ValueType Ty, Operand immTy>
: N3V<0,1,0b11,{?,?,?,?},1,0, (outs QPR:$Vd),
- (ins QPR:$Vn, QPR:$Vm, i32imm:$index), NVExtFrm,
+ (ins QPR:$Vn, QPR:$Vm, imm0_15:$index), NVExtFrm,
IIC_VEXTQ, OpcodeStr, Dt, "$Vd, $Vn, $Vm, $index", "",
[(set QPR:$Vd, (Ty (NEONvext (Ty QPR:$Vn),
- (Ty QPR:$Vm), imm:$index)))]> {
+ (Ty QPR:$Vm), imm:$index)))]> {
bits<4> index;
let Inst{11-8} = index{3-0};
}
-def VEXTd8 : VEXTd<"vext", "8", v8i8> {
+def VEXTd8 : VEXTd<"vext", "8", v8i8, imm0_7> {
let Inst{11-8} = index{3-0};
}
-def VEXTd16 : VEXTd<"vext", "16", v4i16> {
+def VEXTd16 : VEXTd<"vext", "16", v4i16, imm0_3> {
let Inst{11-9} = index{2-0};
let Inst{8} = 0b0;
}
-def VEXTd32 : VEXTd<"vext", "32", v2i32> {
+def VEXTd32 : VEXTd<"vext", "32", v2i32, imm0_1> {
let Inst{11-10} = index{1-0};
let Inst{9-8} = 0b00;
}
(i32 imm:$index))),
(VEXTd32 DPR:$Vn, DPR:$Vm, imm:$index)>;
-def VEXTq8 : VEXTq<"vext", "8", v16i8> {
+def VEXTq8 : VEXTq<"vext", "8", v16i8, imm0_15> {
let Inst{11-8} = index{3-0};
}
-def VEXTq16 : VEXTq<"vext", "16", v8i16> {
+def VEXTq16 : VEXTq<"vext", "16", v8i16, imm0_7> {
let Inst{11-9} = index{2-0};
let Inst{8} = 0b0;
}
-def VEXTq32 : VEXTq<"vext", "32", v4i32> {
+def VEXTq32 : VEXTq<"vext", "32", v4i32, imm0_3> {
let Inst{11-10} = index{1-0};
let Inst{9-8} = 0b00;
}
+def VEXTq64 : VEXTq<"vext", "64", v2i64, imm0_1> {
+ let Inst{11} = index{0};
+ let Inst{10-8} = 0b000;
+}
def : Pat<(v4f32 (NEONvext (v4f32 QPR:$Vn),
(v4f32 QPR:$Vm),
(i32 imm:$index))),
let hasExtraSrcRegAllocReq = 1 in {
def VTBL2
: N3V<1,1,0b11,0b1001,0,0, (outs DPR:$Vd),
- (ins DPR:$Vn, DPR:$tbl2, DPR:$Vm), NVTBLFrm, IIC_VTB2,
- "vtbl", "8", "$Vd, \\{$Vn, $tbl2\\}, $Vm", "", []>;
+ (ins VecListTwoD:$Vn, DPR:$Vm), NVTBLFrm, IIC_VTB2,
+ "vtbl", "8", "$Vd, $Vn, $Vm", "", []>;
def VTBL3
: N3V<1,1,0b11,0b1010,0,0, (outs DPR:$Vd),
- (ins DPR:$Vn, DPR:$tbl2, DPR:$tbl3, DPR:$Vm), NVTBLFrm, IIC_VTB3,
- "vtbl", "8", "$Vd, \\{$Vn, $tbl2, $tbl3\\}, $Vm", "", []>;
+ (ins VecListThreeD:$Vn, DPR:$Vm), NVTBLFrm, IIC_VTB3,
+ "vtbl", "8", "$Vd, $Vn, $Vm", "", []>;
def VTBL4
: N3V<1,1,0b11,0b1011,0,0, (outs DPR:$Vd),
- (ins DPR:$Vn, DPR:$tbl2, DPR:$tbl3, DPR:$tbl4, DPR:$Vm),
+ (ins VecListFourD:$Vn, DPR:$Vm),
NVTBLFrm, IIC_VTB4,
- "vtbl", "8", "$Vd, \\{$Vn, $tbl2, $tbl3, $tbl4\\}, $Vm", "", []>;
+ "vtbl", "8", "$Vd, $Vn, $Vm", "", []>;
} // hasExtraSrcRegAllocReq = 1
def VTBL2Pseudo
let hasExtraSrcRegAllocReq = 1 in {
def VTBX2
: N3V<1,1,0b11,0b1001,1,0, (outs DPR:$Vd),
- (ins DPR:$orig, DPR:$Vn, DPR:$tbl2, DPR:$Vm), NVTBLFrm, IIC_VTBX2,
- "vtbx", "8", "$Vd, \\{$Vn, $tbl2\\}, $Vm", "$orig = $Vd", []>;
+ (ins DPR:$orig, VecListTwoD:$Vn, DPR:$Vm), NVTBLFrm, IIC_VTBX2,
+ "vtbx", "8", "$Vd, $Vn, $Vm", "$orig = $Vd", []>;
def VTBX3
: N3V<1,1,0b11,0b1010,1,0, (outs DPR:$Vd),
- (ins DPR:$orig, DPR:$Vn, DPR:$tbl2, DPR:$tbl3, DPR:$Vm),
+ (ins DPR:$orig, VecListThreeD:$Vn, DPR:$Vm),
NVTBLFrm, IIC_VTBX3,
- "vtbx", "8", "$Vd, \\{$Vn, $tbl2, $tbl3\\}, $Vm",
+ "vtbx", "8", "$Vd, $Vn, $Vm",
"$orig = $Vd", []>;
def VTBX4
- : N3V<1,1,0b11,0b1011,1,0, (outs DPR:$Vd), (ins DPR:$orig, DPR:$Vn,
- DPR:$tbl2, DPR:$tbl3, DPR:$tbl4, DPR:$Vm), NVTBLFrm, IIC_VTBX4,
- "vtbx", "8", "$Vd, \\{$Vn, $tbl2, $tbl3, $tbl4\\}, $Vm",
+ : N3V<1,1,0b11,0b1011,1,0, (outs DPR:$Vd),
+ (ins DPR:$orig, VecListFourD:$Vn, DPR:$Vm), NVTBLFrm, IIC_VTBX4,
+ "vtbx", "8", "$Vd, $Vn, $Vm",
"$orig = $Vd", []>;
} // hasExtraSrcRegAllocReq = 1
// Assembler aliases
//
-// VAND/VEOR/VORR accept but do not require a type suffix.
+def : VFP2InstAlias<"fmdhr${p} $Dd, $Rn",
+ (VSETLNi32 DPR:$Dd, GPR:$Rn, 1, pred:$p)>;
+def : VFP2InstAlias<"fmdlr${p} $Dd, $Rn",
+ (VSETLNi32 DPR:$Dd, GPR:$Rn, 0, pred:$p)>;
+
+
+// VADD two-operand aliases.
+def : NEONInstAlias<"vadd${p}.i8 $Vdn, $Vm",
+ (VADDv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vadd${p}.i16 $Vdn, $Vm",
+ (VADDv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vadd${p}.i32 $Vdn, $Vm",
+ (VADDv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vadd${p}.i64 $Vdn, $Vm",
+ (VADDv2i64 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vadd${p}.i8 $Vdn, $Vm",
+ (VADDv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vadd${p}.i16 $Vdn, $Vm",
+ (VADDv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vadd${p}.i32 $Vdn, $Vm",
+ (VADDv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vadd${p}.i64 $Vdn, $Vm",
+ (VADDv1i64 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vadd${p}.f32 $Vdn, $Vm",
+ (VADDfd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vadd${p}.f32 $Vdn, $Vm",
+ (VADDfq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+// VSUB two-operand aliases.
+def : NEONInstAlias<"vsub${p}.i8 $Vdn, $Vm",
+ (VSUBv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vsub${p}.i16 $Vdn, $Vm",
+ (VSUBv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vsub${p}.i32 $Vdn, $Vm",
+ (VSUBv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vsub${p}.i64 $Vdn, $Vm",
+ (VSUBv2i64 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vsub${p}.i8 $Vdn, $Vm",
+ (VSUBv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vsub${p}.i16 $Vdn, $Vm",
+ (VSUBv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vsub${p}.i32 $Vdn, $Vm",
+ (VSUBv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vsub${p}.i64 $Vdn, $Vm",
+ (VSUBv1i64 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vsub${p}.f32 $Vdn, $Vm",
+ (VSUBfd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vsub${p}.f32 $Vdn, $Vm",
+ (VSUBfq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+// VADDW two-operand aliases.
+def : NEONInstAlias<"vaddw${p}.s8 $Vdn, $Vm",
+ (VADDWsv8i16 QPR:$Vdn, QPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vaddw${p}.s16 $Vdn, $Vm",
+ (VADDWsv4i32 QPR:$Vdn, QPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vaddw${p}.s32 $Vdn, $Vm",
+ (VADDWsv2i64 QPR:$Vdn, QPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vaddw${p}.u8 $Vdn, $Vm",
+ (VADDWuv8i16 QPR:$Vdn, QPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vaddw${p}.u16 $Vdn, $Vm",
+ (VADDWuv4i32 QPR:$Vdn, QPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vaddw${p}.u32 $Vdn, $Vm",
+ (VADDWuv2i64 QPR:$Vdn, QPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+// VAND/VBIC/VEOR/VORR accept but do not require a type suffix.
defm : VFPDTAnyInstAlias<"vand${p}", "$Vd, $Vn, $Vm",
(VANDd DPR:$Vd, DPR:$Vn, DPR:$Vm, pred:$p)>;
defm : VFPDTAnyInstAlias<"vand${p}", "$Vd, $Vn, $Vm",
(VANDq QPR:$Vd, QPR:$Vn, QPR:$Vm, pred:$p)>;
+defm : VFPDTAnyInstAlias<"vbic${p}", "$Vd, $Vn, $Vm",
+ (VBICd DPR:$Vd, DPR:$Vn, DPR:$Vm, pred:$p)>;
+defm : VFPDTAnyInstAlias<"vbic${p}", "$Vd, $Vn, $Vm",
+ (VBICq QPR:$Vd, QPR:$Vn, QPR:$Vm, pred:$p)>;
defm : VFPDTAnyInstAlias<"veor${p}", "$Vd, $Vn, $Vm",
(VEORd DPR:$Vd, DPR:$Vn, DPR:$Vm, pred:$p)>;
defm : VFPDTAnyInstAlias<"veor${p}", "$Vd, $Vn, $Vm",
(VORRd DPR:$Vd, DPR:$Vn, DPR:$Vm, pred:$p)>;
defm : VFPDTAnyInstAlias<"vorr${p}", "$Vd, $Vn, $Vm",
(VORRq QPR:$Vd, QPR:$Vn, QPR:$Vm, pred:$p)>;
+// ... two-operand aliases
+def : NEONInstAlias<"vand${p} $Vdn, $Vm",
+ (VANDd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vand${p} $Vdn, $Vm",
+ (VANDq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vbic${p} $Vdn, $Vm",
+ (VBICd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vbic${p} $Vdn, $Vm",
+ (VBICq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"veor${p} $Vdn, $Vm",
+ (VEORd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"veor${p} $Vdn, $Vm",
+ (VEORq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vorr${p} $Vdn, $Vm",
+ (VORRd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vorr${p} $Vdn, $Vm",
+ (VORRq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+defm : VFPDTAnyInstAlias<"vand${p}", "$Vdn, $Vm",
+ (VANDd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+defm : VFPDTAnyInstAlias<"vand${p}", "$Vdn, $Vm",
+ (VANDq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+defm : VFPDTAnyInstAlias<"veor${p}", "$Vdn, $Vm",
+ (VEORd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+defm : VFPDTAnyInstAlias<"veor${p}", "$Vdn, $Vm",
+ (VEORq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+defm : VFPDTAnyInstAlias<"vorr${p}", "$Vdn, $Vm",
+ (VORRd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+defm : VFPDTAnyInstAlias<"vorr${p}", "$Vdn, $Vm",
+ (VORRq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+// VMUL two-operand aliases.
+def : NEONInstAlias<"vmul${p}.p8 $Qdn, $Qm",
+ (VMULpq QPR:$Qdn, QPR:$Qdn, QPR:$Qm, pred:$p)>;
+def : NEONInstAlias<"vmul${p}.i8 $Qdn, $Qm",
+ (VMULv16i8 QPR:$Qdn, QPR:$Qdn, QPR:$Qm, pred:$p)>;
+def : NEONInstAlias<"vmul${p}.i16 $Qdn, $Qm",
+ (VMULv8i16 QPR:$Qdn, QPR:$Qdn, QPR:$Qm, pred:$p)>;
+def : NEONInstAlias<"vmul${p}.i32 $Qdn, $Qm",
+ (VMULv4i32 QPR:$Qdn, QPR:$Qdn, QPR:$Qm, pred:$p)>;
+
+def : NEONInstAlias<"vmul${p}.p8 $Ddn, $Dm",
+ (VMULpd DPR:$Ddn, DPR:$Ddn, DPR:$Dm, pred:$p)>;
+def : NEONInstAlias<"vmul${p}.i8 $Ddn, $Dm",
+ (VMULv8i8 DPR:$Ddn, DPR:$Ddn, DPR:$Dm, pred:$p)>;
+def : NEONInstAlias<"vmul${p}.i16 $Ddn, $Dm",
+ (VMULv4i16 DPR:$Ddn, DPR:$Ddn, DPR:$Dm, pred:$p)>;
+def : NEONInstAlias<"vmul${p}.i32 $Ddn, $Dm",
+ (VMULv2i32 DPR:$Ddn, DPR:$Ddn, DPR:$Dm, pred:$p)>;
+
+def : NEONInstAlias<"vmul${p}.f32 $Qdn, $Qm",
+ (VMULfq QPR:$Qdn, QPR:$Qdn, QPR:$Qm, pred:$p)>;
+def : NEONInstAlias<"vmul${p}.f32 $Ddn, $Dm",
+ (VMULfd DPR:$Ddn, DPR:$Ddn, DPR:$Dm, pred:$p)>;
+
+def : NEONInstAlias<"vmul${p}.i16 $Ddn, $Dm$lane",
+ (VMULslv4i16 DPR:$Ddn, DPR:$Ddn, DPR_8:$Dm,
+ VectorIndex16:$lane, pred:$p)>;
+def : NEONInstAlias<"vmul${p}.i16 $Qdn, $Dm$lane",
+ (VMULslv8i16 QPR:$Qdn, QPR:$Qdn, DPR_8:$Dm,
+ VectorIndex16:$lane, pred:$p)>;
+
+def : NEONInstAlias<"vmul${p}.i32 $Ddn, $Dm$lane",
+ (VMULslv2i32 DPR:$Ddn, DPR:$Ddn, DPR_VFP2:$Dm,
+ VectorIndex32:$lane, pred:$p)>;
+def : NEONInstAlias<"vmul${p}.i32 $Qdn, $Dm$lane",
+ (VMULslv4i32 QPR:$Qdn, QPR:$Qdn, DPR_VFP2:$Dm,
+ VectorIndex32:$lane, pred:$p)>;
+
+def : NEONInstAlias<"vmul${p}.f32 $Ddn, $Dm$lane",
+ (VMULslfd DPR:$Ddn, DPR:$Ddn, DPR_VFP2:$Dm,
+ VectorIndex32:$lane, pred:$p)>;
+def : NEONInstAlias<"vmul${p}.f32 $Qdn, $Dm$lane",
+ (VMULslfq QPR:$Qdn, QPR:$Qdn, DPR_VFP2:$Dm,
+ VectorIndex32:$lane, pred:$p)>;
+
+// VQADD (register) two-operand aliases.
+def : NEONInstAlias<"vqadd${p}.s8 $Vdn, $Vm",
+ (VQADDsv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.s16 $Vdn, $Vm",
+ (VQADDsv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.s32 $Vdn, $Vm",
+ (VQADDsv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.s64 $Vdn, $Vm",
+ (VQADDsv1i64 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.u8 $Vdn, $Vm",
+ (VQADDuv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.u16 $Vdn, $Vm",
+ (VQADDuv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.u32 $Vdn, $Vm",
+ (VQADDuv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.u64 $Vdn, $Vm",
+ (VQADDuv1i64 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vqadd${p}.s8 $Vdn, $Vm",
+ (VQADDsv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.s16 $Vdn, $Vm",
+ (VQADDsv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.s32 $Vdn, $Vm",
+ (VQADDsv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.s64 $Vdn, $Vm",
+ (VQADDsv2i64 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.u8 $Vdn, $Vm",
+ (VQADDuv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.u16 $Vdn, $Vm",
+ (VQADDuv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.u32 $Vdn, $Vm",
+ (VQADDuv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqadd${p}.u64 $Vdn, $Vm",
+ (VQADDuv2i64 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+// VSHL (immediate) two-operand aliases.
+def : NEONInstAlias<"vshl${p}.i8 $Vdn, $imm",
+ (VSHLiv8i8 DPR:$Vdn, DPR:$Vdn, imm0_7:$imm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.i16 $Vdn, $imm",
+ (VSHLiv4i16 DPR:$Vdn, DPR:$Vdn, imm0_15:$imm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.i32 $Vdn, $imm",
+ (VSHLiv2i32 DPR:$Vdn, DPR:$Vdn, imm0_31:$imm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.i64 $Vdn, $imm",
+ (VSHLiv1i64 DPR:$Vdn, DPR:$Vdn, imm0_63:$imm, pred:$p)>;
+
+def : NEONInstAlias<"vshl${p}.i8 $Vdn, $imm",
+ (VSHLiv16i8 QPR:$Vdn, QPR:$Vdn, imm0_7:$imm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.i16 $Vdn, $imm",
+ (VSHLiv8i16 QPR:$Vdn, QPR:$Vdn, imm0_15:$imm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.i32 $Vdn, $imm",
+ (VSHLiv4i32 QPR:$Vdn, QPR:$Vdn, imm0_31:$imm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.i64 $Vdn, $imm",
+ (VSHLiv2i64 QPR:$Vdn, QPR:$Vdn, imm0_63:$imm, pred:$p)>;
+
+// VSHL (register) two-operand aliases.
+def : NEONInstAlias<"vshl${p}.s8 $Vdn, $Vm",
+ (VSHLsv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.s16 $Vdn, $Vm",
+ (VSHLsv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.s32 $Vdn, $Vm",
+ (VSHLsv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.s64 $Vdn, $Vm",
+ (VSHLsv1i64 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.u8 $Vdn, $Vm",
+ (VSHLuv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.u16 $Vdn, $Vm",
+ (VSHLuv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.u32 $Vdn, $Vm",
+ (VSHLuv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.u64 $Vdn, $Vm",
+ (VSHLuv1i64 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vshl${p}.s8 $Vdn, $Vm",
+ (VSHLsv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.s16 $Vdn, $Vm",
+ (VSHLsv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.s32 $Vdn, $Vm",
+ (VSHLsv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.s64 $Vdn, $Vm",
+ (VSHLsv2i64 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.u8 $Vdn, $Vm",
+ (VSHLuv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.u16 $Vdn, $Vm",
+ (VSHLuv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.u32 $Vdn, $Vm",
+ (VSHLuv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vshl${p}.u64 $Vdn, $Vm",
+ (VSHLuv2i64 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+// VSHL (immediate) two-operand aliases.
+def : NEONInstAlias<"vshr${p}.s8 $Vdn, $imm",
+ (VSHRsv8i8 DPR:$Vdn, DPR:$Vdn, shr_imm8:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.s16 $Vdn, $imm",
+ (VSHRsv4i16 DPR:$Vdn, DPR:$Vdn, shr_imm16:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.s32 $Vdn, $imm",
+ (VSHRsv2i32 DPR:$Vdn, DPR:$Vdn, shr_imm32:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.s64 $Vdn, $imm",
+ (VSHRsv1i64 DPR:$Vdn, DPR:$Vdn, shr_imm64:$imm, pred:$p)>;
+
+def : NEONInstAlias<"vshr${p}.s8 $Vdn, $imm",
+ (VSHRsv16i8 QPR:$Vdn, QPR:$Vdn, shr_imm8:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.s16 $Vdn, $imm",
+ (VSHRsv8i16 QPR:$Vdn, QPR:$Vdn, shr_imm16:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.s32 $Vdn, $imm",
+ (VSHRsv4i32 QPR:$Vdn, QPR:$Vdn, shr_imm32:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.s64 $Vdn, $imm",
+ (VSHRsv2i64 QPR:$Vdn, QPR:$Vdn, shr_imm64:$imm, pred:$p)>;
+
+def : NEONInstAlias<"vshr${p}.u8 $Vdn, $imm",
+ (VSHRuv8i8 DPR:$Vdn, DPR:$Vdn, shr_imm8:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.u16 $Vdn, $imm",
+ (VSHRuv4i16 DPR:$Vdn, DPR:$Vdn, shr_imm16:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.u32 $Vdn, $imm",
+ (VSHRuv2i32 DPR:$Vdn, DPR:$Vdn, shr_imm32:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.u64 $Vdn, $imm",
+ (VSHRuv1i64 DPR:$Vdn, DPR:$Vdn, shr_imm64:$imm, pred:$p)>;
+
+def : NEONInstAlias<"vshr${p}.u8 $Vdn, $imm",
+ (VSHRuv16i8 QPR:$Vdn, QPR:$Vdn, shr_imm8:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.u16 $Vdn, $imm",
+ (VSHRuv8i16 QPR:$Vdn, QPR:$Vdn, shr_imm16:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.u32 $Vdn, $imm",
+ (VSHRuv4i32 QPR:$Vdn, QPR:$Vdn, shr_imm32:$imm, pred:$p)>;
+def : NEONInstAlias<"vshr${p}.u64 $Vdn, $imm",
+ (VSHRuv2i64 QPR:$Vdn, QPR:$Vdn, shr_imm64:$imm, pred:$p)>;
+
+// VLD1 single-lane pseudo-instructions. These need special handling for
+// the lane index that an InstAlias can't handle, so we use these instead.
+defm VLD1LNdAsm : NEONDT8AsmPseudoInst<"vld1${p}", "$list, $addr",
+ (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD1LNdAsm : NEONDT16AsmPseudoInst<"vld1${p}", "$list, $addr",
+ (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD1LNdAsm : NEONDT32AsmPseudoInst<"vld1${p}", "$list, $addr",
+ (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+
+defm VLD1LNdWB_fixed_Asm : NEONDT8AsmPseudoInst<"vld1${p}", "$list, $addr!",
+ (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD1LNdWB_fixed_Asm : NEONDT16AsmPseudoInst<"vld1${p}", "$list, $addr!",
+ (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD1LNdWB_fixed_Asm : NEONDT32AsmPseudoInst<"vld1${p}", "$list, $addr!",
+ (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD1LNdWB_register_Asm :
+ NEONDT8AsmPseudoInst<"vld1${p}", "$list, $addr, $Rm",
+ (ins VecListOneDByteIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VLD1LNdWB_register_Asm :
+ NEONDT16AsmPseudoInst<"vld1${p}", "$list, $addr, $Rm",
+ (ins VecListOneDHWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VLD1LNdWB_register_Asm :
+ NEONDT32AsmPseudoInst<"vld1${p}", "$list, $addr, $Rm",
+ (ins VecListOneDWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+
+
+// VST1 single-lane pseudo-instructions. These need special handling for
+// the lane index that an InstAlias can't handle, so we use these instead.
+defm VST1LNdAsm : NEONDT8AsmPseudoInst<"vst1${p}", "$list, $addr",
+ (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST1LNdAsm : NEONDT16AsmPseudoInst<"vst1${p}", "$list, $addr",
+ (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST1LNdAsm : NEONDT32AsmPseudoInst<"vst1${p}", "$list, $addr",
+ (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+
+defm VST1LNdWB_fixed_Asm : NEONDT8AsmPseudoInst<"vst1${p}", "$list, $addr!",
+ (ins VecListOneDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST1LNdWB_fixed_Asm : NEONDT16AsmPseudoInst<"vst1${p}", "$list, $addr!",
+ (ins VecListOneDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST1LNdWB_fixed_Asm : NEONDT32AsmPseudoInst<"vst1${p}", "$list, $addr!",
+ (ins VecListOneDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST1LNdWB_register_Asm :
+ NEONDT8AsmPseudoInst<"vst1${p}", "$list, $addr, $Rm",
+ (ins VecListOneDByteIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VST1LNdWB_register_Asm :
+ NEONDT16AsmPseudoInst<"vst1${p}", "$list, $addr, $Rm",
+ (ins VecListOneDHWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VST1LNdWB_register_Asm :
+ NEONDT32AsmPseudoInst<"vst1${p}", "$list, $addr, $Rm",
+ (ins VecListOneDWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+
+// VLD2 single-lane pseudo-instructions. These need special handling for
+// the lane index that an InstAlias can't handle, so we use these instead.
+defm VLD2LNdAsm : NEONDT8AsmPseudoInst<"vld2${p}", "$list, $addr",
+ (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD2LNdAsm : NEONDT16AsmPseudoInst<"vld2${p}", "$list, $addr",
+ (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD2LNdAsm : NEONDT32AsmPseudoInst<"vld2${p}", "$list, $addr",
+ (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD2LNqAsm : NEONDT16AsmPseudoInst<"vld2${p}", "$list, $addr",
+ (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD2LNqAsm : NEONDT32AsmPseudoInst<"vld2${p}", "$list, $addr",
+ (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+
+defm VLD2LNdWB_fixed_Asm : NEONDT8AsmPseudoInst<"vld2${p}", "$list, $addr!",
+ (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD2LNdWB_fixed_Asm : NEONDT16AsmPseudoInst<"vld2${p}", "$list, $addr!",
+ (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD2LNdWB_fixed_Asm : NEONDT32AsmPseudoInst<"vld2${p}", "$list, $addr!",
+ (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD2LNqWB_fixed_Asm : NEONDT16AsmPseudoInst<"vld2${p}", "$list, $addr!",
+ (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD2LNqWB_fixed_Asm : NEONDT32AsmPseudoInst<"vld2${p}", "$list, $addr!",
+ (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VLD2LNdWB_register_Asm :
+ NEONDT8AsmPseudoInst<"vld2${p}", "$list, $addr, $Rm",
+ (ins VecListTwoDByteIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VLD2LNdWB_register_Asm :
+ NEONDT16AsmPseudoInst<"vld2${p}", "$list, $addr, $Rm",
+ (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VLD2LNdWB_register_Asm :
+ NEONDT32AsmPseudoInst<"vld2${p}", "$list, $addr, $Rm",
+ (ins VecListTwoDWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VLD2LNqWB_register_Asm :
+ NEONDT16AsmPseudoInst<"vld2${p}", "$list, $addr, $Rm",
+ (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VLD2LNqWB_register_Asm :
+ NEONDT32AsmPseudoInst<"vld2${p}", "$list, $addr, $Rm",
+ (ins VecListTwoQWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+
+
+// VST2 single-lane pseudo-instructions. These need special handling for
+// the lane index that an InstAlias can't handle, so we use these instead.
+defm VST2LNdAsm : NEONDT8AsmPseudoInst<"vst2${p}", "$list, $addr",
+ (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST2LNdAsm : NEONDT16AsmPseudoInst<"vst2${p}", "$list, $addr",
+ (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST2LNdAsm : NEONDT32AsmPseudoInst<"vst2${p}", "$list, $addr",
+ (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST2LNqAsm : NEONDT16AsmPseudoInst<"vst2${p}", "$list, $addr",
+ (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST2LNqAsm : NEONDT32AsmPseudoInst<"vst2${p}", "$list, $addr",
+ (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+
+defm VST2LNdWB_fixed_Asm : NEONDT8AsmPseudoInst<"vst2${p}", "$list, $addr!",
+ (ins VecListTwoDByteIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST2LNdWB_fixed_Asm : NEONDT16AsmPseudoInst<"vst2${p}", "$list, $addr!",
+ (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST2LNdWB_fixed_Asm : NEONDT32AsmPseudoInst<"vst2${p}", "$list, $addr!",
+ (ins VecListTwoDWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST2LNqWB_fixed_Asm : NEONDT16AsmPseudoInst<"vst2${p}", "$list, $addr!",
+ (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST2LNqWB_fixed_Asm : NEONDT32AsmPseudoInst<"vst2${p}", "$list, $addr!",
+ (ins VecListTwoQWordIndexed:$list, addrmode6:$addr, pred:$p)>;
+defm VST2LNdWB_register_Asm :
+ NEONDT8AsmPseudoInst<"vst2${p}", "$list, $addr, $Rm",
+ (ins VecListTwoDByteIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VST2LNdWB_register_Asm :
+ NEONDT16AsmPseudoInst<"vst2${p}", "$list, $addr, $Rm",
+ (ins VecListTwoDHWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VST2LNdWB_register_Asm :
+ NEONDT32AsmPseudoInst<"vst2${p}", "$list, $addr, $Rm",
+ (ins VecListTwoDWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VST2LNqWB_register_Asm :
+ NEONDT16AsmPseudoInst<"vst2${p}", "$list, $addr, $Rm",
+ (ins VecListTwoQHWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+defm VST2LNqWB_register_Asm :
+ NEONDT32AsmPseudoInst<"vst2${p}", "$list, $addr, $Rm",
+ (ins VecListTwoQWordIndexed:$list, addrmode6:$addr,
+ rGPR:$Rm, pred:$p)>;
+
+// VMOV takes an optional datatype suffix
+defm : VFPDTAnyInstAlias<"vmov${p}", "$Vd, $Vm",
+ (VORRd DPR:$Vd, DPR:$Vm, DPR:$Vm, pred:$p)>;
+defm : VFPDTAnyInstAlias<"vmov${p}", "$Vd, $Vm",
+ (VORRq QPR:$Vd, QPR:$Vm, QPR:$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, $Dm",
+ (VCGEsv8i8 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.s16 $Dd, $Dn, $Dm",
+ (VCGEsv4i16 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.s32 $Dd, $Dn, $Dm",
+ (VCGEsv2i32 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.u8 $Dd, $Dn, $Dm",
+ (VCGEuv8i8 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.u16 $Dd, $Dn, $Dm",
+ (VCGEuv4i16 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.u32 $Dd, $Dn, $Dm",
+ (VCGEuv2i32 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.f32 $Dd, $Dn, $Dm",
+ (VCGEfd DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+// Q-register versions.
+def : NEONInstAlias<"vcle${p}.s8 $Qd, $Qn, $Qm",
+ (VCGEsv16i8 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.s16 $Qd, $Qn, $Qm",
+ (VCGEsv8i16 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.s32 $Qd, $Qn, $Qm",
+ (VCGEsv4i32 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.u8 $Qd, $Qn, $Qm",
+ (VCGEuv16i8 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.u16 $Qd, $Qn, $Qm",
+ (VCGEuv8i16 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.u32 $Qd, $Qn, $Qm",
+ (VCGEuv4i32 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vcle${p}.f32 $Qd, $Qn, $Qm",
+ (VCGEfq QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+
+// VCLT (register) is an assembler alias for VCGT w/ the operands reversed.
+// D-register versions.
+def : NEONInstAlias<"vclt${p}.s8 $Dd, $Dn, $Dm",
+ (VCGTsv8i8 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.s16 $Dd, $Dn, $Dm",
+ (VCGTsv4i16 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.s32 $Dd, $Dn, $Dm",
+ (VCGTsv2i32 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.u8 $Dd, $Dn, $Dm",
+ (VCGTuv8i8 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.u16 $Dd, $Dn, $Dm",
+ (VCGTuv4i16 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.u32 $Dd, $Dn, $Dm",
+ (VCGTuv2i32 DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.f32 $Dd, $Dn, $Dm",
+ (VCGTfd DPR:$Dd, DPR:$Dm, DPR:$Dn, pred:$p)>;
+// Q-register versions.
+def : NEONInstAlias<"vclt${p}.s8 $Qd, $Qn, $Qm",
+ (VCGTsv16i8 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.s16 $Qd, $Qn, $Qm",
+ (VCGTsv8i16 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.s32 $Qd, $Qn, $Qm",
+ (VCGTsv4i32 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.u8 $Qd, $Qn, $Qm",
+ (VCGTuv16i8 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.u16 $Qd, $Qn, $Qm",
+ (VCGTuv8i16 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.u32 $Qd, $Qn, $Qm",
+ (VCGTuv4i32 QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+def : NEONInstAlias<"vclt${p}.f32 $Qd, $Qn, $Qm",
+ (VCGTfq QPR:$Qd, QPR:$Qm, QPR:$Qn, pred:$p)>;
+
+// Two-operand variants for VEXT
+def : NEONInstAlias<"vext${p}.8 $Vdn, $Vm, $imm",
+ (VEXTd8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, imm0_7:$imm, pred:$p)>;
+def : NEONInstAlias<"vext${p}.16 $Vdn, $Vm, $imm",
+ (VEXTd16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, imm0_3:$imm, pred:$p)>;
+def : NEONInstAlias<"vext${p}.32 $Vdn, $Vm, $imm",
+ (VEXTd32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, imm0_1:$imm, pred:$p)>;
+
+def : NEONInstAlias<"vext${p}.8 $Vdn, $Vm, $imm",
+ (VEXTq8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, imm0_15:$imm, pred:$p)>;
+def : NEONInstAlias<"vext${p}.16 $Vdn, $Vm, $imm",
+ (VEXTq16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, imm0_7:$imm, pred:$p)>;
+def : NEONInstAlias<"vext${p}.32 $Vdn, $Vm, $imm",
+ (VEXTq32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, imm0_3:$imm, pred:$p)>;
+def : NEONInstAlias<"vext${p}.64 $Vdn, $Vm, $imm",
+ (VEXTq64 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, imm0_1:$imm, pred:$p)>;
+
+// Two-operand variants for VQDMULH
+def : NEONInstAlias<"vqdmulh${p}.s16 $Vdn, $Vm",
+ (VQDMULHv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqdmulh${p}.s32 $Vdn, $Vm",
+ (VQDMULHv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vqdmulh${p}.s16 $Vdn, $Vm",
+ (VQDMULHv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vqdmulh${p}.s32 $Vdn, $Vm",
+ (VQDMULHv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+// Two-operand variants for VMAX.
+def : NEONInstAlias<"vmax${p}.s8 $Vdn, $Vm",
+ (VMAXsv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.s16 $Vdn, $Vm",
+ (VMAXsv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.s32 $Vdn, $Vm",
+ (VMAXsv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.u8 $Vdn, $Vm",
+ (VMAXuv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.u16 $Vdn, $Vm",
+ (VMAXuv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.u32 $Vdn, $Vm",
+ (VMAXuv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.f32 $Vdn, $Vm",
+ (VMAXfd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vmax${p}.s8 $Vdn, $Vm",
+ (VMAXsv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.s16 $Vdn, $Vm",
+ (VMAXsv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.s32 $Vdn, $Vm",
+ (VMAXsv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.u8 $Vdn, $Vm",
+ (VMAXuv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.u16 $Vdn, $Vm",
+ (VMAXuv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.u32 $Vdn, $Vm",
+ (VMAXuv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmax${p}.f32 $Vdn, $Vm",
+ (VMAXfq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+// Two-operand variants for VMIN.
+def : NEONInstAlias<"vmin${p}.s8 $Vdn, $Vm",
+ (VMINsv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.s16 $Vdn, $Vm",
+ (VMINsv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.s32 $Vdn, $Vm",
+ (VMINsv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.u8 $Vdn, $Vm",
+ (VMINuv8i8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.u16 $Vdn, $Vm",
+ (VMINuv4i16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.u32 $Vdn, $Vm",
+ (VMINuv2i32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.f32 $Vdn, $Vm",
+ (VMINfd DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+def : NEONInstAlias<"vmin${p}.s8 $Vdn, $Vm",
+ (VMINsv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.s16 $Vdn, $Vm",
+ (VMINsv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.s32 $Vdn, $Vm",
+ (VMINsv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.u8 $Vdn, $Vm",
+ (VMINuv16i8 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.u16 $Vdn, $Vm",
+ (VMINuv8i16 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.u32 $Vdn, $Vm",
+ (VMINuv4i32 QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vmin${p}.f32 $Vdn, $Vm",
+ (VMINfq QPR:$Vdn, QPR:$Vdn, QPR:$Vm, pred:$p)>;
+
+// Two-operand variants for VPADD.
+def : NEONInstAlias<"vpadd${p}.i8 $Vdn, $Vm",
+ (VPADDi8 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vpadd${p}.i16 $Vdn, $Vm",
+ (VPADDi16 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vpadd${p}.i32 $Vdn, $Vm",
+ (VPADDi32 DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+def : NEONInstAlias<"vpadd${p}.f32 $Vdn, $Vm",
+ (VPADDf DPR:$Vdn, DPR:$Vdn, DPR:$Vm, pred:$p)>;
+
+// VSWP allows, but does not require, a type suffix.
+defm : VFPDTAnyInstAlias<"vswp${p}", "$Vd, $Vm",
+ (VSWPd DPR:$Vd, DPR:$Vm, pred:$p)>;
+defm : VFPDTAnyInstAlias<"vswp${p}", "$Vd, $Vm",
+ (VSWPq QPR:$Vd, QPR:$Vm, pred:$p)>;
+
+// "vmov Rd, #-imm" can be handled via "vmvn".
+def : NEONInstAlias<"vmov${p}.i32 $Vd, $imm",
+ (VMVNv2i32 DPR:$Vd, nImmVMOVI32Neg:$imm, pred:$p)>;
+def : NEONInstAlias<"vmov${p}.i32 $Vd, $imm",
+ (VMVNv4i32 QPR:$Vd, nImmVMOVI32Neg:$imm, pred:$p)>;
+def : NEONInstAlias<"vmvn${p}.i32 $Vd, $imm",
+ (VMOVv2i32 DPR:$Vd, nImmVMOVI32Neg:$imm, pred:$p)>;
+def : NEONInstAlias<"vmvn${p}.i32 $Vd, $imm",
+ (VMOVv4i32 QPR:$Vd, nImmVMOVI32Neg:$imm, pred:$p)>;
+
+// 'gas' compatibility aliases for quad-word instructions. Strictly speaking,
+// these should restrict to just the Q register variants, but the register
+// classes are enough to match correctly regardless, so we keep it simple
+// and just use MnemonicAlias.
+def : NEONMnemonicAlias<"vbicq", "vbic">;
+def : NEONMnemonicAlias<"vandq", "vand">;
+def : NEONMnemonicAlias<"veorq", "veor">;
+def : NEONMnemonicAlias<"vorrq", "vorr">;
+
+def : NEONMnemonicAlias<"vmovq", "vmov">;
+def : NEONMnemonicAlias<"vmvnq", "vmvn">;
+// Explicit versions for floating point so that the FPImm variants get
+// handled early. The parser gets confused otherwise.
+def : NEONMnemonicAlias<"vmovq.f32", "vmov.f32">;
+def : NEONMnemonicAlias<"vmovq.f64", "vmov.f64">;
+
+def : NEONMnemonicAlias<"vaddq", "vadd">;
+def : NEONMnemonicAlias<"vsubq", "vsub">;
+
+def : NEONMnemonicAlias<"vminq", "vmin">;
+def : NEONMnemonicAlias<"vmaxq", "vmax">;
+
+def : NEONMnemonicAlias<"vmulq", "vmul">;
+
+def : NEONMnemonicAlias<"vabsq", "vabs">;
+
+def : NEONMnemonicAlias<"vshlq", "vshl">;
+def : NEONMnemonicAlias<"vshrq", "vshr">;
+
+def : NEONMnemonicAlias<"vcvtq", "vcvt">;
+
+def : NEONMnemonicAlias<"vcleq", "vcle">;
+def : NEONMnemonicAlias<"vceqq", "vceq">;
+
+def : NEONMnemonicAlias<"vzipq", "vzip">;
+def : NEONMnemonicAlias<"vswpq", "vswp">;
+
+def : NEONMnemonicAlias<"vrecpeq.f32", "vrecpe.f32">;
+def : NEONMnemonicAlias<"vrecpeq.u32", "vrecpe.u32">;
projects / oota-llvm.git / blobdiff