def NEONvgetlaneu : SDNode<"ARMISD::VGETLANEu", SDTARMVGETLN>;
def NEONvgetlanes : SDNode<"ARMISD::VGETLANEs", SDTARMVGETLN>;
+def SDTARMVMOVIMM : SDTypeProfile<1, 1, [SDTCisVec<0>, SDTCisVT<1, i32>]>;
+def NEONvmovImm : SDNode<"ARMISD::VMOVIMM", SDTARMVMOVIMM>;
+def NEONvmvnImm : SDNode<"ARMISD::VMVNIMM", SDTARMVMOVIMM>;
+
def NEONvdup : SDNode<"ARMISD::VDUP", SDTypeProfile<1, 1, [SDTCisVec<0>]>>;
// VDUPLANE can produce a quad-register result from a double-register source,
def NEONuzp : SDNode<"ARMISD::VUZP", SDTARMVSHUF2>;
def NEONtrn : SDNode<"ARMISD::VTRN", SDTARMVSHUF2>;
+def SDTARMVMULL : SDTypeProfile<1, 2, [SDTCisInt<0>, SDTCisInt<1>,
+ SDTCisSameAs<1, 2>]>;
+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;
+ uint64_t EltVal = ARM_AM::decodeNEONModImm(ConstVal->getZExtValue(), EltBits);
+ return (EltBits == 32 && EltVal == 0);
+}]>;
+
+def NEONimmAllOnesV: PatLeaf<(NEONvmovImm (i32 timm)), [{
+ ConstantSDNode *ConstVal = cast<ConstantSDNode>(N->getOperand(0));
+ unsigned EltBits = 0;
+ uint64_t EltVal = ARM_AM::decodeNEONModImm(ConstVal->getZExtValue(), EltBits);
+ return (EltBits == 8 && EltVal == 0xff);
+}]>;
+
//===----------------------------------------------------------------------===//
// NEON operand definitions
//===----------------------------------------------------------------------===//
-def h8imm : Operand<i8> {
- let PrintMethod = "printHex8ImmOperand";
-}
-def h16imm : Operand<i16> {
- let PrintMethod = "printHex16ImmOperand";
-}
-def h32imm : Operand<i32> {
- let PrintMethod = "printHex32ImmOperand";
-}
-def h64imm : Operand<i64> {
- let PrintMethod = "printHex64ImmOperand";
+def nModImm : Operand<i32> {
+ let PrintMethod = "printNEONModImmOperand";
}
//===----------------------------------------------------------------------===//
// NEON load / store instructions
//===----------------------------------------------------------------------===//
-let mayLoad = 1 in {
// Use vldmia to load a Q register as a D register pair.
// This is equivalent to VLDMD except that it has a Q register operand
// instead of a pair of D registers.
def VLDMQ
- : AXDI5<(outs QPR:$dst), (ins addrmode5:$addr, pred:$p),
+ : AXDI4<(outs QPR:$dst), (ins addrmode4:$addr, pred:$p),
IndexModeNone, IIC_fpLoadm,
- "vldm${addr:submode}${p}\t${addr:base}, ${dst:dregpair}", "", []>;
-def VLDMQ_UPD
- : AXDI5<(outs QPR:$dst, GPR:$wb), (ins addrmode5:$addr, pred:$p),
- IndexModeUpd, IIC_fpLoadm,
- "vldm${addr:submode}${p}\t${addr:base}!, ${dst:dregpair}",
- "$addr.base = $wb", []>;
-} // mayLoad = 1
-
-let mayStore = 1 in {
+ "vldm${addr:submode}${p}\t$addr, ${dst:dregpair}", "",
+ [(set QPR:$dst, (v2f64 (load addrmode4:$addr)))]>;
+
+let mayLoad = 1, neverHasSideEffects = 1 in {
+// Use vld1 to load a Q register as a D register pair.
+// This alternative to VLDMQ allows an alignment to be specified.
+// This is equivalent to VLD1q64 except that it has a Q register operand.
+def VLD1q
+ : NLdSt<0,0b10,0b1010,0b1100, (outs QPR:$dst), (ins addrmode6:$addr),
+ IIC_VLD1, "vld1", "64", "${dst:dregpair}, $addr", "", []>;
+} // mayLoad = 1, neverHasSideEffects = 1
+
// Use vstmia to store a Q register as a D register pair.
// This is equivalent to VSTMD except that it has a Q register operand
// instead of a pair of D registers.
def VSTMQ
- : AXDI5<(outs), (ins QPR:$src, addrmode5:$addr, pred:$p),
+ : AXDI4<(outs), (ins QPR:$src, addrmode4:$addr, pred:$p),
IndexModeNone, IIC_fpStorem,
- "vstm${addr:submode}${p}\t${addr:base}, ${src:dregpair}", "", []>;
-def VSTMQ_UPD
- : AXDI5<(outs GPR:$wb), (ins QPR:$src, addrmode5:$addr, pred:$p),
- IndexModeUpd, IIC_fpStorem,
- "vstm${addr:submode}${p}\t${addr:base}!, ${src:dregpair}",
- "$addr.base = $wb", []>;
-} // mayStore = 1
-
-let mayLoad = 1, hasExtraDefRegAllocReq = 1 in {
+ "vstm${addr:submode}${p}\t$addr, ${src:dregpair}", "",
+ [(store (v2f64 QPR:$src), addrmode4:$addr)]>;
+
+let mayStore = 1, neverHasSideEffects = 1 in {
+// Use vst1 to store a Q register as a D register pair.
+// This alternative to VSTMQ allows an alignment to be specified.
+// This is equivalent to VST1q64 except that it has a Q register operand.
+def VST1q
+ : NLdSt<0,0b00,0b1010,0b1100, (outs), (ins addrmode6:$addr, QPR:$src),
+ IIC_VST, "vst1", "64", "${src:dregpair}, $addr", "", []>;
+} // mayStore = 1, neverHasSideEffects = 1
+
+let mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1 in {
+
+// Classes for VLD* pseudo-instructions with multi-register operands.
+// These are expanded to real instructions after register allocation.
+class VLDQPseudo
+ : PseudoNLdSt<(outs QPR:$dst), (ins addrmode6:$addr), IIC_VLD2, "">;
+class VLDQWBPseudo
+ : PseudoNLdSt<(outs QPR:$dst, GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset), IIC_VLD2,
+ "$addr.addr = $wb">;
+class VLDQQPseudo
+ : PseudoNLdSt<(outs QQPR:$dst), (ins addrmode6:$addr), IIC_VLD4, "">;
+class VLDQQWBPseudo
+ : PseudoNLdSt<(outs QQPR:$dst, GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset), IIC_VLD4,
+ "$addr.addr = $wb">;
+class VLDQQQQWBPseudo
+ : PseudoNLdSt<(outs QQQQPR:$dst, GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset, QQQQPR:$src), IIC_VLD4,
+ "$addr.addr = $wb, $src = $dst">;
// VLD1 : Vector Load (multiple single elements)
class VLD1D<bits<4> op7_4, string Dt>
def VLD1q32 : VLD1Q<0b1000, "32">;
def VLD1q64 : VLD1Q<0b1100, "64">;
+def VLD1q8Pseudo : VLDQPseudo;
+def VLD1q16Pseudo : VLDQPseudo;
+def VLD1q32Pseudo : VLDQPseudo;
+def VLD1q64Pseudo : VLDQPseudo;
+
// ...with address register writeback:
class VLD1DWB<bits<4> op7_4, string Dt>
: NLdSt<0,0b10,0b0111,op7_4, (outs DPR:$dst, GPR:$wb),
def VLD1q32_UPD : VLD1QWB<0b1000, "32">;
def VLD1q64_UPD : VLD1QWB<0b1100, "64">;
+def VLD1q8Pseudo_UPD : VLDQWBPseudo;
+def VLD1q16Pseudo_UPD : VLDQWBPseudo;
+def VLD1q32Pseudo_UPD : VLDQWBPseudo;
+def VLD1q64Pseudo_UPD : VLDQWBPseudo;
+
// ...with 3 registers (some of these are only for the disassembler):
class VLD1D3<bits<4> op7_4, string Dt>
: NLdSt<0,0b10,0b0110,op7_4, (outs DPR:$dst1, DPR:$dst2, DPR:$dst3),
def VLD1d32T_UPD : VLD1D3WB<0b1000, "32">;
def VLD1d64T_UPD : VLD1D3WB<0b1100, "64">;
+def VLD1d64TPseudo : VLDQQPseudo;
+def VLD1d64TPseudo_UPD : VLDQQWBPseudo;
+
// ...with 4 registers (some of these are only for the disassembler):
class VLD1D4<bits<4> op7_4, string Dt>
: NLdSt<0,0b10,0b0010,op7_4,(outs DPR:$dst1, DPR:$dst2, DPR:$dst3, DPR:$dst4),
def VLD1d32Q_UPD : VLD1D4WB<0b1000, "32">;
def VLD1d64Q_UPD : VLD1D4WB<0b1100, "64">;
+def VLD1d64QPseudo : VLDQQPseudo;
+def VLD1d64QPseudo_UPD : VLDQQWBPseudo;
+
// VLD2 : Vector Load (multiple 2-element structures)
class VLD2D<bits<4> op11_8, bits<4> op7_4, string Dt>
: NLdSt<0, 0b10, op11_8, op7_4, (outs DPR:$dst1, DPR:$dst2),
def VLD2q16 : VLD2Q<0b0100, "16">;
def VLD2q32 : VLD2Q<0b1000, "32">;
+def VLD2d8Pseudo : VLDQPseudo;
+def VLD2d16Pseudo : VLDQPseudo;
+def VLD2d32Pseudo : VLDQPseudo;
+
+def VLD2q8Pseudo : VLDQQPseudo;
+def VLD2q16Pseudo : VLDQQPseudo;
+def VLD2q32Pseudo : VLDQQPseudo;
+
// ...with address register writeback:
class VLD2DWB<bits<4> op11_8, bits<4> op7_4, string Dt>
: NLdSt<0, 0b10, op11_8, op7_4, (outs DPR:$dst1, DPR:$dst2, GPR:$wb),
def VLD2q16_UPD : VLD2QWB<0b0100, "16">;
def VLD2q32_UPD : VLD2QWB<0b1000, "32">;
+def VLD2d8Pseudo_UPD : VLDQWBPseudo;
+def VLD2d16Pseudo_UPD : VLDQWBPseudo;
+def VLD2d32Pseudo_UPD : VLDQWBPseudo;
+
+def VLD2q8Pseudo_UPD : VLDQQWBPseudo;
+def VLD2q16Pseudo_UPD : VLDQQWBPseudo;
+def VLD2q32Pseudo_UPD : VLDQQWBPseudo;
+
// ...with double-spaced registers (for disassembly only):
def VLD2b8 : VLD2D<0b1001, 0b0000, "8">;
def VLD2b16 : VLD2D<0b1001, 0b0100, "16">;
def VLD3d16 : VLD3D<0b0100, 0b0100, "16">;
def VLD3d32 : VLD3D<0b0100, 0b1000, "32">;
+def VLD3d8Pseudo : VLDQQPseudo;
+def VLD3d16Pseudo : VLDQQPseudo;
+def VLD3d32Pseudo : VLDQQPseudo;
+
// ...with address register writeback:
class VLD3DWB<bits<4> op11_8, bits<4> op7_4, string Dt>
: NLdSt<0, 0b10, op11_8, op7_4,
def VLD3d16_UPD : VLD3DWB<0b0100, 0b0100, "16">;
def VLD3d32_UPD : VLD3DWB<0b0100, 0b1000, "32">;
+def VLD3d8Pseudo_UPD : VLDQQWBPseudo;
+def VLD3d16Pseudo_UPD : VLDQQWBPseudo;
+def VLD3d32Pseudo_UPD : VLDQQWBPseudo;
+
// ...with double-spaced registers (non-updating versions for disassembly only):
def VLD3q8 : VLD3D<0b0101, 0b0000, "8">;
def VLD3q16 : VLD3D<0b0101, 0b0100, "16">;
def VLD3q16_UPD : VLD3DWB<0b0101, 0b0100, "16">;
def VLD3q32_UPD : VLD3DWB<0b0101, 0b1000, "32">;
+def VLD3q8Pseudo_UPD : VLDQQQQWBPseudo;
+def VLD3q16Pseudo_UPD : VLDQQQQWBPseudo;
+def VLD3q32Pseudo_UPD : VLDQQQQWBPseudo;
+
// ...alternate versions to be allocated odd register numbers:
-def VLD3q8odd_UPD : VLD3DWB<0b0101, 0b0000, "8">;
-def VLD3q16odd_UPD : VLD3DWB<0b0101, 0b0100, "16">;
-def VLD3q32odd_UPD : VLD3DWB<0b0101, 0b1000, "32">;
+def VLD3q8oddPseudo_UPD : VLDQQQQWBPseudo;
+def VLD3q16oddPseudo_UPD : VLDQQQQWBPseudo;
+def VLD3q32oddPseudo_UPD : VLDQQQQWBPseudo;
// VLD4 : Vector Load (multiple 4-element structures)
class VLD4D<bits<4> op11_8, bits<4> op7_4, string Dt>
def VLD4d16 : VLD4D<0b0000, 0b0100, "16">;
def VLD4d32 : VLD4D<0b0000, 0b1000, "32">;
+def VLD4d8Pseudo : VLDQQPseudo;
+def VLD4d16Pseudo : VLDQQPseudo;
+def VLD4d32Pseudo : VLDQQPseudo;
+
// ...with address register writeback:
class VLD4DWB<bits<4> op11_8, bits<4> op7_4, string Dt>
: NLdSt<0, 0b10, op11_8, op7_4,
def VLD4d16_UPD : VLD4DWB<0b0000, 0b0100, "16">;
def VLD4d32_UPD : VLD4DWB<0b0000, 0b1000, "32">;
+def VLD4d8Pseudo_UPD : VLDQQWBPseudo;
+def VLD4d16Pseudo_UPD : VLDQQWBPseudo;
+def VLD4d32Pseudo_UPD : VLDQQWBPseudo;
+
// ...with double-spaced registers (non-updating versions for disassembly only):
def VLD4q8 : VLD4D<0b0001, 0b0000, "8">;
def VLD4q16 : VLD4D<0b0001, 0b0100, "16">;
def VLD4q16_UPD : VLD4DWB<0b0001, 0b0100, "16">;
def VLD4q32_UPD : VLD4DWB<0b0001, 0b1000, "32">;
+def VLD4q8Pseudo_UPD : VLDQQQQWBPseudo;
+def VLD4q16Pseudo_UPD : VLDQQQQWBPseudo;
+def VLD4q32Pseudo_UPD : VLDQQQQWBPseudo;
+
// ...alternate versions to be allocated odd register numbers:
-def VLD4q8odd_UPD : VLD4DWB<0b0001, 0b0000, "8">;
-def VLD4q16odd_UPD : VLD4DWB<0b0001, 0b0100, "16">;
-def VLD4q32odd_UPD : VLD4DWB<0b0001, 0b1000, "32">;
+def VLD4q8oddPseudo_UPD : VLDQQQQWBPseudo;
+def VLD4q16oddPseudo_UPD : VLDQQQQWBPseudo;
+def VLD4q32oddPseudo_UPD : VLDQQQQWBPseudo;
+
+// Classes for VLD*LN pseudo-instructions with multi-register operands.
+// These are expanded to real instructions after register allocation.
+class VLDQLNPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs QPR:$dst),
+ (ins addrmode6:$addr, QPR:$src, nohash_imm:$lane),
+ itin, "$src = $dst">;
+class VLDQLNWBPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs QPR:$dst, GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset, QPR:$src,
+ nohash_imm:$lane), itin, "$addr.addr = $wb, $src = $dst">;
+class VLDQQLNPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs QQPR:$dst),
+ (ins addrmode6:$addr, QQPR:$src, nohash_imm:$lane),
+ itin, "$src = $dst">;
+class VLDQQLNWBPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs QQPR:$dst, GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset, QQPR:$src,
+ nohash_imm:$lane), itin, "$addr.addr = $wb, $src = $dst">;
+class VLDQQQQLNPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs QQQQPR:$dst),
+ (ins addrmode6:$addr, QQQQPR:$src, nohash_imm:$lane),
+ itin, "$src = $dst">;
+class VLDQQQQLNWBPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs QQQQPR:$dst, GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset, QQQQPR:$src,
+ nohash_imm:$lane), itin, "$addr.addr = $wb, $src = $dst">;
// VLD1LN : Vector Load (single element to one lane)
// FIXME: Not yet implemented.
def VLD2LNd16 : VLD2LN<0b0101, {?,?,0,?}, "16">;
def VLD2LNd32 : VLD2LN<0b1001, {?,0,?,?}, "32">;
+def VLD2LNd8Pseudo : VLDQLNPseudo<IIC_VLD2>;
+def VLD2LNd16Pseudo : VLDQLNPseudo<IIC_VLD2>;
+def VLD2LNd32Pseudo : VLDQLNPseudo<IIC_VLD2>;
+
// ...with double-spaced registers:
def VLD2LNq16 : VLD2LN<0b0101, {?,?,1,?}, "16">;
def VLD2LNq32 : VLD2LN<0b1001, {?,1,?,?}, "32">;
-// ...alternate versions to be allocated odd register numbers:
-def VLD2LNq16odd : VLD2LN<0b0101, {?,?,1,?}, "16">;
-def VLD2LNq32odd : VLD2LN<0b1001, {?,1,?,?}, "32">;
+def VLD2LNq16Pseudo : VLDQQLNPseudo<IIC_VLD2>;
+def VLD2LNq32Pseudo : VLDQQLNPseudo<IIC_VLD2>;
// ...with address register writeback:
class VLD2LNWB<bits<4> op11_8, bits<4> op7_4, string Dt>
def VLD2LNd16_UPD : VLD2LNWB<0b0101, {?,?,0,?}, "16">;
def VLD2LNd32_UPD : VLD2LNWB<0b1001, {?,0,?,?}, "32">;
+def VLD2LNd8Pseudo_UPD : VLDQLNWBPseudo<IIC_VLD2>;
+def VLD2LNd16Pseudo_UPD : VLDQLNWBPseudo<IIC_VLD2>;
+def VLD2LNd32Pseudo_UPD : VLDQLNWBPseudo<IIC_VLD2>;
+
def VLD2LNq16_UPD : VLD2LNWB<0b0101, {?,?,1,?}, "16">;
def VLD2LNq32_UPD : VLD2LNWB<0b1001, {?,1,?,?}, "32">;
+def VLD2LNq16Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD2>;
+def VLD2LNq32Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD2>;
+
// VLD3LN : Vector Load (single 3-element structure to one lane)
class VLD3LN<bits<4> op11_8, bits<4> op7_4, string Dt>
: NLdSt<1, 0b10, op11_8, op7_4, (outs DPR:$dst1, DPR:$dst2, DPR:$dst3),
def VLD3LNd16 : VLD3LN<0b0110, {?,?,0,0}, "16">;
def VLD3LNd32 : VLD3LN<0b1010, {?,0,0,0}, "32">;
+def VLD3LNd8Pseudo : VLDQQLNPseudo<IIC_VLD3>;
+def VLD3LNd16Pseudo : VLDQQLNPseudo<IIC_VLD3>;
+def VLD3LNd32Pseudo : VLDQQLNPseudo<IIC_VLD3>;
+
// ...with double-spaced registers:
def VLD3LNq16 : VLD3LN<0b0110, {?,?,1,0}, "16">;
def VLD3LNq32 : VLD3LN<0b1010, {?,1,0,0}, "32">;
-// ...alternate versions to be allocated odd register numbers:
-def VLD3LNq16odd : VLD3LN<0b0110, {?,?,1,0}, "16">;
-def VLD3LNq32odd : VLD3LN<0b1010, {?,1,0,0}, "32">;
+def VLD3LNq16Pseudo : VLDQQQQLNPseudo<IIC_VLD3>;
+def VLD3LNq32Pseudo : VLDQQQQLNPseudo<IIC_VLD3>;
// ...with address register writeback:
class VLD3LNWB<bits<4> op11_8, bits<4> op7_4, string Dt>
def VLD3LNd16_UPD : VLD3LNWB<0b0110, {?,?,0,0}, "16">;
def VLD3LNd32_UPD : VLD3LNWB<0b1010, {?,0,0,0}, "32">;
+def VLD3LNd8Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD3>;
+def VLD3LNd16Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD3>;
+def VLD3LNd32Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD3>;
+
def VLD3LNq16_UPD : VLD3LNWB<0b0110, {?,?,1,0}, "16">;
def VLD3LNq32_UPD : VLD3LNWB<0b1010, {?,1,0,0}, "32">;
+def VLD3LNq16Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD3>;
+def VLD3LNq32Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD3>;
+
// VLD4LN : Vector Load (single 4-element structure to one lane)
class VLD4LN<bits<4> op11_8, bits<4> op7_4, string Dt>
: NLdSt<1, 0b10, op11_8, op7_4,
def VLD4LNd16 : VLD4LN<0b0111, {?,?,0,?}, "16">;
def VLD4LNd32 : VLD4LN<0b1011, {?,0,?,?}, "32">;
+def VLD4LNd8Pseudo : VLDQQLNPseudo<IIC_VLD4>;
+def VLD4LNd16Pseudo : VLDQQLNPseudo<IIC_VLD4>;
+def VLD4LNd32Pseudo : VLDQQLNPseudo<IIC_VLD4>;
+
// ...with double-spaced registers:
def VLD4LNq16 : VLD4LN<0b0111, {?,?,1,?}, "16">;
def VLD4LNq32 : VLD4LN<0b1011, {?,1,?,?}, "32">;
-// ...alternate versions to be allocated odd register numbers:
-def VLD4LNq16odd : VLD4LN<0b0111, {?,?,1,?}, "16">;
-def VLD4LNq32odd : VLD4LN<0b1011, {?,1,?,?}, "32">;
+def VLD4LNq16Pseudo : VLDQQQQLNPseudo<IIC_VLD4>;
+def VLD4LNq32Pseudo : VLDQQQQLNPseudo<IIC_VLD4>;
// ...with address register writeback:
class VLD4LNWB<bits<4> op11_8, bits<4> op7_4, string Dt>
def VLD4LNd16_UPD : VLD4LNWB<0b0111, {?,?,0,?}, "16">;
def VLD4LNd32_UPD : VLD4LNWB<0b1011, {?,0,?,?}, "32">;
+def VLD4LNd8Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD4>;
+def VLD4LNd16Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD4>;
+def VLD4LNd32Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD4>;
+
def VLD4LNq16_UPD : VLD4LNWB<0b0111, {?,?,1,?}, "16">;
def VLD4LNq32_UPD : VLD4LNWB<0b1011, {?,1,?,?}, "32">;
+def VLD4LNq16Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD4>;
+def VLD4LNq32Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD4>;
+
// VLD1DUP : Vector Load (single element to all lanes)
// VLD2DUP : Vector Load (single 2-element structure to all lanes)
// VLD3DUP : Vector Load (single 3-element structure to all lanes)
// VLD4DUP : Vector Load (single 4-element structure to all lanes)
// FIXME: Not yet implemented.
-} // mayLoad = 1, hasExtraDefRegAllocReq = 1
-
-let mayStore = 1, hasExtraSrcRegAllocReq = 1 in {
+} // mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1
+
+let mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1 in {
+
+// Classes for VST* pseudo-instructions with multi-register operands.
+// These are expanded to real instructions after register allocation.
+class VSTQPseudo
+ : PseudoNLdSt<(outs), (ins addrmode6:$addr, QPR:$src), IIC_VST, "">;
+class VSTQWBPseudo
+ : PseudoNLdSt<(outs GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset, QPR:$src), IIC_VST,
+ "$addr.addr = $wb">;
+class VSTQQPseudo
+ : PseudoNLdSt<(outs), (ins addrmode6:$addr, QQPR:$src), IIC_VST, "">;
+class VSTQQWBPseudo
+ : PseudoNLdSt<(outs GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset, QQPR:$src), IIC_VST,
+ "$addr.addr = $wb">;
+class VSTQQQQWBPseudo
+ : PseudoNLdSt<(outs GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset, QQQQPR:$src), IIC_VST,
+ "$addr.addr = $wb">;
// VST1 : Vector Store (multiple single elements)
class VST1D<bits<4> op7_4, string Dt>
def VST1q32 : VST1Q<0b1000, "32">;
def VST1q64 : VST1Q<0b1100, "64">;
+def VST1q8Pseudo : VSTQPseudo;
+def VST1q16Pseudo : VSTQPseudo;
+def VST1q32Pseudo : VSTQPseudo;
+def VST1q64Pseudo : VSTQPseudo;
+
// ...with address register writeback:
class VST1DWB<bits<4> op7_4, string Dt>
: NLdSt<0, 0b00, 0b0111, op7_4, (outs GPR:$wb),
def VST1q32_UPD : VST1QWB<0b1000, "32">;
def VST1q64_UPD : VST1QWB<0b1100, "64">;
+def VST1q8Pseudo_UPD : VSTQWBPseudo;
+def VST1q16Pseudo_UPD : VSTQWBPseudo;
+def VST1q32Pseudo_UPD : VSTQWBPseudo;
+def VST1q64Pseudo_UPD : VSTQWBPseudo;
+
// ...with 3 registers (some of these are only for the disassembler):
class VST1D3<bits<4> op7_4, string Dt>
: NLdSt<0, 0b00, 0b0110, op7_4, (outs),
def VST1d32T_UPD : VST1D3WB<0b1000, "32">;
def VST1d64T_UPD : VST1D3WB<0b1100, "64">;
+def VST1d64TPseudo : VSTQQPseudo;
+def VST1d64TPseudo_UPD : VSTQQWBPseudo;
+
// ...with 4 registers (some of these are only for the disassembler):
class VST1D4<bits<4> op7_4, string Dt>
: NLdSt<0, 0b00, 0b0010, op7_4, (outs),
def VST1d32Q_UPD : VST1D4WB<0b1000, "32">;
def VST1d64Q_UPD : VST1D4WB<0b1100, "64">;
+def VST1d64QPseudo : VSTQQPseudo;
+def VST1d64QPseudo_UPD : VSTQQWBPseudo;
+
// 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),
def VST2q16 : VST2Q<0b0100, "16">;
def VST2q32 : VST2Q<0b1000, "32">;
+def VST2d8Pseudo : VSTQPseudo;
+def VST2d16Pseudo : VSTQPseudo;
+def VST2d32Pseudo : VSTQPseudo;
+
+def VST2q8Pseudo : VSTQQPseudo;
+def VST2q16Pseudo : VSTQQPseudo;
+def VST2q32Pseudo : VSTQQPseudo;
+
// ...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:$wb),
def VST2q16_UPD : VST2QWB<0b0100, "16">;
def VST2q32_UPD : VST2QWB<0b1000, "32">;
+def VST2d8Pseudo_UPD : VSTQWBPseudo;
+def VST2d16Pseudo_UPD : VSTQWBPseudo;
+def VST2d32Pseudo_UPD : VSTQWBPseudo;
+
+def VST2q8Pseudo_UPD : VSTQQWBPseudo;
+def VST2q16Pseudo_UPD : VSTQQWBPseudo;
+def VST2q32Pseudo_UPD : VSTQQWBPseudo;
+
// ...with double-spaced registers (for disassembly only):
def VST2b8 : VST2D<0b1001, 0b0000, "8">;
def VST2b16 : VST2D<0b1001, 0b0100, "16">;
def VST3d16 : VST3D<0b0100, 0b0100, "16">;
def VST3d32 : VST3D<0b0100, 0b1000, "32">;
+def VST3d8Pseudo : VSTQQPseudo;
+def VST3d16Pseudo : VSTQQPseudo;
+def VST3d32Pseudo : VSTQQPseudo;
+
// ...with address register writeback:
class VST3DWB<bits<4> op11_8, bits<4> op7_4, string Dt>
: NLdSt<0, 0b00, op11_8, op7_4, (outs GPR:$wb),
def VST3d16_UPD : VST3DWB<0b0100, 0b0100, "16">;
def VST3d32_UPD : VST3DWB<0b0100, 0b1000, "32">;
+def VST3d8Pseudo_UPD : VSTQQWBPseudo;
+def VST3d16Pseudo_UPD : VSTQQWBPseudo;
+def VST3d32Pseudo_UPD : VSTQQWBPseudo;
+
// ...with double-spaced registers (non-updating versions for disassembly only):
def VST3q8 : VST3D<0b0101, 0b0000, "8">;
def VST3q16 : VST3D<0b0101, 0b0100, "16">;
def VST3q16_UPD : VST3DWB<0b0101, 0b0100, "16">;
def VST3q32_UPD : VST3DWB<0b0101, 0b1000, "32">;
+def VST3q8Pseudo_UPD : VSTQQQQWBPseudo;
+def VST3q16Pseudo_UPD : VSTQQQQWBPseudo;
+def VST3q32Pseudo_UPD : VSTQQQQWBPseudo;
+
// ...alternate versions to be allocated odd register numbers:
-def VST3q8odd_UPD : VST3DWB<0b0101, 0b0000, "8">;
-def VST3q16odd_UPD : VST3DWB<0b0101, 0b0100, "16">;
-def VST3q32odd_UPD : VST3DWB<0b0101, 0b1000, "32">;
+def VST3q8oddPseudo_UPD : VSTQQQQWBPseudo;
+def VST3q16oddPseudo_UPD : VSTQQQQWBPseudo;
+def VST3q32oddPseudo_UPD : VSTQQQQWBPseudo;
// VST4 : Vector Store (multiple 4-element structures)
class VST4D<bits<4> op11_8, bits<4> op7_4, string Dt>
def VST4d16 : VST4D<0b0000, 0b0100, "16">;
def VST4d32 : VST4D<0b0000, 0b1000, "32">;
+def VST4d8Pseudo : VSTQQPseudo;
+def VST4d16Pseudo : VSTQQPseudo;
+def VST4d32Pseudo : VSTQQPseudo;
+
// ...with address register writeback:
class VST4DWB<bits<4> op11_8, bits<4> op7_4, string Dt>
: NLdSt<0, 0b00, op11_8, op7_4, (outs GPR:$wb),
def VST4d16_UPD : VST4DWB<0b0000, 0b0100, "16">;
def VST4d32_UPD : VST4DWB<0b0000, 0b1000, "32">;
+def VST4d8Pseudo_UPD : VSTQQWBPseudo;
+def VST4d16Pseudo_UPD : VSTQQWBPseudo;
+def VST4d32Pseudo_UPD : VSTQQWBPseudo;
+
// ...with double-spaced registers (non-updating versions for disassembly only):
def VST4q8 : VST4D<0b0001, 0b0000, "8">;
def VST4q16 : VST4D<0b0001, 0b0100, "16">;
def VST4q16_UPD : VST4DWB<0b0001, 0b0100, "16">;
def VST4q32_UPD : VST4DWB<0b0001, 0b1000, "32">;
+def VST4q8Pseudo_UPD : VSTQQQQWBPseudo;
+def VST4q16Pseudo_UPD : VSTQQQQWBPseudo;
+def VST4q32Pseudo_UPD : VSTQQQQWBPseudo;
+
// ...alternate versions to be allocated odd register numbers:
-def VST4q8odd_UPD : VST4DWB<0b0001, 0b0000, "8">;
-def VST4q16odd_UPD : VST4DWB<0b0001, 0b0100, "16">;
-def VST4q32odd_UPD : VST4DWB<0b0001, 0b1000, "32">;
+def VST4q8oddPseudo_UPD : VSTQQQQWBPseudo;
+def VST4q16oddPseudo_UPD : VSTQQQQWBPseudo;
+def VST4q32oddPseudo_UPD : VSTQQQQWBPseudo;
+
+// Classes for VST*LN pseudo-instructions with multi-register operands.
+// These are expanded to real instructions after register allocation.
+class VSTQLNPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs), (ins addrmode6:$addr, QPR:$src, nohash_imm:$lane),
+ itin, "">;
+class VSTQLNWBPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset, QPR:$src,
+ nohash_imm:$lane), itin, "$addr.addr = $wb">;
+class VSTQQLNPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs), (ins addrmode6:$addr, QQPR:$src, nohash_imm:$lane),
+ itin, "">;
+class VSTQQLNWBPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset, QQPR:$src,
+ nohash_imm:$lane), itin, "$addr.addr = $wb">;
+class VSTQQQQLNPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs), (ins addrmode6:$addr, QQQQPR:$src, nohash_imm:$lane),
+ itin, "">;
+class VSTQQQQLNWBPseudo<InstrItinClass itin>
+ : PseudoNLdSt<(outs GPR:$wb),
+ (ins addrmode6:$addr, am6offset:$offset, QQQQPR:$src,
+ nohash_imm:$lane), itin, "$addr.addr = $wb">;
// VST1LN : Vector Store (single element from one lane)
// FIXME: Not yet implemented.
def VST2LNd16 : VST2LN<0b0101, {?,?,0,?}, "16">;
def VST2LNd32 : VST2LN<0b1001, {?,0,?,?}, "32">;
+def VST2LNd8Pseudo : VSTQLNPseudo<IIC_VST>;
+def VST2LNd16Pseudo : VSTQLNPseudo<IIC_VST>;
+def VST2LNd32Pseudo : VSTQLNPseudo<IIC_VST>;
+
// ...with double-spaced registers:
def VST2LNq16 : VST2LN<0b0101, {?,?,1,?}, "16">;
def VST2LNq32 : VST2LN<0b1001, {?,1,?,?}, "32">;
-// ...alternate versions to be allocated odd register numbers:
-def VST2LNq16odd : VST2LN<0b0101, {?,?,1,?}, "16">;
-def VST2LNq32odd : VST2LN<0b1001, {?,1,?,?}, "32">;
+def VST2LNq16Pseudo : VSTQQLNPseudo<IIC_VST>;
+def VST2LNq32Pseudo : VSTQQLNPseudo<IIC_VST>;
// ...with address register writeback:
class VST2LNWB<bits<4> op11_8, bits<4> op7_4, string Dt>
def VST2LNd16_UPD : VST2LNWB<0b0101, {?,?,0,?}, "16">;
def VST2LNd32_UPD : VST2LNWB<0b1001, {?,0,?,?}, "32">;
+def VST2LNd8Pseudo_UPD : VSTQLNWBPseudo<IIC_VST>;
+def VST2LNd16Pseudo_UPD : VSTQLNWBPseudo<IIC_VST>;
+def VST2LNd32Pseudo_UPD : VSTQLNWBPseudo<IIC_VST>;
+
def VST2LNq16_UPD : VST2LNWB<0b0101, {?,?,1,?}, "16">;
def VST2LNq32_UPD : VST2LNWB<0b1001, {?,1,?,?}, "32">;
+def VST2LNq16Pseudo_UPD : VSTQQLNWBPseudo<IIC_VST>;
+def VST2LNq32Pseudo_UPD : VSTQQLNWBPseudo<IIC_VST>;
+
// VST3LN : Vector Store (single 3-element structure from one lane)
class VST3LN<bits<4> op11_8, bits<4> op7_4, string Dt>
: NLdSt<1, 0b00, op11_8, op7_4, (outs),
def VST3LNd16 : VST3LN<0b0110, {?,?,0,0}, "16">;
def VST3LNd32 : VST3LN<0b1010, {?,0,0,0}, "32">;
+def VST3LNd8Pseudo : VSTQQLNPseudo<IIC_VST>;
+def VST3LNd16Pseudo : VSTQQLNPseudo<IIC_VST>;
+def VST3LNd32Pseudo : VSTQQLNPseudo<IIC_VST>;
+
// ...with double-spaced registers:
def VST3LNq16 : VST3LN<0b0110, {?,?,1,0}, "16">;
def VST3LNq32 : VST3LN<0b1010, {?,1,0,0}, "32">;
-// ...alternate versions to be allocated odd register numbers:
-def VST3LNq16odd : VST3LN<0b0110, {?,?,1,0}, "16">;
-def VST3LNq32odd : VST3LN<0b1010, {?,1,0,0}, "32">;
+def VST3LNq16Pseudo : VSTQQQQLNPseudo<IIC_VST>;
+def VST3LNq32Pseudo : VSTQQQQLNPseudo<IIC_VST>;
// ...with address register writeback:
class VST3LNWB<bits<4> op11_8, bits<4> op7_4, string Dt>
def VST3LNd16_UPD : VST3LNWB<0b0110, {?,?,0,0}, "16">;
def VST3LNd32_UPD : VST3LNWB<0b1010, {?,0,0,0}, "32">;
+def VST3LNd8Pseudo_UPD : VSTQQLNWBPseudo<IIC_VST>;
+def VST3LNd16Pseudo_UPD : VSTQQLNWBPseudo<IIC_VST>;
+def VST3LNd32Pseudo_UPD : VSTQQLNWBPseudo<IIC_VST>;
+
def VST3LNq16_UPD : VST3LNWB<0b0110, {?,?,1,0}, "16">;
def VST3LNq32_UPD : VST3LNWB<0b1010, {?,1,0,0}, "32">;
+def VST3LNq16Pseudo_UPD : VSTQQQQLNWBPseudo<IIC_VST>;
+def VST3LNq32Pseudo_UPD : VSTQQQQLNWBPseudo<IIC_VST>;
+
// VST4LN : Vector Store (single 4-element structure from one lane)
class VST4LN<bits<4> op11_8, bits<4> op7_4, string Dt>
: NLdSt<1, 0b00, op11_8, op7_4, (outs),
def VST4LNd16 : VST4LN<0b0111, {?,?,0,?}, "16">;
def VST4LNd32 : VST4LN<0b1011, {?,0,?,?}, "32">;
+def VST4LNd8Pseudo : VSTQQLNPseudo<IIC_VST>;
+def VST4LNd16Pseudo : VSTQQLNPseudo<IIC_VST>;
+def VST4LNd32Pseudo : VSTQQLNPseudo<IIC_VST>;
+
// ...with double-spaced registers:
def VST4LNq16 : VST4LN<0b0111, {?,?,1,?}, "16">;
def VST4LNq32 : VST4LN<0b1011, {?,1,?,?}, "32">;
-// ...alternate versions to be allocated odd register numbers:
-def VST4LNq16odd : VST4LN<0b0111, {?,?,1,?}, "16">;
-def VST4LNq32odd : VST4LN<0b1011, {?,1,?,?}, "32">;
+def VST4LNq16Pseudo : VSTQQQQLNPseudo<IIC_VST>;
+def VST4LNq32Pseudo : VSTQQQQLNPseudo<IIC_VST>;
// ...with address register writeback:
class VST4LNWB<bits<4> op11_8, bits<4> op7_4, string Dt>
def VST4LNd16_UPD : VST4LNWB<0b0111, {?,?,0,?}, "16">;
def VST4LNd32_UPD : VST4LNWB<0b1011, {?,0,?,?}, "32">;
+def VST4LNd8Pseudo_UPD : VSTQQLNWBPseudo<IIC_VST>;
+def VST4LNd16Pseudo_UPD : VSTQQLNWBPseudo<IIC_VST>;
+def VST4LNd32Pseudo_UPD : VSTQQLNWBPseudo<IIC_VST>;
+
def VST4LNq16_UPD : VST4LNWB<0b0111, {?,?,1,?}, "16">;
def VST4LNq32_UPD : VST4LNWB<0b1011, {?,1,?,?}, "32">;
-} // mayStore = 1, hasExtraSrcRegAllocReq = 1
+def VST4LNq16Pseudo_UPD : VSTQQQQLNWBPseudo<IIC_VST>;
+def VST4LNq32Pseudo_UPD : VSTQQQQLNWBPseudo<IIC_VST>;
+
+} // mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1
//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//
// Extract D sub-registers of Q registers.
-// (arm_dsubreg_0 is 5; arm_dsubreg_1 is 6)
def DSubReg_i8_reg : SDNodeXForm<imm, [{
- return CurDAG->getTargetConstant(5 + N->getZExtValue() / 8, MVT::i32);
+ assert(ARM::dsub_7 == ARM::dsub_0+7 && "Unexpected subreg numbering");
+ return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue()/8, MVT::i32);
}]>;
def DSubReg_i16_reg : SDNodeXForm<imm, [{
- return CurDAG->getTargetConstant(5 + N->getZExtValue() / 4, MVT::i32);
+ assert(ARM::dsub_7 == ARM::dsub_0+7 && "Unexpected subreg numbering");
+ return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue()/4, MVT::i32);
}]>;
def DSubReg_i32_reg : SDNodeXForm<imm, [{
- return CurDAG->getTargetConstant(5 + N->getZExtValue() / 2, MVT::i32);
+ assert(ARM::dsub_7 == ARM::dsub_0+7 && "Unexpected subreg numbering");
+ return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue()/2, MVT::i32);
}]>;
def DSubReg_f64_reg : SDNodeXForm<imm, [{
- return CurDAG->getTargetConstant(5 + N->getZExtValue(), MVT::i32);
-}]>;
-def DSubReg_f64_other_reg : SDNodeXForm<imm, [{
- return CurDAG->getTargetConstant(5 + (1 - N->getZExtValue()), MVT::i32);
+ assert(ARM::dsub_7 == ARM::dsub_0+7 && "Unexpected subreg numbering");
+ return CurDAG->getTargetConstant(ARM::dsub_0 + N->getZExtValue(), MVT::i32);
}]>;
// Extract S sub-registers of Q/D registers.
-// (arm_ssubreg_0 is 1; arm_ssubreg_1 is 2; etc.)
def SSubReg_f32_reg : SDNodeXForm<imm, [{
- return CurDAG->getTargetConstant(1 + N->getZExtValue(), MVT::i32);
+ assert(ARM::ssub_3 == ARM::ssub_0+3 && "Unexpected subreg numbering");
+ return CurDAG->getTargetConstant(ARM::ssub_0 + N->getZExtValue(), MVT::i32);
}]>;
// Translate lane numbers from Q registers to D subregs.
(ins QPR:$src), itin, OpcodeStr, Dt, "$dst, $src", "",
[(set QPR:$dst, (ResTy (IntOp (OpTy QPR:$src))))]>;
+// Narrow 2-register operations.
+class N2VN<bits<2> op24_23, bits<2> op21_20, bits<2> op19_18,
+ bits<2> op17_16, bits<5> op11_7, bit op6, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType TyD, ValueType TyQ, SDNode OpNode>
+ : N2V<op24_23, op21_20, op19_18, op17_16, op11_7, op6, op4, (outs DPR:$dst),
+ (ins QPR:$src), itin, OpcodeStr, Dt, "$dst, $src", "",
+ [(set DPR:$dst, (TyD (OpNode (TyQ QPR:$src))))]>;
+
// Narrow 2-register intrinsics.
class N2VNInt<bits<2> op24_23, bits<2> op21_20, bits<2> op19_18,
bits<2> op17_16, bits<5> op11_7, bit op6, bit op4,
(ins QPR:$src), itin, OpcodeStr, Dt, "$dst, $src", "",
[(set DPR:$dst, (TyD (IntOp (TyQ QPR:$src))))]>;
-// Long 2-register intrinsics (currently only used for VMOVL).
-class N2VLInt<bits<2> op24_23, bits<2> op21_20, bits<2> op19_18,
- bits<2> op17_16, bits<5> op11_7, bit op6, bit op4,
- InstrItinClass itin, string OpcodeStr, string Dt,
- ValueType TyQ, ValueType TyD, Intrinsic IntOp>
+// Long 2-register operations (currently only used for VMOVL).
+class N2VL<bits<2> op24_23, bits<2> op21_20, bits<2> op19_18,
+ bits<2> op17_16, bits<5> op11_7, bit op6, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType TyQ, ValueType TyD, SDNode OpNode>
: N2V<op24_23, op21_20, op19_18, op17_16, op11_7, op6, op4, (outs QPR:$dst),
(ins DPR:$src), itin, OpcodeStr, Dt, "$dst, $src", "",
- [(set QPR:$dst, (TyQ (IntOp (TyD DPR:$src))))]>;
+ [(set QPR:$dst, (TyQ (OpNode (TyD DPR:$src))))]>;
// 2-register shuffles (VTRN/VZIP/VUZP), both double- and quad-register.
class N2VDShuffle<bits<2> op19_18, bits<5> op11_7, string OpcodeStr, string Dt>
(ResTy (NEONvduplane (OpTy DPR_8:$src3),
imm:$lane)))))))]>;
+// Neon Intrinsic-Op instructions (VABA): double- and quad-register.
+class N3VDIntOp<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType Ty, Intrinsic IntOp, SDNode OpNode>
+ : N3V<op24, op23, op21_20, op11_8, 0, op4,
+ (outs DPR:$dst), (ins DPR:$src1, DPR:$src2, DPR:$src3), N3RegFrm, itin,
+ OpcodeStr, Dt, "$dst, $src2, $src3", "$src1 = $dst",
+ [(set DPR:$dst, (Ty (OpNode DPR:$src1,
+ (Ty (IntOp (Ty DPR:$src2), (Ty DPR:$src3))))))]>;
+class N3VQIntOp<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType Ty, Intrinsic IntOp, SDNode OpNode>
+ : N3V<op24, op23, op21_20, op11_8, 1, op4,
+ (outs QPR:$dst), (ins QPR:$src1, QPR:$src2, QPR:$src3), N3RegFrm, itin,
+ OpcodeStr, Dt, "$dst, $src2, $src3", "$src1 = $dst",
+ [(set QPR:$dst, (Ty (OpNode QPR:$src1,
+ (Ty (IntOp (Ty QPR:$src2), (Ty QPR:$src3))))))]>;
+
// Neon 3-argument intrinsics, both double- and quad-register.
// The destination register is also used as the first source operand register.
class N3VDInt3<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
[(set QPR:$dst, (ResTy (IntOp (OpTy QPR:$src1),
(OpTy QPR:$src2), (OpTy QPR:$src3))))]>;
+// Long Multiply-Add/Sub operations.
+class N3VLMulOp<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType TyQ, ValueType TyD, SDNode MulOp, SDNode OpNode>
+ : N3V<op24, op23, op21_20, op11_8, 0, op4,
+ (outs QPR:$dst), (ins QPR:$src1, DPR:$src2, DPR:$src3), N3RegFrm, itin,
+ OpcodeStr, Dt, "$dst, $src2, $src3", "$src1 = $dst",
+ [(set QPR:$dst, (OpNode (TyQ QPR:$src1),
+ (TyQ (MulOp (TyD DPR:$src2),
+ (TyD DPR:$src3)))))]>;
+class N3VLMulOpSL<bit op24, bits<2> op21_20, bits<4> op11_8,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType TyQ, ValueType TyD, SDNode MulOp, SDNode OpNode>
+ : N3V<op24, 1, op21_20, op11_8, 1, 0, (outs QPR:$dst),
+ (ins QPR:$src1, DPR:$src2, DPR_VFP2:$src3, nohash_imm:$lane),
+ NVMulSLFrm, itin,
+ OpcodeStr, Dt, "$dst, $src2, $src3[$lane]", "$src1 = $dst",
+ [(set QPR:$dst,
+ (OpNode (TyQ QPR:$src1),
+ (TyQ (MulOp (TyD DPR:$src2),
+ (TyD (NEONvduplane (TyD DPR_VFP2:$src3),
+ imm:$lane))))))]>;
+class N3VLMulOpSL16<bit op24, bits<2> op21_20, bits<4> op11_8,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType TyQ, ValueType TyD, SDNode MulOp, SDNode OpNode>
+ : N3V<op24, 1, op21_20, op11_8, 1, 0, (outs QPR:$dst),
+ (ins QPR:$src1, DPR:$src2, DPR_8:$src3, nohash_imm:$lane),
+ NVMulSLFrm, itin,
+ OpcodeStr, Dt, "$dst, $src2, $src3[$lane]", "$src1 = $dst",
+ [(set QPR:$dst,
+ (OpNode (TyQ QPR:$src1),
+ (TyQ (MulOp (TyD DPR:$src2),
+ (TyD (NEONvduplane (TyD DPR_8:$src3),
+ imm:$lane))))))]>;
+
+// Long Intrinsic-Op vector operations with explicit extend (VABAL).
+class N3VLIntExtOp<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType TyQ, ValueType TyD, Intrinsic IntOp, SDNode ExtOp,
+ SDNode OpNode>
+ : N3V<op24, op23, op21_20, op11_8, 0, op4,
+ (outs QPR:$dst), (ins QPR:$src1, DPR:$src2, DPR:$src3), N3RegFrm, itin,
+ OpcodeStr, Dt, "$dst, $src2, $src3", "$src1 = $dst",
+ [(set QPR:$dst, (OpNode (TyQ QPR:$src1),
+ (TyQ (ExtOp (TyD (IntOp (TyD DPR:$src2),
+ (TyD DPR:$src3)))))))]>;
+
// Neon Long 3-argument intrinsic. The destination register is
// a quad-register and is also used as the first source operand register.
class N3VLInt3<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
let isCommutable = Commutable;
}
+// Long 3-register operations.
+class N3VL<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType TyQ, ValueType TyD, SDNode OpNode, bit Commutable>
+ : N3V<op24, op23, op21_20, op11_8, 0, op4,
+ (outs QPR:$dst), (ins DPR:$src1, DPR:$src2), N3RegFrm, itin,
+ OpcodeStr, Dt, "$dst, $src1, $src2", "",
+ [(set QPR:$dst, (TyQ (OpNode (TyD DPR:$src1), (TyD DPR:$src2))))]> {
+ 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>
+ : N3V<op24, 1, op21_20, op11_8, 1, 0,
+ (outs QPR:$dst), (ins DPR:$src1, DPR_VFP2:$src2, nohash_imm:$lane),
+ NVMulSLFrm, itin, OpcodeStr, Dt, "$dst, $src1, $src2[$lane]", "",
+ [(set QPR:$dst,
+ (TyQ (OpNode (TyD DPR:$src1),
+ (TyD (NEONvduplane (TyD DPR_VFP2:$src2),imm:$lane)))))]>;
+class N3VLSL16<bit op24, bits<2> op21_20, bits<4> op11_8,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType TyQ, ValueType TyD, SDNode OpNode>
+ : N3V<op24, 1, op21_20, op11_8, 1, 0,
+ (outs QPR:$dst), (ins DPR:$src1, DPR_8:$src2, nohash_imm:$lane),
+ NVMulSLFrm, itin, OpcodeStr, Dt, "$dst, $src1, $src2[$lane]", "",
+ [(set QPR:$dst,
+ (TyQ (OpNode (TyD DPR:$src1),
+ (TyD (NEONvduplane (TyD DPR_8:$src2), imm:$lane)))))]>;
+
+// Long 3-register operations with explicitly extended operands.
+class N3VLExt<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType TyQ, ValueType TyD, SDNode OpNode, SDNode ExtOp,
+ bit Commutable>
+ : N3V<op24, op23, op21_20, op11_8, 0, op4,
+ (outs QPR:$dst), (ins DPR:$src1, DPR:$src2), N3RegFrm, itin,
+ OpcodeStr, Dt, "$dst, $src1, $src2", "",
+ [(set QPR:$dst, (OpNode (TyQ (ExtOp (TyD DPR:$src1))),
+ (TyQ (ExtOp (TyD DPR:$src2)))))]> {
+ let isCommutable = Commutable;
+}
+
+// Long 3-register intrinsics with explicit extend (VABDL).
+class N3VLIntExt<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ ValueType TyQ, ValueType TyD, Intrinsic IntOp, SDNode ExtOp,
+ bit Commutable>
+ : N3V<op24, op23, op21_20, op11_8, 0, op4,
+ (outs QPR:$dst), (ins DPR:$src1, DPR:$src2), N3RegFrm, itin,
+ OpcodeStr, Dt, "$dst, $src1, $src2", "",
+ [(set QPR:$dst, (TyQ (ExtOp (TyD (IntOp (TyD DPR:$src1),
+ (TyD DPR:$src2))))))]> {
+ let isCommutable = Commutable;
+}
+
// Long 3-register intrinsics.
class N3VLInt<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
InstrItinClass itin, string OpcodeStr, string Dt,
(OpTy (NEONvduplane (OpTy DPR_8:$src2),
imm:$lane)))))]>;
-// Wide 3-register intrinsics.
-class N3VWInt<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
- string OpcodeStr, string Dt, ValueType TyQ, ValueType TyD,
- Intrinsic IntOp, bit Commutable>
+// Wide 3-register operations.
+class N3VW<bit op24, bit op23, bits<2> op21_20, bits<4> op11_8, bit op4,
+ string OpcodeStr, string Dt, ValueType TyQ, ValueType TyD,
+ SDNode OpNode, SDNode ExtOp, bit Commutable>
: N3V<op24, op23, op21_20, op11_8, 0, op4,
(outs QPR:$dst), (ins QPR:$src1, DPR:$src2), N3RegFrm, IIC_VSUBiD,
OpcodeStr, Dt, "$dst, $src1, $src2", "",
- [(set QPR:$dst, (TyQ (IntOp (TyQ QPR:$src1), (TyD DPR:$src2))))]> {
+ [(set QPR:$dst, (OpNode (TyQ QPR:$src1),
+ (TyQ (ExtOp (TyD DPR:$src2)))))]> {
let isCommutable = Commutable;
}
}
+// Neon Narrowing 2-register vector operations,
+// source operand element sizes of 16, 32 and 64 bits:
+multiclass N2VN_HSD<bits<2> op24_23, bits<2> op21_20, bits<2> op17_16,
+ bits<5> op11_7, bit op6, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ SDNode OpNode> {
+ def v8i8 : N2VN<op24_23, op21_20, 0b00, op17_16, op11_7, op6, op4,
+ itin, OpcodeStr, !strconcat(Dt, "16"),
+ v8i8, v8i16, OpNode>;
+ def v4i16 : N2VN<op24_23, op21_20, 0b01, op17_16, op11_7, op6, op4,
+ itin, OpcodeStr, !strconcat(Dt, "32"),
+ v4i16, v4i32, OpNode>;
+ def v2i32 : N2VN<op24_23, op21_20, 0b10, op17_16, op11_7, op6, op4,
+ itin, OpcodeStr, !strconcat(Dt, "64"),
+ v2i32, v2i64, OpNode>;
+}
+
// Neon Narrowing 2-register vector intrinsics,
// source operand element sizes of 16, 32 and 64 bits:
multiclass N2VNInt_HSD<bits<2> op24_23, bits<2> op21_20, bits<2> op17_16,
// Neon Lengthening 2-register vector intrinsic (currently specific to VMOVL).
// source operand element sizes of 16, 32 and 64 bits:
-multiclass N2VLInt_QHS<bits<2> op24_23, bits<5> op11_7, bit op6, bit op4,
- string OpcodeStr, string Dt, Intrinsic IntOp> {
- def v8i16 : N2VLInt<op24_23, 0b00, 0b10, 0b00, op11_7, op6, op4, IIC_VQUNAiD,
- OpcodeStr, !strconcat(Dt, "8"), v8i16, v8i8, IntOp>;
- def v4i32 : N2VLInt<op24_23, 0b01, 0b00, 0b00, op11_7, op6, op4, IIC_VQUNAiD,
- OpcodeStr, !strconcat(Dt, "16"), v4i32, v4i16, IntOp>;
- def v2i64 : N2VLInt<op24_23, 0b10, 0b00, 0b00, op11_7, op6, op4, IIC_VQUNAiD,
- OpcodeStr, !strconcat(Dt, "32"), v2i64, v2i32, IntOp>;
+multiclass N2VL_QHS<bits<2> op24_23, bits<5> op11_7, bit op6, bit op4,
+ string OpcodeStr, string Dt, SDNode OpNode> {
+ def v8i16 : N2VL<op24_23, 0b00, 0b10, 0b00, op11_7, op6, op4, IIC_VQUNAiD,
+ OpcodeStr, !strconcat(Dt, "8"), v8i16, v8i8, OpNode>;
+ def v4i32 : N2VL<op24_23, 0b01, 0b00, 0b00, op11_7, op6, op4, IIC_VQUNAiD,
+ OpcodeStr, !strconcat(Dt, "16"), v4i32, v4i16, OpNode>;
+ def v2i64 : N2VL<op24_23, 0b10, 0b00, 0b00, op11_7, op6, op4, IIC_VQUNAiD,
+ OpcodeStr, !strconcat(Dt, "32"), v2i64, v2i32, OpNode>;
}
}
+// Neon Long 3-register vector operations.
+
+multiclass N3VL_QHS<bit op24, bit op23, bits<4> op11_8, bit op4,
+ InstrItinClass itin16, InstrItinClass itin32,
+ string OpcodeStr, string Dt,
+ SDNode OpNode, bit Commutable = 0> {
+ def v8i16 : N3VL<op24, op23, 0b00, op11_8, op4, itin16,
+ OpcodeStr, !strconcat(Dt, "8"),
+ v8i16, v8i8, OpNode, Commutable>;
+ def v4i32 : N3VL<op24, op23, 0b01, op11_8, op4, itin16,
+ OpcodeStr, !strconcat(Dt, "16"),
+ v4i32, v4i16, OpNode, Commutable>;
+ def v2i64 : N3VL<op24, op23, 0b10, op11_8, op4, itin32,
+ OpcodeStr, !strconcat(Dt, "32"),
+ v2i64, v2i32, OpNode, Commutable>;
+}
+
+multiclass N3VLSL_HS<bit op24, bits<4> op11_8,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ SDNode OpNode> {
+ def v4i16 : N3VLSL16<op24, 0b01, op11_8, itin, OpcodeStr,
+ !strconcat(Dt, "16"), v4i32, v4i16, OpNode>;
+ def v2i32 : N3VLSL<op24, 0b10, op11_8, itin, OpcodeStr,
+ !strconcat(Dt, "32"), v2i64, v2i32, OpNode>;
+}
+
+multiclass N3VLExt_QHS<bit op24, bit op23, bits<4> op11_8, bit op4,
+ InstrItinClass itin16, InstrItinClass itin32,
+ string OpcodeStr, string Dt,
+ SDNode OpNode, SDNode ExtOp, bit Commutable = 0> {
+ def v8i16 : N3VLExt<op24, op23, 0b00, op11_8, op4, itin16,
+ OpcodeStr, !strconcat(Dt, "8"),
+ v8i16, v8i8, OpNode, ExtOp, Commutable>;
+ def v4i32 : N3VLExt<op24, op23, 0b01, op11_8, op4, itin16,
+ OpcodeStr, !strconcat(Dt, "16"),
+ v4i32, v4i16, OpNode, ExtOp, Commutable>;
+ def v2i64 : N3VLExt<op24, op23, 0b10, op11_8, op4, itin32,
+ OpcodeStr, !strconcat(Dt, "32"),
+ v2i64, v2i32, OpNode, ExtOp, Commutable>;
+}
+
// Neon Long 3-register vector intrinsics.
// First with only element sizes of 16 and 32 bits:
v8i16, v8i8, IntOp, Commutable>;
}
+// ....with explicit extend (VABDL).
+multiclass N3VLIntExt_QHS<bit op24, bit op23, bits<4> op11_8, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ Intrinsic IntOp, SDNode ExtOp, bit Commutable = 0> {
+ def v8i16 : N3VLIntExt<op24, op23, 0b00, op11_8, op4, itin,
+ OpcodeStr, !strconcat(Dt, "8"),
+ v8i16, v8i8, IntOp, ExtOp, Commutable>;
+ def v4i32 : N3VLIntExt<op24, op23, 0b01, op11_8, op4, itin,
+ OpcodeStr, !strconcat(Dt, "16"),
+ v4i32, v4i16, IntOp, ExtOp, Commutable>;
+ def v2i64 : N3VLIntExt<op24, op23, 0b10, op11_8, op4, itin,
+ OpcodeStr, !strconcat(Dt, "32"),
+ v2i64, v2i32, IntOp, ExtOp, Commutable>;
+}
+
// Neon Wide 3-register vector intrinsics,
// source operand element sizes of 8, 16 and 32 bits:
-multiclass N3VWInt_QHS<bit op24, bit op23, bits<4> op11_8, bit op4,
- string OpcodeStr, string Dt,
- Intrinsic IntOp, bit Commutable = 0> {
- def v8i16 : N3VWInt<op24, op23, 0b00, op11_8, op4,
- OpcodeStr, !strconcat(Dt, "8"),
- v8i16, v8i8, IntOp, Commutable>;
- def v4i32 : N3VWInt<op24, op23, 0b01, op11_8, op4,
- OpcodeStr, !strconcat(Dt, "16"),
- v4i32, v4i16, IntOp, Commutable>;
- def v2i64 : N3VWInt<op24, op23, 0b10, op11_8, op4,
- OpcodeStr, !strconcat(Dt, "32"),
- v2i64, v2i32, IntOp, Commutable>;
+multiclass N3VW_QHS<bit op24, bit op23, bits<4> op11_8, bit op4,
+ string OpcodeStr, string Dt,
+ SDNode OpNode, SDNode ExtOp, bit Commutable = 0> {
+ def v8i16 : N3VW<op24, op23, 0b00, op11_8, op4,
+ OpcodeStr, !strconcat(Dt, "8"),
+ v8i16, v8i8, OpNode, ExtOp, Commutable>;
+ def v4i32 : N3VW<op24, op23, 0b01, op11_8, op4,
+ OpcodeStr, !strconcat(Dt, "16"),
+ v4i32, v4i16, OpNode, ExtOp, Commutable>;
+ def v2i64 : N3VW<op24, op23, 0b10, op11_8, op4,
+ OpcodeStr, !strconcat(Dt, "32"),
+ v2i64, v2i32, OpNode, ExtOp, Commutable>;
}
mul, ShOp>;
}
+// Neon Intrinsic-Op vector operations,
+// element sizes of 8, 16 and 32 bits:
+multiclass N3VIntOp_QHS<bit op24, bit op23, bits<4> op11_8, bit op4,
+ InstrItinClass itinD, InstrItinClass itinQ,
+ string OpcodeStr, string Dt, Intrinsic IntOp,
+ SDNode OpNode> {
+ // 64-bit vector types.
+ def v8i8 : N3VDIntOp<op24, op23, 0b00, op11_8, op4, itinD,
+ OpcodeStr, !strconcat(Dt, "8"), v8i8, IntOp, OpNode>;
+ def v4i16 : N3VDIntOp<op24, op23, 0b01, op11_8, op4, itinD,
+ OpcodeStr, !strconcat(Dt, "16"), v4i16, IntOp, OpNode>;
+ def v2i32 : N3VDIntOp<op24, op23, 0b10, op11_8, op4, itinD,
+ OpcodeStr, !strconcat(Dt, "32"), v2i32, IntOp, OpNode>;
+
+ // 128-bit vector types.
+ def v16i8 : N3VQIntOp<op24, op23, 0b00, op11_8, op4, itinQ,
+ OpcodeStr, !strconcat(Dt, "8"), v16i8, IntOp, OpNode>;
+ def v8i16 : N3VQIntOp<op24, op23, 0b01, op11_8, op4, itinQ,
+ OpcodeStr, !strconcat(Dt, "16"), v8i16, IntOp, OpNode>;
+ def v4i32 : N3VQIntOp<op24, op23, 0b10, op11_8, op4, itinQ,
+ OpcodeStr, !strconcat(Dt, "32"), v4i32, IntOp, OpNode>;
+}
+
// 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,
}
+// Neon Long Multiply-Op vector operations,
+// element sizes of 8, 16 and 32 bits:
+multiclass N3VLMulOp_QHS<bit op24, bit op23, bits<4> op11_8, bit op4,
+ InstrItinClass itin16, InstrItinClass itin32,
+ string OpcodeStr, string Dt, SDNode MulOp,
+ SDNode OpNode> {
+ def v8i16 : N3VLMulOp<op24, op23, 0b00, op11_8, op4, itin16, OpcodeStr,
+ !strconcat(Dt, "8"), v8i16, v8i8, MulOp, OpNode>;
+ def v4i32 : N3VLMulOp<op24, op23, 0b01, op11_8, op4, itin16, OpcodeStr,
+ !strconcat(Dt, "16"), v4i32, v4i16, MulOp, OpNode>;
+ def v2i64 : N3VLMulOp<op24, op23, 0b10, op11_8, op4, itin32, OpcodeStr,
+ !strconcat(Dt, "32"), v2i64, v2i32, MulOp, OpNode>;
+}
+
+multiclass N3VLMulOpSL_HS<bit op24, bits<4> op11_8, string OpcodeStr,
+ string Dt, SDNode MulOp, SDNode OpNode> {
+ def v4i16 : N3VLMulOpSL16<op24, 0b01, op11_8, IIC_VMACi16D, OpcodeStr,
+ !strconcat(Dt,"16"), v4i32, v4i16, MulOp, OpNode>;
+ def v2i32 : N3VLMulOpSL<op24, 0b10, op11_8, IIC_VMACi32D, OpcodeStr,
+ !strconcat(Dt, "32"), v2i64, v2i32, MulOp, OpNode>;
+}
+
+
// Neon Long 3-argument intrinsics.
// First with only element sizes of 16 and 32 bits:
OpcodeStr, !strconcat(Dt, "8"), v8i16, v8i8, IntOp>;
}
+// ....with explicit extend (VABAL).
+multiclass N3VLIntExtOp_QHS<bit op24, bit op23, bits<4> op11_8, bit op4,
+ InstrItinClass itin, string OpcodeStr, string Dt,
+ Intrinsic IntOp, SDNode ExtOp, SDNode OpNode> {
+ def v8i16 : N3VLIntExtOp<op24, op23, 0b00, op11_8, op4, itin,
+ OpcodeStr, !strconcat(Dt, "8"), v8i16, v8i8,
+ IntOp, ExtOp, OpNode>;
+ def v4i32 : N3VLIntExtOp<op24, op23, 0b01, op11_8, op4, itin,
+ OpcodeStr, !strconcat(Dt, "16"), v4i32, v4i16,
+ IntOp, ExtOp, OpNode>;
+ def v2i64 : N3VLIntExtOp<op24, op23, 0b10, op11_8, op4, itin,
+ OpcodeStr, !strconcat(Dt, "32"), v2i64, v2i32,
+ IntOp, ExtOp, OpNode>;
+}
+
// Neon 2-register vector intrinsics,
// element sizes of 8, 16 and 32 bits:
def VADDfq : N3VQ<0, 0, 0b00, 0b1101, 0, IIC_VBINQ, "vadd", "f32",
v4f32, v4f32, fadd, 1>;
// VADDL : Vector Add Long (Q = D + D)
-defm VADDLs : N3VLInt_QHS<0,1,0b0000,0, IIC_VSHLiD, IIC_VSHLiD,
- "vaddl", "s", int_arm_neon_vaddls, 1>;
-defm VADDLu : N3VLInt_QHS<1,1,0b0000,0, IIC_VSHLiD, IIC_VSHLiD,
- "vaddl", "u", int_arm_neon_vaddlu, 1>;
+defm VADDLs : N3VLExt_QHS<0,1,0b0000,0, IIC_VSHLiD, IIC_VSHLiD,
+ "vaddl", "s", add, sext, 1>;
+defm VADDLu : N3VLExt_QHS<1,1,0b0000,0, IIC_VSHLiD, IIC_VSHLiD,
+ "vaddl", "u", add, zext, 1>;
// VADDW : Vector Add Wide (Q = Q + D)
-defm VADDWs : N3VWInt_QHS<0,1,0b0001,0, "vaddw", "s", int_arm_neon_vaddws, 0>;
-defm VADDWu : N3VWInt_QHS<1,1,0b0001,0, "vaddw", "u", int_arm_neon_vaddwu, 0>;
+defm VADDWs : N3VW_QHS<0,1,0b0001,0, "vaddw", "s", add, sext, 0>;
+defm VADDWu : N3VW_QHS<1,1,0b0001,0, "vaddw", "u", add, zext, 0>;
// VHADD : Vector Halving Add
defm VHADDs : N3VInt_QHS<0, 0, 0b0000, 0, N3RegFrm,
IIC_VBINi4D, IIC_VBINi4D, IIC_VBINi4Q, IIC_VBINi4Q,
(SubReg_i32_lane imm:$lane)))>;
// VMULL : Vector Multiply Long (integer and polynomial) (Q = D * D)
-defm VMULLs : N3VLInt_QHS<0,1,0b1100,0, IIC_VMULi16D, IIC_VMULi32D,
- "vmull", "s", int_arm_neon_vmulls, 1>;
-defm VMULLu : N3VLInt_QHS<1,1,0b1100,0, IIC_VMULi16D, IIC_VMULi32D,
- "vmull", "u", int_arm_neon_vmullu, 1>;
+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>;
-defm VMULLsls : N3VLIntSL_HS<0, 0b1010, IIC_VMULi16D, "vmull", "s",
- int_arm_neon_vmulls>;
-defm VMULLslu : N3VLIntSL_HS<1, 0b1010, IIC_VMULi16D, "vmull", "u",
- int_arm_neon_vmullu>;
+defm VMULLsls : N3VLSL_HS<0, 0b1010, IIC_VMULi16D, "vmull", "s", NEONvmulls>;
+defm VMULLslu : N3VLSL_HS<1, 0b1010, IIC_VMULi16D, "vmull", "u", NEONvmullu>;
// VQDMULL : Vector Saturating Doubling Multiply Long (Q = D * D)
defm VQDMULL : N3VLInt_HS<0,1,0b1101,0, IIC_VMULi16D, IIC_VMULi32D,
(SubReg_i32_lane imm:$lane)))>;
// VMLAL : Vector Multiply Accumulate Long (Q += D * D)
-defm VMLALs : N3VLInt3_QHS<0,1,0b1000,0, IIC_VMACi16D, IIC_VMACi32D,
- "vmlal", "s", int_arm_neon_vmlals>;
-defm VMLALu : N3VLInt3_QHS<1,1,0b1000,0, IIC_VMACi16D, IIC_VMACi32D,
- "vmlal", "u", int_arm_neon_vmlalu>;
+defm VMLALs : N3VLMulOp_QHS<0,1,0b1000,0, IIC_VMACi16D, IIC_VMACi32D,
+ "vmlal", "s", NEONvmulls, add>;
+defm VMLALu : N3VLMulOp_QHS<1,1,0b1000,0, IIC_VMACi16D, IIC_VMACi32D,
+ "vmlal", "u", NEONvmullu, add>;
-defm VMLALsls : N3VLInt3SL_HS<0, 0b0010, "vmlal", "s", int_arm_neon_vmlals>;
-defm VMLALslu : N3VLInt3SL_HS<1, 0b0010, "vmlal", "u", int_arm_neon_vmlalu>;
+defm VMLALsls : N3VLMulOpSL_HS<0, 0b0010, "vmlal", "s", NEONvmulls, add>;
+defm VMLALslu : N3VLMulOpSL_HS<1, 0b0010, "vmlal", "u", NEONvmullu, add>;
// VQDMLAL : Vector Saturating Doubling Multiply Accumulate Long (Q += D * D)
defm VQDMLAL : N3VLInt3_HS<0, 1, 0b1001, 0, IIC_VMACi16D, IIC_VMACi32D,
(SubReg_i32_lane imm:$lane)))>;
// VMLSL : Vector Multiply Subtract Long (Q -= D * D)
-defm VMLSLs : N3VLInt3_QHS<0,1,0b1010,0, IIC_VMACi16D, IIC_VMACi32D,
- "vmlsl", "s", int_arm_neon_vmlsls>;
-defm VMLSLu : N3VLInt3_QHS<1,1,0b1010,0, IIC_VMACi16D, IIC_VMACi32D,
- "vmlsl", "u", int_arm_neon_vmlslu>;
+defm VMLSLs : N3VLMulOp_QHS<0,1,0b1010,0, IIC_VMACi16D, IIC_VMACi32D,
+ "vmlsl", "s", NEONvmulls, sub>;
+defm VMLSLu : N3VLMulOp_QHS<1,1,0b1010,0, IIC_VMACi16D, IIC_VMACi32D,
+ "vmlsl", "u", NEONvmullu, sub>;
-defm VMLSLsls : N3VLInt3SL_HS<0, 0b0110, "vmlsl", "s", int_arm_neon_vmlsls>;
-defm VMLSLslu : N3VLInt3SL_HS<1, 0b0110, "vmlsl", "u", int_arm_neon_vmlslu>;
+defm VMLSLsls : N3VLMulOpSL_HS<0, 0b0110, "vmlsl", "s", NEONvmulls, sub>;
+defm VMLSLslu : N3VLMulOpSL_HS<1, 0b0110, "vmlsl", "u", NEONvmullu, sub>;
// VQDMLSL : Vector Saturating Doubling Multiply Subtract Long (Q -= D * D)
defm VQDMLSL : N3VLInt3_HS<0, 1, 0b1011, 0, IIC_VMACi16D, IIC_VMACi32D,
def VSUBfq : N3VQ<0, 0, 0b10, 0b1101, 0, IIC_VBINQ, "vsub", "f32",
v4f32, v4f32, fsub, 0>;
// VSUBL : Vector Subtract Long (Q = D - D)
-defm VSUBLs : N3VLInt_QHS<0,1,0b0010,0, IIC_VSHLiD, IIC_VSHLiD,
- "vsubl", "s", int_arm_neon_vsubls, 1>;
-defm VSUBLu : N3VLInt_QHS<1,1,0b0010,0, IIC_VSHLiD, IIC_VSHLiD,
- "vsubl", "u", int_arm_neon_vsublu, 1>;
+defm VSUBLs : N3VLExt_QHS<0,1,0b0010,0, IIC_VSHLiD, IIC_VSHLiD,
+ "vsubl", "s", sub, sext, 0>;
+defm VSUBLu : N3VLExt_QHS<1,1,0b0010,0, IIC_VSHLiD, IIC_VSHLiD,
+ "vsubl", "u", sub, zext, 0>;
// VSUBW : Vector Subtract Wide (Q = Q - D)
-defm VSUBWs : N3VWInt_QHS<0,1,0b0011,0, "vsubw", "s", int_arm_neon_vsubws, 0>;
-defm VSUBWu : N3VWInt_QHS<1,1,0b0011,0, "vsubw", "u", int_arm_neon_vsubwu, 0>;
+defm VSUBWs : N3VW_QHS<0,1,0b0011,0, "vsubw", "s", sub, sext, 0>;
+defm VSUBWu : N3VW_QHS<1,1,0b0011,0, "vsubw", "u", sub, zext, 0>;
// VHSUB : Vector Halving Subtract
defm VHSUBs : N3VInt_QHS<0, 0, 0b0010, 0, N3RegFrm,
IIC_VSUBi4D, IIC_VSUBi4D, IIC_VSUBi4Q, IIC_VSUBi4Q,
NEONvceq, 1>;
// For disassembly only.
defm VCEQz : N2V_QHS_cmp<0b11, 0b11, 0b01, 0b00010, 0, "vceq", "i",
- "$dst, $src, #0">;
+ "$dst, $src, #0">;
// VCGE : Vector Compare Greater Than or Equal
defm VCGEs : N3V_QHS<0, 0, 0b0011, 1, IIC_VSUBi4D, IIC_VSUBi4D, IIC_VSUBi4Q,
// Vector Bitwise Operations.
-def vnot8 : PatFrag<(ops node:$in),
- (xor node:$in, (bitconvert (v8i8 immAllOnesV)))>;
-def vnot16 : PatFrag<(ops node:$in),
- (xor node:$in, (bitconvert (v16i8 immAllOnesV)))>;
+def vnotd : PatFrag<(ops node:$in),
+ (xor node:$in, (bitconvert (v8i8 NEONimmAllOnesV)))>;
+def vnotq : PatFrag<(ops node:$in),
+ (xor node:$in, (bitconvert (v16i8 NEONimmAllOnesV)))>;
// VAND : Vector Bitwise AND
(ins DPR:$src1, DPR:$src2), N3RegFrm, IIC_VBINiD,
"vbic", "$dst, $src1, $src2", "",
[(set DPR:$dst, (v2i32 (and DPR:$src1,
- (vnot8 DPR:$src2))))]>;
+ (vnotd DPR:$src2))))]>;
def VBICq : N3VX<0, 0, 0b01, 0b0001, 1, 1, (outs QPR:$dst),
(ins QPR:$src1, QPR:$src2), N3RegFrm, IIC_VBINiQ,
"vbic", "$dst, $src1, $src2", "",
[(set QPR:$dst, (v4i32 (and QPR:$src1,
- (vnot16 QPR:$src2))))]>;
+ (vnotq QPR:$src2))))]>;
// VORN : Vector Bitwise OR NOT
def VORNd : N3VX<0, 0, 0b11, 0b0001, 0, 1, (outs DPR:$dst),
(ins DPR:$src1, DPR:$src2), N3RegFrm, IIC_VBINiD,
"vorn", "$dst, $src1, $src2", "",
[(set DPR:$dst, (v2i32 (or DPR:$src1,
- (vnot8 DPR:$src2))))]>;
+ (vnotd DPR:$src2))))]>;
def VORNq : N3VX<0, 0, 0b11, 0b0001, 1, 1, (outs QPR:$dst),
(ins QPR:$src1, QPR:$src2), N3RegFrm, IIC_VBINiQ,
"vorn", "$dst, $src1, $src2", "",
[(set QPR:$dst, (v4i32 (or QPR:$src1,
- (vnot16 QPR:$src2))))]>;
+ (vnotq QPR:$src2))))]>;
+
+// VMVN : Vector Bitwise NOT (Immediate)
+
+let isReMaterializable = 1 in {
+def VMVNv4i16 : N1ModImm<1, 0b000, {1,0,?,0}, 0, 0, 1, 1, (outs DPR:$dst),
+ (ins nModImm:$SIMM), IIC_VMOVImm,
+ "vmvn", "i16", "$dst, $SIMM", "",
+ [(set DPR:$dst, (v4i16 (NEONvmvnImm timm:$SIMM)))]>;
+def VMVNv8i16 : N1ModImm<1, 0b000, {1,0,?,0}, 0, 1, 1, 1, (outs QPR:$dst),
+ (ins nModImm:$SIMM), IIC_VMOVImm,
+ "vmvn", "i16", "$dst, $SIMM", "",
+ [(set QPR:$dst, (v8i16 (NEONvmvnImm timm:$SIMM)))]>;
+
+def VMVNv2i32 : N1ModImm<1, 0b000, {?,?,?,?}, 0, 0, 1, 1, (outs DPR:$dst),
+ (ins nModImm:$SIMM), IIC_VMOVImm,
+ "vmvn", "i32", "$dst, $SIMM", "",
+ [(set DPR:$dst, (v2i32 (NEONvmvnImm timm:$SIMM)))]>;
+def VMVNv4i32 : N1ModImm<1, 0b000, {?,?,?,?}, 0, 1, 1, 1, (outs QPR:$dst),
+ (ins nModImm:$SIMM), IIC_VMOVImm,
+ "vmvn", "i32", "$dst, $SIMM", "",
+ [(set QPR:$dst, (v4i32 (NEONvmvnImm timm:$SIMM)))]>;
+}
// VMVN : Vector Bitwise NOT
def VMVNd : N2VX<0b11, 0b11, 0b00, 0b00, 0b01011, 0, 0,
(outs DPR:$dst), (ins DPR:$src), IIC_VSUBiD,
"vmvn", "$dst, $src", "",
- [(set DPR:$dst, (v2i32 (vnot8 DPR:$src)))]>;
+ [(set DPR:$dst, (v2i32 (vnotd DPR:$src)))]>;
def VMVNq : N2VX<0b11, 0b11, 0b00, 0b00, 0b01011, 1, 0,
(outs QPR:$dst), (ins QPR:$src), IIC_VSUBiD,
"vmvn", "$dst, $src", "",
- [(set QPR:$dst, (v4i32 (vnot16 QPR:$src)))]>;
-def : Pat<(v2i32 (vnot8 DPR:$src)), (VMVNd DPR:$src)>;
-def : Pat<(v4i32 (vnot16 QPR:$src)), (VMVNq QPR:$src)>;
+ [(set QPR:$dst, (v4i32 (vnotq QPR:$src)))]>;
+def : Pat<(v2i32 (vnotd DPR:$src)), (VMVNd DPR:$src)>;
+def : Pat<(v4i32 (vnotq QPR:$src)), (VMVNq QPR:$src)>;
// VBSL : Vector Bitwise Select
def VBSLd : N3VX<1, 0, 0b01, 0b0001, 0, 1, (outs DPR:$dst),
"vbsl", "$dst, $src2, $src3", "$src1 = $dst",
[(set DPR:$dst,
(v2i32 (or (and DPR:$src2, DPR:$src1),
- (and DPR:$src3, (vnot8 DPR:$src1)))))]>;
+ (and DPR:$src3, (vnotd DPR:$src1)))))]>;
def VBSLq : N3VX<1, 0, 0b01, 0b0001, 1, 1, (outs QPR:$dst),
(ins QPR:$src1, QPR:$src2, QPR:$src3),
N3RegFrm, IIC_VCNTiQ,
"vbsl", "$dst, $src2, $src3", "$src1 = $dst",
[(set QPR:$dst,
(v4i32 (or (and QPR:$src2, QPR:$src1),
- (and QPR:$src3, (vnot16 QPR:$src1)))))]>;
+ (and QPR:$src3, (vnotq QPR:$src1)))))]>;
// VBIF : Vector Bitwise Insert if False
// like VBSL but with: "vbif $dst, $src3, $src1", "$src2 = $dst",
// VABD : Vector Absolute Difference
defm VABDs : N3VInt_QHS<0, 0, 0b0111, 0, N3RegFrm,
IIC_VSUBi4D, IIC_VSUBi4D, IIC_VSUBi4Q, IIC_VSUBi4Q,
- "vabd", "s", int_arm_neon_vabds, 0>;
+ "vabd", "s", int_arm_neon_vabds, 1>;
defm VABDu : N3VInt_QHS<1, 0, 0b0111, 0, N3RegFrm,
IIC_VSUBi4D, IIC_VSUBi4D, IIC_VSUBi4Q, IIC_VSUBi4Q,
- "vabd", "u", int_arm_neon_vabdu, 0>;
+ "vabd", "u", int_arm_neon_vabdu, 1>;
def VABDfd : N3VDInt<1, 0, 0b10, 0b1101, 0, N3RegFrm, IIC_VBIND,
- "vabd", "f32", v2f32, v2f32, int_arm_neon_vabds, 0>;
+ "vabd", "f32", v2f32, v2f32, int_arm_neon_vabds, 1>;
def VABDfq : N3VQInt<1, 0, 0b10, 0b1101, 0, N3RegFrm, IIC_VBINQ,
- "vabd", "f32", v4f32, v4f32, int_arm_neon_vabds, 0>;
+ "vabd", "f32", v4f32, v4f32, int_arm_neon_vabds, 1>;
// VABDL : Vector Absolute Difference Long (Q = | D - D |)
-defm VABDLs : N3VLInt_QHS<0,1,0b0111,0, IIC_VSUBi4Q, IIC_VSUBi4Q,
- "vabdl", "s", int_arm_neon_vabdls, 0>;
-defm VABDLu : N3VLInt_QHS<1,1,0b0111,0, IIC_VSUBi4Q, IIC_VSUBi4Q,
- "vabdl", "u", int_arm_neon_vabdlu, 0>;
+defm VABDLs : N3VLIntExt_QHS<0,1,0b0111,0, IIC_VSUBi4Q,
+ "vabdl", "s", int_arm_neon_vabds, zext, 1>;
+defm VABDLu : N3VLIntExt_QHS<1,1,0b0111,0, IIC_VSUBi4Q,
+ "vabdl", "u", int_arm_neon_vabdu, zext, 1>;
// VABA : Vector Absolute Difference and Accumulate
-defm VABAs : N3VInt3_QHS<0,0,0b0111,1, IIC_VABAD, IIC_VABAQ,
- "vaba", "s", int_arm_neon_vabas>;
-defm VABAu : N3VInt3_QHS<1,0,0b0111,1, IIC_VABAD, IIC_VABAQ,
- "vaba", "u", int_arm_neon_vabau>;
+defm VABAs : N3VIntOp_QHS<0,0,0b0111,1, IIC_VABAD, IIC_VABAQ,
+ "vaba", "s", int_arm_neon_vabds, add>;
+defm VABAu : N3VIntOp_QHS<1,0,0b0111,1, IIC_VABAD, IIC_VABAQ,
+ "vaba", "u", int_arm_neon_vabdu, add>;
// VABAL : Vector Absolute Difference and Accumulate Long (Q += | D - D |)
-defm VABALs : N3VLInt3_QHS<0,1,0b0101,0, IIC_VABAD, IIC_VABAD,
- "vabal", "s", int_arm_neon_vabals>;
-defm VABALu : N3VLInt3_QHS<1,1,0b0101,0, IIC_VABAD, IIC_VABAD,
- "vabal", "u", int_arm_neon_vabalu>;
+defm VABALs : N3VLIntExtOp_QHS<0,1,0b0101,0, IIC_VABAD,
+ "vabal", "s", int_arm_neon_vabds, zext, add>;
+defm VABALu : N3VLIntExtOp_QHS<1,1,0b0101,0, IIC_VABAD,
+ "vabal", "u", int_arm_neon_vabdu, zext, add>;
// Vector Maximum and Minimum.
// Vector Negate.
-def vneg : PatFrag<(ops node:$in), (sub immAllZerosV, node:$in)>;
-def vneg8 : PatFrag<(ops node:$in),
- (sub (bitconvert (v8i8 immAllZerosV)), node:$in)>;
-def vneg16 : PatFrag<(ops node:$in),
- (sub (bitconvert (v16i8 immAllZerosV)), node:$in)>;
+def vnegd : PatFrag<(ops node:$in),
+ (sub (bitconvert (v2i32 NEONimmAllZerosV)), node:$in)>;
+def vnegq : PatFrag<(ops node:$in),
+ (sub (bitconvert (v4i32 NEONimmAllZerosV)), node:$in)>;
class VNEGD<bits<2> size, string OpcodeStr, string Dt, ValueType Ty>
: N2V<0b11, 0b11, size, 0b01, 0b00111, 0, 0, (outs DPR:$dst), (ins DPR:$src),
IIC_VSHLiD, OpcodeStr, Dt, "$dst, $src", "",
- [(set DPR:$dst, (Ty (vneg8 DPR:$src)))]>;
+ [(set DPR:$dst, (Ty (vnegd DPR:$src)))]>;
class VNEGQ<bits<2> size, string OpcodeStr, string Dt, ValueType Ty>
: N2V<0b11, 0b11, size, 0b01, 0b00111, 1, 0, (outs QPR:$dst), (ins QPR:$src),
IIC_VSHLiD, OpcodeStr, Dt, "$dst, $src", "",
- [(set QPR:$dst, (Ty (vneg16 QPR:$src)))]>;
+ [(set QPR:$dst, (Ty (vnegq QPR:$src)))]>;
// VNEG : Vector Negate (integer)
def VNEGs8d : VNEGD<0b00, "vneg", "s8", v8i8>;
"vneg", "f32", "$dst, $src", "",
[(set QPR:$dst, (v4f32 (fneg QPR:$src)))]>;
-def : Pat<(v8i8 (vneg8 DPR:$src)), (VNEGs8d DPR:$src)>;
-def : Pat<(v4i16 (vneg8 DPR:$src)), (VNEGs16d DPR:$src)>;
-def : Pat<(v2i32 (vneg8 DPR:$src)), (VNEGs32d DPR:$src)>;
-def : Pat<(v16i8 (vneg16 QPR:$src)), (VNEGs8q QPR:$src)>;
-def : Pat<(v8i16 (vneg16 QPR:$src)), (VNEGs16q QPR:$src)>;
-def : Pat<(v4i32 (vneg16 QPR:$src)), (VNEGs32q QPR:$src)>;
+def : Pat<(v8i8 (vnegd DPR:$src)), (VNEGs8d DPR:$src)>;
+def : Pat<(v4i16 (vnegd DPR:$src)), (VNEGs16d DPR:$src)>;
+def : Pat<(v2i32 (vnegd DPR:$src)), (VNEGs32d DPR:$src)>;
+def : Pat<(v16i8 (vnegq QPR:$src)), (VNEGs8q QPR:$src)>;
+def : Pat<(v8i16 (vnegq QPR:$src)), (VNEGs16q QPR:$src)>;
+def : Pat<(v4i32 (vnegq QPR:$src)), (VNEGs32q QPR:$src)>;
// VQNEG : Vector Saturating Negate
defm VQNEG : N2VInt_QHS<0b11, 0b11, 0b00, 0b01111, 0,
// VMOV : Vector Move (Register)
+let neverHasSideEffects = 1 in {
def VMOVDneon: N3VX<0, 0, 0b10, 0b0001, 0, 1, (outs DPR:$dst), (ins DPR:$src),
N3RegFrm, IIC_VMOVD, "vmov", "$dst, $src", "", []>;
def VMOVQ : N3VX<0, 0, 0b10, 0b0001, 1, 1, (outs QPR:$dst), (ins QPR:$src),
N3RegFrm, IIC_VMOVD, "vmov", "$dst, $src", "", []>;
-// Pseudo vector move instruction for QQ (a pair of Q) registers. This should
+// Pseudo vector move instructions for QQ and QQQQ registers. This should
// be expanded after register allocation is completed.
def VMOVQQ : PseudoInst<(outs QQPR:$dst), (ins QQPR:$src),
- NoItinerary, "@ vmov\t$dst, $src", []>;
-
-// VMOV : Vector Move (Immediate)
-
-// VMOV_get_imm8 xform function: convert build_vector to VMOV.i8 imm.
-def VMOV_get_imm8 : SDNodeXForm<build_vector, [{
- return ARM::getVMOVImm(N, 1, *CurDAG);
-}]>;
-def vmovImm8 : PatLeaf<(build_vector), [{
- return ARM::getVMOVImm(N, 1, *CurDAG).getNode() != 0;
-}], VMOV_get_imm8>;
-
-// VMOV_get_imm16 xform function: convert build_vector to VMOV.i16 imm.
-def VMOV_get_imm16 : SDNodeXForm<build_vector, [{
- return ARM::getVMOVImm(N, 2, *CurDAG);
-}]>;
-def vmovImm16 : PatLeaf<(build_vector), [{
- return ARM::getVMOVImm(N, 2, *CurDAG).getNode() != 0;
-}], VMOV_get_imm16>;
-
-// VMOV_get_imm32 xform function: convert build_vector to VMOV.i32 imm.
-def VMOV_get_imm32 : SDNodeXForm<build_vector, [{
- return ARM::getVMOVImm(N, 4, *CurDAG);
-}]>;
-def vmovImm32 : PatLeaf<(build_vector), [{
- return ARM::getVMOVImm(N, 4, *CurDAG).getNode() != 0;
-}], VMOV_get_imm32>;
+ NoItinerary, "${:comment} vmov\t$dst, $src", []>;
-// VMOV_get_imm64 xform function: convert build_vector to VMOV.i64 imm.
-def VMOV_get_imm64 : SDNodeXForm<build_vector, [{
- return ARM::getVMOVImm(N, 8, *CurDAG);
-}]>;
-def vmovImm64 : PatLeaf<(build_vector), [{
- return ARM::getVMOVImm(N, 8, *CurDAG).getNode() != 0;
-}], VMOV_get_imm64>;
+def VMOVQQQQ : PseudoInst<(outs QQQQPR:$dst), (ins QQQQPR:$src),
+ NoItinerary, "${:comment} vmov\t$dst, $src", []>;
+} // neverHasSideEffects
-// Note: Some of the cmode bits in the following VMOV instructions need to
-// be encoded based on the immed values.
+// VMOV : Vector Move (Immediate)
+let isReMaterializable = 1 in {
def VMOVv8i8 : N1ModImm<1, 0b000, 0b1110, 0, 0, 0, 1, (outs DPR:$dst),
- (ins h8imm:$SIMM), IIC_VMOVImm,
+ (ins nModImm:$SIMM), IIC_VMOVImm,
"vmov", "i8", "$dst, $SIMM", "",
- [(set DPR:$dst, (v8i8 vmovImm8:$SIMM))]>;
+ [(set DPR:$dst, (v8i8 (NEONvmovImm timm:$SIMM)))]>;
def VMOVv16i8 : N1ModImm<1, 0b000, 0b1110, 0, 1, 0, 1, (outs QPR:$dst),
- (ins h8imm:$SIMM), IIC_VMOVImm,
+ (ins nModImm:$SIMM), IIC_VMOVImm,
"vmov", "i8", "$dst, $SIMM", "",
- [(set QPR:$dst, (v16i8 vmovImm8:$SIMM))]>;
+ [(set QPR:$dst, (v16i8 (NEONvmovImm timm:$SIMM)))]>;
-def VMOVv4i16 : N1ModImm<1, 0b000, {1,0,?,?}, 0, 0, {?}, 1, (outs DPR:$dst),
- (ins h16imm:$SIMM), IIC_VMOVImm,
+def VMOVv4i16 : N1ModImm<1, 0b000, {1,0,?,0}, 0, 0, 0, 1, (outs DPR:$dst),
+ (ins nModImm:$SIMM), IIC_VMOVImm,
"vmov", "i16", "$dst, $SIMM", "",
- [(set DPR:$dst, (v4i16 vmovImm16:$SIMM))]>;
-def VMOVv8i16 : N1ModImm<1, 0b000, {1,0,?,?}, 0, 1, {?}, 1, (outs QPR:$dst),
- (ins h16imm:$SIMM), IIC_VMOVImm,
+ [(set DPR:$dst, (v4i16 (NEONvmovImm timm:$SIMM)))]>;
+def VMOVv8i16 : N1ModImm<1, 0b000, {1,0,?,0}, 0, 1, 0, 1, (outs QPR:$dst),
+ (ins nModImm:$SIMM), IIC_VMOVImm,
"vmov", "i16", "$dst, $SIMM", "",
- [(set QPR:$dst, (v8i16 vmovImm16:$SIMM))]>;
+ [(set QPR:$dst, (v8i16 (NEONvmovImm timm:$SIMM)))]>;
-def VMOVv2i32 : N1ModImm<1, 0b000, {?,?,?,?}, 0, 0, {?}, 1, (outs DPR:$dst),
- (ins h32imm:$SIMM), IIC_VMOVImm,
+def VMOVv2i32 : N1ModImm<1, 0b000, {?,?,?,?}, 0, 0, 0, 1, (outs DPR:$dst),
+ (ins nModImm:$SIMM), IIC_VMOVImm,
"vmov", "i32", "$dst, $SIMM", "",
- [(set DPR:$dst, (v2i32 vmovImm32:$SIMM))]>;
-def VMOVv4i32 : N1ModImm<1, 0b000, {?,?,?,?}, 0, 1, {?}, 1, (outs QPR:$dst),
- (ins h32imm:$SIMM), IIC_VMOVImm,
+ [(set DPR:$dst, (v2i32 (NEONvmovImm timm:$SIMM)))]>;
+def VMOVv4i32 : N1ModImm<1, 0b000, {?,?,?,?}, 0, 1, 0, 1, (outs QPR:$dst),
+ (ins nModImm:$SIMM), IIC_VMOVImm,
"vmov", "i32", "$dst, $SIMM", "",
- [(set QPR:$dst, (v4i32 vmovImm32:$SIMM))]>;
+ [(set QPR:$dst, (v4i32 (NEONvmovImm timm:$SIMM)))]>;
def VMOVv1i64 : N1ModImm<1, 0b000, 0b1110, 0, 0, 1, 1, (outs DPR:$dst),
- (ins h64imm:$SIMM), IIC_VMOVImm,
+ (ins nModImm:$SIMM), IIC_VMOVImm,
"vmov", "i64", "$dst, $SIMM", "",
- [(set DPR:$dst, (v1i64 vmovImm64:$SIMM))]>;
+ [(set DPR:$dst, (v1i64 (NEONvmovImm timm:$SIMM)))]>;
def VMOVv2i64 : N1ModImm<1, 0b000, 0b1110, 0, 1, 1, 1, (outs QPR:$dst),
- (ins h64imm:$SIMM), IIC_VMOVImm,
+ (ins nModImm:$SIMM), IIC_VMOVImm,
"vmov", "i64", "$dst, $SIMM", "",
- [(set QPR:$dst, (v2i64 vmovImm64:$SIMM))]>;
+ [(set QPR:$dst, (v2i64 (NEONvmovImm timm:$SIMM)))]>;
+} // isReMaterializable
// VMOV : Vector Get Lane (move scalar to ARM core register)
(INSERT_SUBREG QPR:$src1, DPR:$src2, (DSubReg_f64_reg imm:$src3))>;
def : Pat<(v2f32 (scalar_to_vector SPR:$src)),
- (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)), SPR:$src, arm_ssubreg_0)>;
+ (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)), SPR:$src, ssub_0)>;
def : Pat<(v2f64 (scalar_to_vector (f64 DPR:$src))),
- (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), DPR:$src, arm_dsubreg_0)>;
+ (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), DPR:$src, dsub_0)>;
def : Pat<(v4f32 (scalar_to_vector SPR:$src)),
- (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), SPR:$src, arm_ssubreg_0)>;
+ (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), SPR:$src, ssub_0)>;
def : Pat<(v8i8 (scalar_to_vector GPR:$src)),
(VSETLNi8 (v8i8 (IMPLICIT_DEF)), GPR:$src, (i32 0))>;
def : Pat<(v16i8 (scalar_to_vector GPR:$src)),
(INSERT_SUBREG (v16i8 (IMPLICIT_DEF)),
(VSETLNi8 (v8i8 (IMPLICIT_DEF)), GPR:$src, (i32 0)),
- arm_dsubreg_0)>;
+ dsub_0)>;
def : Pat<(v8i16 (scalar_to_vector GPR:$src)),
(INSERT_SUBREG (v8i16 (IMPLICIT_DEF)),
(VSETLNi16 (v4i16 (IMPLICIT_DEF)), GPR:$src, (i32 0)),
- arm_dsubreg_0)>;
+ dsub_0)>;
def : Pat<(v4i32 (scalar_to_vector GPR:$src)),
(INSERT_SUBREG (v4i32 (IMPLICIT_DEF)),
(VSETLNi32 (v2i32 (IMPLICIT_DEF)), GPR:$src, (i32 0)),
- arm_dsubreg_0)>;
+ dsub_0)>;
// VDUP : Vector Duplicate (from ARM core register to all elements)
IIC_VMOVD, "vdup", "32", "$dst, ${src:lane}", "",
[(set QPR:$dst, (v4f32 (NEONvdup (f32 SPR:$src))))]>;
-def : Pat<(v2i64 (NEONvduplane (v2i64 QPR:$src), imm:$lane)),
- (INSERT_SUBREG QPR:$src,
- (i64 (EXTRACT_SUBREG QPR:$src,
- (DSubReg_f64_reg imm:$lane))),
- (DSubReg_f64_other_reg imm:$lane))>;
-def : Pat<(v2f64 (NEONvduplane (v2f64 QPR:$src), imm:$lane)),
- (INSERT_SUBREG QPR:$src,
- (f64 (EXTRACT_SUBREG QPR:$src,
- (DSubReg_f64_reg imm:$lane))),
- (DSubReg_f64_other_reg imm:$lane))>;
-
// VMOVN : Vector Narrowing Move
-defm VMOVN : N2VNInt_HSD<0b11,0b11,0b10,0b00100,0,0, IIC_VMOVD,
- "vmovn", "i", int_arm_neon_vmovn>;
+defm VMOVN : N2VN_HSD<0b11,0b11,0b10,0b00100,0,0, IIC_VMOVD,
+ "vmovn", "i", trunc>;
// VQMOVN : Vector Saturating Narrowing Move
defm VQMOVNs : N2VNInt_HSD<0b11,0b11,0b10,0b00101,0,0, IIC_VQUNAiD,
"vqmovn", "s", int_arm_neon_vqmovns>;
defm VQMOVNsu : N2VNInt_HSD<0b11,0b11,0b10,0b00100,1,0, IIC_VQUNAiD,
"vqmovun", "s", int_arm_neon_vqmovnsu>;
// VMOVL : Vector Lengthening Move
-defm VMOVLs : N2VLInt_QHS<0b01,0b10100,0,1, "vmovl", "s",
- int_arm_neon_vmovls>;
-defm VMOVLu : N2VLInt_QHS<0b11,0b10100,0,1, "vmovl", "u",
- int_arm_neon_vmovlu>;
+defm VMOVLs : N2VL_QHS<0b01,0b10100,0,1, "vmovl", "s", sext>;
+defm VMOVLu : N2VL_QHS<0b11,0b10100,0,1, "vmovl", "u", zext>;
// Vector Conversions.
def VTBL2
: N3V<1,1,0b11,0b1001,0,0, (outs DPR:$dst),
(ins DPR:$tbl1, DPR:$tbl2, DPR:$src), NVTBLFrm, IIC_VTB2,
- "vtbl", "8", "$dst, \\{$tbl1, $tbl2\\}, $src", "",
- [(set DPR:$dst, (v8i8 (int_arm_neon_vtbl2
- DPR:$tbl1, DPR:$tbl2, DPR:$src)))]>;
+ "vtbl", "8", "$dst, \\{$tbl1, $tbl2\\}, $src", "", []>;
def VTBL3
: N3V<1,1,0b11,0b1010,0,0, (outs DPR:$dst),
(ins DPR:$tbl1, DPR:$tbl2, DPR:$tbl3, DPR:$src), NVTBLFrm, IIC_VTB3,
- "vtbl", "8", "$dst, \\{$tbl1, $tbl2, $tbl3\\}, $src", "",
- [(set DPR:$dst, (v8i8 (int_arm_neon_vtbl3
- DPR:$tbl1, DPR:$tbl2, DPR:$tbl3, DPR:$src)))]>;
+ "vtbl", "8", "$dst, \\{$tbl1, $tbl2, $tbl3\\}, $src", "", []>;
def VTBL4
: N3V<1,1,0b11,0b1011,0,0, (outs DPR:$dst),
(ins DPR:$tbl1, DPR:$tbl2, DPR:$tbl3, DPR:$tbl4, DPR:$src),
NVTBLFrm, IIC_VTB4,
- "vtbl", "8", "$dst, \\{$tbl1, $tbl2, $tbl3, $tbl4\\}, $src", "",
- [(set DPR:$dst, (v8i8 (int_arm_neon_vtbl4 DPR:$tbl1, DPR:$tbl2,
- DPR:$tbl3, DPR:$tbl4, DPR:$src)))]>;
+ "vtbl", "8", "$dst, \\{$tbl1, $tbl2, $tbl3, $tbl4\\}, $src", "", []>;
} // hasExtraSrcRegAllocReq = 1
+def VTBL2Pseudo
+ : PseudoNeonI<(outs DPR:$dst), (ins QPR:$tbl, DPR:$src), IIC_VTB2, "">;
+def VTBL3Pseudo
+ : PseudoNeonI<(outs DPR:$dst), (ins QQPR:$tbl, DPR:$src), IIC_VTB3, "">;
+def VTBL4Pseudo
+ : PseudoNeonI<(outs DPR:$dst), (ins QQPR:$tbl, DPR:$src), IIC_VTB4, "">;
+
// VTBX : Vector Table Extension
def VTBX1
: N3V<1,1,0b11,0b1000,1,0, (outs DPR:$dst),
def VTBX2
: N3V<1,1,0b11,0b1001,1,0, (outs DPR:$dst),
(ins DPR:$orig, DPR:$tbl1, DPR:$tbl2, DPR:$src), NVTBLFrm, IIC_VTBX2,
- "vtbx", "8", "$dst, \\{$tbl1, $tbl2\\}, $src", "$orig = $dst",
- [(set DPR:$dst, (v8i8 (int_arm_neon_vtbx2
- DPR:$orig, DPR:$tbl1, DPR:$tbl2, DPR:$src)))]>;
+ "vtbx", "8", "$dst, \\{$tbl1, $tbl2\\}, $src", "$orig = $dst", []>;
def VTBX3
: N3V<1,1,0b11,0b1010,1,0, (outs DPR:$dst),
(ins DPR:$orig, DPR:$tbl1, DPR:$tbl2, DPR:$tbl3, DPR:$src),
NVTBLFrm, IIC_VTBX3,
- "vtbx", "8", "$dst, \\{$tbl1, $tbl2, $tbl3\\}, $src", "$orig = $dst",
- [(set DPR:$dst, (v8i8 (int_arm_neon_vtbx3 DPR:$orig, DPR:$tbl1,
- DPR:$tbl2, DPR:$tbl3, DPR:$src)))]>;
+ "vtbx", "8", "$dst, \\{$tbl1, $tbl2, $tbl3\\}, $src",
+ "$orig = $dst", []>;
def VTBX4
: N3V<1,1,0b11,0b1011,1,0, (outs DPR:$dst), (ins DPR:$orig, DPR:$tbl1,
DPR:$tbl2, DPR:$tbl3, DPR:$tbl4, DPR:$src), NVTBLFrm, IIC_VTBX4,
"vtbx", "8", "$dst, \\{$tbl1, $tbl2, $tbl3, $tbl4\\}, $src",
- "$orig = $dst",
- [(set DPR:$dst, (v8i8 (int_arm_neon_vtbx4 DPR:$orig, DPR:$tbl1,
- DPR:$tbl2, DPR:$tbl3, DPR:$tbl4, DPR:$src)))]>;
+ "$orig = $dst", []>;
} // hasExtraSrcRegAllocReq = 1
+def VTBX2Pseudo
+ : PseudoNeonI<(outs DPR:$dst), (ins DPR:$orig, QPR:$tbl, DPR:$src),
+ IIC_VTBX2, "$orig = $dst">;
+def VTBX3Pseudo
+ : PseudoNeonI<(outs DPR:$dst), (ins DPR:$orig, QQPR:$tbl, DPR:$src),
+ IIC_VTBX3, "$orig = $dst">;
+def VTBX4Pseudo
+ : PseudoNeonI<(outs DPR:$dst), (ins DPR:$orig, QQPR:$tbl, DPR:$src),
+ IIC_VTBX4, "$orig = $dst">;
+
//===----------------------------------------------------------------------===//
// NEON instructions for single-precision FP math
//===----------------------------------------------------------------------===//
class N2VSPat<SDNode OpNode, ValueType ResTy, ValueType OpTy, NeonI Inst>
: NEONFPPat<(ResTy (OpNode SPR:$a)),
(EXTRACT_SUBREG (OpTy (Inst (INSERT_SUBREG (OpTy (IMPLICIT_DEF)),
- SPR:$a, arm_ssubreg_0))),
- arm_ssubreg_0)>;
+ SPR:$a, ssub_0))),
+ ssub_0)>;
class N3VSPat<SDNode OpNode, NeonI Inst>
: NEONFPPat<(f32 (OpNode SPR:$a, SPR:$b)),
(EXTRACT_SUBREG (v2f32
(Inst (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
- SPR:$a, arm_ssubreg_0),
+ SPR:$a, ssub_0),
(INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
- SPR:$b, arm_ssubreg_0))),
- arm_ssubreg_0)>;
+ SPR:$b, ssub_0))),
+ ssub_0)>;
class N3VSMulOpPat<SDNode MulNode, SDNode OpNode, NeonI Inst>
: NEONFPPat<(f32 (OpNode SPR:$acc, (f32 (MulNode SPR:$a, SPR:$b)))),
(EXTRACT_SUBREG (Inst (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
- SPR:$acc, arm_ssubreg_0),
+ SPR:$acc, ssub_0),
(INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
- SPR:$a, arm_ssubreg_0),
+ SPR:$a, ssub_0),
(INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
- SPR:$b, arm_ssubreg_0)),
- arm_ssubreg_0)>;
+ SPR:$b, ssub_0)),
+ ssub_0)>;
// These need separate instructions because they must use DPR_VFP2 register
// class which have SPR sub-registers.
projects / oota-llvm.git / blobdiff