X-Git-Url: http://plrg.eecs.uci.edu/git/?p=oota-llvm.git;a=blobdiff_plain;f=lib%2FTarget%2FARM%2FARMInstrInfo.td;h=bda6c944ce3bfb6060e80bde1565601fa639bc2f;hp=3d38cc98960c28f2d518f5c1ba7665ad5552846b;hb=1f7a90d7936a9a6278365ea8b0521c7ff17432c3;hpb=793e2aaa73dfd19ee223ccad24f8b7a6de53812a diff --git a/lib/Target/ARM/ARMInstrInfo.td b/lib/Target/ARM/ARMInstrInfo.td index 3d38cc98960..bda6c944ce3 100644 --- a/lib/Target/ARM/ARMInstrInfo.td +++ b/lib/Target/ARM/ARMInstrInfo.td @@ -186,7 +186,8 @@ def ARMvminnm : SDNode<"ARMISD::VMINNM", SDT_ARMVMINNM, []>; def HasV4T : Predicate<"Subtarget->hasV4TOps()">, AssemblerPredicate<"HasV4TOps", "armv4t">; def NoV4T : Predicate<"!Subtarget->hasV4TOps()">; -def HasV5T : Predicate<"Subtarget->hasV5TOps()">; +def HasV5T : Predicate<"Subtarget->hasV5TOps()">, + AssemblerPredicate<"HasV5TOps", "armv5t">; def HasV5TE : Predicate<"Subtarget->hasV5TEOps()">, AssemblerPredicate<"HasV5TEOps", "armv5te">; def HasV6 : Predicate<"Subtarget->hasV6Ops()">, @@ -198,12 +199,17 @@ def HasV6M : Predicate<"Subtarget->hasV6MOps()">, def HasV6T2 : Predicate<"Subtarget->hasV6T2Ops()">, AssemblerPredicate<"HasV6T2Ops", "armv6t2">; def NoV6T2 : Predicate<"!Subtarget->hasV6T2Ops()">; +def HasV6K : Predicate<"Subtarget->hasV6KOps()">, + AssemblerPredicate<"HasV6KOps", "armv6k">; +def NoV6K : Predicate<"!Subtarget->hasV6KOps()">; def HasV7 : Predicate<"Subtarget->hasV7Ops()">, AssemblerPredicate<"HasV7Ops", "armv7">; def HasV8 : Predicate<"Subtarget->hasV8Ops()">, AssemblerPredicate<"HasV8Ops", "armv8">; def PreV8 : Predicate<"!Subtarget->hasV8Ops()">, AssemblerPredicate<"!HasV8Ops", "armv7 or earlier">; +def HasV8_1a : Predicate<"Subtarget->hasV8_1aOps()">, + AssemblerPredicate<"HasV8_1aOps", "armv8.1a">; def NoVFP : Predicate<"!Subtarget->hasVFP2()">; def HasVFP2 : Predicate<"Subtarget->hasVFP2()">, AssemblerPredicate<"FeatureVFP2", "VFP2">; @@ -240,9 +246,13 @@ def HasDB : Predicate<"Subtarget->hasDataBarrier()">, def HasMP : Predicate<"Subtarget->hasMPExtension()">, AssemblerPredicate<"FeatureMP", "mp-extensions">; +def HasVirtualization: Predicate<"false">, + AssemblerPredicate<"FeatureVirtualization", + "virtualization-extensions">; def HasTrustZone : Predicate<"Subtarget->hasTrustZone()">, AssemblerPredicate<"FeatureTrustZone", "TrustZone">; +def HasZCZ : Predicate<"Subtarget->hasZeroCycleZeroing()">; def UseNEONForFP : Predicate<"Subtarget->useNEONForSinglePrecisionFP()">; def DontUseNEONForFP : Predicate<"!Subtarget->useNEONForSinglePrecisionFP()">; def IsThumb : Predicate<"Subtarget->isThumb()">, @@ -258,8 +268,6 @@ def IsNotMClass : Predicate<"!Subtarget->isMClass()">, "!armv*m">; def IsARM : Predicate<"!Subtarget->isThumb()">, AssemblerPredicate<"!ModeThumb", "arm-mode">; -def IsIOS : Predicate<"Subtarget->isTargetIOS()">; -def IsNotIOS : Predicate<"!Subtarget->isTargetIOS()">; def IsMachO : Predicate<"Subtarget->isTargetMachO()">; def IsNotMachO : Predicate<"!Subtarget->isTargetMachO()">; def IsNaCl : Predicate<"Subtarget->isTargetNaCl()">; @@ -268,8 +276,8 @@ def UseNaClTrap : Predicate<"Subtarget->useNaClTrap()">, def DontUseNaClTrap : Predicate<"!Subtarget->useNaClTrap()">; // FIXME: Eventually this will be just "hasV6T2Ops". -def UseMovt : Predicate<"Subtarget->useMovt()">; -def DontUseMovt : Predicate<"!Subtarget->useMovt()">; +def UseMovt : Predicate<"Subtarget->useMovt(*MF)">; +def DontUseMovt : Predicate<"!Subtarget->useMovt(*MF)">; def UseFPVMLx : Predicate<"Subtarget->useFPVMLx()">; def UseMulOps : Predicate<"Subtarget->useMulOps()">; @@ -315,12 +323,12 @@ class RegConstraint { // imm_neg_XFORM - Return the negation of an i32 immediate value. def imm_neg_XFORM : SDNodeXFormgetTargetConstant(-(int)N->getZExtValue(), MVT::i32); + return CurDAG->getTargetConstant(-(int)N->getZExtValue(), SDLoc(N), MVT::i32); }]>; // imm_not_XFORM - Return the complement of a i32 immediate value. def imm_not_XFORM : SDNodeXFormgetTargetConstant(~(int)N->getZExtValue(), MVT::i32); + return CurDAG->getTargetConstant(~(int)N->getZExtValue(), SDLoc(N), MVT::i32); }]>; /// imm16_31 predicate - True if the 32-bit immediate is in the range [16,31]. @@ -328,24 +336,6 @@ def imm16_31 : ImmLeaf= 16 && (int32_t)Imm < 32; }]>; -def so_imm_neg_asmoperand : AsmOperandClass { let Name = "ARMSOImmNeg"; } -def so_imm_neg : Operand, PatLeaf<(imm), [{ - unsigned Value = -(unsigned)N->getZExtValue(); - return Value && ARM_AM::getSOImmVal(Value) != -1; - }], imm_neg_XFORM> { - let ParserMatchClass = so_imm_neg_asmoperand; -} - -// Note: this pattern doesn't require an encoder method and such, as it's -// only used on aliases (Pat<> and InstAlias<>). The actual encoding -// is handled by the destination instructions, which use so_imm. -def so_imm_not_asmoperand : AsmOperandClass { let Name = "ARMSOImmNot"; } -def so_imm_not : Operand, PatLeaf<(imm), [{ - return ARM_AM::getSOImmVal(~(uint32_t)N->getZExtValue()) != -1; - }], imm_not_XFORM> { - let ParserMatchClass = so_imm_not_asmoperand; -} - // sext_16_node predicate - True if the SDNode is sign-extended 16 or more bits. def sext_16_node : PatLeaf<(i32 GPR:$a), [{ return CurDAG->ComputeNumSignBits(SDValue(N,0)) >= 17; @@ -353,7 +343,8 @@ def sext_16_node : PatLeaf<(i32 GPR:$a), [{ /// Split a 32-bit immediate into two 16 bit parts. def hi16 : SDNodeXFormgetTargetConstant((uint32_t)N->getZExtValue() >> 16, MVT::i32); + return CurDAG->getTargetConstant((uint32_t)N->getZExtValue() >> 16, SDLoc(N), + MVT::i32); }]>; def lo16AllZero : PatLeaf<(i32 imm), [{ @@ -398,6 +389,9 @@ def fsub_mlx : PatFrag<(ops node:$lhs, node:$rhs),(fsub node:$lhs, node:$rhs),[{ // Immediate operands with a shared generic asm render method. class ImmAsmOperand : AsmOperandClass { let RenderMethod = "addImmOperands"; } +// Operands that are part of a memory addressing mode. +class MemOperand : Operand { let OperandType = "OPERAND_MEMORY"; } + // Branch target. // FIXME: rename brtarget to t2_brtarget def brtarget : Operand { @@ -491,11 +485,11 @@ def neon_vcvt_imm32 : Operand { // rot_imm: An integer that encodes a rotate amount. Must be 8, 16, or 24. def rot_imm_XFORM: SDNodeXFormgetZExtValue()){ - default: assert(0); - case 0: return CurDAG->getTargetConstant(0, MVT::i32); - case 8: return CurDAG->getTargetConstant(1, MVT::i32); - case 16: return CurDAG->getTargetConstant(2, MVT::i32); - case 24: return CurDAG->getTargetConstant(3, MVT::i32); + default: llvm_unreachable(nullptr); + case 0: return CurDAG->getTargetConstant(0, SDLoc(N), MVT::i32); + case 8: return CurDAG->getTargetConstant(1, SDLoc(N), MVT::i32); + case 16: return CurDAG->getTargetConstant(2, SDLoc(N), MVT::i32); + case 24: return CurDAG->getTargetConstant(3, SDLoc(N), MVT::i32); } }]>; def RotImmAsmOperand : AsmOperandClass { @@ -525,7 +519,7 @@ def shift_imm : Operand { let ParserMatchClass = ShifterImmAsmOperand; } -// shifter_operand operands: so_reg_reg, so_reg_imm, and so_imm. +// shifter_operand operands: so_reg_reg, so_reg_imm, and mod_imm. def ShiftedRegAsmOperand : AsmOperandClass { let Name = "RegShiftedReg"; } def so_reg_reg : Operand, // reg reg imm ComplexPattern, // reg reg imm let MIOperandInfo = (ops GPR, i32imm); } - -// so_imm - Match a 32-bit shifter_operand immediate operand, which is an -// 8-bit immediate rotated by an arbitrary number of bits. -def SOImmAsmOperand: ImmAsmOperand { let Name = "ARMSOImm"; } -def so_imm : Operand, ImmLeaf, ImmLeaf { - let EncoderMethod = "getSOImmOpValue"; - let ParserMatchClass = SOImmAsmOperand; - let DecoderMethod = "DecodeSOImmOperand"; + let EncoderMethod = "getModImmOpValue"; + let PrintMethod = "printModImmOperand"; + let ParserMatchClass = ModImmAsmOperand; } -// Break so_imm's up into two pieces. This handles immediates with up to 16 -// bits set in them. This uses so_imm2part to match and so_imm2part_[12] to -// get the first/second pieces. -def so_imm2part : PatLeaf<(imm), [{ - return ARM_AM::isSOImmTwoPartVal((unsigned)N->getZExtValue()); -}]>; +// Note: the patterns mod_imm_not and mod_imm_neg do not require an encoder +// method and such, as they are only used on aliases (Pat<> and InstAlias<>). +// The actual parsing, encoding, decoding are handled by the destination +// instructions, which use mod_imm. -/// arm_i32imm - True for +V6T2, or true only if so_imm2part is true. -/// +def ModImmNotAsmOperand : AsmOperandClass { let Name = "ModImmNot"; } +def mod_imm_not : Operand, PatLeaf<(imm), [{ + return ARM_AM::getSOImmVal(~(uint32_t)N->getZExtValue()) != -1; + }], imm_not_XFORM> { + let ParserMatchClass = ModImmNotAsmOperand; +} + +def ModImmNegAsmOperand : AsmOperandClass { let Name = "ModImmNeg"; } +def mod_imm_neg : Operand, PatLeaf<(imm), [{ + unsigned Value = -(unsigned)N->getZExtValue(); + return Value && ARM_AM::getSOImmVal(Value) != -1; + }], imm_neg_XFORM> { + let ParserMatchClass = ModImmNegAsmOperand; +} + +/// arm_i32imm - True for +V6T2, or when isSOImmTwoParVal() def arm_i32imm : PatLeaf<(imm), [{ - if (Subtarget->hasV6T2Ops()) + if (Subtarget->useMovt(*MF)) return true; return ARM_AM::isSOImmTwoPartVal((unsigned)N->getZExtValue()); }]>; @@ -631,6 +641,8 @@ def imm32 : Operand, ImmLeaf { let ParserMatchClass = Imm32AsmOperand; } +def imm8_or_16 : ImmLeaf; + /// imm1_7 predicate - Immediate in the range [1,7]. def Imm1_7AsmOperand: ImmAsmOperand { let Name = "Imm1_7"; } def imm1_7 : Operand, ImmLeaf 0 && Imm < 8; }]> { @@ -756,7 +768,8 @@ def bf_inv_mask_imm : Operand, } def imm1_32_XFORM: SDNodeXFormgetTargetConstant((int)N->getZExtValue() - 1, MVT::i32); + return CurDAG->getTargetConstant((int)N->getZExtValue() - 1, SDLoc(N), + MVT::i32); }]>; def Imm1_32AsmOperand: AsmOperandClass { let Name = "Imm1_32"; } def imm1_32 : Operand, PatLeaf<(imm), [{ @@ -769,7 +782,8 @@ def imm1_32 : Operand, PatLeaf<(imm), [{ } def imm1_16_XFORM: SDNodeXFormgetTargetConstant((int)N->getZExtValue() - 1, MVT::i32); + return CurDAG->getTargetConstant((int)N->getZExtValue() - 1, SDLoc(N), + MVT::i32); }]>; def Imm1_16AsmOperand: AsmOperandClass { let Name = "Imm1_16"; } def imm1_16 : Operand, PatLeaf<(imm), [{ return Imm > 0 && Imm <= 16; }], @@ -782,7 +796,7 @@ def imm1_16 : Operand, PatLeaf<(imm), [{ return Imm > 0 && Imm <= 16; }], // addrmode_imm12 := reg +/- imm12 // def MemImm12OffsetAsmOperand : AsmOperandClass { let Name = "MemImm12Offset"; } -class AddrMode_Imm12 : Operand, +class AddrMode_Imm12 : MemOperand, ComplexPattern { // 12-bit immediate operand. Note that instructions using this encode // #0 and #-0 differently. We flag #-0 as the magic value INT32_MIN. All other @@ -805,7 +819,7 @@ def addrmode_imm12_pre : AddrMode_Imm12 { // ldst_so_reg := reg +/- reg shop imm // def MemRegOffsetAsmOperand : AsmOperandClass { let Name = "MemRegOffset"; } -def ldst_so_reg : Operand, +def ldst_so_reg : MemOperand, ComplexPattern { let EncoderMethod = "getLdStSORegOpValue"; // FIXME: Simplify the printer @@ -821,7 +835,7 @@ def ldst_so_reg : Operand, // {8} 1 is imm8 is non-negative. 0 otherwise. // {7-0} [0,255] imm8 value. def PostIdxImm8AsmOperand : AsmOperandClass { let Name = "PostIdxImm8"; } -def postidx_imm8 : Operand { +def postidx_imm8 : MemOperand { let PrintMethod = "printPostIdxImm8Operand"; let ParserMatchClass = PostIdxImm8AsmOperand; let MIOperandInfo = (ops i32imm); @@ -833,7 +847,7 @@ def postidx_imm8 : Operand { // {8} 1 is imm8 is non-negative. 0 otherwise. // {7-0} [0,255] imm8 value, scaled by 4. def PostIdxImm8s4AsmOperand : AsmOperandClass { let Name = "PostIdxImm8s4"; } -def postidx_imm8s4 : Operand { +def postidx_imm8s4 : MemOperand { let PrintMethod = "printPostIdxImm8s4Operand"; let ParserMatchClass = PostIdxImm8s4AsmOperand; let MIOperandInfo = (ops i32imm); @@ -846,7 +860,7 @@ def PostIdxRegAsmOperand : AsmOperandClass { let Name = "PostIdxReg"; let ParserMethod = "parsePostIdxReg"; } -def postidx_reg : Operand { +def postidx_reg : MemOperand { let EncoderMethod = "getPostIdxRegOpValue"; let DecoderMethod = "DecodePostIdxReg"; let PrintMethod = "printPostIdxRegOperand"; @@ -861,7 +875,7 @@ def postidx_reg : Operand { // FIXME: addrmode2 should be refactored the rest of the way to always // use explicit imm vs. reg versions above (addrmode_imm12 and ldst_so_reg). def AddrMode2AsmOperand : AsmOperandClass { let Name = "AddrMode2"; } -def addrmode2 : Operand, +def addrmode2 : MemOperand, ComplexPattern { let EncoderMethod = "getAddrMode2OpValue"; let PrintMethod = "printAddrMode2Operand"; @@ -873,7 +887,7 @@ def PostIdxRegShiftedAsmOperand : AsmOperandClass { let Name = "PostIdxRegShifted"; let ParserMethod = "parsePostIdxReg"; } -def am2offset_reg : Operand, +def am2offset_reg : MemOperand, ComplexPattern { let EncoderMethod = "getAddrMode2OffsetOpValue"; @@ -886,7 +900,7 @@ def am2offset_reg : Operand, // FIXME: am2offset_imm should only need the immediate, not the GPR. Having // the GPR is purely vestigal at this point. def AM2OffsetImmAsmOperand : AsmOperandClass { let Name = "AM2OffsetImm"; } -def am2offset_imm : Operand, +def am2offset_imm : MemOperand, ComplexPattern { let EncoderMethod = "getAddrMode2OffsetOpValue"; @@ -901,7 +915,7 @@ def am2offset_imm : Operand, // // FIXME: split into imm vs. reg versions. def AddrMode3AsmOperand : AsmOperandClass { let Name = "AddrMode3"; } -class AddrMode3 : Operand, +class AddrMode3 : MemOperand, ComplexPattern { let EncoderMethod = "getAddrMode3OpValue"; let ParserMatchClass = AddrMode3AsmOperand; @@ -924,7 +938,7 @@ def AM3OffsetAsmOperand : AsmOperandClass { let Name = "AM3Offset"; let ParserMethod = "parseAM3Offset"; } -def am3offset : Operand, +def am3offset : MemOperand, ComplexPattern { let EncoderMethod = "getAddrMode3OffsetOpValue"; @@ -943,7 +957,7 @@ def ldstm_mode : OptionalDefOperand { // addrmode5 := reg +/- imm8*4 // def AddrMode5AsmOperand : AsmOperandClass { let Name = "AddrMode5"; } -class AddrMode5 : Operand, +class AddrMode5 : MemOperand, ComplexPattern { let EncoderMethod = "getAddrMode5OpValue"; let DecoderMethod = "DecodeAddrMode5Operand"; @@ -962,7 +976,7 @@ def addrmode5_pre : AddrMode5 { // addrmode6 := reg with optional alignment // def AddrMode6AsmOperand : AsmOperandClass { let Name = "AlignedMemory"; } -def addrmode6 : Operand, +def addrmode6 : MemOperand, ComplexPattern{ let PrintMethod = "printAddrMode6Operand"; let MIOperandInfo = (ops GPR:$addr, i32imm:$align); @@ -971,7 +985,7 @@ def addrmode6 : Operand, let ParserMatchClass = AddrMode6AsmOperand; } -def am6offset : Operand, +def am6offset : MemOperand, ComplexPattern { let PrintMethod = "printAddrMode6OffsetOperand"; @@ -982,16 +996,91 @@ def am6offset : Operand, // Special version of addrmode6 to handle alignment encoding for VST1/VLD1 // (single element from one lane) for size 32. -def addrmode6oneL32 : Operand, +def addrmode6oneL32 : MemOperand, ComplexPattern{ let PrintMethod = "printAddrMode6Operand"; let MIOperandInfo = (ops GPR:$addr, i32imm); let EncoderMethod = "getAddrMode6OneLane32AddressOpValue"; } +// Base class for addrmode6 with specific alignment restrictions. +class AddrMode6Align : MemOperand, + ComplexPattern{ + let PrintMethod = "printAddrMode6Operand"; + let MIOperandInfo = (ops GPR:$addr, i32imm:$align); + let EncoderMethod = "getAddrMode6AddressOpValue"; + let DecoderMethod = "DecodeAddrMode6Operand"; +} + +// Special version of addrmode6 to handle no allowed alignment encoding for +// VLD/VST instructions and checking the alignment is not specified. +def AddrMode6AlignNoneAsmOperand : AsmOperandClass { + let Name = "AlignedMemoryNone"; + let DiagnosticType = "AlignedMemoryRequiresNone"; +} +def addrmode6alignNone : AddrMode6Align { + // The alignment specifier can only be omitted. + let ParserMatchClass = AddrMode6AlignNoneAsmOperand; +} + +// Special version of addrmode6 to handle 16-bit alignment encoding for +// VLD/VST instructions and checking the alignment value. +def AddrMode6Align16AsmOperand : AsmOperandClass { + let Name = "AlignedMemory16"; + let DiagnosticType = "AlignedMemoryRequires16"; +} +def addrmode6align16 : AddrMode6Align { + // The alignment specifier can only be 16 or omitted. + let ParserMatchClass = AddrMode6Align16AsmOperand; +} + +// Special version of addrmode6 to handle 32-bit alignment encoding for +// VLD/VST instructions and checking the alignment value. +def AddrMode6Align32AsmOperand : AsmOperandClass { + let Name = "AlignedMemory32"; + let DiagnosticType = "AlignedMemoryRequires32"; +} +def addrmode6align32 : AddrMode6Align { + // The alignment specifier can only be 32 or omitted. + let ParserMatchClass = AddrMode6Align32AsmOperand; +} + +// Special version of addrmode6 to handle 64-bit alignment encoding for +// VLD/VST instructions and checking the alignment value. +def AddrMode6Align64AsmOperand : AsmOperandClass { + let Name = "AlignedMemory64"; + let DiagnosticType = "AlignedMemoryRequires64"; +} +def addrmode6align64 : AddrMode6Align { + // The alignment specifier can only be 64 or omitted. + let ParserMatchClass = AddrMode6Align64AsmOperand; +} + +// Special version of addrmode6 to handle 64-bit or 128-bit alignment encoding +// for VLD/VST instructions and checking the alignment value. +def AddrMode6Align64or128AsmOperand : AsmOperandClass { + let Name = "AlignedMemory64or128"; + let DiagnosticType = "AlignedMemoryRequires64or128"; +} +def addrmode6align64or128 : AddrMode6Align { + // The alignment specifier can only be 64, 128 or omitted. + let ParserMatchClass = AddrMode6Align64or128AsmOperand; +} + +// Special version of addrmode6 to handle 64-bit, 128-bit or 256-bit alignment +// encoding for VLD/VST instructions and checking the alignment value. +def AddrMode6Align64or128or256AsmOperand : AsmOperandClass { + let Name = "AlignedMemory64or128or256"; + let DiagnosticType = "AlignedMemoryRequires64or128or256"; +} +def addrmode6align64or128or256 : AddrMode6Align { + // The alignment specifier can only be 64, 128, 256 or omitted. + let ParserMatchClass = AddrMode6Align64or128or256AsmOperand; +} + // Special version of addrmode6 to handle alignment encoding for VLD-dup // instructions, specifically VLD4-dup. -def addrmode6dup : Operand, +def addrmode6dup : MemOperand, ComplexPattern{ let PrintMethod = "printAddrMode6Operand"; let MIOperandInfo = (ops GPR:$addr, i32imm); @@ -1001,9 +1090,72 @@ def addrmode6dup : Operand, let ParserMatchClass = AddrMode6AsmOperand; } +// Base class for addrmode6dup with specific alignment restrictions. +class AddrMode6DupAlign : MemOperand, + ComplexPattern{ + let PrintMethod = "printAddrMode6Operand"; + let MIOperandInfo = (ops GPR:$addr, i32imm); + let EncoderMethod = "getAddrMode6DupAddressOpValue"; +} + +// Special version of addrmode6 to handle no allowed alignment encoding for +// VLD-dup instruction and checking the alignment is not specified. +def AddrMode6dupAlignNoneAsmOperand : AsmOperandClass { + let Name = "DupAlignedMemoryNone"; + let DiagnosticType = "DupAlignedMemoryRequiresNone"; +} +def addrmode6dupalignNone : AddrMode6DupAlign { + // The alignment specifier can only be omitted. + let ParserMatchClass = AddrMode6dupAlignNoneAsmOperand; +} + +// Special version of addrmode6 to handle 16-bit alignment encoding for VLD-dup +// instruction and checking the alignment value. +def AddrMode6dupAlign16AsmOperand : AsmOperandClass { + let Name = "DupAlignedMemory16"; + let DiagnosticType = "DupAlignedMemoryRequires16"; +} +def addrmode6dupalign16 : AddrMode6DupAlign { + // The alignment specifier can only be 16 or omitted. + let ParserMatchClass = AddrMode6dupAlign16AsmOperand; +} + +// Special version of addrmode6 to handle 32-bit alignment encoding for VLD-dup +// instruction and checking the alignment value. +def AddrMode6dupAlign32AsmOperand : AsmOperandClass { + let Name = "DupAlignedMemory32"; + let DiagnosticType = "DupAlignedMemoryRequires32"; +} +def addrmode6dupalign32 : AddrMode6DupAlign { + // The alignment specifier can only be 32 or omitted. + let ParserMatchClass = AddrMode6dupAlign32AsmOperand; +} + +// Special version of addrmode6 to handle 64-bit alignment encoding for VLD +// instructions and checking the alignment value. +def AddrMode6dupAlign64AsmOperand : AsmOperandClass { + let Name = "DupAlignedMemory64"; + let DiagnosticType = "DupAlignedMemoryRequires64"; +} +def addrmode6dupalign64 : AddrMode6DupAlign { + // The alignment specifier can only be 64 or omitted. + let ParserMatchClass = AddrMode6dupAlign64AsmOperand; +} + +// Special version of addrmode6 to handle 64-bit or 128-bit alignment encoding +// for VLD instructions and checking the alignment value. +def AddrMode6dupAlign64or128AsmOperand : AsmOperandClass { + let Name = "DupAlignedMemory64or128"; + let DiagnosticType = "DupAlignedMemoryRequires64or128"; +} +def addrmode6dupalign64or128 : AddrMode6DupAlign { + // The alignment specifier can only be 64, 128 or omitted. + let ParserMatchClass = AddrMode6dupAlign64or128AsmOperand; +} + // addrmodepc := pc + reg // -def addrmodepc : Operand, +def addrmodepc : MemOperand, ComplexPattern { let PrintMethod = "printAddrModePCOperand"; let MIOperandInfo = (ops GPR, i32imm); @@ -1012,7 +1164,7 @@ def addrmodepc : Operand, // addr_offset_none := reg // def MemNoOffsetAsmOperand : AsmOperandClass { let Name = "MemNoOffset"; } -def addr_offset_none : Operand, +def addr_offset_none : MemOperand, ComplexPattern { let PrintMethod = "printAddrMode7Operand"; let DecoderMethod = "DecodeAddrMode7Operand"; @@ -1059,7 +1211,7 @@ include "ARMInstrFormats.td" // Multiclass helpers... // -/// AsI1_bin_irs - Defines a set of (op r, {so_imm|r|so_reg}) patterns for a +/// AsI1_bin_irs - Defines a set of (op r, {mod_imm|r|so_reg}) patterns for a /// binop that produces a value. let TwoOperandAliasConstraint = "$Rn = $Rd" in multiclass AsI1_bin_irs opcod, string opc, @@ -1068,9 +1220,9 @@ multiclass AsI1_bin_irs opcod, string opc, // The register-immediate version is re-materializable. This is useful // in particular for taking the address of a local. let isReMaterializable = 1 in { - def ri : AsI1, + [(set GPR:$Rd, (opnode GPR:$Rn, mod_imm:$imm))]>, Sched<[WriteALU, ReadALU]> { bits<4> Rd; bits<4> Rn; @@ -1141,9 +1293,9 @@ multiclass AsI1_rbin_irs opcod, string opc, // The register-immediate version is re-materializable. This is useful // in particular for taking the address of a local. let isReMaterializable = 1 in { - def ri : AsI1, + [(set GPR:$Rd, (opnode mod_imm:$imm, GPR:$Rn))]>, Sched<[WriteALU, ReadALU]> { bits<4> Rd; bits<4> Rn; @@ -1211,9 +1363,9 @@ let hasPostISelHook = 1, Defs = [CPSR] in { multiclass AsI1_bin_s_irs { - def ri : ARMPseudoInst<(outs GPR:$Rd), (ins GPR:$Rn, so_imm:$imm, pred:$p), + def ri : ARMPseudoInst<(outs GPR:$Rd), (ins GPR:$Rn, mod_imm:$imm, pred:$p), 4, iii, - [(set GPR:$Rd, CPSR, (opnode GPR:$Rn, so_imm:$imm))]>, + [(set GPR:$Rd, CPSR, (opnode GPR:$Rn, mod_imm:$imm))]>, Sched<[WriteALU, ReadALU]>; def rr : ARMPseudoInst<(outs GPR:$Rd), (ins GPR:$Rn, GPR:$Rm, pred:$p), @@ -1244,9 +1396,9 @@ let hasPostISelHook = 1, Defs = [CPSR] in { multiclass AsI1_rbin_s_is { - def ri : ARMPseudoInst<(outs GPR:$Rd), (ins GPR:$Rn, so_imm:$imm, pred:$p), + def ri : ARMPseudoInst<(outs GPR:$Rd), (ins GPR:$Rn, mod_imm:$imm, pred:$p), 4, iii, - [(set GPR:$Rd, CPSR, (opnode so_imm:$imm, GPR:$Rn))]>, + [(set GPR:$Rd, CPSR, (opnode mod_imm:$imm, GPR:$Rn))]>, Sched<[WriteALU, ReadALU]>; def rsi : ARMPseudoInst<(outs GPR:$Rd), @@ -1265,16 +1417,17 @@ multiclass AsI1_rbin_s_is opcod, string opc, InstrItinClass iii, InstrItinClass iir, InstrItinClass iis, - PatFrag opnode, bit Commutable = 0> { - def ri : AI1 { + def ri : AI1, + [(opnode GPR:$Rn, mod_imm:$imm)]>, Sched<[WriteCMP, ReadALU]> { bits<4> Rn; bits<12> imm; @@ -1299,6 +1452,7 @@ multiclass AI1_cmp_irs opcod, string opc, let Inst{15-12} = 0b0000; let Inst{11-4} = 0b00000000; let Inst{3-0} = Rm; + let DecoderMethod = rrDecoderMethod; let Unpredictable{15-12} = 0b1111; } @@ -1402,9 +1556,9 @@ let TwoOperandAliasConstraint = "$Rn = $Rd" in multiclass AI1_adde_sube_irs opcod, string opc, PatFrag opnode, bit Commutable = 0> { let hasPostISelHook = 1, Defs = [CPSR], Uses = [CPSR] in { - def ri : AsI1, + [(set GPR:$Rd, CPSR, (opnode GPR:$Rn, mod_imm:$imm, CPSR))]>, Requires<[IsARM]>, Sched<[WriteALU, ReadALU]> { bits<4> Rd; @@ -1472,9 +1626,9 @@ multiclass AI1_adde_sube_irs opcod, string opc, PatFrag opnode, let TwoOperandAliasConstraint = "$Rn = $Rd" in multiclass AI1_rsc_irs opcod, string opc, PatFrag opnode> { let hasPostISelHook = 1, Defs = [CPSR], Uses = [CPSR] in { - def ri : AsI1, + [(set GPR:$Rd, CPSR, (opnode mod_imm:$imm, GPR:$Rn, CPSR))]>, Requires<[IsARM]>, Sched<[WriteALU, ReadALU]> { bits<4> Rd; @@ -1668,7 +1822,7 @@ multiclass AI_str1nopc; @@ -1687,21 +1841,20 @@ PseudoInst<(outs), (ins i32imm:$amt, pred:$p), NoItinerary, } def HINT : AI<(outs), (ins imm0_239:$imm), MiscFrm, NoItinerary, - "hint", "\t$imm", []>, Requires<[IsARM, HasV6]> { + "hint", "\t$imm", [(int_arm_hint imm0_239:$imm)]>, + Requires<[IsARM, HasV6]> { bits<8> imm; let Inst{27-8} = 0b00110010000011110000; let Inst{7-0} = imm; } -def : InstAlias<"nop$p", (HINT 0, pred:$p)>, Requires<[IsARM, HasV6T2]>; -def : InstAlias<"yield$p", (HINT 1, pred:$p)>, Requires<[IsARM, HasV6T2]>; -def : InstAlias<"wfe$p", (HINT 2, pred:$p)>, Requires<[IsARM, HasV6T2]>; -def : InstAlias<"wfi$p", (HINT 3, pred:$p)>, Requires<[IsARM, HasV6T2]>; -def : InstAlias<"sev$p", (HINT 4, pred:$p)>, Requires<[IsARM, HasV6T2]>; +def : InstAlias<"nop$p", (HINT 0, pred:$p)>, Requires<[IsARM, HasV6K]>; +def : InstAlias<"yield$p", (HINT 1, pred:$p)>, Requires<[IsARM, HasV6K]>; +def : InstAlias<"wfe$p", (HINT 2, pred:$p)>, Requires<[IsARM, HasV6K]>; +def : InstAlias<"wfi$p", (HINT 3, pred:$p)>, Requires<[IsARM, HasV6K]>; +def : InstAlias<"sev$p", (HINT 4, pred:$p)>, Requires<[IsARM, HasV6K]>; def : InstAlias<"sevl$p", (HINT 5, pred:$p)>, Requires<[IsARM, HasV8]>; -def : Pat<(int_arm_sevl), (HINT 5)>; - def SEL : AI<(outs GPR:$Rd), (ins GPR:$Rn, GPR:$Rm), DPFrm, NoItinerary, "sel", "\t$Rd, $Rn, $Rm", []>, Requires<[IsARM, HasV6]> { bits<4> Rd; @@ -1774,8 +1927,8 @@ let imod = 0, iflags = 0, M = 1 in // Preload signals the memory system of possible future data/instruction access. multiclass APreLoad read, bits<1> data, string opc> { - def i12 : AXI<(outs), (ins addrmode_imm12:$addr), MiscFrm, IIC_Preload, - !strconcat(opc, "\t$addr"), + def i12 : AXIM<(outs), (ins addrmode_imm12:$addr), AddrMode_i12, MiscFrm, + IIC_Preload, !strconcat(opc, "\t$addr"), [(ARMPreload addrmode_imm12:$addr, (i32 read), (i32 data))]>, Sched<[WritePreLd]> { bits<4> Rt; @@ -1822,12 +1975,24 @@ def SETEND : AXI<(outs), (ins setend_op:$end), MiscFrm, NoItinerary, } def DBG : AI<(outs), (ins imm0_15:$opt), MiscFrm, NoItinerary, "dbg", "\t$opt", - []>, Requires<[IsARM, HasV7]> { + [(int_arm_dbg imm0_15:$opt)]>, Requires<[IsARM, HasV7]> { bits<4> opt; let Inst{27-4} = 0b001100100000111100001111; let Inst{3-0} = opt; } +// A8.8.247 UDF - Undefined (Encoding A1) +def UDF : AInoP<(outs), (ins imm0_65535:$imm16), MiscFrm, NoItinerary, + "udf", "\t$imm16", [(int_arm_undefined imm0_65535:$imm16)]> { + bits<16> imm16; + let Inst{31-28} = 0b1110; // AL + let Inst{27-25} = 0b011; + let Inst{24-20} = 0b11111; + let Inst{19-8} = imm16{15-4}; + let Inst{7-4} = 0b1111; + let Inst{3-0} = imm16{3-0}; +} + /* * A5.4 Permanently UNDEFINED instructions. * @@ -1901,7 +2066,7 @@ def PICSTRB : ARMPseudoInst<(outs), (ins GPR:$src, addrmodepc:$addr, pred:$p), // LEApcrel - Load a pc-relative address into a register without offending the // assembler. -let neverHasSideEffects = 1, isReMaterializable = 1 in +let hasSideEffects = 0, isReMaterializable = 1 in // The 'adr' mnemonic encodes differently if the label is before or after // the instruction. The {24-21} opcode bits are set by the fixup, as we don't // know until then which form of the instruction will be used. @@ -2231,6 +2396,33 @@ def RFEIB_UPD : RFEI<1, "rfeib\t$Rn!"> { let Inst{24-23} = 0b11; } +// Hypervisor Call is a system instruction +let isCall = 1 in { +def HVC : AInoP< (outs), (ins imm0_65535:$imm), BrFrm, NoItinerary, + "hvc", "\t$imm", []>, + Requires<[IsARM, HasVirtualization]> { + bits<16> imm; + + // Even though HVC isn't predicable, it's encoding includes a condition field. + // The instruction is undefined if the condition field is 0xf otherwise it is + // unpredictable if it isn't condition AL (0xe). + let Inst{31-28} = 0b1110; + let Unpredictable{31-28} = 0b1111; + let Inst{27-24} = 0b0001; + let Inst{23-20} = 0b0100; + let Inst{19-8} = imm{15-4}; + let Inst{7-4} = 0b0111; + let Inst{3-0} = imm{3-0}; +} +} + +// Return from exception in Hypervisor mode. +let isReturn = 1, isBarrier = 1, isTerminator = 1, Defs = [PC] in +def ERET : ABI<0b0001, (outs), (ins), NoItinerary, "eret", "", []>, + Requires<[IsARM, HasVirtualization]> { + let Inst{23-0} = 0b011000000000000001101110; +} + //===----------------------------------------------------------------------===// // Load / Store Instructions. // @@ -2248,7 +2440,7 @@ defm STRB : AI_str1nopc<1, "strb", IIC_iStore_bh_r, IIC_iStore_bh_si, BinOpFrag<(truncstorei8 node:$LHS, node:$RHS)>>; // Special LDR for loads from non-pc-relative constpools. -let canFoldAsLoad = 1, mayLoad = 1, neverHasSideEffects = 1, +let canFoldAsLoad = 1, mayLoad = 1, hasSideEffects = 0, isReMaterializable = 1, isCodeGenOnly = 1 in def LDRcp : AI2ldst<0b010, 1, 0, (outs GPR:$Rt), (ins addrmode_imm12:$addr), AddrMode_i12, LdFrm, IIC_iLoad_r, "ldr", "\t$Rt, $addr", @@ -2275,17 +2467,11 @@ def LDRSB : AI3ld<0b1101, 1, (outs GPR:$Rt), (ins addrmode3:$addr), LdMiscFrm, IIC_iLoad_bh_r, "ldrsb", "\t$Rt, $addr", [(set GPR:$Rt, (sextloadi8 addrmode3:$addr))]>; -let mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1 in { +let mayLoad = 1, hasSideEffects = 0, hasExtraDefRegAllocReq = 1 in { // Load doubleword def LDRD : AI3ld<0b1101, 0, (outs GPR:$Rt, GPR:$Rt2), (ins addrmode3:$addr), LdMiscFrm, IIC_iLoad_d_r, "ldrd", "\t$Rt, $Rt2, $addr", []>, Requires<[IsARM, HasV5TE]>; - - // GNU Assembler extension (compatibility) - let isAsmParserOnly = 1 in - def LDRD_PAIR : AI3ld<0b1101, 0, (outs GPRPairOp:$Rt), (ins addrmode3:$addr), - LdMiscFrm, IIC_iLoad_d_r, "ldrd", "\t$Rt, $addr", []>, - Requires<[IsARM, HasV5TE]>; } def LDA : AIldracq<0b00, (outs GPR:$Rt), (ins addr_offset_none:$addr), @@ -2358,7 +2544,7 @@ multiclass AI2_ldridx; @@ -2394,7 +2580,7 @@ multiclass AI3_ldridx op, string opc, InstrItinClass itin> { } } -let mayLoad = 1, neverHasSideEffects = 1 in { +let mayLoad = 1, hasSideEffects = 0 in { defm LDRH : AI3_ldridx<0b1011, "ldrh", IIC_iLoad_bh_ru>; defm LDRSH : AI3_ldridx<0b1111, "ldrsh", IIC_iLoad_bh_ru>; defm LDRSB : AI3_ldridx<0b1101, "ldrsb", IIC_iLoad_bh_ru>; @@ -2427,10 +2613,10 @@ def LDRD_POST: AI3ldstidx<0b1101, 0, 0, (outs GPR:$Rt, GPR:$Rt2, GPR:$Rn_wb), let DecoderMethod = "DecodeAddrMode3Instruction"; } } // hasExtraDefRegAllocReq = 1 -} // mayLoad = 1, neverHasSideEffects = 1 +} // mayLoad = 1, hasSideEffects = 0 // LDRT, LDRBT, LDRSBT, LDRHT, LDRSHT. -let mayLoad = 1, neverHasSideEffects = 1 in { +let mayLoad = 1, hasSideEffects = 0 in { def LDRT_POST_REG : AI2ldstidx<1, 0, 0, (outs GPR:$Rt, GPR:$Rn_wb), (ins addr_offset_none:$addr, am2offset_reg:$offset), IndexModePost, LdFrm, IIC_iLoad_ru, @@ -2450,30 +2636,23 @@ def LDRT_POST_REG : AI2ldstidx<1, 0, 0, (outs GPR:$Rt, GPR:$Rn_wb), let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } -class LDRTImmediate - : AI2ldstidx<1, 0, 0, (outs GPR:$Rt, GPR:$Rn_wb), iops, +def LDRT_POST_IMM + : AI2ldstidx<1, 0, 0, (outs GPR:$Rt, GPR:$Rn_wb), + (ins addr_offset_none:$addr, am2offset_imm:$offset), IndexModePost, LdFrm, IIC_iLoad_ru, - "ldrt", args, "$addr.base = $Rn_wb", []> { + "ldrt", "\t$Rt, $addr, $offset", "$addr.base = $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; bits<4> addr; let Inst{25} = 0; - let Inst{23} = !if(has_offset, offset{12}, 1); + let Inst{23} = offset{12}; let Inst{21} = 1; // overwrite let Inst{19-16} = addr; - let Inst{11-0} = !if(has_offset, offset{11-0}, 0); -} - -def LDRT_POST_IMM - : LDRTImmediate<1, "\t$Rt, $addr, $offset", - (ins addr_offset_none:$addr, am2offset_imm:$offset)> { + let Inst{11-0} = offset{11-0}; let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } -def LDRT_POST_IMM_0 - : LDRTImmediate<0, "\t$Rt, $addr", (ins addr_offset_none:$addr)>; - def LDRBT_POST_REG : AI2ldstidx<1, 1, 0, (outs GPR:$Rt, GPR:$Rn_wb), (ins addr_offset_none:$addr, am2offset_reg:$offset), IndexModePost, LdFrm, IIC_iLoad_bh_ru, @@ -2493,30 +2672,23 @@ def LDRBT_POST_REG : AI2ldstidx<1, 1, 0, (outs GPR:$Rt, GPR:$Rn_wb), let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } -class LDRBTImmediate - : AI2ldstidx<1, 1, 0, (outs GPR:$Rt, GPR:$Rn_wb), iops, - IndexModePost, LdFrm, IIC_iLoad_bh_ru, - "ldrbt", args, "$addr.base = $Rn_wb", []> { +def LDRBT_POST_IMM + : AI2ldstidx<1, 1, 0, (outs GPR:$Rt, GPR:$Rn_wb), + (ins addr_offset_none:$addr, am2offset_imm:$offset), + IndexModePost, LdFrm, IIC_iLoad_bh_ru, + "ldrbt", "\t$Rt, $addr, $offset", "$addr.base = $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; bits<4> addr; let Inst{25} = 0; - let Inst{23} = !if(has_offset, offset{12}, 1); + let Inst{23} = offset{12}; let Inst{21} = 1; // overwrite let Inst{19-16} = addr; - let Inst{11-0} = !if(has_offset, offset{11-0}, 0); -} - -def LDRBT_POST_IMM - : LDRBTImmediate<1, "\t$Rt, $addr, $offset", - (ins addr_offset_none:$addr, am2offset_imm:$offset)> { + let Inst{11-0} = offset{11-0}; let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } -def LDRBT_POST_IMM_0 - : LDRBTImmediate<0, "\t$Rt, $addr", (ins addr_offset_none:$addr)>; - multiclass AI3ldrT op, string opc> { def i : AI3ldstidxT; defm LDRSHT : AI3ldrT<0b1111, "ldrsht">; } +def LDRT_POST + : ARMAsmPseudo<"ldrt${q} $Rt, $addr", (ins addr_offset_none:$addr, pred:$q), + (outs GPR:$Rt)>; + +def LDRBT_POST + : ARMAsmPseudo<"ldrbt${q} $Rt, $addr", (ins addr_offset_none:$addr, pred:$q), + (outs GPR:$Rt)>; + // Store // Stores with truncate @@ -2555,20 +2735,12 @@ def STRH : AI3str<0b1011, (outs), (ins GPR:$Rt, addrmode3:$addr), StMiscFrm, [(truncstorei16 GPR:$Rt, addrmode3:$addr)]>; // Store doubleword -let mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1 in { +let mayStore = 1, hasSideEffects = 0, hasExtraSrcRegAllocReq = 1 in { def STRD : AI3str<0b1111, (outs), (ins GPR:$Rt, GPR:$Rt2, addrmode3:$addr), StMiscFrm, IIC_iStore_d_r, "strd", "\t$Rt, $Rt2, $addr", []>, Requires<[IsARM, HasV5TE]> { let Inst{21} = 0; } - - // GNU Assembler extension (compatibility) - let isAsmParserOnly = 1 in - def STRD_PAIR : AI3str<0b1111, (outs), (ins GPRPairOp:$Rt, addrmode3:$addr), - StMiscFrm, IIC_iStore_d_r, "strd", "\t$Rt, $addr", []>, - Requires<[IsARM, HasV5TE]> { - let Inst{21} = 0; - } } // Indexed stores @@ -2577,7 +2749,8 @@ multiclass AI2_stridx { + opc, "\t$Rt, $addr!", + "$addr.base = $Rn_wb,@earlyclobber $Rn_wb", []> { bits<17> addr; let Inst{25} = 0; let Inst{23} = addr{12}; // U (add = ('U' == 1)) @@ -2589,7 +2762,8 @@ multiclass AI2_stridx { + opc, "\t$Rt, $addr!", + "$addr.base = $Rn_wb,@earlyclobber $Rn_wb", []> { bits<17> addr; let Inst{25} = 1; let Inst{23} = addr{12}; // U (add = ('U' == 1)) @@ -2602,7 +2776,7 @@ multiclass AI2_stridx { + "$addr.base = $Rn_wb,@earlyclobber $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; @@ -2620,7 +2794,7 @@ multiclass AI2_stridx { + "$addr.base = $Rn_wb,@earlyclobber $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; @@ -2634,7 +2808,7 @@ multiclass AI2_stridx; @@ -2697,7 +2871,8 @@ def STRH_preidx: ARMPseudoInst<(outs GPR:$Rn_wb), def STRH_PRE : AI3ldstidx<0b1011, 0, 1, (outs GPR:$Rn_wb), (ins GPR:$Rt, addrmode3_pre:$addr), IndexModePre, StMiscFrm, IIC_iStore_bh_ru, - "strh", "\t$Rt, $addr!", "$addr.base = $Rn_wb", []> { + "strh", "\t$Rt, $addr!", + "$addr.base = $Rn_wb,@earlyclobber $Rn_wb", []> { bits<14> addr; let Inst{23} = addr{8}; // U bit let Inst{22} = addr{13}; // 1 == imm8, 0 == Rm @@ -2710,7 +2885,8 @@ def STRH_PRE : AI3ldstidx<0b1011, 0, 1, (outs GPR:$Rn_wb), def STRH_POST : AI3ldstidx<0b1011, 0, 0, (outs GPR:$Rn_wb), (ins GPR:$Rt, addr_offset_none:$addr, am3offset:$offset), IndexModePost, StMiscFrm, IIC_iStore_bh_ru, - "strh", "\t$Rt, $addr, $offset", "$addr.base = $Rn_wb", + "strh", "\t$Rt, $addr, $offset", + "$addr.base = $Rn_wb,@earlyclobber $Rn_wb", [(set GPR:$Rn_wb, (post_truncsti16 GPR:$Rt, addr_offset_none:$addr, am3offset:$offset))]> { @@ -2724,7 +2900,7 @@ def STRH_POST : AI3ldstidx<0b1011, 0, 0, (outs GPR:$Rn_wb), let DecoderMethod = "DecodeAddrMode3Instruction"; } -let mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1 in { +let mayStore = 1, hasSideEffects = 0, hasExtraSrcRegAllocReq = 1 in { def STRD_PRE : AI3ldstidx<0b1111, 0, 1, (outs GPR:$Rn_wb), (ins GPR:$Rt, GPR:$Rt2, addrmode3_pre:$addr), IndexModePre, StMiscFrm, IIC_iStore_d_ru, @@ -2754,7 +2930,7 @@ def STRD_POST: AI3ldstidx<0b1111, 0, 0, (outs GPR:$Rn_wb), let Inst{3-0} = offset{3-0}; // imm3_0/Rm let DecoderMethod = "DecodeAddrMode3Instruction"; } -} // mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1 +} // mayStore = 1, hasSideEffects = 0, hasExtraSrcRegAllocReq = 1 // STRT, STRBT, and STRHT @@ -2777,31 +2953,28 @@ def STRBT_POST_REG : AI2ldstidx<0, 1, 0, (outs GPR:$Rn_wb), let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } -class STRBTImmediate - : AI2ldstidx<0, 1, 0, (outs GPR:$Rn_wb), iops, IndexModePost, StFrm, - IIC_iStore_bh_ru, "strbt", args, "$addr.base = $Rn_wb", []> { +def STRBT_POST_IMM + : AI2ldstidx<0, 1, 0, (outs GPR:$Rn_wb), + (ins GPR:$Rt, addr_offset_none:$addr, am2offset_imm:$offset), + IndexModePost, StFrm, IIC_iStore_bh_ru, + "strbt", "\t$Rt, $addr, $offset", "$addr.base = $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; bits<4> addr; let Inst{25} = 0; - let Inst{23} = !if(has_offset, offset{12}, 1); + let Inst{23} = offset{12}; let Inst{21} = 1; // overwrite let Inst{19-16} = addr; - let Inst{11-0} = !if(has_offset, offset{11-0}, 0); -} - -def STRBT_POST_IMM - : STRBTImmediate<1, "\t$Rt, $addr, $offset", - (ins GPR:$Rt, addr_offset_none:$addr, am2offset_imm:$offset)> { + let Inst{11-0} = offset{11-0}; let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } +def STRBT_POST + : ARMAsmPseudo<"strbt${q} $Rt, $addr", + (ins GPR:$Rt, addr_offset_none:$addr, pred:$q)>; -def STRBT_POST_IMM_0 - : STRBTImmediate<0, "\t$Rt, $addr", (ins GPR:$Rt, addr_offset_none:$addr)>; - -let mayStore = 1, neverHasSideEffects = 1 in { +let mayStore = 1, hasSideEffects = 0 in { def STRT_POST_REG : AI2ldstidx<0, 0, 0, (outs GPR:$Rn_wb), (ins GPR:$Rt, addr_offset_none:$addr, am2offset_reg:$offset), IndexModePost, StFrm, IIC_iStore_ru, @@ -2821,30 +2994,27 @@ def STRT_POST_REG : AI2ldstidx<0, 0, 0, (outs GPR:$Rn_wb), let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } -class STRTImmediate - : AI2ldstidx<0, 0, 0, (outs GPR:$Rn_wb), iops, IndexModePost, StFrm, - IIC_iStore_ru, "strt", args, "$addr.base = $Rn_wb", []> { +def STRT_POST_IMM + : AI2ldstidx<0, 0, 0, (outs GPR:$Rn_wb), + (ins GPR:$Rt, addr_offset_none:$addr, am2offset_imm:$offset), + IndexModePost, StFrm, IIC_iStore_ru, + "strt", "\t$Rt, $addr, $offset", "$addr.base = $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; bits<4> addr; let Inst{25} = 0; - let Inst{23} = !if(has_offset, offset{12}, 1); + let Inst{23} = offset{12}; let Inst{21} = 1; // overwrite let Inst{19-16} = addr; - let Inst{11-0} = !if(has_offset, offset{11-0}, 0); -} - -def STRT_POST_IMM - : STRTImmediate<1, "\t$Rt, $addr, $offset", - (ins GPR:$Rt, addr_offset_none:$addr, am2offset_imm:$offset)> { + let Inst{11-0} = offset{11-0}; let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } - -def STRT_POST_IMM_0 - : STRTImmediate<0, "\t$Rt, $addr", (ins GPR:$Rt, addr_offset_none:$addr)>; } +def STRT_POST + : ARMAsmPseudo<"strt${q} $Rt, $addr", + (ins GPR:$Rt, addr_offset_none:$addr, pred:$q)>; multiclass AI3strT op, string opc> { def i : AI3ldstidxT; + IIC_iLoad_mu>, ComplexDeprecationPredicate<"ARMLoad">; let mayStore = 1, hasExtraSrcRegAllocReq = 1 in defm STM : arm_ldst_mult<"stm", "", 0, 0, LdStMulFrm, IIC_iStore_m, - IIC_iStore_mu>; + IIC_iStore_mu>, + ComplexDeprecationPredicate<"ARMStore">; -} // neverHasSideEffects +} // hasSideEffects // FIXME: remove when we have a way to marking a MI with these properties. // FIXME: Should pc be an implicit operand like PICADD, etc? @@ -3005,7 +3176,7 @@ defm sysSTM : arm_ldst_mult<"stm", " ^", 0, 1, LdStMulFrm, IIC_iStore_m, // Move Instructions. // -let neverHasSideEffects = 1 in +let hasSideEffects = 0 in def MOVr : AsI1<0b1101, (outs GPR:$Rd), (ins GPR:$Rm), DPFrm, IIC_iMOVr, "mov", "\t$Rd, $Rm", []>, UnaryDP, Sched<[WriteALU]> { bits<4> Rd; @@ -3019,7 +3190,7 @@ def MOVr : AsI1<0b1101, (outs GPR:$Rd), (ins GPR:$Rm), DPFrm, IIC_iMOVr, } // A version for the smaller set of tail call registers. -let neverHasSideEffects = 1 in +let hasSideEffects = 0 in def MOVr_TC : AsI1<0b1101, (outs tcGPR:$Rd), (ins tcGPR:$Rm), DPFrm, IIC_iMOVr, "mov", "\t$Rd, $Rm", []>, UnaryDP, Sched<[WriteALU]> { bits<4> Rd; @@ -3063,8 +3234,8 @@ def MOVsi : AsI1<0b1101, (outs GPR:$Rd), (ins shift_so_reg_imm:$src), } let isReMaterializable = 1, isAsCheapAsAMove = 1, isMoveImm = 1 in -def MOVi : AsI1<0b1101, (outs GPR:$Rd), (ins so_imm:$imm), DPFrm, IIC_iMOVi, - "mov", "\t$Rd, $imm", [(set GPR:$Rd, so_imm:$imm)]>, UnaryDP, +def MOVi : AsI1<0b1101, (outs GPR:$Rd), (ins mod_imm:$imm), DPFrm, IIC_iMOVi, + "mov", "\t$Rd, $imm", [(set GPR:$Rd, mod_imm:$imm)]>, UnaryDP, Sched<[WriteALU]> { bits<4> Rd; bits<12> imm; @@ -3209,8 +3380,8 @@ def SBFX : I<(outs GPRnopc:$Rd), let Inst{3-0} = Rn; } -def UBFX : I<(outs GPR:$Rd), - (ins GPR:$Rn, imm0_31:$lsb, imm1_32:$width), +def UBFX : I<(outs GPRnopc:$Rd), + (ins GPRnopc:$Rn, imm0_31:$lsb, imm1_32:$width), AddrMode1, 4, IndexModeNone, DPFrm, IIC_iUNAsi, "ubfx", "\t$Rd, $Rn, $lsb, $width", "", []>, Requires<[IsARM, HasV6T2]> { @@ -3274,10 +3445,10 @@ defm RSC : AI1_rsc_irs<0b0111, "rsc", // assume opposite meanings of the carry flag (i.e., carry == !borrow). // See the definition of AddWithCarry() in the ARM ARM A2.2.1 for the gory // details. -def : ARMPat<(add GPR:$src, so_imm_neg:$imm), - (SUBri GPR:$src, so_imm_neg:$imm)>; -def : ARMPat<(ARMaddc GPR:$src, so_imm_neg:$imm), - (SUBSri GPR:$src, so_imm_neg:$imm)>; +def : ARMPat<(add GPR:$src, mod_imm_neg:$imm), + (SUBri GPR:$src, mod_imm_neg:$imm)>; +def : ARMPat<(ARMaddc GPR:$src, mod_imm_neg:$imm), + (SUBSri GPR:$src, mod_imm_neg:$imm)>; def : ARMPat<(add GPR:$src, imm0_65535_neg:$imm), (SUBrr GPR:$src, (MOVi16 (imm_neg_XFORM imm:$imm)))>, @@ -3289,10 +3460,11 @@ def : ARMPat<(ARMaddc GPR:$src, imm0_65535_neg:$imm), // The with-carry-in form matches bitwise not instead of the negation. // Effectively, the inverse interpretation of the carry flag already accounts // for part of the negation. -def : ARMPat<(ARMadde GPR:$src, so_imm_not:$imm, CPSR), - (SBCri GPR:$src, so_imm_not:$imm)>; +def : ARMPat<(ARMadde GPR:$src, mod_imm_not:$imm, CPSR), + (SBCri GPR:$src, mod_imm_not:$imm)>; def : ARMPat<(ARMadde GPR:$src, imm0_65535_neg:$imm, CPSR), - (SBCrr GPR:$src, (MOVi16 (imm_not_XFORM imm:$imm)))>; + (SBCrr GPR:$src, (MOVi16 (imm_not_XFORM imm:$imm)))>, + Requires<[IsARM, HasV6T2]>; // Note: These are implemented in C++ code, because they have to generate // ADD/SUBrs instructions, which use a complex pattern that a xform function @@ -3570,9 +3742,9 @@ def MVNsr : AsI1<0b1111, (outs GPR:$Rd), (ins so_reg_reg:$shift), let Inst{3-0} = shift{3-0}; } let isReMaterializable = 1, isAsCheapAsAMove = 1, isMoveImm = 1 in -def MVNi : AsI1<0b1111, (outs GPR:$Rd), (ins so_imm:$imm), DPFrm, +def MVNi : AsI1<0b1111, (outs GPR:$Rd), (ins mod_imm:$imm), DPFrm, IIC_iMVNi, "mvn", "\t$Rd, $imm", - [(set GPR:$Rd, so_imm_not:$imm)]>,UnaryDP, Sched<[WriteALU]> { + [(set GPR:$Rd, mod_imm_not:$imm)]>,UnaryDP, Sched<[WriteALU]> { bits<4> Rd; bits<12> imm; let Inst{25} = 1; @@ -3581,8 +3753,8 @@ def MVNi : AsI1<0b1111, (outs GPR:$Rd), (ins so_imm:$imm), DPFrm, let Inst{11-0} = imm; } -def : ARMPat<(and GPR:$src, so_imm_not:$imm), - (BICri GPR:$src, so_imm_not:$imm)>; +def : ARMPat<(and GPR:$src, mod_imm_not:$imm), + (BICri GPR:$src, mod_imm_not:$imm)>; //===----------------------------------------------------------------------===// // Multiply Instructions. @@ -3644,21 +3816,22 @@ def MULv5: ARMPseudoExpand<(outs GPRnopc:$Rd), (ins GPRnopc:$Rn, GPRnopc:$Rm, Requires<[IsARM, NoV6, UseMulOps]>; } -def MLA : AsMul1I32<0b0000001, (outs GPR:$Rd), (ins GPR:$Rn, GPR:$Rm, GPR:$Ra), +def MLA : AsMul1I32<0b0000001, (outs GPRnopc:$Rd), + (ins GPRnopc:$Rn, GPRnopc:$Rm, GPRnopc:$Ra), IIC_iMAC32, "mla", "\t$Rd, $Rn, $Rm, $Ra", - [(set GPR:$Rd, (add (mul GPR:$Rn, GPR:$Rm), GPR:$Ra))]>, - Requires<[IsARM, HasV6, UseMulOps]> { + [(set GPRnopc:$Rd, (add (mul GPRnopc:$Rn, GPRnopc:$Rm), GPRnopc:$Ra))]>, + Requires<[IsARM, HasV6, UseMulOps]> { bits<4> Ra; let Inst{15-12} = Ra; } let Constraints = "@earlyclobber $Rd" in -def MLAv5: ARMPseudoExpand<(outs GPR:$Rd), - (ins GPR:$Rn, GPR:$Rm, GPR:$Ra, pred:$p, cc_out:$s), - 4, IIC_iMAC32, - [(set GPR:$Rd, (add (mul GPR:$Rn, GPR:$Rm), GPR:$Ra))], - (MLA GPR:$Rd, GPR:$Rn, GPR:$Rm, GPR:$Ra, pred:$p, cc_out:$s)>, - Requires<[IsARM, NoV6]>; +def MLAv5: ARMPseudoExpand<(outs GPRnopc:$Rd), + (ins GPRnopc:$Rn, GPRnopc:$Rm, GPRnopc:$Ra, + pred:$p, cc_out:$s), 4, IIC_iMAC32, + [(set GPRnopc:$Rd, (add (mul GPRnopc:$Rn, GPRnopc:$Rm), GPRnopc:$Ra))], + (MLA GPRnopc:$Rd, GPRnopc:$Rn, GPRnopc:$Rm, GPRnopc:$Ra, pred:$p, cc_out:$s)>, + Requires<[IsARM, NoV6]>; def MLS : AMul1I<0b0000011, (outs GPR:$Rd), (ins GPR:$Rn, GPR:$Rm, GPR:$Ra), IIC_iMAC32, "mls", "\t$Rd, $Rn, $Rm, $Ra", @@ -3675,7 +3848,7 @@ def MLS : AMul1I<0b0000011, (outs GPR:$Rd), (ins GPR:$Rn, GPR:$Rm, GPR:$Ra), } // Extra precision multiplies with low / high results -let neverHasSideEffects = 1 in { +let hasSideEffects = 0 in { let isCommutable = 1 in { def SMULL : AsMul1I64<0b0000110, (outs GPR:$RdLo, GPR:$RdHi), (ins GPR:$Rn, GPR:$Rm), IIC_iMUL64, @@ -3726,7 +3899,8 @@ def UMAAL : AMul1I <0b0000010, (outs GPR:$RdLo, GPR:$RdHi), let Inst{3-0} = Rn; } -let Constraints = "$RLo = $RdLo,$RHi = $RdHi" in { +let Constraints = + "@earlyclobber $RdLo,@earlyclobber $RdHi,$RLo = $RdLo,$RHi = $RdHi" in { def SMLALv5 : ARMPseudoExpand<(outs GPR:$RdLo, GPR:$RdHi), (ins GPR:$Rn, GPR:$Rm, GPR:$RLo, GPR:$RHi, pred:$p, cc_out:$s), 4, IIC_iMAC64, [], @@ -3741,15 +3915,7 @@ def UMLALv5 : ARMPseudoExpand<(outs GPR:$RdLo, GPR:$RdHi), Requires<[IsARM, NoV6]>; } -let Constraints = "@earlyclobber $RdLo,@earlyclobber $RdHi" in { -def UMAALv5 : ARMPseudoExpand<(outs GPR:$RdLo, GPR:$RdHi), - (ins GPR:$Rn, GPR:$Rm, pred:$p), - 4, IIC_iMAC64, [], - (UMAAL GPR:$RdLo, GPR:$RdHi, GPR:$Rn, GPR:$Rm, pred:$p)>, - Requires<[IsARM, NoV6]>; -} - -} // neverHasSideEffects +} // hasSideEffects // Most significant word multiply def SMMUL : AMul2I <0b0111010, 0b0001, (outs GPR:$Rd), (ins GPR:$Rn, GPR:$Rm), @@ -3813,14 +3979,12 @@ multiclass AI_smul { def WB : AMulxyI<0b0001001, 0b01, (outs GPR:$Rd), (ins GPR:$Rn, GPR:$Rm), IIC_iMUL16, !strconcat(opc, "wb"), "\t$Rd, $Rn, $Rm", - [(set GPR:$Rd, (sra (opnode GPR:$Rn, - (sext_inreg GPR:$Rm, i16)), (i32 16)))]>, + []>, Requires<[IsARM, HasV5TE]>; def WT : AMulxyI<0b0001001, 0b11, (outs GPR:$Rd), (ins GPR:$Rn, GPR:$Rm), IIC_iMUL16, !strconcat(opc, "wt"), "\t$Rd, $Rn, $Rm", - [(set GPR:$Rd, (sra (opnode GPR:$Rn, - (sra GPR:$Rm, (i32 16))), (i32 16)))]>, + []>, Requires<[IsARM, HasV5TE]>; } @@ -3862,17 +4026,13 @@ multiclass AI_smla { def WB : AMulxyIa<0b0001001, 0b00, (outs GPRnopc:$Rd), (ins GPRnopc:$Rn, GPRnopc:$Rm, GPR:$Ra), IIC_iMAC16, !strconcat(opc, "wb"), "\t$Rd, $Rn, $Rm, $Ra", - [(set GPRnopc:$Rd, - (add GPR:$Ra, (sra (opnode GPRnopc:$Rn, - (sext_inreg GPRnopc:$Rm, i16)), (i32 16))))]>, + []>, Requires<[IsARM, HasV5TE, UseMulOps]>; def WT : AMulxyIa<0b0001001, 0b10, (outs GPRnopc:$Rd), (ins GPRnopc:$Rn, GPRnopc:$Rm, GPR:$Ra), IIC_iMAC16, !strconcat(opc, "wt"), "\t$Rd, $Rn, $Rm, $Ra", - [(set GPRnopc:$Rd, - (add GPR:$Ra, (sra (opnode GPRnopc:$Rn, - (sra GPRnopc:$Rm, (i32 16))), (i32 16))))]>, + []>, Requires<[IsARM, HasV5TE, UseMulOps]>; } } @@ -3992,7 +4152,7 @@ def UDIV : ADivA1I<0b011, (outs GPR:$Rd), (ins GPR:$Rn, GPR:$Rm), IIC_iDIV, // Misc. Arithmetic Instructions. // -def CLZ : AMiscA1I<0b000010110, 0b0001, (outs GPR:$Rd), (ins GPR:$Rm), +def CLZ : AMiscA1I<0b00010110, 0b0001, (outs GPR:$Rd), (ins GPR:$Rm), IIC_iUNAr, "clz", "\t$Rd, $Rm", [(set GPR:$Rd, (ctlz GPR:$Rm))]>, Requires<[IsARM, HasV5T]>, Sched<[WriteALU]>; @@ -4015,6 +4175,11 @@ def REV16 : AMiscA1I<0b01101011, 0b1011, (outs GPR:$Rd), (ins GPR:$Rm), Requires<[IsARM, HasV6]>, Sched<[WriteALU]>; +def : ARMV6Pat<(srl (bswap (extloadi16 addrmode3:$addr)), (i32 16)), + (REV16 (LDRH addrmode3:$addr))>; +def : ARMV6Pat<(truncstorei16 (srl (bswap GPR:$Rn), (i32 16)), addrmode3:$addr), + (STRH (REV16 GPR:$Rn), addrmode3:$addr)>; + let AddedComplexity = 5 in def REVSH : AMiscA1I<0b01101111, 0b1011, (outs GPR:$Rd), (ins GPR:$Rm), IIC_iUNAr, "revsh", "\t$Rd, $Rm", @@ -4105,6 +4270,30 @@ def CRC32CH : AI_crc32<1, 0b01, "ch", int_arm_crc32ch>; def CRC32W : AI_crc32<0, 0b10, "w", int_arm_crc32w>; def CRC32CW : AI_crc32<1, 0b10, "cw", int_arm_crc32cw>; +//===----------------------------------------------------------------------===// +// ARMv8.1a Privilege Access Never extension +// +// SETPAN #imm1 + +def SETPAN : AInoP<(outs), (ins imm0_1:$imm), MiscFrm, NoItinerary, "setpan", + "\t$imm", []>, Requires<[IsARM, HasV8, HasV8_1a]> { + bits<1> imm; + + let Inst{31-28} = 0b1111; + let Inst{27-20} = 0b00010001; + let Inst{19-16} = 0b0000; + let Inst{15-10} = 0b000000; + let Inst{9} = imm; + let Inst{8} = 0b0; + let Inst{7-4} = 0b0000; + let Inst{3-0} = 0b0000; + + let Unpredictable{19-16} = 0b1111; + let Unpredictable{15-10} = 0b111111; + let Unpredictable{8} = 0b1; + let Unpredictable{3-0} = 0b1111; +} + //===----------------------------------------------------------------------===// // Comparison Instructions... // @@ -4114,8 +4303,8 @@ defm CMP : AI1_cmp_irs<0b1010, "cmp", BinOpFrag<(ARMcmp node:$LHS, node:$RHS)>>; // ARMcmpZ can re-use the above instruction definitions. -def : ARMPat<(ARMcmpZ GPR:$src, so_imm:$imm), - (CMPri GPR:$src, so_imm:$imm)>; +def : ARMPat<(ARMcmpZ GPR:$src, mod_imm:$imm), + (CMPri GPR:$src, mod_imm:$imm)>; def : ARMPat<(ARMcmpZ GPR:$src, GPR:$rhs), (CMPrr GPR:$src, GPR:$rhs)>; def : ARMPat<(ARMcmpZ GPR:$src, so_reg_imm:$rhs), @@ -4125,9 +4314,9 @@ def : ARMPat<(ARMcmpZ GPR:$src, so_reg_reg:$rhs), // CMN register-integer let isCompare = 1, Defs = [CPSR] in { -def CMNri : AI1<0b1011, (outs), (ins GPR:$Rn, so_imm:$imm), DPFrm, IIC_iCMPi, +def CMNri : AI1<0b1011, (outs), (ins GPR:$Rn, mod_imm:$imm), DPFrm, IIC_iCMPi, "cmn", "\t$Rn, $imm", - [(ARMcmn GPR:$Rn, so_imm:$imm)]>, + [(ARMcmn GPR:$Rn, mod_imm:$imm)]>, Sched<[WriteCMP, ReadALU]> { bits<4> Rn; bits<12> imm; @@ -4200,16 +4389,17 @@ def CMNzrsr : AI1<0b1011, (outs), } -def : ARMPat<(ARMcmp GPR:$src, so_imm_neg:$imm), - (CMNri GPR:$src, so_imm_neg:$imm)>; +def : ARMPat<(ARMcmp GPR:$src, mod_imm_neg:$imm), + (CMNri GPR:$src, mod_imm_neg:$imm)>; -def : ARMPat<(ARMcmpZ GPR:$src, so_imm_neg:$imm), - (CMNri GPR:$src, so_imm_neg:$imm)>; +def : ARMPat<(ARMcmpZ GPR:$src, mod_imm_neg:$imm), + (CMNri GPR:$src, mod_imm_neg:$imm)>; // Note that TST/TEQ don't set all the same flags that CMP does! defm TST : AI1_cmp_irs<0b1000, "tst", IIC_iTSTi, IIC_iTSTr, IIC_iTSTsr, - BinOpFrag<(ARMcmpZ (and_su node:$LHS, node:$RHS), 0)>, 1>; + BinOpFrag<(ARMcmpZ (and_su node:$LHS, node:$RHS), 0)>, 1, + "DecodeTSTInstruction">; defm TEQ : AI1_cmp_irs<0b1001, "teq", IIC_iTSTi, IIC_iTSTr, IIC_iTSTsr, BinOpFrag<(ARMcmpZ (xor_su node:$LHS, node:$RHS), 0)>, 1>; @@ -4231,7 +4421,7 @@ def BCCZi64 : PseudoInst<(outs), // Conditional moves -let neverHasSideEffects = 1 in { +let hasSideEffects = 0 in { let isCommutable = 1, isSelect = 1 in def MOVCCr : ARMPseudoInst<(outs GPR:$Rd), @@ -4268,9 +4458,9 @@ def MOVCCi16 let isMoveImm = 1 in def MOVCCi : ARMPseudoInst<(outs GPR:$Rd), - (ins GPR:$false, so_imm:$imm, cmovpred:$p), + (ins GPR:$false, mod_imm:$imm, cmovpred:$p), 4, IIC_iCMOVi, - [(set GPR:$Rd, (ARMcmov GPR:$false, so_imm:$imm, + [(set GPR:$Rd, (ARMcmov GPR:$false, mod_imm:$imm, cmovpred:$p))]>, RegConstraint<"$false = $Rd">, Sched<[WriteALU]>; @@ -4286,13 +4476,13 @@ def MOVCCi32imm let isMoveImm = 1 in def MVNCCi : ARMPseudoInst<(outs GPR:$Rd), - (ins GPR:$false, so_imm:$imm, cmovpred:$p), + (ins GPR:$false, mod_imm:$imm, cmovpred:$p), 4, IIC_iCMOVi, - [(set GPR:$Rd, (ARMcmov GPR:$false, so_imm_not:$imm, + [(set GPR:$Rd, (ARMcmov GPR:$false, mod_imm_not:$imm, cmovpred:$p))]>, RegConstraint<"$false = $Rd">, Sched<[WriteALU]>; -} // neverHasSideEffects +} // hasSideEffects //===----------------------------------------------------------------------===// @@ -4319,7 +4509,7 @@ def instsyncb_opt : Operand { let DecoderMethod = "DecodeInstSyncBarrierOption"; } -// memory barriers protect the atomic sequences +// Memory barriers protect the atomic sequences let hasSideEffects = 1 in { def DMB : AInoP<(outs), (ins memb_opt:$opt), MiscFrm, NoItinerary, "dmb", "\t$opt", [(int_arm_dmb (i32 imm0_15:$opt))]>, @@ -4328,7 +4518,6 @@ def DMB : AInoP<(outs), (ins memb_opt:$opt), MiscFrm, NoItinerary, let Inst{31-4} = 0xf57ff05; let Inst{3-0} = opt; } -} def DSB : AInoP<(outs), (ins memb_opt:$opt), MiscFrm, NoItinerary, "dsb", "\t$opt", [(int_arm_dsb (i32 imm0_15:$opt))]>, @@ -4340,226 +4529,19 @@ def DSB : AInoP<(outs), (ins memb_opt:$opt), MiscFrm, NoItinerary, // ISB has only full system option def ISB : AInoP<(outs), (ins instsyncb_opt:$opt), MiscFrm, NoItinerary, - "isb", "\t$opt", []>, + "isb", "\t$opt", [(int_arm_isb (i32 imm0_15:$opt))]>, Requires<[IsARM, HasDB]> { bits<4> opt; let Inst{31-4} = 0xf57ff06; let Inst{3-0} = opt; } +} let usesCustomInserter = 1, Defs = [CPSR] in { // Pseudo instruction that combines movs + predicated rsbmi // to implement integer ABS def ABS : ARMPseudoInst<(outs GPR:$dst), (ins GPR:$src), 8, NoItinerary, []>; - -// Atomic pseudo-insts which will be lowered to ldrex/strex loops. -// (64-bit pseudos use a hand-written selection code). - let mayLoad = 1, mayStore = 1 in { - def ATOMIC_LOAD_ADD_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_SUB_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_AND_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_OR_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_XOR_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_NAND_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_MIN_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_MAX_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_UMIN_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_UMAX_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_SWAP_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$new, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_CMP_SWAP_I8 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$old, GPR:$new, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_ADD_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_SUB_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_AND_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_OR_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_XOR_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_NAND_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_MIN_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_MAX_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_UMIN_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_UMAX_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_SWAP_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$new, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_CMP_SWAP_I16 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$old, GPR:$new, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_ADD_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_SUB_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_AND_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_OR_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_XOR_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_NAND_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$incr, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_MIN_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_MAX_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_UMIN_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_UMAX_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$val, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_SWAP_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$new, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_CMP_SWAP_I32 : PseudoInst< - (outs GPR:$dst), - (ins GPR:$ptr, GPR:$old, GPR:$new, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_ADD_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_SUB_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_AND_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_OR_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_XOR_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_NAND_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_MIN_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_MAX_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_UMIN_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_LOAD_UMAX_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_SWAP_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; - def ATOMIC_CMP_SWAP_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$cmp1, GPR:$cmp2, - GPR:$set1, GPR:$set2, i32imm:$ordering), - NoItinerary, []>; - } - let mayLoad = 1 in - def ATOMIC_LOAD_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, i32imm:$ordering), - NoItinerary, []>; - let mayStore = 1 in - def ATOMIC_STORE_I64 : PseudoInst< - (outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2, i32imm:$ordering), - NoItinerary, []>; } let usesCustomInserter = 1 in { @@ -4596,6 +4578,33 @@ def strex_4 : PatFrag<(ops node:$val, node:$ptr), return cast(N)->getMemoryVT() == MVT::i32; }]>; +def ldaex_1 : PatFrag<(ops node:$ptr), (int_arm_ldaex node:$ptr), [{ + return cast(N)->getMemoryVT() == MVT::i8; +}]>; + +def ldaex_2 : PatFrag<(ops node:$ptr), (int_arm_ldaex node:$ptr), [{ + return cast(N)->getMemoryVT() == MVT::i16; +}]>; + +def ldaex_4 : PatFrag<(ops node:$ptr), (int_arm_ldaex node:$ptr), [{ + return cast(N)->getMemoryVT() == MVT::i32; +}]>; + +def stlex_1 : PatFrag<(ops node:$val, node:$ptr), + (int_arm_stlex node:$val, node:$ptr), [{ + return cast(N)->getMemoryVT() == MVT::i8; +}]>; + +def stlex_2 : PatFrag<(ops node:$val, node:$ptr), + (int_arm_stlex node:$val, node:$ptr), [{ + return cast(N)->getMemoryVT() == MVT::i16; +}]>; + +def stlex_4 : PatFrag<(ops node:$val, node:$ptr), + (int_arm_stlex node:$val, node:$ptr), [{ + return cast(N)->getMemoryVT() == MVT::i32; +}]>; + let mayLoad = 1 in { def LDREXB : AIldrex<0b10, (outs GPR:$Rt), (ins addr_offset_none:$addr), NoItinerary, "ldrexb", "\t$Rt, $addr", @@ -4613,11 +4622,14 @@ def LDREXD : AIldrex<0b01, (outs GPRPairOp:$Rt),(ins addr_offset_none:$addr), } def LDAEXB : AIldaex<0b10, (outs GPR:$Rt), (ins addr_offset_none:$addr), - NoItinerary, "ldaexb", "\t$Rt, $addr", []>; + NoItinerary, "ldaexb", "\t$Rt, $addr", + [(set GPR:$Rt, (ldaex_1 addr_offset_none:$addr))]>; def LDAEXH : AIldaex<0b11, (outs GPR:$Rt), (ins addr_offset_none:$addr), - NoItinerary, "ldaexh", "\t$Rt, $addr", []>; + NoItinerary, "ldaexh", "\t$Rt, $addr", + [(set GPR:$Rt, (ldaex_2 addr_offset_none:$addr))]>; def LDAEX : AIldaex<0b00, (outs GPR:$Rt), (ins addr_offset_none:$addr), - NoItinerary, "ldaex", "\t$Rt, $addr", []>; + NoItinerary, "ldaex", "\t$Rt, $addr", + [(set GPR:$Rt, (ldaex_4 addr_offset_none:$addr))]>; let hasExtraDefRegAllocReq = 1 in def LDAEXD : AIldaex<0b01, (outs GPRPairOp:$Rt),(ins addr_offset_none:$addr), NoItinerary, "ldaexd", "\t$Rt, $addr", []> { @@ -4628,13 +4640,16 @@ def LDAEXD : AIldaex<0b01, (outs GPRPairOp:$Rt),(ins addr_offset_none:$addr), let mayStore = 1, Constraints = "@earlyclobber $Rd" in { def STREXB: AIstrex<0b10, (outs GPR:$Rd), (ins GPR:$Rt, addr_offset_none:$addr), NoItinerary, "strexb", "\t$Rd, $Rt, $addr", - [(set GPR:$Rd, (strex_1 GPR:$Rt, addr_offset_none:$addr))]>; + [(set GPR:$Rd, (strex_1 GPR:$Rt, + addr_offset_none:$addr))]>; def STREXH: AIstrex<0b11, (outs GPR:$Rd), (ins GPR:$Rt, addr_offset_none:$addr), NoItinerary, "strexh", "\t$Rd, $Rt, $addr", - [(set GPR:$Rd, (strex_2 GPR:$Rt, addr_offset_none:$addr))]>; + [(set GPR:$Rd, (strex_2 GPR:$Rt, + addr_offset_none:$addr))]>; def STREX : AIstrex<0b00, (outs GPR:$Rd), (ins GPR:$Rt, addr_offset_none:$addr), NoItinerary, "strex", "\t$Rd, $Rt, $addr", - [(set GPR:$Rd, (strex_4 GPR:$Rt, addr_offset_none:$addr))]>; + [(set GPR:$Rd, (strex_4 GPR:$Rt, + addr_offset_none:$addr))]>; let hasExtraSrcRegAllocReq = 1 in def STREXD : AIstrex<0b01, (outs GPR:$Rd), (ins GPRPairOp:$Rt, addr_offset_none:$addr), @@ -4643,13 +4658,16 @@ def STREXD : AIstrex<0b01, (outs GPR:$Rd), } def STLEXB: AIstlex<0b10, (outs GPR:$Rd), (ins GPR:$Rt, addr_offset_none:$addr), NoItinerary, "stlexb", "\t$Rd, $Rt, $addr", - []>; + [(set GPR:$Rd, + (stlex_1 GPR:$Rt, addr_offset_none:$addr))]>; def STLEXH: AIstlex<0b11, (outs GPR:$Rd), (ins GPR:$Rt, addr_offset_none:$addr), NoItinerary, "stlexh", "\t$Rd, $Rt, $addr", - []>; + [(set GPR:$Rd, + (stlex_2 GPR:$Rt, addr_offset_none:$addr))]>; def STLEX : AIstlex<0b00, (outs GPR:$Rd), (ins GPR:$Rt, addr_offset_none:$addr), NoItinerary, "stlex", "\t$Rd, $Rt, $addr", - []>; + [(set GPR:$Rd, + (stlex_4 GPR:$Rt, addr_offset_none:$addr))]>; let hasExtraSrcRegAllocReq = 1 in def STLEXD : AIstlex<0b01, (outs GPR:$Rd), (ins GPRPairOp:$Rt, addr_offset_none:$addr), @@ -4664,19 +4682,20 @@ def CLREX : AXI<(outs), (ins), MiscFrm, NoItinerary, "clrex", let Inst{31-0} = 0b11110101011111111111000000011111; } -def : ARMPat<(and (ldrex_1 addr_offset_none:$addr), 0xff), - (LDREXB addr_offset_none:$addr)>; -def : ARMPat<(and (ldrex_2 addr_offset_none:$addr), 0xffff), - (LDREXH addr_offset_none:$addr)>; def : ARMPat<(strex_1 (and GPR:$Rt, 0xff), addr_offset_none:$addr), (STREXB GPR:$Rt, addr_offset_none:$addr)>; def : ARMPat<(strex_2 (and GPR:$Rt, 0xffff), addr_offset_none:$addr), (STREXH GPR:$Rt, addr_offset_none:$addr)>; +def : ARMPat<(stlex_1 (and GPR:$Rt, 0xff), addr_offset_none:$addr), + (STLEXB GPR:$Rt, addr_offset_none:$addr)>; +def : ARMPat<(stlex_2 (and GPR:$Rt, 0xffff), addr_offset_none:$addr), + (STLEXH GPR:$Rt, addr_offset_none:$addr)>; + class acquiring_load : PatFrag<(ops node:$ptr), (base node:$ptr), [{ AtomicOrdering Ordering = cast(N)->getOrdering(); - return Ordering == Acquire || Ordering == SequentiallyConsistent; + return isAtLeastAcquire(Ordering); }]>; def atomic_load_acquire_8 : acquiring_load; @@ -4686,7 +4705,7 @@ def atomic_load_acquire_32 : acquiring_load; class releasing_store : PatFrag<(ops node:$ptr, node:$val), (base node:$ptr, node:$val), [{ AtomicOrdering Ordering = cast(N)->getOrdering(); - return Ordering == Release || Ordering == SequentiallyConsistent; + return isAtLeastRelease(Ordering); }]>; def atomic_store_release_8 : releasing_store; @@ -5003,7 +5022,7 @@ def MCR2 : MovRCopro2<"mcr2", 0 /* from ARM core register to coprocessor */, [(int_arm_mcr2 imm:$cop, imm:$opc1, GPR:$Rt, imm:$CRn, imm:$CRm, imm:$opc2)]>, Requires<[PreV8]>; -def : ARMInstAlias<"mcr2$ $cop, $opc1, $Rt, $CRn, $CRm", +def : ARMInstAlias<"mcr2 $cop, $opc1, $Rt, $CRn, $CRm", (MCR2 p_imm:$cop, imm0_7:$opc1, GPR:$Rt, c_imm:$CRn, c_imm:$CRm, 0)>; def MRC2 : MovRCopro2<"mrc2", 1 /* from coprocessor to ARM core register */, @@ -5011,7 +5030,7 @@ def MRC2 : MovRCopro2<"mrc2", 1 /* from coprocessor to ARM core register */, (ins p_imm:$cop, imm0_7:$opc1, c_imm:$CRn, c_imm:$CRm, imm0_7:$opc2), []>, Requires<[PreV8]>; -def : ARMInstAlias<"mrc2$ $cop, $opc1, $Rt, $CRn, $CRm", +def : ARMInstAlias<"mrc2 $cop, $opc1, $Rt, $CRn, $CRm", (MRC2 GPRwithAPSR:$Rt, p_imm:$cop, imm0_7:$opc1, c_imm:$CRn, c_imm:$CRm, 0)>; @@ -5107,12 +5126,31 @@ def MRSsys : ABI<0b0001, (outs GPRnopc:$Rd), (ins), NoItinerary, let Unpredictable{11-0} = 0b110100001111; } +// However, the MRS (banked register) system instruction (ARMv7VE) *does* have a +// separate encoding (distinguished by bit 5. +def MRSbanked : ABI<0b0001, (outs GPRnopc:$Rd), (ins banked_reg:$banked), + NoItinerary, "mrs", "\t$Rd, $banked", []>, + Requires<[IsARM, HasVirtualization]> { + bits<6> banked; + bits<4> Rd; + + let Inst{23} = 0; + let Inst{22} = banked{5}; // R bit + let Inst{21-20} = 0b00; + let Inst{19-16} = banked{3-0}; + let Inst{15-12} = Rd; + let Inst{11-9} = 0b001; + let Inst{8} = banked{4}; + let Inst{7-0} = 0b00000000; +} + // Move from ARM core register to Special Register // -// No need to have both system and application versions, the encodings are the -// same and the assembly parser has no way to distinguish between them. The mask -// operand contains the special register (R Bit) in bit 4 and bits 3-0 contains -// the mask with the fields to be accessed in the special register. +// No need to have both system and application versions of MSR (immediate) or +// MSR (register), the encodings are the same and the assembly parser has no way +// to distinguish between them. The mask operand contains the special register +// (R Bit) in bit 4 and bits 3-0 contains the mask with the fields to be +// accessed in the special register. def MSR : ABI<0b0001, (outs), (ins msr_mask:$mask, GPR:$Rn), NoItinerary, "msr", "\t$mask, $Rn", []> { bits<5> mask; @@ -5127,19 +5165,51 @@ def MSR : ABI<0b0001, (outs), (ins msr_mask:$mask, GPR:$Rn), NoItinerary, let Inst{3-0} = Rn; } -def MSRi : ABI<0b0011, (outs), (ins msr_mask:$mask, so_imm:$a), NoItinerary, - "msr", "\t$mask, $a", []> { +def MSRi : ABI<0b0011, (outs), (ins msr_mask:$mask, mod_imm:$imm), NoItinerary, + "msr", "\t$mask, $imm", []> { bits<5> mask; - bits<12> a; + bits<12> imm; let Inst{23} = 0; let Inst{22} = mask{4}; // R bit let Inst{21-20} = 0b10; let Inst{19-16} = mask{3-0}; let Inst{15-12} = 0b1111; - let Inst{11-0} = a; + let Inst{11-0} = imm; +} + +// However, the MSR (banked register) system instruction (ARMv7VE) *does* have a +// separate encoding (distinguished by bit 5. +def MSRbanked : ABI<0b0001, (outs), (ins banked_reg:$banked, GPRnopc:$Rn), + NoItinerary, "msr", "\t$banked, $Rn", []>, + Requires<[IsARM, HasVirtualization]> { + bits<6> banked; + bits<4> Rn; + + let Inst{23} = 0; + let Inst{22} = banked{5}; // R bit + let Inst{21-20} = 0b10; + let Inst{19-16} = banked{3-0}; + let Inst{15-12} = 0b1111; + let Inst{11-9} = 0b001; + let Inst{8} = banked{4}; + let Inst{7-4} = 0b0000; + let Inst{3-0} = Rn; } +// Dynamic stack allocation yields a _chkstk for Windows targets. These calls +// are needed to probe the stack when allocating more than +// 4k bytes in one go. Touching the stack at 4K increments is necessary to +// ensure that the guard pages used by the OS virtual memory manager are +// allocated in correct sequence. +// The main point of having separate instruction are extra unmodelled effects +// (compared to ordinary calls) like stack pointer change. + +def win__chkstk : SDNode<"ARMISD::WIN__CHKSTK", SDTNone, + [SDNPHasChain, SDNPSideEffect]>; +let usesCustomInserter = 1, Uses = [R4], Defs = [R4, SP] in + def WIN__CHKSTK : PseudoInst<(outs), (ins), NoItinerary, [(win__chkstk)]>; + //===----------------------------------------------------------------------===// // TLS Instructions // @@ -5147,9 +5217,11 @@ def MSRi : ABI<0b0011, (outs), (ins msr_mask:$mask, so_imm:$a), NoItinerary, // __aeabi_read_tp preserves the registers r1-r3. // This is a pseudo inst so that we can get the encoding right, // complete with fixup for the aeabi_read_tp function. +// TPsoft is valid for ARM mode only, in case of Thumb mode a tTPsoft pattern +// is defined in "ARMInstrThumb.td". let isCall = 1, Defs = [R0, R12, LR, CPSR], Uses = [SP] in { - def TPsoft : PseudoInst<(outs), (ins), IIC_Br, + def TPsoft : ARMPseudoInst<(outs), (ins), 4, IIC_Br, [(set R0, ARMthread_pointer)]>, Sched<[WriteBr]>; } @@ -5194,7 +5266,7 @@ let isBarrier = 1, hasSideEffects = 1, isTerminator = 1, def Int_eh_sjlj_longjmp : PseudoInst<(outs), (ins GPR:$src, GPR:$scratch), NoItinerary, [(ARMeh_sjlj_longjmp GPR:$src, GPR:$scratch)]>, - Requires<[IsARM, IsIOS]>; + Requires<[IsARM]>; } // eh.sjlj.dispatchsetup pseudo-instruction. @@ -5218,7 +5290,7 @@ let isBranch = 1, isTerminator = 1, isBarrier = 1, isIndirectBranch = 1 in // Large immediate handling. -// 32-bit immediate using two piece so_imms or movw + movt. +// 32-bit immediate using two piece mod_imms or movw + movt. // This is a single pseudo instruction, the benefit is that it can be remat'd // as a single unit instead of having to handle reg inputs. // FIXME: Remove this when we can do generalized remat. @@ -5247,6 +5319,7 @@ def LDRLIT_ga_pcrel : PseudoInst<(outs GPR:$dst), (ins i32imm:$addr), (ARMWrapperPIC tglobaladdr:$addr))]>, Requires<[IsARM, DontUseMovt]>; +let AddedComplexity = 10 in def LDRLIT_ga_pcrel_ldr : PseudoInst<(outs GPR:$dst), (ins i32imm:$addr), NoItinerary, [(set GPR:$dst, @@ -5310,11 +5383,6 @@ def : ARMV5TEPat<(mul (sra GPR:$a, (i32 16)), (SMULTB GPR:$a, GPR:$b)>; def : ARMV5TEPat<(mul (sra GPR:$a, (i32 16)), sext_16_node:$b), (SMULTB GPR:$a, GPR:$b)>; -def : ARMV5TEPat<(sra (mul GPR:$a, (sra (shl GPR:$b, (i32 16)), (i32 16))), - (i32 16)), - (SMULWB GPR:$a, GPR:$b)>; -def : ARMV5TEPat<(sra (mul GPR:$a, sext_16_node:$b), (i32 16)), - (SMULWB GPR:$a, GPR:$b)>; def : ARMV5MOPat<(add GPR:$acc, (mul (sra (shl GPR:$a, (i32 16)), (i32 16)), @@ -5337,13 +5405,6 @@ def : ARMV5MOPat<(add GPR:$acc, def : ARMV5MOPat<(add GPR:$acc, (mul (sra GPR:$a, (i32 16)), sext_16_node:$b)), (SMLATB GPR:$a, GPR:$b, GPR:$acc)>; -def : ARMV5MOPat<(add GPR:$acc, - (sra (mul GPR:$a, (sra (shl GPR:$b, (i32 16)), (i32 16))), - (i32 16))), - (SMLAWB GPR:$a, GPR:$b, GPR:$acc)>; -def : ARMV5MOPat<(add GPR:$acc, - (sra (mul GPR:$a, sext_16_node:$b), (i32 16))), - (SMLAWB GPR:$a, GPR:$b, GPR:$acc)>; // Pre-v7 uses MCR for synchronization barriers. @@ -5521,36 +5582,36 @@ def : MnemonicAlias<"uqsubaddx", "uqsax">; // USAX == USUBADDX def : MnemonicAlias<"usubaddx", "usax">; -// "mov Rd, so_imm_not" can be handled via "mvn" in assembly, just like +// "mov Rd, mod_imm_not" can be handled via "mvn" in assembly, just like // for isel. def : ARMInstAlias<"mov${s}${p} $Rd, $imm", - (MVNi rGPR:$Rd, so_imm_not:$imm, pred:$p, cc_out:$s)>; + (MVNi rGPR:$Rd, mod_imm_not:$imm, pred:$p, cc_out:$s)>; def : ARMInstAlias<"mvn${s}${p} $Rd, $imm", - (MOVi rGPR:$Rd, so_imm_not:$imm, pred:$p, cc_out:$s)>; + (MOVi rGPR:$Rd, mod_imm_not:$imm, pred:$p, cc_out:$s)>; // Same for AND <--> BIC def : ARMInstAlias<"bic${s}${p} $Rd, $Rn, $imm", - (ANDri rGPR:$Rd, rGPR:$Rn, so_imm_not:$imm, + (ANDri rGPR:$Rd, rGPR:$Rn, mod_imm_not:$imm, pred:$p, cc_out:$s)>; def : ARMInstAlias<"bic${s}${p} $Rdn, $imm", - (ANDri rGPR:$Rdn, rGPR:$Rdn, so_imm_not:$imm, + (ANDri rGPR:$Rdn, rGPR:$Rdn, mod_imm_not:$imm, pred:$p, cc_out:$s)>; def : ARMInstAlias<"and${s}${p} $Rd, $Rn, $imm", - (BICri rGPR:$Rd, rGPR:$Rn, so_imm_not:$imm, + (BICri rGPR:$Rd, rGPR:$Rn, mod_imm_not:$imm, pred:$p, cc_out:$s)>; def : ARMInstAlias<"and${s}${p} $Rdn, $imm", - (BICri rGPR:$Rdn, rGPR:$Rdn, so_imm_not:$imm, + (BICri rGPR:$Rdn, rGPR:$Rdn, mod_imm_not:$imm, pred:$p, cc_out:$s)>; -// Likewise, "add Rd, so_imm_neg" -> sub +// Likewise, "add Rd, mod_imm_neg" -> sub def : ARMInstAlias<"add${s}${p} $Rd, $Rn, $imm", - (SUBri GPR:$Rd, GPR:$Rn, so_imm_neg:$imm, pred:$p, cc_out:$s)>; + (SUBri GPR:$Rd, GPR:$Rn, mod_imm_neg:$imm, pred:$p, cc_out:$s)>; def : ARMInstAlias<"add${s}${p} $Rd, $imm", - (SUBri GPR:$Rd, GPR:$Rd, so_imm_neg:$imm, pred:$p, cc_out:$s)>; -// Same for CMP <--> CMN via so_imm_neg + (SUBri GPR:$Rd, GPR:$Rd, mod_imm_neg:$imm, pred:$p, cc_out:$s)>; +// Same for CMP <--> CMN via mod_imm_neg def : ARMInstAlias<"cmp${p} $Rd, $imm", - (CMNri rGPR:$Rd, so_imm_neg:$imm, pred:$p)>; + (CMNri rGPR:$Rd, mod_imm_neg:$imm, pred:$p)>; def : ARMInstAlias<"cmn${p} $Rd, $imm", - (CMPri rGPR:$Rd, so_imm_neg:$imm, pred:$p)>; + (CMPri rGPR:$Rd, mod_imm_neg:$imm, pred:$p)>; // The shifter forms of the MOV instruction are aliased to the ASR, LSL, // LSR, ROR, and RRX instructions. @@ -5596,9 +5657,22 @@ def : ARMInstAlias<"neg${s}${p} $Rd, $Rm", def : InstAlias<"nop${p}", (MOVr R0, R0, pred:$p, zero_reg)>, Requires<[IsARM, NoV6]>; -// UMULL/SMULL are available on all arches, but the instruction definitions -// need difference constraints pre-v6. Use these aliases for the assembly -// parsing on pre-v6. +// MUL/UMLAL/SMLAL/UMULL/SMULL are available on all arches, but +// the instruction definitions need difference constraints pre-v6. +// Use these aliases for the assembly parsing on pre-v6. +def : InstAlias<"mul${s}${p} $Rd, $Rn, $Rm", + (MUL GPRnopc:$Rd, GPRnopc:$Rn, GPRnopc:$Rm, pred:$p, cc_out:$s)>, + Requires<[IsARM, NoV6]>; +def : InstAlias<"mla${s}${p} $Rd, $Rn, $Rm, $Ra", + (MLA GPRnopc:$Rd, GPRnopc:$Rn, GPRnopc:$Rm, GPRnopc:$Ra, + pred:$p, cc_out:$s)>, + Requires<[IsARM, NoV6]>; +def : InstAlias<"smlal${s}${p} $RdLo, $RdHi, $Rn, $Rm", + (SMLAL GPR:$RdLo, GPR:$RdHi, GPR:$Rn, GPR:$Rm, pred:$p, cc_out:$s)>, + Requires<[IsARM, NoV6]>; +def : InstAlias<"umlal${s}${p} $RdLo, $RdHi, $Rn, $Rm", + (UMLAL GPR:$RdLo, GPR:$RdHi, GPR:$Rn, GPR:$Rm, pred:$p, cc_out:$s)>, + Requires<[IsARM, NoV6]>; def : InstAlias<"smull${s}${p} $RdLo, $RdHi, $Rn, $Rm", (SMULL GPR:$RdLo, GPR:$RdHi, GPR:$Rn, GPR:$Rm, pred:$p, cc_out:$s)>, Requires<[IsARM, NoV6]>; @@ -5610,3 +5684,8 @@ def : InstAlias<"umull${s}${p} $RdLo, $RdHi, $Rn, $Rm", // is discarded. def ITasm : ARMAsmPseudo<"it$mask $cc", (ins it_pred:$cc, it_mask:$mask)>, ComplexDeprecationPredicate<"IT">; + +let mayLoad = 1, mayStore =1, hasSideEffects = 1 in +def SPACE : PseudoInst<(outs GPR:$Rd), (ins i32imm:$size, GPR:$Rn), + NoItinerary, + [(set GPR:$Rd, (int_arm_space imm:$size, GPR:$Rn))]>;