X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTarget%2FARM%2FARMInstrInfo.td;h=1c2f1900bcb0526daa4312334c8e32bc5d5fe68f;hb=12af22e8cc217827cf4f118b0f5e4ebbda9925ae;hp=c2434023f8f2f420f958dc76a186dc1f965731a4;hpb=fd79485dfa4fee67467299720efac2d0c21d846c;p=oota-llvm.git diff --git a/lib/Target/ARM/ARMInstrInfo.td b/lib/Target/ARM/ARMInstrInfo.td index c2434023f8f..1c2f1900bcb 100644 --- a/lib/Target/ARM/ARMInstrInfo.td +++ b/lib/Target/ARM/ARMInstrInfo.td @@ -71,6 +71,9 @@ def SDT_ARMTCRET : SDTypeProfile<0, 1, [SDTCisPtrTy<0>]>; def SDT_ARMBFI : SDTypeProfile<1, 3, [SDTCisVT<0, i32>, SDTCisVT<1, i32>, SDTCisVT<2, i32>, SDTCisVT<3, i32>]>; +def SDT_ARMVMAXNM : SDTypeProfile<1, 2, [SDTCisFP<0>, SDTCisFP<1>, SDTCisFP<2>]>; +def SDT_ARMVMINNM : SDTypeProfile<1, 2, [SDTCisFP<0>, SDTCisFP<1>, SDTCisFP<2>]>; + def SDTBinaryArithWithFlags : SDTypeProfile<2, 2, [SDTCisSameAs<0, 2>, SDTCisSameAs<0, 3>, @@ -92,7 +95,6 @@ def ARMSmlal : SDNode<"ARMISD::SMLAL", SDT_ARM64bitmlal>; // Node definitions. def ARMWrapper : SDNode<"ARMISD::Wrapper", SDTIntUnaryOp>; -def ARMWrapperDYN : SDNode<"ARMISD::WrapperDYN", SDTIntUnaryOp>; def ARMWrapperPIC : SDNode<"ARMISD::WrapperPIC", SDTIntUnaryOp>; def ARMWrapperJT : SDNode<"ARMISD::WrapperJT", SDTIntBinOp>; @@ -118,7 +120,8 @@ def ARMcall_nolink : SDNode<"ARMISD::CALL_NOLINK", SDT_ARMcall, def ARMretflag : SDNode<"ARMISD::RET_FLAG", SDTNone, [SDNPHasChain, SDNPOptInGlue, SDNPVariadic]>; - +def ARMintretflag : SDNode<"ARMISD::INTRET_FLAG", SDT_ARMcall, + [SDNPHasChain, SDNPOptInGlue, SDNPVariadic]>; def ARMcmov : SDNode<"ARMISD::CMOV", SDT_ARMCMov, [SDNPInGlue]>; @@ -162,8 +165,6 @@ def ARMeh_sjlj_longjmp: SDNode<"ARMISD::EH_SJLJ_LONGJMP", SDT_ARMEH_SJLJ_Longjmp, [SDNPHasChain, SDNPSideEffect]>; -def ARMMemBarrier : SDNode<"ARMISD::MEMBARRIER", SDT_ARMMEMBARRIER, - [SDNPHasChain, SDNPSideEffect]>; def ARMMemBarrierMCR : SDNode<"ARMISD::MEMBARRIER_MCR", SDT_ARMMEMBARRIER, [SDNPHasChain, SDNPSideEffect]>; def ARMPreload : SDNode<"ARMISD::PRELOAD", SDT_ARMPREFETCH, @@ -174,21 +175,27 @@ def ARMrbit : SDNode<"ARMISD::RBIT", SDTIntUnaryOp>; def ARMtcret : SDNode<"ARMISD::TC_RETURN", SDT_ARMTCRET, [SDNPHasChain, SDNPOptInGlue, SDNPVariadic]>; - def ARMbfi : SDNode<"ARMISD::BFI", SDT_ARMBFI>; +def ARMvmaxnm : SDNode<"ARMISD::VMAXNM", SDT_ARMVMAXNM, []>; +def ARMvminnm : SDNode<"ARMISD::VMINNM", SDT_ARMVMINNM, []>; + //===----------------------------------------------------------------------===// // ARM Instruction Predicate Definitions. // 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()">, AssemblerPredicate<"HasV6Ops", "armv6">; def NoV6 : Predicate<"!Subtarget->hasV6Ops()">; +def HasV6M : Predicate<"Subtarget->hasV6MOps()">, + AssemblerPredicate<"HasV6MOps", + "armv6m or armv6t2">; def HasV6T2 : Predicate<"Subtarget->hasV6T2Ops()">, AssemblerPredicate<"HasV6T2Ops", "armv6t2">; def NoV6T2 : Predicate<"!Subtarget->hasV6T2Ops()">; @@ -196,6 +203,8 @@ 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 NoVFP : Predicate<"!Subtarget->hasVFP2()">; def HasVFP2 : Predicate<"Subtarget->hasVFP2()">, AssemblerPredicate<"FeatureVFP2", "VFP2">; @@ -203,16 +212,23 @@ def HasVFP3 : Predicate<"Subtarget->hasVFP3()">, AssemblerPredicate<"FeatureVFP3", "VFP3">; def HasVFP4 : Predicate<"Subtarget->hasVFP4()">, AssemblerPredicate<"FeatureVFP4", "VFP4">; -def HasV8FP : Predicate<"Subtarget->hasV8FP()">, - AssemblerPredicate<"FeatureV8FP", "V8FP">; +def HasDPVFP : Predicate<"!Subtarget->isFPOnlySP()">, + AssemblerPredicate<"!FeatureVFPOnlySP", + "double precision VFP">; +def HasFPARMv8 : Predicate<"Subtarget->hasFPARMv8()">, + AssemblerPredicate<"FeatureFPARMv8", "FPARMv8">; def HasNEON : Predicate<"Subtarget->hasNEON()">, AssemblerPredicate<"FeatureNEON", "NEON">; +def HasCrypto : Predicate<"Subtarget->hasCrypto()">, + AssemblerPredicate<"FeatureCrypto", "crypto">; +def HasCRC : Predicate<"Subtarget->hasCRC()">, + AssemblerPredicate<"FeatureCRC", "crc">; def HasFP16 : Predicate<"Subtarget->hasFP16()">, AssemblerPredicate<"FeatureFP16","half-float">; def HasDivide : Predicate<"Subtarget->hasDivide()">, - AssemblerPredicate<"FeatureHWDiv", "divide">; + AssemblerPredicate<"FeatureHWDiv", "divide in THUMB">; def HasDivideInARM : Predicate<"Subtarget->hasDivideInARMMode()">, - AssemblerPredicate<"FeatureHWDivARM">; + AssemblerPredicate<"FeatureHWDivARM", "divide in ARM">; def HasT2ExtractPack : Predicate<"Subtarget->hasT2ExtractPack()">, AssemblerPredicate<"FeatureT2XtPk", "pack/extract">; @@ -225,9 +241,13 @@ 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()">, @@ -237,22 +257,24 @@ def IsThumb2 : Predicate<"Subtarget->isThumb2()">, AssemblerPredicate<"ModeThumb,FeatureThumb2", "thumb2">; def IsMClass : Predicate<"Subtarget->isMClass()">, - AssemblerPredicate<"FeatureMClass", "armv7m">; -def IsARClass : Predicate<"!Subtarget->isMClass()">, + AssemblerPredicate<"FeatureMClass", "armv*m">; +def IsNotMClass : Predicate<"!Subtarget->isMClass()">, AssemblerPredicate<"!FeatureMClass", - "armv7a/r">; + "!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()">; def UseNaClTrap : Predicate<"Subtarget->useNaClTrap()">, AssemblerPredicate<"FeatureNaClTrap", "NaCl">; 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()">; @@ -260,7 +282,8 @@ def UseMulOps : Predicate<"Subtarget->useMulOps()">; // But only select them if more precision in FP computation is allowed. // Do not use them for Darwin platforms. def UseFusedMAC : Predicate<"(TM.Options.AllowFPOpFusion ==" - " FPOpFusion::Fast) && " + " FPOpFusion::Fast && " + " Subtarget->hasVFP4()) && " "!Subtarget->isTargetDarwin()">; def DontUseFusedMAC : Predicate<"!(TM.Options.AllowFPOpFusion ==" " FPOpFusion::Fast &&" @@ -473,7 +496,7 @@ 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); + default: llvm_unreachable(nullptr); case 0: return CurDAG->getTargetConstant(0, MVT::i32); case 8: return CurDAG->getTargetConstant(1, MVT::i32); case 16: return CurDAG->getTargetConstant(2, MVT::i32); @@ -574,7 +597,7 @@ def so_imm2part : PatLeaf<(imm), [{ /// arm_i32imm - True for +V6T2, or true only if so_imm2part is true. /// def arm_i32imm : PatLeaf<(imm), [{ - if (Subtarget->hasV6T2Ops()) + if (Subtarget->useMovt(*MF)) return true; return ARM_AM::isSOImmTwoPartVal((unsigned)N->getZExtValue()); }]>; @@ -587,17 +610,6 @@ def imm0_1 : Operand { let ParserMatchClass = Imm0_1AsmOperand; } def Imm0_3AsmOperand: ImmAsmOperand { let Name = "Imm0_3"; } def imm0_3 : Operand { let ParserMatchClass = Imm0_3AsmOperand; } -/// imm0_4 predicate - Immediate in the range [0,4]. -def Imm0_4AsmOperand : ImmAsmOperand -{ - let Name = "Imm0_4"; - let DiagnosticType = "ImmRange0_4"; -} -def imm0_4 : Operand, ImmLeaf= 0 && Imm < 5; }]> { - let ParserMatchClass = Imm0_4AsmOperand; - let DecoderMethod = "DecodeImm0_4"; -} - /// imm0_7 predicate - Immediate in the range [0,7]. def Imm0_7AsmOperand: ImmAsmOperand { let Name = "Imm0_7"; } def imm0_7 : Operand, ImmLeaf, 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; }]> { @@ -677,6 +691,15 @@ def imm0_63 : Operand, ImmLeaf, ImmLeaf= 0 && Imm < 240; }]> { + let ParserMatchClass = Imm0_239AsmOperand; +} + /// imm0_255 predicate - Immediate in the range [0,255]. def Imm0_255AsmOperand : ImmAsmOperand { let Name = "Imm0_255"; } def imm0_255 : Operand, ImmLeaf= 0 && Imm < 256; }]> { @@ -708,6 +731,11 @@ def imm0_65535_expr : Operand { let ParserMatchClass = Imm0_65535ExprAsmOperand; } +def Imm256_65535ExprAsmOperand: ImmAsmOperand { let Name = "Imm256_65535Expr"; } +def imm256_65535_expr : Operand { + let ParserMatchClass = Imm256_65535ExprAsmOperand; +} + /// imm24b - True if the 32-bit immediate is encodable in 24 bits. def Imm24bitAsmOperand: ImmAsmOperand { let Name = "Imm24bit"; } def imm24b : Operand, ImmLeaf, let EncoderMethod = "getAddrMode6OneLane32AddressOpValue"; } +// Base class for addrmode6 with specific alignment restrictions. +class AddrMode6Align : Operand, + 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, @@ -980,6 +1083,69 @@ def addrmode6dup : Operand, let ParserMatchClass = AddrMode6AsmOperand; } +// Base class for addrmode6dup with specific alignment restrictions. +class AddrMode6DupAlign : Operand, + 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, @@ -1665,53 +1831,12 @@ PseudoInst<(outs), (ins i32imm:$amt, pred:$p), NoItinerary, [(ARMcallseq_start timm:$amt)]>; } -// Atomic pseudo-insts which will be lowered to ldrexd/strexd loops. -// (These pseudos use a hand-written selection code). -let usesCustomInserter = 1, Defs = [CPSR], mayLoad = 1, mayStore = 1 in { -def ATOMOR6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2), - NoItinerary, []>; -def ATOMXOR6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2), - NoItinerary, []>; -def ATOMADD6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2), - NoItinerary, []>; -def ATOMSUB6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2), - NoItinerary, []>; -def ATOMNAND6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2), - NoItinerary, []>; -def ATOMAND6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2), - NoItinerary, []>; -def ATOMSWAP6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2), - NoItinerary, []>; -def ATOMCMPXCHG6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$cmp1, GPR:$cmp2, - GPR:$set1, GPR:$set2), - NoItinerary, []>; -def ATOMMIN6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2), - NoItinerary, []>; -def ATOMUMIN6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2), - NoItinerary, []>; -def ATOMMAX6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2), - NoItinerary, []>; -def ATOMUMAX6432 : PseudoInst<(outs GPR:$dst1, GPR:$dst2), - (ins GPR:$addr, GPR:$src1, GPR:$src2), - NoItinerary, []>; -} - -def HINT : AI<(outs), (ins imm0_4:$imm), MiscFrm, NoItinerary, - "hint", "\t$imm", []>, Requires<[IsARM, HasV6]> { - bits<3> imm; - let Inst{27-3} = 0b0011001000001111000000000; - let Inst{2-0} = imm; +def HINT : AI<(outs), (ins imm0_239:$imm), MiscFrm, NoItinerary, + "hint", "\t$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:$p)>, Requires<[IsARM, HasV6T2]>; @@ -1719,6 +1844,7 @@ def : InstAlias<"yield$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:$p)>, Requires<[IsARM, HasV6T2]>; +def : InstAlias<"sevl$p", (HINT 5, pred:$p)>, Requires<[IsARM, HasV8]>; def SEL : AI<(outs GPR:$Rd), (ins GPR:$Rn, GPR:$Rm), DPFrm, NoItinerary, "sel", "\t$Rd, $Rn, $Rm", []>, Requires<[IsARM, HasV6]> { @@ -1745,6 +1871,18 @@ def BKPT : AInoP<(outs), (ins imm0_65535:$val), MiscFrm, NoItinerary, let Inst{31-28} = 0xe; // AL let Inst{7-4} = 0b0111; } +// default immediate for breakpoint mnemonic +def : InstAlias<"bkpt", (BKPT 0)>, Requires<[IsARM]>; + +def HLT : AInoP<(outs), (ins imm0_65535:$val), MiscFrm, NoItinerary, + "hlt", "\t$val", []>, Requires<[IsARM, HasV8]> { + bits<16> val; + let Inst{3-0} = val{3-0}; + let Inst{19-8} = val{15-4}; + let Inst{27-20} = 0b00010000; + let Inst{31-28} = 0xe; // AL + let Inst{7-4} = 0b0111; +} // Change Processor State // FIXME: We should use InstAlias to handle the optional operands. @@ -1780,8 +1918,8 @@ 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:$addr, (i32 read), (i32 data))]>, Sched<[WritePreLd]> { bits<4> Rt; @@ -1820,7 +1958,7 @@ defm PLDW : APreLoad<0, 1, "pldw">, Requires<[IsARM,HasV7,HasMP]>; defm PLI : APreLoad<1, 0, "pli">, Requires<[IsARM,HasV7]>; def SETEND : AXI<(outs), (ins setend_op:$end), MiscFrm, NoItinerary, - "setend\t$end", []>, Requires<[IsARM]> { + "setend\t$end", []>, Requires<[IsARM]>, Deprecated { bits<1> end; let Inst{31-10} = 0b1111000100000001000000; let Inst{9} = end; @@ -1828,12 +1966,24 @@ 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:$imm16)]> { + bits<16> imm16; + let Inst{31-28} = 0b1110; // AL + let Inst{27-25} = 0b011; + let Inst{24-20} = 0b11111; + let Inst{19-8} = imm16{15-4}; + let Inst{7-4} = 0b1111; + let Inst{3-0} = imm16{3-0}; +} + /* * A5.4 Permanently UNDEFINED instructions. * @@ -1953,6 +2103,12 @@ let isReturn = 1, isTerminator = 1, isBarrier = 1 in { Requires<[IsARM, NoV4T]>, Sched<[WriteBr]> { let Inst{27-0} = 0b0001101000001111000000001110; } + + // Exception return: N.b. doesn't set CPSR as far as we're concerned (it sets + // the user-space one). + def SUBS_PC_LR : ARMPseudoInst<(outs), (ins i32imm:$offset, pred:$p), + 4, IIC_Br, + [(ARMintretflag imm:$offset)]>; } // Indirect branches @@ -2276,13 +2432,19 @@ def LDRSB : AI3ld<0b1101, 1, (outs GPR:$Rt), (ins addrmode3:$addr), LdMiscFrm, [(set GPR:$Rt, (sextloadi8 addrmode3:$addr))]>; let mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1 in { -// Load doubleword -def LDRD : AI3ld<0b1101, 0, (outs GPR:$Rd, GPR:$dst2), - (ins addrmode3:$addr), LdMiscFrm, - IIC_iLoad_d_r, "ldrd", "\t$Rd, $dst2, $addr", - []>, Requires<[IsARM, HasV5TE]>; + // Load doubleword + def LDRD : AI3ld<0b1101, 0, (outs GPR:$Rt, GPR:$Rt2), (ins addrmode3:$addr), + LdMiscFrm, IIC_iLoad_d_r, "ldrd", "\t$Rt, $Rt2, $addr", []>, + Requires<[IsARM, HasV5TE]>; } +def LDA : AIldracq<0b00, (outs GPR:$Rt), (ins addr_offset_none:$addr), + NoItinerary, "lda", "\t$Rt, $addr", []>; +def LDAB : AIldracq<0b10, (outs GPR:$Rt), (ins addr_offset_none:$addr), + NoItinerary, "ldab", "\t$Rt, $addr", []>; +def LDAH : AIldracq<0b11, (outs GPR:$Rt), (ins addr_offset_none:$addr), + NoItinerary, "ldah", "\t$Rt, $addr", []>; + // Indexed loads multiclass AI2_ldridx { @@ -2438,11 +2600,11 @@ def LDRT_POST_REG : AI2ldstidx<1, 0, 0, (outs GPR:$Rt, GPR:$Rn_wb), let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } -def LDRT_POST_IMM : AI2ldstidx<1, 0, 0, (outs GPR:$Rt, GPR:$Rn_wb), - (ins addr_offset_none:$addr, am2offset_imm:$offset), - IndexModePost, LdFrm, IIC_iLoad_ru, - "ldrt", "\t$Rt, $addr, $offset", - "$addr.base = $Rn_wb", []> { +def LDRT_POST_IMM + : AI2ldstidx<1, 0, 0, (outs GPR:$Rt, GPR:$Rn_wb), + (ins addr_offset_none:$addr, am2offset_imm:$offset), + IndexModePost, LdFrm, IIC_iLoad_ru, + "ldrt", "\t$Rt, $addr, $offset", "$addr.base = $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; @@ -2474,11 +2636,11 @@ def LDRBT_POST_REG : AI2ldstidx<1, 1, 0, (outs GPR:$Rt, GPR:$Rn_wb), let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } -def LDRBT_POST_IMM : AI2ldstidx<1, 1, 0, (outs GPR:$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 = $Rn_wb", []> { +def LDRBT_POST_IMM + : AI2ldstidx<1, 1, 0, (outs GPR:$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 = $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; @@ -2521,6 +2683,14 @@ defm LDRHT : AI3ldrT<0b1011, "ldrht">; defm LDRSHT : AI3ldrT<0b1111, "ldrsht">; } +def LDRT_POST + : ARMAsmPseudo<"ldrt${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 @@ -2529,12 +2699,12 @@ 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 -def STRD : AI3str<0b1111, (outs), (ins GPR:$Rt, GPR:$src2, addrmode3:$addr), - StMiscFrm, IIC_iStore_d_r, - "strd", "\t$Rt, $src2, $addr", []>, - Requires<[IsARM, HasV5TE]> { - let Inst{21} = 0; +let mayStore = 1, neverHasSideEffects = 1, hasExtraSrcRegAllocReq = 1 in { + def STRD : AI3str<0b1111, (outs), (ins GPR:$Rt, GPR:$Rt2, addrmode3:$addr), + StMiscFrm, IIC_iStore_d_r, "strd", "\t$Rt, $Rt2, $addr", []>, + Requires<[IsARM, HasV5TE]> { + let Inst{21} = 0; + } } // Indexed stores @@ -2543,7 +2713,8 @@ multiclass AI2_stridx { + opc, "\t$Rt, $addr!", + "$addr.base = $Rn_wb,@earlyclobber $Rn_wb", []> { bits<17> addr; let Inst{25} = 0; let Inst{23} = addr{12}; // U (add = ('U' == 1)) @@ -2555,7 +2726,8 @@ multiclass AI2_stridx { + opc, "\t$Rt, $addr!", + "$addr.base = $Rn_wb,@earlyclobber $Rn_wb", []> { bits<17> addr; let Inst{25} = 1; let Inst{23} = addr{12}; // U (add = ('U' == 1)) @@ -2568,7 +2740,7 @@ multiclass AI2_stridx { + "$addr.base = $Rn_wb,@earlyclobber $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; @@ -2586,7 +2758,7 @@ multiclass AI2_stridx { + "$addr.base = $Rn_wb,@earlyclobber $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; @@ -2663,7 +2835,8 @@ def STRH_preidx: ARMPseudoInst<(outs GPR:$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 $Rn_wb", []> { bits<14> addr; let Inst{23} = addr{8}; // U bit let Inst{22} = addr{13}; // 1 == imm8, 0 == Rm @@ -2676,7 +2849,8 @@ def STRH_PRE : AI3ldstidx<0b1011, 0, 1, (outs GPR:$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:$offset))]> { @@ -2743,11 +2917,11 @@ def STRBT_POST_REG : AI2ldstidx<0, 1, 0, (outs GPR:$Rn_wb), let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } -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 = $Rn_wb", []> { +def STRBT_POST_IMM + : AI2ldstidx<0, 1, 0, (outs GPR:$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 = $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; @@ -2760,6 +2934,10 @@ def STRBT_POST_IMM : AI2ldstidx<0, 1, 0, (outs GPR:$Rn_wb), let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } +def STRBT_POST + : ARMAsmPseudo<"strbt${q} $Rt, $addr", + (ins GPR:$Rt, addr_offset_none:$addr, pred:$q)>; + let mayStore = 1, neverHasSideEffects = 1 in { def STRT_POST_REG : AI2ldstidx<0, 0, 0, (outs GPR:$Rn_wb), (ins GPR:$Rt, addr_offset_none:$addr, am2offset_reg:$offset), @@ -2780,11 +2958,11 @@ def STRT_POST_REG : AI2ldstidx<0, 0, 0, (outs GPR:$Rn_wb), let DecoderMethod = "DecodeAddrMode2IdxInstruction"; } -def STRT_POST_IMM : AI2ldstidx<0, 0, 0, (outs GPR:$Rn_wb), - (ins GPR:$Rt, addr_offset_none:$addr, am2offset_imm:$offset), - IndexModePost, StFrm, IIC_iStore_ru, - "strt", "\t$Rt, $addr, $offset", - "$addr.base = $Rn_wb", []> { +def STRT_POST_IMM + : AI2ldstidx<0, 0, 0, (outs GPR:$Rn_wb), + (ins GPR:$Rt, addr_offset_none:$addr, am2offset_imm:$offset), + IndexModePost, StFrm, IIC_iStore_ru, + "strt", "\t$Rt, $addr, $offset", "$addr.base = $Rn_wb", []> { // {12} isAdd // {11-0} imm12/Rm bits<14> offset; @@ -2798,6 +2976,9 @@ def STRT_POST_IMM : AI2ldstidx<0, 0, 0, (outs GPR:$Rn_wb), } } +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 op, string opc> { defm STRHT : AI3strT<0b1011, "strht">; +def STL : AIstrrel<0b00, (outs), (ins GPR:$Rt, addr_offset_none:$addr), + NoItinerary, "stl", "\t$Rt, $addr", []>; +def STLB : AIstrrel<0b10, (outs), (ins GPR:$Rt, addr_offset_none:$addr), + NoItinerary, "stlb", "\t$Rt, $addr", []>; +def STLH : AIstrrel<0b11, (outs), (ins GPR:$Rt, addr_offset_none:$addr), + NoItinerary, "stlh", "\t$Rt, $addr", []>; //===----------------------------------------------------------------------===// // Load / store multiple Instructions. @@ -3156,8 +3343,8 @@ def SBFX : I<(outs GPRnopc:$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, $width", "", []>, Requires<[IsARM, HasV6T2]> { @@ -3239,7 +3426,8 @@ def : ARMPat<(ARMaddc GPR:$src, imm0_65535_neg:$imm), def : ARMPat<(ARMadde GPR:$src, so_imm_not:$imm, CPSR), (SBCri GPR:$src, so_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:$imm)))>, + Requires<[IsARM, HasV6T2]>; // Note: These are implemented in C++ code, because they have to generate // ADD/SUBrs instructions, which use a complex pattern that a xform function @@ -3591,21 +3779,22 @@ def MULv5: ARMPseudoExpand<(outs GPRnopc:$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, $Ra", @@ -3673,7 +3862,8 @@ def UMAAL : AMul1I <0b0000010, (outs GPR:$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:$s), 4, IIC_iMAC64, [], @@ -3688,14 +3878,6 @@ def UMLALv5 : ARMPseudoExpand<(outs GPR:$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:$p)>, - Requires<[IsARM, NoV6]>; -} - } // neverHasSideEffects // Most significant word multiply @@ -3939,7 +4121,7 @@ def UDIV : ADivA1I<0b011, (outs GPR:$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:$Rm))]>, Requires<[IsARM, HasV5T]>, Sched<[WriteALU]>; @@ -3962,6 +4144,11 @@ def REV16 : AMiscA1I<0b01101011, 0b1011, (outs GPR:$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, $Rm", @@ -4013,6 +4200,45 @@ def : ARMV6Pat<(or (and GPRnopc:$src1, 0xFFFF0000), (and (srl GPRnopc:$src2, imm1_15:$sh), 0xFFFF)), (PKHTB GPRnopc:$src1, GPRnopc:$src2, imm1_15:$sh)>; +//===----------------------------------------------------------------------===// +// CRC Instructions +// +// Polynomials: +// + CRC32{B,H,W} 0x04C11DB7 +// + CRC32C{B,H,W} 0x1EDC6F41 +// + +class AI_crc32 sz, string suffix, SDPatternOperator builtin> + : AInoP<(outs GPRnopc:$Rd), (ins GPRnopc:$Rn, GPRnopc:$Rm), MiscFrm, NoItinerary, + !strconcat("crc32", suffix), "\t$Rd, $Rn, $Rm", + [(set GPRnopc:$Rd, (builtin GPRnopc:$Rn, GPRnopc:$Rm))]>, + Requires<[IsARM, HasV8, HasCRC]> { + bits<4> Rd; + bits<4> Rn; + bits<4> Rm; + + let Inst{31-28} = 0b1110; + let Inst{27-23} = 0b00010; + let Inst{22-21} = sz; + let Inst{20} = 0; + let Inst{19-16} = Rn; + let Inst{15-12} = Rd; + let Inst{11-10} = 0b00; + let Inst{9} = C; + let Inst{8} = 0; + let Inst{7-4} = 0b0100; + let Inst{3-0} = Rm; + + let Unpredictable{11-8} = 0b1101; +} + +def CRC32B : AI_crc32<0, 0b00, "b", int_arm_crc32b>; +def CRC32CB : AI_crc32<1, 0b00, "cb", int_arm_crc32cb>; +def CRC32H : AI_crc32<0, 0b01, "h", int_arm_crc32h>; +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>; + //===----------------------------------------------------------------------===// // Comparison Instructions... // @@ -4139,56 +4365,65 @@ def BCCZi64 : PseudoInst<(outs), // Conditional moves -// FIXME: should be able to write a pattern for ARMcmov, but can't use -// a two-value operand where a dag node expects two operands. :( let neverHasSideEffects = 1 in { let isCommutable = 1, isSelect = 1 in -def MOVCCr : ARMPseudoInst<(outs GPR:$Rd), (ins GPR:$false, GPR:$Rm, pred:$p), +def MOVCCr : ARMPseudoInst<(outs GPR:$Rd), + (ins GPR:$false, GPR:$Rm, cmovpred:$p), 4, IIC_iCMOVr, - [/*(set GPR:$Rd, (ARMcmov GPR:$false, GPR:$Rm, imm:$cc, CCR:$ccr))*/]>, - RegConstraint<"$false = $Rd">, Sched<[WriteALU]>; + [(set GPR:$Rd, (ARMcmov GPR:$false, GPR:$Rm, + cmovpred:$p))]>, + RegConstraint<"$false = $Rd">, Sched<[WriteALU]>; def MOVCCsi : ARMPseudoInst<(outs GPR:$Rd), - (ins GPR:$false, so_reg_imm:$shift, pred:$p), - 4, IIC_iCMOVsr, - [/*(set GPR:$Rd, (ARMcmov GPR:$false, so_reg_imm:$shift, - imm:$cc, CCR:$ccr))*/]>, + (ins GPR:$false, so_reg_imm:$shift, cmovpred:$p), + 4, IIC_iCMOVsr, + [(set GPR:$Rd, + (ARMcmov GPR:$false, so_reg_imm:$shift, + cmovpred:$p))]>, RegConstraint<"$false = $Rd">, Sched<[WriteALU]>; def MOVCCsr : ARMPseudoInst<(outs GPR:$Rd), - (ins GPR:$false, so_reg_reg:$shift, pred:$p), + (ins GPR:$false, so_reg_reg:$shift, cmovpred:$p), 4, IIC_iCMOVsr, - [/*(set GPR:$Rd, (ARMcmov GPR:$false, so_reg_reg:$shift, - imm:$cc, CCR:$ccr))*/]>, + [(set GPR:$Rd, (ARMcmov GPR:$false, so_reg_reg:$shift, + cmovpred:$p))]>, RegConstraint<"$false = $Rd">, Sched<[WriteALU]>; let isMoveImm = 1 in -def MOVCCi16 : ARMPseudoInst<(outs GPR:$Rd), - (ins GPR:$false, imm0_65535_expr:$imm, pred:$p), - 4, IIC_iMOVi, - []>, +def MOVCCi16 + : ARMPseudoInst<(outs GPR:$Rd), + (ins GPR:$false, imm0_65535_expr:$imm, cmovpred:$p), + 4, IIC_iMOVi, + [(set GPR:$Rd, (ARMcmov GPR:$false, imm0_65535:$imm, + cmovpred:$p))]>, RegConstraint<"$false = $Rd">, Requires<[IsARM, HasV6T2]>, Sched<[WriteALU]>; let isMoveImm = 1 in def MOVCCi : ARMPseudoInst<(outs GPR:$Rd), - (ins GPR:$false, so_imm:$imm, pred:$p), + (ins GPR:$false, so_imm:$imm, cmovpred:$p), 4, IIC_iCMOVi, - [/*(set GPR:$Rd, (ARMcmov GPR:$false, so_imm:$imm, imm:$cc, CCR:$ccr))*/]>, + [(set GPR:$Rd, (ARMcmov GPR:$false, so_imm:$imm, + cmovpred:$p))]>, RegConstraint<"$false = $Rd">, Sched<[WriteALU]>; // Two instruction predicate mov immediate. let isMoveImm = 1 in -def MOVCCi32imm : ARMPseudoInst<(outs GPR:$Rd), - (ins GPR:$false, i32imm:$src, pred:$p), - 8, IIC_iCMOVix2, []>, RegConstraint<"$false = $Rd">; +def MOVCCi32imm + : ARMPseudoInst<(outs GPR:$Rd), + (ins GPR:$false, i32imm:$src, cmovpred:$p), + 8, IIC_iCMOVix2, + [(set GPR:$Rd, (ARMcmov GPR:$false, imm:$src, + cmovpred:$p))]>, + RegConstraint<"$false = $Rd">, Requires<[IsARM, HasV6T2]>; let isMoveImm = 1 in def MVNCCi : ARMPseudoInst<(outs GPR:$Rd), - (ins GPR:$false, so_imm:$imm, pred:$p), + (ins GPR:$false, so_imm:$imm, cmovpred:$p), 4, IIC_iCMOVi, - [/*(set GPR:$Rd, (ARMcmov GPR:$false, so_imm_not:$imm, imm:$cc, CCR:$ccr))*/]>, + [(set GPR:$Rd, (ARMcmov GPR:$false, so_imm_not:$imm, + cmovpred:$p))]>, RegConstraint<"$false = $Rd">, Sched<[WriteALU]>; } // neverHasSideEffects @@ -4218,19 +4453,18 @@ def instsyncb_opt : Operand { let DecoderMethod = "DecodeInstSyncBarrierOption"; } -// memory barriers protect the atomic sequences +// Memory barriers protect the atomic sequences let hasSideEffects = 1 in { def DMB : AInoP<(outs), (ins memb_opt:$opt), MiscFrm, NoItinerary, - "dmb", "\t$opt", [(ARMMemBarrier (i32 imm:$opt))]>, + "dmb", "\t$opt", [(int_arm_dmb (i32 imm0_15:$opt))]>, Requires<[IsARM, HasDB]> { bits<4> opt; let Inst{31-4} = 0xf57ff05; let Inst{3-0} = opt; } -} def DSB : AInoP<(outs), (ins memb_opt:$opt), MiscFrm, NoItinerary, - "dsb", "\t$opt", []>, + "dsb", "\t$opt", [(int_arm_dsb (i32 imm0_15:$opt))]>, Requires<[IsARM, HasDB]> { bits<4> opt; let Inst{31-4} = 0xf57ff04; @@ -4239,131 +4473,19 @@ 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:$opt))]>, Requires<[IsARM, HasDB]> { bits<4> opt; let Inst{31-4} = 0xf57ff06; let Inst{3-0} = opt; } +} + +let usesCustomInserter = 1, Defs = [CPSR] in { // Pseudo instruction that combines movs + predicated rsbmi // to implement integer ABS -let usesCustomInserter = 1, Defs = [CPSR] in -def ABS : ARMPseudoInst<(outs GPR:$dst), (ins GPR:$src), 8, NoItinerary, []>; - -let usesCustomInserter = 1 in { - let Defs = [CPSR] in { - def ATOMIC_LOAD_ADD_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_add_8 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_SUB_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_sub_8 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_AND_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_and_8 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_OR_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_or_8 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_XOR_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_xor_8 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_NAND_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_nand_8 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_MIN_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_min_8 GPR:$ptr, GPR:$val))]>; - def ATOMIC_LOAD_MAX_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_max_8 GPR:$ptr, GPR:$val))]>; - def ATOMIC_LOAD_UMIN_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_umin_8 GPR:$ptr, GPR:$val))]>; - def ATOMIC_LOAD_UMAX_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_umax_8 GPR:$ptr, GPR:$val))]>; - def ATOMIC_LOAD_ADD_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_add_16 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_SUB_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_sub_16 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_AND_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_and_16 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_OR_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_or_16 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_XOR_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_xor_16 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_NAND_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_nand_16 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_MIN_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_min_16 GPR:$ptr, GPR:$val))]>; - def ATOMIC_LOAD_MAX_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_max_16 GPR:$ptr, GPR:$val))]>; - def ATOMIC_LOAD_UMIN_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_umin_16 GPR:$ptr, GPR:$val))]>; - def ATOMIC_LOAD_UMAX_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_umax_16 GPR:$ptr, GPR:$val))]>; - def ATOMIC_LOAD_ADD_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_add_32 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_SUB_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_sub_32 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_AND_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_and_32 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_OR_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_or_32 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_XOR_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_xor_32 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_NAND_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$incr), NoItinerary, - [(set GPR:$dst, (atomic_load_nand_32 GPR:$ptr, GPR:$incr))]>; - def ATOMIC_LOAD_MIN_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_min_32 GPR:$ptr, GPR:$val))]>; - def ATOMIC_LOAD_MAX_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_max_32 GPR:$ptr, GPR:$val))]>; - def ATOMIC_LOAD_UMIN_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_umin_32 GPR:$ptr, GPR:$val))]>; - def ATOMIC_LOAD_UMAX_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$val), NoItinerary, - [(set GPR:$dst, (atomic_load_umax_32 GPR:$ptr, GPR:$val))]>; - - def ATOMIC_SWAP_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$new), NoItinerary, - [(set GPR:$dst, (atomic_swap_8 GPR:$ptr, GPR:$new))]>; - def ATOMIC_SWAP_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$new), NoItinerary, - [(set GPR:$dst, (atomic_swap_16 GPR:$ptr, GPR:$new))]>; - def ATOMIC_SWAP_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$new), NoItinerary, - [(set GPR:$dst, (atomic_swap_32 GPR:$ptr, GPR:$new))]>; - - def ATOMIC_CMP_SWAP_I8 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$old, GPR:$new), NoItinerary, - [(set GPR:$dst, (atomic_cmp_swap_8 GPR:$ptr, GPR:$old, GPR:$new))]>; - def ATOMIC_CMP_SWAP_I16 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$old, GPR:$new), NoItinerary, - [(set GPR:$dst, (atomic_cmp_swap_16 GPR:$ptr, GPR:$old, GPR:$new))]>; - def ATOMIC_CMP_SWAP_I32 : PseudoInst< - (outs GPR:$dst), (ins GPR:$ptr, GPR:$old, GPR:$new), NoItinerary, - [(set GPR:$dst, (atomic_cmp_swap_32 GPR:$ptr, GPR:$old, GPR:$new))]>; -} + def ABS : ARMPseudoInst<(outs GPR:$dst), (ins GPR:$src), 8, NoItinerary, []>; } let usesCustomInserter = 1 in { @@ -4400,10 +4522,36 @@ def strex_4 : PatFrag<(ops node:$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, $addr", + NoItinerary, "ldrexb", "\t$Rt, $addr", [(set GPR:$Rt, (ldrex_1 addr_offset_none:$addr))]>; def LDREXH : AIldrex<0b11, (outs GPR:$Rt), (ins addr_offset_none:$addr), NoItinerary, "ldrexh", "\t$Rt, $addr", @@ -4412,52 +4560,119 @@ def LDREX : AIldrex<0b00, (outs GPR:$Rt), (ins addr_offset_none:$addr), NoItinerary, "ldrex", "\t$Rt, $addr", [(set GPR:$Rt, (ldrex_4 addr_offset_none:$addr))]>; let hasExtraDefRegAllocReq = 1 in -def LDREXD: AIldrex<0b01, (outs GPRPairOp:$Rt),(ins addr_offset_none:$addr), +def LDREXD : AIldrex<0b01, (outs GPRPairOp:$Rt),(ins addr_offset_none:$addr), NoItinerary, "ldrexd", "\t$Rt, $addr", []> { let DecoderMethod = "DecodeDoubleRegLoad"; } + +def LDAEXB : AIldaex<0b10, (outs GPR:$Rt), (ins addr_offset_none:$addr), + NoItinerary, "ldaexb", "\t$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, $addr", + [(set GPR:$Rt, (ldaex_2 addr_offset_none:$addr))]>; +def LDAEX : AIldaex<0b00, (outs GPR:$Rt), (ins addr_offset_none:$addr), + NoItinerary, "ldaex", "\t$Rt, $addr", + [(set GPR:$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", []> { + let DecoderMethod = "DecodeDoubleRegLoad"; +} } 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), NoItinerary, "strexd", "\t$Rd, $Rt, $addr", []> { let DecoderMethod = "DecodeDoubleRegStore"; } +def STLEXB: AIstlex<0b10, (outs GPR:$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:$addr), + NoItinerary, "stlexd", "\t$Rd, $Rt, $addr", []> { + let DecoderMethod = "DecodeDoubleRegStore"; +} } - def CLREX : AXI<(outs), (ins), MiscFrm, NoItinerary, "clrex", [(int_arm_clrex)]>, Requires<[IsARM, HasV7]> { let Inst{31-0} = 0b11110101011111111111000000011111; } -def : ARMPat<(and (ldrex_1 addr_offset_none:$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:$ptr), [{ + AtomicOrdering Ordering = cast(N)->getOrdering(); + return isAtLeastAcquire(Ordering); +}]>; + +def atomic_load_acquire_8 : acquiring_load; +def atomic_load_acquire_16 : acquiring_load; +def atomic_load_acquire_32 : acquiring_load; + +class releasing_store + : PatFrag<(ops node:$ptr, node:$val), (base node:$ptr, node:$val), [{ + AtomicOrdering Ordering = cast(N)->getOrdering(); + return isAtLeastRelease(Ordering); +}]>; + +def atomic_store_release_8 : releasing_store; +def atomic_store_release_16 : releasing_store; +def atomic_store_release_32 : releasing_store; + +let AddedComplexity = 8 in { + def : ARMPat<(atomic_load_acquire_8 addr_offset_none:$addr), (LDAB addr_offset_none:$addr)>; + def : ARMPat<(atomic_load_acquire_16 addr_offset_none:$addr), (LDAH addr_offset_none:$addr)>; + def : ARMPat<(atomic_load_acquire_32 addr_offset_none:$addr), (LDA addr_offset_none:$addr)>; + def : ARMPat<(atomic_store_release_8 addr_offset_none:$addr, GPR:$val), (STLB GPR:$val, addr_offset_none:$addr)>; + def : ARMPat<(atomic_store_release_16 addr_offset_none:$addr, GPR:$val), (STLH GPR:$val, addr_offset_none:$addr)>; + def : ARMPat<(atomic_store_release_32 addr_offset_none:$addr, GPR:$val), (STL GPR:$val, addr_offset_none:$addr)>; +} + // SWP/SWPB are deprecated in V6/V7. let mayLoad = 1, mayStore = 1 in { def SWP : AIswp<0, (outs GPRnopc:$Rt), - (ins GPRnopc:$Rt2, addr_offset_none:$addr), "swp", []>; + (ins GPRnopc:$Rt2, addr_offset_none:$addr), "swp", []>, + Requires<[PreV8]>; def SWPB: AIswp<1, (outs GPRnopc:$Rt), - (ins GPRnopc:$Rt2, addr_offset_none:$addr), "swpb", []>; + (ins GPRnopc:$Rt2, addr_offset_none:$addr), "swpb", []>, + Requires<[PreV8]>; } //===----------------------------------------------------------------------===// @@ -4468,7 +4683,8 @@ def CDP : ABI<0b1110, (outs), (ins p_imm:$cop, imm0_15:$opc1, c_imm:$CRd, c_imm:$CRn, c_imm:$CRm, imm0_7:$opc2), NoItinerary, "cdp", "\t$cop, $opc1, $CRd, $CRn, $CRm, $opc2", [(int_arm_cdp imm:$cop, imm:$opc1, imm:$CRd, imm:$CRn, - imm:$CRm, imm:$opc2)]> { + imm:$CRm, imm:$opc2)]>, + Requires<[PreV8]> { bits<4> opc1; bits<4> CRn; bits<4> CRd; @@ -4489,7 +4705,8 @@ def CDP2 : ABXI<0b1110, (outs), (ins p_imm:$cop, imm0_15:$opc1, c_imm:$CRd, c_imm:$CRn, c_imm:$CRm, imm0_7:$opc2), NoItinerary, "cdp2\t$cop, $opc1, $CRd, $CRn, $CRm, $opc2", [(int_arm_cdp2 imm:$cop, imm:$opc1, imm:$CRd, imm:$CRn, - imm:$CRm, imm:$opc2)]> { + imm:$CRm, imm:$opc2)]>, + Requires<[PreV8]> { let Inst{31-28} = 0b1111; bits<4> opc1; bits<4> CRn; @@ -4667,10 +4884,10 @@ defm LDC : LdStCop <1, 0, "ldc">; defm LDCL : LdStCop <1, 1, "ldcl">; defm STC : LdStCop <0, 0, "stc">; defm STCL : LdStCop <0, 1, "stcl">; -defm LDC2 : LdSt2Cop<1, 0, "ldc2">; -defm LDC2L : LdSt2Cop<1, 1, "ldc2l">; -defm STC2 : LdSt2Cop<0, 0, "stc2">; -defm STC2L : LdSt2Cop<0, 1, "stc2l">; +defm LDC2 : LdSt2Cop<1, 0, "ldc2">, Requires<[PreV8]>; +defm LDC2L : LdSt2Cop<1, 1, "ldc2l">, Requires<[PreV8]>; +defm STC2 : LdSt2Cop<0, 0, "stc2">, Requires<[PreV8]>; +defm STC2L : LdSt2Cop<0, 1, "stc2l">, Requires<[PreV8]>; //===----------------------------------------------------------------------===// // Move between coprocessor and ARM core register. @@ -4703,7 +4920,8 @@ def MCR : MovRCopro<"mcr", 0 /* from ARM core register to coprocessor */, (ins p_imm:$cop, imm0_7:$opc1, GPR:$Rt, c_imm:$CRn, c_imm:$CRm, imm0_7:$opc2), [(int_arm_mcr imm:$cop, imm:$opc1, GPR:$Rt, imm:$CRn, - imm:$CRm, imm:$opc2)]>; + imm:$CRm, imm:$opc2)]>, + ComplexDeprecationPredicate<"MCR">; def : ARMInstAlias<"mcr${p} $cop, $opc1, $Rt, $CRn, $CRm", (MCR p_imm:$cop, imm0_7:$opc1, GPR:$Rt, c_imm:$CRn, c_imm:$CRm, 0, pred:$p)>; @@ -4746,15 +4964,17 @@ def MCR2 : MovRCopro2<"mcr2", 0 /* from ARM core register to coprocessor */, (ins p_imm:$cop, imm0_7:$opc1, GPR:$Rt, c_imm:$CRn, c_imm:$CRm, imm0_7:$opc2), [(int_arm_mcr2 imm:$cop, imm:$opc1, GPR:$Rt, imm:$CRn, - imm:$CRm, imm:$opc2)]>; -def : ARMInstAlias<"mcr2$ $cop, $opc1, $Rt, $CRn, $CRm", + imm:$CRm, imm:$opc2)]>, + Requires<[PreV8]>; +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 */, (outs GPRwithAPSR:$Rt), (ins p_imm:$cop, imm0_7:$opc1, c_imm:$CRn, c_imm:$CRm, - imm0_7:$opc2), []>; -def : ARMInstAlias<"mrc2$ $cop, $opc1, $Rt, $CRn, $CRm", + imm0_7:$opc2), []>, + Requires<[PreV8]>; +def : ARMInstAlias<"mrc2 $cop, $opc1, $Rt, $CRn, $CRm", (MRC2 GPRwithAPSR:$Rt, p_imm:$cop, imm0_7:$opc1, c_imm:$CRn, c_imm:$CRm, 0)>; @@ -4790,7 +5010,8 @@ def MRRC : MovRRCopro<"mrrc", 1 /* from coprocessor to ARM core register */>; class MovRRCopro2 pattern = []> : ABXI<0b1100, (outs), (ins p_imm:$cop, imm0_15:$opc1, GPRnopc:$Rt, GPRnopc:$Rt2, c_imm:$CRm), NoItinerary, - !strconcat(opc, "\t$cop, $opc1, $Rt, $Rt2, $CRm"), pattern> { + !strconcat(opc, "\t$cop, $opc1, $Rt, $Rt2, $CRm"), pattern>, + Requires<[PreV8]> { let Inst{31-28} = 0b1111; let Inst{23-21} = 0b010; let Inst{20} = direction; @@ -4849,12 +5070,31 @@ def MRSsys : ABI<0b0001, (outs GPRnopc:$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, $banked", []>, + Requires<[IsARM, HasVirtualization]> { + bits<6> banked; + bits<4> Rd; + + let Inst{23} = 0; + let Inst{22} = banked{5}; // R bit + let Inst{21-20} = 0b10; + let Inst{19-16} = banked{3-0}; + let Inst{15-12} = Rd; + let Inst{11-9} = 0b001; + let Inst{8} = banked{4}; + let Inst{7-0} = 0b00000000; +} + // Move from ARM core register to Special Register // -// No need to have both system and application versions, the encodings are the -// same and the assembly parser has no way to distinguish between them. The mask -// operand contains the special register (R Bit) in bit 4 and bits 3-0 contains -// the mask with the fields to be accessed in the special register. +// No need to have both system and application versions of MSR (immediate) or +// MSR (register), the encodings are the same and the assembly parser has no way +// to distinguish between them. The mask operand contains the special register +// (R Bit) in bit 4 and bits 3-0 contains the mask with the fields to be +// accessed in the special register. def MSR : ABI<0b0001, (outs), (ins msr_mask:$mask, GPR:$Rn), NoItinerary, "msr", "\t$mask, $Rn", []> { bits<5> mask; @@ -4882,6 +5122,38 @@ def MSRi : ABI<0b0011, (outs), (ins msr_mask:$mask, so_imm:$a), NoItinerary, let Inst{11-0} = a; } +// However, the MSR (banked register) system instruction (ARMv7VE) *does* have a +// separate encoding (distinguished by bit 5. +def MSRbanked : ABI<0b0001, (outs), (ins banked_reg:$banked, GPRnopc:$Rn), + NoItinerary, "msr", "\t$banked, $Rn", []>, + Requires<[IsARM, HasVirtualization]> { + bits<6> banked; + bits<4> Rn; + + let Inst{23} = 0; + let Inst{22} = banked{5}; // R bit + let Inst{21-20} = 0b10; + let Inst{19-16} = banked{3-0}; + let Inst{15-12} = 0b1111; + let Inst{11-9} = 0b001; + let Inst{8} = banked{4}; + let Inst{7-4} = 0b0000; + let Inst{3-0} = Rn; +} + +// Dynamic stack allocation yields a _chkstk for Windows targets. These calls +// are needed to probe the stack when allocating more than +// 4k bytes in one go. Touching the stack at 4K increments is necessary to +// ensure that the guard pages used by the OS virtual memory manager are +// allocated in correct sequence. +// The main point of having separate instruction are extra unmodelled effects +// (compared to ordinary calls) like stack pointer change. + +def win__chkstk : SDNode<"ARMISD::WIN__CHKSTK", SDTNone, + [SDNPHasChain, SDNPSideEffect]>; +let usesCustomInserter = 1, Uses = [R4], Defs = [R4, SP] in + def WIN__CHKSTK : PseudoInst<(outs), (ins), NoItinerary, [(win__chkstk)]>; + //===----------------------------------------------------------------------===// // TLS Instructions // @@ -4889,9 +5161,11 @@ def MSRi : ABI<0b0011, (outs), (ins msr_mask:$mask, so_imm:$a), NoItinerary, // __aeabi_read_tp preserves the registers r1-r3. // This is a pseudo inst so that we can get the encoding right, // complete with fixup for the aeabi_read_tp function. +// TPsoft is valid for ARM mode only, in case of Thumb mode a tTPsoft pattern +// is defined in "ARMInstrThumb.td". let isCall = 1, Defs = [R0, R12, LR, CPSR], Uses = [SP] in { - def TPsoft : PseudoInst<(outs), (ins), IIC_Br, + def TPsoft : ARMPseudoInst<(outs), (ins), 4, IIC_Br, [(set R0, ARMthread_pointer)]>, Sched<[WriteBr]>; } @@ -4969,6 +5243,10 @@ def MOVi32imm : PseudoInst<(outs GPR:$dst), (ins i32imm:$src), IIC_iMOVix2, [(set GPR:$dst, (arm_i32imm:$src))]>, Requires<[IsARM]>; +def LDRLIT_ga_abs : PseudoInst<(outs GPR:$dst), (ins i32imm:$src), IIC_iLoad_i, + [(set GPR:$dst, (ARMWrapper tglobaladdr:$src))]>, + Requires<[IsARM, DontUseMovt]>; + // Pseudo instruction that combines movw + movt + add pc (if PIC). // It also makes it possible to rematerialize the instructions. // FIXME: Remove this when we can do generalized remat and when machine licm @@ -4979,10 +5257,17 @@ def MOV_ga_pcrel : PseudoInst<(outs GPR:$dst), (ins i32imm:$addr), [(set GPR:$dst, (ARMWrapperPIC tglobaladdr:$addr))]>, Requires<[IsARM, UseMovt]>; -def MOV_ga_dyn : PseudoInst<(outs GPR:$dst), (ins i32imm:$addr), - IIC_iMOVix2, - [(set GPR:$dst, (ARMWrapperDYN tglobaladdr:$addr))]>, - Requires<[IsARM, UseMovt]>; +def LDRLIT_ga_pcrel : PseudoInst<(outs GPR:$dst), (ins i32imm:$addr), + IIC_iLoadiALU, + [(set GPR:$dst, + (ARMWrapperPIC tglobaladdr:$addr))]>, + Requires<[IsARM, DontUseMovt]>; + +def LDRLIT_ga_pcrel_ldr : PseudoInst<(outs GPR:$dst), (ins i32imm:$addr), + NoItinerary, + [(set GPR:$dst, + (load (ARMWrapperPIC tglobaladdr:$addr)))]>, + Requires<[IsARM, DontUseMovt]>; let AddedComplexity = 10 in def MOV_ga_pcrel_ldr : PseudoInst<(outs GPR:$dst), (ins i32imm:$addr), @@ -4992,8 +5277,6 @@ def MOV_ga_pcrel_ldr : PseudoInst<(outs GPR:$dst), (ins i32imm:$addr), } // isReMaterializable // ConstantPool, GlobalAddress, and JumpTable -def : ARMPat<(ARMWrapper tglobaladdr :$dst), (LEApcrel tglobaladdr :$dst)>, - Requires<[IsARM, DontUseMovt]>; def : ARMPat<(ARMWrapper tconstpool :$dst), (LEApcrel tconstpool :$dst)>; def : ARMPat<(ARMWrapper tglobaladdr :$dst), (MOVi32imm tglobaladdr :$dst)>, Requires<[IsARM, UseMovt]>; @@ -5329,9 +5612,22 @@ def : ARMInstAlias<"neg${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:$s)>, Requires<[IsARM, NoV6]>; @@ -5341,4 +5637,5 @@ def : InstAlias<"umull${s}${p} $RdLo, $RdHi, $Rn, $Rm", // 'it' blocks in ARM mode just validate the predicates. The IT itself // is discarded. -def ITasm : ARMAsmPseudo<"it$mask $cc", (ins it_pred:$cc, it_mask:$mask)>; +def ITasm : ARMAsmPseudo<"it$mask $cc", (ins it_pred:$cc, it_mask:$mask)>, + ComplexDeprecationPredicate<"IT">;