diff options
Diffstat (limited to 'lib/Target/PowerPC')
38 files changed, 4733 insertions, 2301 deletions
diff --git a/lib/Target/PowerPC/InstPrinter/PPCInstPrinter.cpp b/lib/Target/PowerPC/InstPrinter/PPCInstPrinter.cpp index 3d58306..bacc108 100644 --- a/lib/Target/PowerPC/InstPrinter/PPCInstPrinter.cpp +++ b/lib/Target/PowerPC/InstPrinter/PPCInstPrinter.cpp @@ -13,7 +13,7 @@ #define DEBUG_TYPE "asm-printer" #include "PPCInstPrinter.h" -#include "MCTargetDesc/PPCBaseInfo.h" +#include "MCTargetDesc/PPCMCTargetDesc.h" #include "MCTargetDesc/PPCPredicates.h" #include "llvm/MC/MCExpr.h" #include "llvm/MC/MCInst.h" @@ -87,35 +87,9 @@ void PPCInstPrinter::printPredicateOperand(const MCInst *MI, unsigned OpNo, raw_ostream &O, const char *Modifier) { unsigned Code = MI->getOperand(OpNo).getImm(); - if (!Modifier) { - unsigned CCReg = MI->getOperand(OpNo+1).getReg(); - unsigned RegNo; - switch (CCReg) { - default: llvm_unreachable("Unknown CR register"); - case PPC::CR0: RegNo = 0; break; - case PPC::CR1: RegNo = 1; break; - case PPC::CR2: RegNo = 2; break; - case PPC::CR3: RegNo = 3; break; - case PPC::CR4: RegNo = 4; break; - case PPC::CR5: RegNo = 5; break; - case PPC::CR6: RegNo = 6; break; - case PPC::CR7: RegNo = 7; break; - } - - // Print the CR bit number. The Code is ((BI << 5) | BO) for a - // BCC, but we must have the positive form here (BO == 12) - unsigned BI = Code >> 5; - assert((Code & 0xF) == 12 && - "BO in predicate bit must have the positive form"); - - unsigned Value = 4*RegNo + BI; - O << Value; - return; - } if (StringRef(Modifier) == "cc") { switch ((PPC::Predicate)Code) { - case PPC::PRED_ALWAYS: return; // Don't print anything for always. case PPC::PRED_LT: O << "lt"; return; case PPC::PRED_LE: O << "le"; return; case PPC::PRED_EQ: O << "eq"; return; @@ -129,8 +103,6 @@ void PPCInstPrinter::printPredicateOperand(const MCInst *MI, unsigned OpNo, assert(StringRef(Modifier) == "reg" && "Need to specify 'cc' or 'reg' as predicate op modifier!"); - // Don't print the register for 'always'. - if (Code == PPC::PRED_ALWAYS) return; printOperand(MI, OpNo+1, O); } diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCAsmBackend.cpp b/lib/Target/PowerPC/MCTargetDesc/PPCAsmBackend.cpp index f24edf6..ec26574 100644 --- a/lib/Target/PowerPC/MCTargetDesc/PPCAsmBackend.cpp +++ b/lib/Target/PowerPC/MCTargetDesc/PPCAsmBackend.cpp @@ -30,13 +30,9 @@ static unsigned adjustFixupValue(unsigned Kind, uint64_t Value) { case FK_Data_2: case FK_Data_4: case FK_Data_8: - case PPC::fixup_ppc_toc: case PPC::fixup_ppc_tlsreg: case PPC::fixup_ppc_nofixup: return Value; - case PPC::fixup_ppc_lo14: - case PPC::fixup_ppc_toc16_ds: - return (Value & 0xffff) << 2; case PPC::fixup_ppc_brcond14: return Value & 0xfffc; case PPC::fixup_ppc_br24: @@ -48,8 +44,9 @@ static unsigned adjustFixupValue(unsigned Kind, uint64_t Value) { case PPC::fixup_ppc_ha16: return ((Value >> 16) + ((Value & 0x8000) ? 1 : 0)) & 0xffff; case PPC::fixup_ppc_lo16: - case PPC::fixup_ppc_toc16: return Value & 0xffff; + case PPC::fixup_ppc_lo16_ds: + return Value & 0xfffc; } } @@ -82,10 +79,7 @@ public: { "fixup_ppc_brcond14", 16, 14, MCFixupKindInfo::FKF_IsPCRel }, { "fixup_ppc_lo16", 16, 16, 0 }, { "fixup_ppc_ha16", 16, 16, 0 }, - { "fixup_ppc_lo14", 16, 14, 0 }, - { "fixup_ppc_toc", 0, 64, 0 }, - { "fixup_ppc_toc16", 16, 16, 0 }, - { "fixup_ppc_toc16_ds", 16, 14, 0 }, + { "fixup_ppc_lo16_ds", 16, 14, 0 }, { "fixup_ppc_tlsreg", 0, 0, 0 }, { "fixup_ppc_nofixup", 0, 0, 0 } }; diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCBaseInfo.h b/lib/Target/PowerPC/MCTargetDesc/PPCBaseInfo.h deleted file mode 100644 index 9c975c0..0000000 --- a/lib/Target/PowerPC/MCTargetDesc/PPCBaseInfo.h +++ /dev/null @@ -1,70 +0,0 @@ -//===-- PPCBaseInfo.h - Top level definitions for PPC -----------*- C++ -*-===// -// -// The LLVM Compiler Infrastructure -// -// This file is distributed under the University of Illinois Open Source -// License. See LICENSE.TXT for details. -// -//===----------------------------------------------------------------------===// -// -// This file contains small standalone helper functions and enum definitions for -// the PPC target useful for the compiler back-end and the MC libraries. -// As such, it deliberately does not include references to LLVM core -// code gen types, passes, etc.. -// -//===----------------------------------------------------------------------===// - -#ifndef PPCBASEINFO_H -#define PPCBASEINFO_H - -#include "PPCMCTargetDesc.h" -#include "llvm/Support/ErrorHandling.h" - -namespace llvm { - -/// getPPCRegisterNumbering - Given the enum value for some register, e.g. -/// PPC::F14, return the number that it corresponds to (e.g. 14). -inline static unsigned getPPCRegisterNumbering(unsigned RegEnum) { - using namespace PPC; - switch (RegEnum) { - case 0: return 0; - case R0 : case X0 : case F0 : case V0 : case CR0: case CR0LT: return 0; - case R1 : case X1 : case F1 : case V1 : case CR1: case CR0GT: return 1; - case R2 : case X2 : case F2 : case V2 : case CR2: case CR0EQ: return 2; - case R3 : case X3 : case F3 : case V3 : case CR3: case CR0UN: return 3; - case R4 : case X4 : case F4 : case V4 : case CR4: case CR1LT: return 4; - case R5 : case X5 : case F5 : case V5 : case CR5: case CR1GT: return 5; - case R6 : case X6 : case F6 : case V6 : case CR6: case CR1EQ: return 6; - case R7 : case X7 : case F7 : case V7 : case CR7: case CR1UN: return 7; - case R8 : case X8 : case F8 : case V8 : case CR2LT: return 8; - case R9 : case X9 : case F9 : case V9 : case CR2GT: return 9; - case R10: case X10: case F10: case V10: case CR2EQ: return 10; - case R11: case X11: case F11: case V11: case CR2UN: return 11; - case R12: case X12: case F12: case V12: case CR3LT: return 12; - case R13: case X13: case F13: case V13: case CR3GT: return 13; - case R14: case X14: case F14: case V14: case CR3EQ: return 14; - case R15: case X15: case F15: case V15: case CR3UN: return 15; - case R16: case X16: case F16: case V16: case CR4LT: return 16; - case R17: case X17: case F17: case V17: case CR4GT: return 17; - case R18: case X18: case F18: case V18: case CR4EQ: return 18; - case R19: case X19: case F19: case V19: case CR4UN: return 19; - case R20: case X20: case F20: case V20: case CR5LT: return 20; - case R21: case X21: case F21: case V21: case CR5GT: return 21; - case R22: case X22: case F22: case V22: case CR5EQ: return 22; - case R23: case X23: case F23: case V23: case CR5UN: return 23; - case R24: case X24: case F24: case V24: case CR6LT: return 24; - case R25: case X25: case F25: case V25: case CR6GT: return 25; - case R26: case X26: case F26: case V26: case CR6EQ: return 26; - case R27: case X27: case F27: case V27: case CR6UN: return 27; - case R28: case X28: case F28: case V28: case CR7LT: return 28; - case R29: case X29: case F29: case V29: case CR7GT: return 29; - case R30: case X30: case F30: case V30: case CR7EQ: return 30; - case R31: case X31: case F31: case V31: case CR7UN: return 31; - default: - llvm_unreachable("Unhandled reg in PPCRegisterInfo::getRegisterNumbering!"); - } -} - -} // end namespace llvm; - -#endif diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp b/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp index 61868d4..81a86dc 100644 --- a/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp +++ b/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp @@ -77,6 +77,9 @@ unsigned PPCELFObjectWriter::getRelocTypeInner(const MCValue &Target, case PPC::fixup_ppc_br24: Type = ELF::R_PPC_REL24; break; + case PPC::fixup_ppc_brcond14: + Type = ELF::R_PPC_REL14; + break; case FK_Data_4: case FK_PCRel_4: Type = ELF::R_PPC_REL32; @@ -133,6 +136,9 @@ unsigned PPCELFObjectWriter::getRelocTypeInner(const MCValue &Target, case MCSymbolRefExpr::VK_None: Type = ELF::R_PPC_ADDR16_LO; break; + case MCSymbolRefExpr::VK_PPC_TOC_ENTRY: + Type = ELF::R_PPC64_TOC16; + break; case MCSymbolRefExpr::VK_PPC_TOC16_LO: Type = ELF::R_PPC64_TOC16_LO; break; @@ -144,35 +150,12 @@ unsigned PPCELFObjectWriter::getRelocTypeInner(const MCValue &Target, break; } break; - case PPC::fixup_ppc_lo14: - Type = ELF::R_PPC_ADDR14; - break; - case PPC::fixup_ppc_toc: - Type = ELF::R_PPC64_TOC; - break; - case PPC::fixup_ppc_toc16: + case PPC::fixup_ppc_lo16_ds: switch (Modifier) { default: llvm_unreachable("Unsupported Modifier"); - case MCSymbolRefExpr::VK_PPC_TPREL16_LO: - Type = ELF::R_PPC64_TPREL16_LO; - break; - case MCSymbolRefExpr::VK_PPC_DTPREL16_LO: - Type = ELF::R_PPC64_DTPREL16_LO; - break; case MCSymbolRefExpr::VK_None: - Type = ELF::R_PPC64_TOC16; - break; - case MCSymbolRefExpr::VK_PPC_TOC16_LO: - Type = ELF::R_PPC64_TOC16_LO; + Type = ELF::R_PPC64_ADDR16_DS; break; - case MCSymbolRefExpr::VK_PPC_GOT_TLSLD16_LO: - Type = ELF::R_PPC64_GOT_TLSLD16_LO; - break; - } - break; - case PPC::fixup_ppc_toc16_ds: - switch (Modifier) { - default: llvm_unreachable("Unsupported Modifier"); case MCSymbolRefExpr::VK_PPC_TOC_ENTRY: Type = ELF::R_PPC64_TOC16_DS; break; @@ -253,8 +236,7 @@ adjustFixupOffset(const MCFixup &Fixup, uint64_t &RelocOffset) { switch ((unsigned)Fixup.getKind()) { case PPC::fixup_ppc_ha16: case PPC::fixup_ppc_lo16: - case PPC::fixup_ppc_toc16: - case PPC::fixup_ppc_toc16_ds: + case PPC::fixup_ppc_lo16_ds: RelocOffset += 2; break; default: diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h b/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h index 709daa4..86c44f5 100644 --- a/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h +++ b/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h @@ -33,19 +33,9 @@ enum Fixups { /// like 'lis'. fixup_ppc_ha16, - /// fixup_ppc_lo14 - A 14-bit fixup corresponding to lo16(_foo) for instrs - /// like 'std'. - fixup_ppc_lo14, - - /// fixup_ppc_toc - Insert value of TOC base (.TOC.). - fixup_ppc_toc, - - /// fixup_ppc_toc16 - A 16-bit signed fixup relative to the TOC base. - fixup_ppc_toc16, - - /// fixup_ppc_toc16_ds - A 14-bit signed fixup relative to the TOC base with - /// implied 2 zero bits - fixup_ppc_toc16_ds, + /// fixup_ppc_lo16_ds - A 14-bit fixup corresponding to lo16(_foo) with + /// implied 2 zero bits for instrs like 'std'. + fixup_ppc_lo16_ds, /// fixup_ppc_tlsreg - Insert thread-pointer register number. fixup_ppc_tlsreg, diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCMCCodeEmitter.cpp b/lib/Target/PowerPC/MCTargetDesc/PPCMCCodeEmitter.cpp index d048426..2223cd6 100644 --- a/lib/Target/PowerPC/MCTargetDesc/PPCMCCodeEmitter.cpp +++ b/lib/Target/PowerPC/MCTargetDesc/PPCMCCodeEmitter.cpp @@ -13,10 +13,10 @@ #define DEBUG_TYPE "mccodeemitter" #include "MCTargetDesc/PPCMCTargetDesc.h" -#include "MCTargetDesc/PPCBaseInfo.h" #include "MCTargetDesc/PPCFixupKinds.h" #include "llvm/ADT/Statistic.h" #include "llvm/MC/MCCodeEmitter.h" +#include "llvm/MC/MCContext.h" #include "llvm/MC/MCExpr.h" #include "llvm/MC/MCInst.h" #include "llvm/MC/MCInstrInfo.h" @@ -33,24 +33,17 @@ class PPCMCCodeEmitter : public MCCodeEmitter { void operator=(const PPCMCCodeEmitter &) LLVM_DELETED_FUNCTION; const MCSubtargetInfo &STI; + const MCContext &CTX; Triple TT; public: PPCMCCodeEmitter(const MCInstrInfo &mcii, const MCSubtargetInfo &sti, MCContext &ctx) - : STI(sti), TT(STI.getTargetTriple()) { + : STI(sti), CTX(ctx), TT(STI.getTargetTriple()) { } ~PPCMCCodeEmitter() {} - bool is64BitMode() const { - return (STI.getFeatureBits() & PPC::Feature64Bit) != 0; - } - - bool isSVR4ABI() const { - return TT.isMacOSX() == 0; - } - unsigned getDirectBrEncoding(const MCInst &MI, unsigned OpNo, SmallVectorImpl<MCFixup> &Fixups) const; unsigned getCondBrEncoding(const MCInst &MI, unsigned OpNo, @@ -81,12 +74,11 @@ public: SmallVectorImpl<MCFixup> &Fixups) const { uint64_t Bits = getBinaryCodeForInstr(MI, Fixups); - // BL8_NOP_ELF, BLA8_NOP_ELF, etc., all have a size of 8 because of the - // following 'nop'. + // BL8_NOP etc. all have a size of 8 because of the following 'nop'. unsigned Size = 4; // FIXME: Have Desc.getSize() return the correct value! unsigned Opcode = MI.getOpcode(); - if (Opcode == PPC::BL8_NOP_ELF || Opcode == PPC::BLA8_NOP_ELF || - Opcode == PPC::BL8_NOP_ELF_TLSGD || Opcode == PPC::BL8_NOP_ELF_TLSLD) + if (Opcode == PPC::BL8_NOP || Opcode == PPC::BLA8_NOP || + Opcode == PPC::BL8_NOP_TLSGD || Opcode == PPC::BL8_NOP_TLSLD) Size = 8; // Output the constant in big endian byte order. @@ -121,11 +113,11 @@ getDirectBrEncoding(const MCInst &MI, unsigned OpNo, (MCFixupKind)PPC::fixup_ppc_br24)); // For special TLS calls, add another fixup for the symbol. Apparently - // BL8_NOP_ELF, BL8_NOP_ELF_TLSGD, and BL8_NOP_ELF_TLSLD are sufficiently + // BL8_NOP, BL8_NOP_TLSGD, and BL8_NOP_TLSLD are sufficiently // similar that TblGen will not generate a separate case for the latter // two, so this is the only way to get the extra fixup generated. unsigned Opcode = MI.getOpcode(); - if (Opcode == PPC::BL8_NOP_ELF_TLSGD || Opcode == PPC::BL8_NOP_ELF_TLSLD) { + if (Opcode == PPC::BL8_NOP_TLSGD || Opcode == PPC::BL8_NOP_TLSLD) { const MCOperand &MO2 = MI.getOperand(OpNo+1); Fixups.push_back(MCFixup::Create(0, MO2.getExpr(), (MCFixupKind)PPC::fixup_ppc_nofixup)); @@ -178,12 +170,8 @@ unsigned PPCMCCodeEmitter::getMemRIEncoding(const MCInst &MI, unsigned OpNo, return (getMachineOpValue(MI, MO, Fixups) & 0xFFFF) | RegBits; // Add a fixup for the displacement field. - if (isSVR4ABI() && is64BitMode()) - Fixups.push_back(MCFixup::Create(0, MO.getExpr(), - (MCFixupKind)PPC::fixup_ppc_toc16)); - else - Fixups.push_back(MCFixup::Create(0, MO.getExpr(), - (MCFixupKind)PPC::fixup_ppc_lo16)); + Fixups.push_back(MCFixup::Create(0, MO.getExpr(), + (MCFixupKind)PPC::fixup_ppc_lo16)); return RegBits; } @@ -199,13 +187,9 @@ unsigned PPCMCCodeEmitter::getMemRIXEncoding(const MCInst &MI, unsigned OpNo, if (MO.isImm()) return (getMachineOpValue(MI, MO, Fixups) & 0x3FFF) | RegBits; - // Add a fixup for the branch target. - if (isSVR4ABI() && is64BitMode()) - Fixups.push_back(MCFixup::Create(0, MO.getExpr(), - (MCFixupKind)PPC::fixup_ppc_toc16_ds)); - else - Fixups.push_back(MCFixup::Create(0, MO.getExpr(), - (MCFixupKind)PPC::fixup_ppc_lo14)); + // Add a fixup for the displacement field. + Fixups.push_back(MCFixup::Create(0, MO.getExpr(), + (MCFixupKind)PPC::fixup_ppc_lo16_ds)); return RegBits; } @@ -220,7 +204,7 @@ unsigned PPCMCCodeEmitter::getTLSRegEncoding(const MCInst &MI, unsigned OpNo, // Return the thread-pointer register's encoding. Fixups.push_back(MCFixup::Create(0, MO.getExpr(), (MCFixupKind)PPC::fixup_ppc_tlsreg)); - return getPPCRegisterNumbering(PPC::X13); + return CTX.getRegisterInfo().getEncodingValue(PPC::X13); } unsigned PPCMCCodeEmitter:: @@ -231,7 +215,7 @@ get_crbitm_encoding(const MCInst &MI, unsigned OpNo, MI.getOpcode() == PPC::MFOCRF || MI.getOpcode() == PPC::MTCRF8) && (MO.getReg() >= PPC::CR0 && MO.getReg() <= PPC::CR7)); - return 0x80 >> getPPCRegisterNumbering(MO.getReg()); + return 0x80 >> CTX.getRegisterInfo().getEncodingValue(MO.getReg()); } @@ -243,7 +227,7 @@ getMachineOpValue(const MCInst &MI, const MCOperand &MO, // The GPR operand should come through here though. assert((MI.getOpcode() != PPC::MTCRF && MI.getOpcode() != PPC::MFOCRF) || MO.getReg() < PPC::CR0 || MO.getReg() > PPC::CR7); - return getPPCRegisterNumbering(MO.getReg()); + return CTX.getRegisterInfo().getEncodingValue(MO.getReg()); } assert(MO.isImm() && diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.cpp b/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.cpp index 12bb0a1..853e505 100644 --- a/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.cpp +++ b/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.cpp @@ -18,7 +18,6 @@ using namespace llvm; PPC::Predicate PPC::InvertPredicate(PPC::Predicate Opcode) { switch (Opcode) { - default: llvm_unreachable("Unknown PPC branch opcode!"); case PPC::PRED_EQ: return PPC::PRED_NE; case PPC::PRED_NE: return PPC::PRED_EQ; case PPC::PRED_LT: return PPC::PRED_GE; @@ -28,4 +27,20 @@ PPC::Predicate PPC::InvertPredicate(PPC::Predicate Opcode) { case PPC::PRED_NU: return PPC::PRED_UN; case PPC::PRED_UN: return PPC::PRED_NU; } + llvm_unreachable("Unknown PPC branch opcode!"); } + +PPC::Predicate PPC::getSwappedPredicate(PPC::Predicate Opcode) { + switch (Opcode) { + case PPC::PRED_EQ: return PPC::PRED_EQ; + case PPC::PRED_NE: return PPC::PRED_NE; + case PPC::PRED_LT: return PPC::PRED_GT; + case PPC::PRED_GE: return PPC::PRED_LE; + case PPC::PRED_GT: return PPC::PRED_LT; + case PPC::PRED_LE: return PPC::PRED_GE; + case PPC::PRED_NU: return PPC::PRED_NU; + case PPC::PRED_UN: return PPC::PRED_UN; + } + llvm_unreachable("Unknown PPC branch opcode!"); +} + diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.h b/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.h index b0680fb..444758c 100644 --- a/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.h +++ b/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.h @@ -25,7 +25,6 @@ namespace llvm { namespace PPC { /// Predicate - These are "(BI << 5) | BO" for various predicates. enum Predicate { - PRED_ALWAYS = (0 << 5) | 20, PRED_LT = (0 << 5) | 12, PRED_LE = (1 << 5) | 4, PRED_EQ = (2 << 5) | 12, @@ -38,6 +37,10 @@ namespace PPC { /// Invert the specified predicate. != -> ==, < -> >=. Predicate InvertPredicate(Predicate Opcode); + + /// Assume the condition register is set by MI(a,b), return the predicate if + /// we modify the instructions such that condition register is set by MI(b,a). + Predicate getSwappedPredicate(Predicate Opcode); } } diff --git a/lib/Target/PowerPC/PPC.h b/lib/Target/PowerPC/PPC.h index f71979f..b4be51a 100644 --- a/lib/Target/PowerPC/PPC.h +++ b/lib/Target/PowerPC/PPC.h @@ -15,7 +15,6 @@ #ifndef LLVM_TARGET_POWERPC_H #define LLVM_TARGET_POWERPC_H -#include "MCTargetDesc/PPCBaseInfo.h" #include "MCTargetDesc/PPCMCTargetDesc.h" #include <string> @@ -32,6 +31,7 @@ namespace llvm { class MCInst; FunctionPass *createPPCCTRLoops(); + FunctionPass *createPPCEarlyReturnPass(); FunctionPass *createPPCBranchSelectionPass(); FunctionPass *createPPCISelDag(PPCTargetMachine &TM); FunctionPass *createPPCJITCodeEmitterPass(PPCTargetMachine &TM, @@ -41,7 +41,7 @@ namespace llvm { /// \brief Creates an PPC-specific Target Transformation Info pass. ImmutablePass *createPPCTargetTransformInfoPass(const PPCTargetMachine *TM); - + namespace PPCII { /// Target Operand Flag enum. diff --git a/lib/Target/PowerPC/PPC.td b/lib/Target/PowerPC/PPC.td index 9929136..649ffc1 100644 --- a/lib/Target/PowerPC/PPC.td +++ b/lib/Target/PowerPC/PPC.td @@ -57,10 +57,30 @@ def FeatureMFOCRF : SubtargetFeature<"mfocrf","HasMFOCRF", "true", "Enable the MFOCRF instruction">; def FeatureFSqrt : SubtargetFeature<"fsqrt","HasFSQRT", "true", "Enable the fsqrt instruction">; +def FeatureFRE : SubtargetFeature<"fre", "HasFRE", "true", + "Enable the fre instruction">; +def FeatureFRES : SubtargetFeature<"fres", "HasFRES", "true", + "Enable the fres instruction">; +def FeatureFRSQRTE : SubtargetFeature<"frsqrte", "HasFRSQRTE", "true", + "Enable the frsqrte instruction">; +def FeatureFRSQRTES : SubtargetFeature<"frsqrtes", "HasFRSQRTES", "true", + "Enable the frsqrtes instruction">; +def FeatureRecipPrec : SubtargetFeature<"recipprec", "HasRecipPrec", "true", + "Assume higher precision reciprocal estimates">; def FeatureSTFIWX : SubtargetFeature<"stfiwx","HasSTFIWX", "true", "Enable the stfiwx instruction">; +def FeatureLFIWAX : SubtargetFeature<"lfiwax","HasLFIWAX", "true", + "Enable the lfiwax instruction">; +def FeatureFPRND : SubtargetFeature<"fprnd", "HasFPRND", "true", + "Enable the fri[mnpz] instructions">; +def FeatureFPCVT : SubtargetFeature<"fpcvt", "HasFPCVT", "true", + "Enable fc[ft]* (unsigned and single-precision) and lfiwzx instructions">; def FeatureISEL : SubtargetFeature<"isel","HasISEL", "true", "Enable the isel instruction">; +def FeaturePOPCNTD : SubtargetFeature<"popcntd","HasPOPCNTD", "true", + "Enable the popcnt[dw] instructions">; +def FeatureLDBRX : SubtargetFeature<"ldbrx","HasLDBRX", "true", + "Enable the ldbrx instruction">; def FeatureBookE : SubtargetFeature<"booke", "IsBookE", "true", "Enable Book E instructions">; def FeatureQPX : SubtargetFeature<"qpx","HasQPX", "true", @@ -71,19 +91,47 @@ def FeatureQPX : SubtargetFeature<"qpx","HasQPX", "true", // // CMPB p6, p6x, p7 cmpb // DFP p6, p6x, p7 decimal floating-point instructions -// FLT_CVT p7 fcfids, fcfidu, fcfidus, fcfiduz, fctiwuz -// FPRND p5x, p6, p6x, p7 frim, frin, frip, friz -// FRE p5 through p7 fre (vs. fres, available since p3) -// FRSQRTES p5 through p7 frsqrtes (vs. frsqrte, available since p3) -// LDBRX p7 load with byte reversal -// LFIWAX p6, p6x, p7 lfiwax -// LFIWZX p7 lfiwzx // POPCNTB p5 through p7 popcntb and related instructions -// POPCNTD p7 popcntd and related instructions -// RECIP_PREC p6, p6x, p7 higher precision reciprocal estimates // VSX p7 vector-scalar instruction set //===----------------------------------------------------------------------===// +// Classes used for relation maps. +//===----------------------------------------------------------------------===// +// RecFormRel - Filter class used to relate non-record-form instructions with +// their record-form variants. +class RecFormRel; + +//===----------------------------------------------------------------------===// +// Relation Map Definitions. +//===----------------------------------------------------------------------===// + +def getRecordFormOpcode : InstrMapping { + let FilterClass = "RecFormRel"; + // Instructions with the same BaseName and Interpretation64Bit values + // form a row. + let RowFields = ["BaseName", "Interpretation64Bit"]; + // Instructions with the same RC value form a column. + let ColFields = ["RC"]; + // The key column are the non-record-form instructions. + let KeyCol = ["0"]; + // Value columns RC=1 + let ValueCols = [["1"]]; +} + +def getNonRecordFormOpcode : InstrMapping { + let FilterClass = "RecFormRel"; + // Instructions with the same BaseName and Interpretation64Bit values + // form a row. + let RowFields = ["BaseName", "Interpretation64Bit"]; + // Instructions with the same RC value form a column. + let ColFields = ["RC"]; + // The key column are the record-form instructions. + let KeyCol = ["1"]; + // Value columns are RC=0 + let ValueCols = [["0"]]; +} + +//===----------------------------------------------------------------------===// // Register File Description //===----------------------------------------------------------------------===// @@ -97,30 +145,46 @@ include "PPCInstrInfo.td" def : Processor<"generic", G3Itineraries, [Directive32]>; def : Processor<"440", PPC440Itineraries, [Directive440, FeatureISEL, + FeatureFRES, FeatureFRSQRTE, FeatureBookE]>; def : Processor<"450", PPC440Itineraries, [Directive440, FeatureISEL, + FeatureFRES, FeatureFRSQRTE, FeatureBookE]>; def : Processor<"601", G3Itineraries, [Directive601]>; def : Processor<"602", G3Itineraries, [Directive602]>; -def : Processor<"603", G3Itineraries, [Directive603]>; -def : Processor<"603e", G3Itineraries, [Directive603]>; -def : Processor<"603ev", G3Itineraries, [Directive603]>; -def : Processor<"604", G3Itineraries, [Directive604]>; -def : Processor<"604e", G3Itineraries, [Directive604]>; -def : Processor<"620", G3Itineraries, [Directive620]>; -def : Processor<"750", G4Itineraries, [Directive750]>; -def : Processor<"g3", G3Itineraries, [Directive750]>; -def : Processor<"7400", G4Itineraries, [Directive7400, FeatureAltivec]>; -def : Processor<"g4", G4Itineraries, [Directive7400, FeatureAltivec]>; -def : Processor<"7450", G4PlusItineraries, [Directive7400, FeatureAltivec]>; -def : Processor<"g4+", G4PlusItineraries, [Directive7400, FeatureAltivec]>; -def : Processor<"970", G5Itineraries, +def : Processor<"603", G3Itineraries, [Directive603, + FeatureFRES, FeatureFRSQRTE]>; +def : Processor<"603e", G3Itineraries, [Directive603, + FeatureFRES, FeatureFRSQRTE]>; +def : Processor<"603ev", G3Itineraries, [Directive603, + FeatureFRES, FeatureFRSQRTE]>; +def : Processor<"604", G3Itineraries, [Directive604, + FeatureFRES, FeatureFRSQRTE]>; +def : Processor<"604e", G3Itineraries, [Directive604, + FeatureFRES, FeatureFRSQRTE]>; +def : Processor<"620", G3Itineraries, [Directive620, + FeatureFRES, FeatureFRSQRTE]>; +def : Processor<"750", G4Itineraries, [Directive750, + FeatureFRES, FeatureFRSQRTE]>; +def : Processor<"g3", G3Itineraries, [Directive750, + FeatureFRES, FeatureFRSQRTE]>; +def : Processor<"7400", G4Itineraries, [Directive7400, FeatureAltivec, + FeatureFRES, FeatureFRSQRTE]>; +def : Processor<"g4", G4Itineraries, [Directive7400, FeatureAltivec, + FeatureFRES, FeatureFRSQRTE]>; +def : Processor<"7450", G4PlusItineraries, [Directive7400, FeatureAltivec, + FeatureFRES, FeatureFRSQRTE]>; +def : Processor<"g4+", G4PlusItineraries, [Directive7400, FeatureAltivec, + FeatureFRES, FeatureFRSQRTE]>; +def : ProcessorModel<"970", G5Model, [Directive970, FeatureAltivec, - FeatureMFOCRF, FeatureFSqrt, FeatureSTFIWX, + FeatureMFOCRF, FeatureFSqrt, + FeatureFRES, FeatureFRSQRTE, FeatureSTFIWX, Feature64Bit /*, Feature64BitRegs */]>; -def : Processor<"g5", G5Itineraries, +def : ProcessorModel<"g5", G5Model, [Directive970, FeatureAltivec, FeatureMFOCRF, FeatureFSqrt, FeatureSTFIWX, + FeatureFRES, FeatureFRSQRTE, Feature64Bit /*, Feature64BitRegs */]>; def : ProcessorModel<"e500mc", PPCE500mcModel, [DirectiveE500mc, FeatureMFOCRF, @@ -128,46 +192,67 @@ def : ProcessorModel<"e500mc", PPCE500mcModel, def : ProcessorModel<"e5500", PPCE5500Model, [DirectiveE5500, FeatureMFOCRF, Feature64Bit, FeatureSTFIWX, FeatureBookE, FeatureISEL]>; -def : Processor<"a2", PPCA2Itineraries, [DirectiveA2, FeatureBookE, - FeatureMFOCRF, FeatureFSqrt, - FeatureSTFIWX, FeatureISEL, - Feature64Bit - /*, Feature64BitRegs */]>; -def : Processor<"a2q", PPCA2Itineraries, [DirectiveA2, FeatureBookE, - FeatureMFOCRF, FeatureFSqrt, - FeatureSTFIWX, FeatureISEL, - Feature64Bit /*, Feature64BitRegs */, - FeatureQPX]>; -def : Processor<"pwr3", G5Itineraries, - [DirectivePwr3, FeatureAltivec, FeatureMFOCRF, +def : ProcessorModel<"a2", PPCA2Model, + [DirectiveA2, FeatureBookE, FeatureMFOCRF, + FeatureFSqrt, FeatureFRE, FeatureFRES, + FeatureFRSQRTE, FeatureFRSQRTES, FeatureRecipPrec, + FeatureSTFIWX, FeatureLFIWAX, + FeatureFPRND, FeatureFPCVT, FeatureISEL, + FeaturePOPCNTD, FeatureLDBRX, Feature64Bit + /*, Feature64BitRegs */]>; +def : ProcessorModel<"a2q", PPCA2Model, + [DirectiveA2, FeatureBookE, FeatureMFOCRF, + FeatureFSqrt, FeatureFRE, FeatureFRES, + FeatureFRSQRTE, FeatureFRSQRTES, FeatureRecipPrec, + FeatureSTFIWX, FeatureLFIWAX, + FeatureFPRND, FeatureFPCVT, FeatureISEL, + FeaturePOPCNTD, FeatureLDBRX, Feature64Bit + /*, Feature64BitRegs */, FeatureQPX]>; +def : ProcessorModel<"pwr3", G5Model, + [DirectivePwr3, FeatureAltivec, + FeatureFRES, FeatureFRSQRTE, FeatureMFOCRF, FeatureSTFIWX, Feature64Bit]>; -def : Processor<"pwr4", G5Itineraries, +def : ProcessorModel<"pwr4", G5Model, [DirectivePwr4, FeatureAltivec, FeatureMFOCRF, - FeatureFSqrt, FeatureSTFIWX, Feature64Bit]>; -def : Processor<"pwr5", G5Itineraries, + FeatureFSqrt, FeatureFRES, FeatureFRSQRTE, + FeatureSTFIWX, Feature64Bit]>; +def : ProcessorModel<"pwr5", G5Model, [DirectivePwr5, FeatureAltivec, FeatureMFOCRF, - FeatureFSqrt, FeatureSTFIWX, Feature64Bit]>; -def : Processor<"pwr5x", G5Itineraries, + FeatureFSqrt, FeatureFRE, FeatureFRES, + FeatureFRSQRTE, FeatureFRSQRTES, + FeatureSTFIWX, Feature64Bit]>; +def : ProcessorModel<"pwr5x", G5Model, [DirectivePwr5x, FeatureAltivec, FeatureMFOCRF, - FeatureFSqrt, FeatureSTFIWX, Feature64Bit]>; -def : Processor<"pwr6", G5Itineraries, + FeatureFSqrt, FeatureFRE, FeatureFRES, + FeatureFRSQRTE, FeatureFRSQRTES, + FeatureSTFIWX, FeatureFPRND, Feature64Bit]>; +def : ProcessorModel<"pwr6", G5Model, [DirectivePwr6, FeatureAltivec, - FeatureMFOCRF, FeatureFSqrt, FeatureSTFIWX, - Feature64Bit /*, Feature64BitRegs */]>; -def : Processor<"pwr6x", G5Itineraries, + FeatureMFOCRF, FeatureFSqrt, FeatureFRE, + FeatureFRES, FeatureFRSQRTE, FeatureFRSQRTES, + FeatureRecipPrec, FeatureSTFIWX, FeatureLFIWAX, + FeatureFPRND, Feature64Bit /*, Feature64BitRegs */]>; +def : ProcessorModel<"pwr6x", G5Model, [DirectivePwr5x, FeatureAltivec, FeatureMFOCRF, - FeatureFSqrt, FeatureSTFIWX, Feature64Bit]>; -def : Processor<"pwr7", G5Itineraries, + FeatureFSqrt, FeatureFRE, FeatureFRES, + FeatureFRSQRTE, FeatureFRSQRTES, FeatureRecipPrec, + FeatureSTFIWX, FeatureLFIWAX, + FeatureFPRND, Feature64Bit]>; +def : ProcessorModel<"pwr7", G5Model, [DirectivePwr7, FeatureAltivec, - FeatureMFOCRF, FeatureFSqrt, FeatureSTFIWX, - FeatureISEL, Feature64Bit /*, Feature64BitRegs */]>; + FeatureMFOCRF, FeatureFSqrt, FeatureFRE, + FeatureFRES, FeatureFRSQRTE, FeatureFRSQRTES, + FeatureRecipPrec, FeatureSTFIWX, FeatureLFIWAX, + FeatureFPRND, FeatureFPCVT, FeatureISEL, + FeaturePOPCNTD, FeatureLDBRX, + Feature64Bit /*, Feature64BitRegs */]>; def : Processor<"ppc", G3Itineraries, [Directive32]>; -def : Processor<"ppc64", G5Itineraries, +def : ProcessorModel<"ppc64", G5Model, [Directive64, FeatureAltivec, - FeatureMFOCRF, FeatureFSqrt, FeatureSTFIWX, + FeatureMFOCRF, FeatureFSqrt, FeatureFRES, + FeatureFRSQRTE, FeatureSTFIWX, Feature64Bit /*, Feature64BitRegs */]>; - //===----------------------------------------------------------------------===// // Calling Conventions //===----------------------------------------------------------------------===// diff --git a/lib/Target/PowerPC/PPCAsmPrinter.cpp b/lib/Target/PowerPC/PPCAsmPrinter.cpp index eae9b7b..3c7cc4e 100644 --- a/lib/Target/PowerPC/PPCAsmPrinter.cpp +++ b/lib/Target/PowerPC/PPCAsmPrinter.cpp @@ -370,7 +370,7 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { MCSymbol *PICBase = MF->getPICBaseSymbol(); // Emit the 'bl'. - OutStreamer.EmitInstruction(MCInstBuilder(PPC::BL_Darwin) // Darwin vs SVR4 doesn't matter here. + OutStreamer.EmitInstruction(MCInstBuilder(PPC::BL) // FIXME: We would like an efficient form for this, so we don't have to do // a lot of extra uniquing. .addExpr(MCSymbolRefExpr::Create(PICBase, OutContext))); @@ -458,11 +458,10 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { // Transform %Xd = LDtocL <ga:@sym>, %Xs LowerPPCMachineInstrToMCInst(MI, TmpInst, *this, Subtarget.isDarwin()); - // Change the opcode to LDrs, which is a form of LD with the offset - // specified by a SymbolLo. If the global address is external, has + // Change the opcode to LD. If the global address is external, has // common linkage, or is a jump table address, then reference the // associated TOC entry. Otherwise reference the symbol directly. - TmpInst.setOpcode(PPC::LDrs); + TmpInst.setOpcode(PPC::LD); const MachineOperand &MO = MI->getOperand(1); assert((MO.isGlobal() || MO.isJTI() || MO.isCPI()) && "Invalid operand for LDtocL!"); @@ -496,10 +495,10 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { // Transform %Xd = ADDItocL %Xs, <ga:@sym> LowerPPCMachineInstrToMCInst(MI, TmpInst, *this, Subtarget.isDarwin()); - // Change the opcode to ADDI8L. If the global address is external, then + // Change the opcode to ADDI8. If the global address is external, then // generate a TOC entry and reference that. Otherwise reference the // symbol directly. - TmpInst.setOpcode(PPC::ADDI8L); + TmpInst.setOpcode(PPC::ADDI8); const MachineOperand &MO = MI->getOperand(2); assert((MO.isGlobal() || MO.isCPI()) && "Invalid operand for ADDItocL"); MCSymbol *MOSymbol = 0; @@ -548,9 +547,8 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { // Transform %Xd = LDgotTprelL <ga:@sym>, %Xs LowerPPCMachineInstrToMCInst(MI, TmpInst, *this, Subtarget.isDarwin()); - // Change the opcode to LDrs, which is a form of LD with the offset - // specified by a SymbolLo. - TmpInst.setOpcode(PPC::LDrs); + // Change the opcode to LD. + TmpInst.setOpcode(PPC::LD); const MachineOperand &MO = MI->getOperand(1); const GlobalValue *GValue = MO.getGlobal(); MCSymbol *MOSymbol = Mang->getSymbol(GValue); @@ -579,7 +577,7 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { } case PPC::ADDItlsgdL: { // Transform: %Xd = ADDItlsgdL %Xs, <ga:@sym> - // Into: %Xd = ADDI8L %Xs, sym@got@tlsgd@l + // Into: %Xd = ADDI8 %Xs, sym@got@tlsgd@l assert(Subtarget.isPPC64() && "Not supported for 32-bit PowerPC"); const MachineOperand &MO = MI->getOperand(2); const GlobalValue *GValue = MO.getGlobal(); @@ -587,7 +585,7 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { const MCExpr *SymGotTlsGD = MCSymbolRefExpr::Create(MOSymbol, MCSymbolRefExpr::VK_PPC_GOT_TLSGD16_LO, OutContext); - OutStreamer.EmitInstruction(MCInstBuilder(PPC::ADDI8L) + OutStreamer.EmitInstruction(MCInstBuilder(PPC::ADDI8) .addReg(MI->getOperand(0).getReg()) .addReg(MI->getOperand(1).getReg()) .addExpr(SymGotTlsGD)); @@ -595,7 +593,7 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { } case PPC::GETtlsADDR: { // Transform: %X3 = GETtlsADDR %X3, <ga:@sym> - // Into: BL8_NOP_ELF_TLSGD __tls_get_addr(sym@tlsgd) + // Into: BL8_NOP_TLSGD __tls_get_addr(sym@tlsgd) assert(Subtarget.isPPC64() && "Not supported for 32-bit PowerPC"); StringRef Name = "__tls_get_addr"; @@ -608,7 +606,7 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { const MCExpr *SymVar = MCSymbolRefExpr::Create(MOSymbol, MCSymbolRefExpr::VK_PPC_TLSGD, OutContext); - OutStreamer.EmitInstruction(MCInstBuilder(PPC::BL8_NOP_ELF_TLSGD) + OutStreamer.EmitInstruction(MCInstBuilder(PPC::BL8_NOP_TLSGD) .addExpr(TlsRef) .addExpr(SymVar)); return; @@ -631,7 +629,7 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { } case PPC::ADDItlsldL: { // Transform: %Xd = ADDItlsldL %Xs, <ga:@sym> - // Into: %Xd = ADDI8L %Xs, sym@got@tlsld@l + // Into: %Xd = ADDI8 %Xs, sym@got@tlsld@l assert(Subtarget.isPPC64() && "Not supported for 32-bit PowerPC"); const MachineOperand &MO = MI->getOperand(2); const GlobalValue *GValue = MO.getGlobal(); @@ -639,7 +637,7 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { const MCExpr *SymGotTlsLD = MCSymbolRefExpr::Create(MOSymbol, MCSymbolRefExpr::VK_PPC_GOT_TLSLD16_LO, OutContext); - OutStreamer.EmitInstruction(MCInstBuilder(PPC::ADDI8L) + OutStreamer.EmitInstruction(MCInstBuilder(PPC::ADDI8) .addReg(MI->getOperand(0).getReg()) .addReg(MI->getOperand(1).getReg()) .addExpr(SymGotTlsLD)); @@ -647,7 +645,7 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { } case PPC::GETtlsldADDR: { // Transform: %X3 = GETtlsldADDR %X3, <ga:@sym> - // Into: BL8_NOP_ELF_TLSLD __tls_get_addr(sym@tlsld) + // Into: BL8_NOP_TLSLD __tls_get_addr(sym@tlsld) assert(Subtarget.isPPC64() && "Not supported for 32-bit PowerPC"); StringRef Name = "__tls_get_addr"; @@ -660,7 +658,7 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { const MCExpr *SymVar = MCSymbolRefExpr::Create(MOSymbol, MCSymbolRefExpr::VK_PPC_TLSLD, OutContext); - OutStreamer.EmitInstruction(MCInstBuilder(PPC::BL8_NOP_ELF_TLSLD) + OutStreamer.EmitInstruction(MCInstBuilder(PPC::BL8_NOP_TLSLD) .addExpr(TlsRef) .addExpr(SymVar)); return; @@ -683,7 +681,7 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { } case PPC::ADDIdtprelL: { // Transform: %Xd = ADDIdtprelL %Xs, <ga:@sym> - // Into: %Xd = ADDI8L %Xs, sym@dtprel@l + // Into: %Xd = ADDI8 %Xs, sym@dtprel@l assert(Subtarget.isPPC64() && "Not supported for 32-bit PowerPC"); const MachineOperand &MO = MI->getOperand(2); const GlobalValue *GValue = MO.getGlobal(); @@ -691,7 +689,7 @@ void PPCAsmPrinter::EmitInstruction(const MachineInstr *MI) { const MCExpr *SymDtprel = MCSymbolRefExpr::Create(MOSymbol, MCSymbolRefExpr::VK_PPC_DTPREL16_LO, OutContext); - OutStreamer.EmitInstruction(MCInstBuilder(PPC::ADDI8L) + OutStreamer.EmitInstruction(MCInstBuilder(PPC::ADDI8) .addReg(MI->getOperand(0).getReg()) .addReg(MI->getOperand(1).getReg()) .addExpr(SymDtprel)); @@ -723,7 +721,7 @@ void PPCLinuxAsmPrinter::EmitFunctionEntryLabel() { return AsmPrinter::EmitFunctionEntryLabel(); // Emit an official procedure descriptor. - const MCSection *Current = OutStreamer.getCurrentSection(); + MCSectionSubPair Current = OutStreamer.getCurrentSection(); const MCSectionELF *Section = OutStreamer.getContext().getELFSection(".opd", ELF::SHT_PROGBITS, ELF::SHF_WRITE | ELF::SHF_ALLOC, SectionKind::getReadOnly()); @@ -743,7 +741,7 @@ void PPCLinuxAsmPrinter::EmitFunctionEntryLabel() { 8/*size*/); // Emit a null environment pointer. OutStreamer.EmitIntValue(0, 8 /* size */); - OutStreamer.SwitchSection(Current); + OutStreamer.SwitchSection(Current.first, Current.second); MCSymbol *RealFnSym = OutContext.GetOrCreateSymbol( ".L." + Twine(CurrentFnSym->getName())); @@ -911,18 +909,19 @@ EmitFunctionStubs(const MachineModuleInfoMachO::SymbolListTy &Stubs) { OutStreamer.EmitLabel(Stub); OutStreamer.EmitSymbolAttribute(RawSym, MCSA_IndirectSymbol); + const MCExpr *Anon = MCSymbolRefExpr::Create(AnonSymbol, OutContext); + // mflr r0 OutStreamer.EmitInstruction(MCInstBuilder(PPC::MFLR).addReg(PPC::R0)); - // FIXME: MCize this. - OutStreamer.EmitRawText("\tbcl 20, 31, " + Twine(AnonSymbol->getName())); + // bcl 20, 31, AnonSymbol + OutStreamer.EmitInstruction(MCInstBuilder(PPC::BCLalways).addExpr(Anon)); OutStreamer.EmitLabel(AnonSymbol); // mflr r11 OutStreamer.EmitInstruction(MCInstBuilder(PPC::MFLR).addReg(PPC::R11)); // addis r11, r11, ha16(LazyPtr - AnonSymbol) const MCExpr *Sub = MCBinaryExpr::CreateSub(MCSymbolRefExpr::Create(LazyPtr, OutContext), - MCSymbolRefExpr::Create(AnonSymbol, OutContext), - OutContext); + Anon, OutContext); OutStreamer.EmitInstruction(MCInstBuilder(PPC::ADDIS) .addReg(PPC::R11) .addReg(PPC::R11) diff --git a/lib/Target/PowerPC/PPCCTRLoops.cpp b/lib/Target/PowerPC/PPCCTRLoops.cpp index ecece8c..81a54d7 100644 --- a/lib/Target/PowerPC/PPCCTRLoops.cpp +++ b/lib/Target/PowerPC/PPCCTRLoops.cpp @@ -396,7 +396,7 @@ CountValue *PPCCTRLoops::getTripCount(MachineLoop *L, // Here we need to look for an immediate load (an li or lis/ori pair). if (DefInstr && (DefInstr->getOpcode() == PPC::ORI8 || DefInstr->getOpcode() == PPC::ORI)) { - int64_t start = (short) DefInstr->getOperand(2).getImm(); + int64_t start = DefInstr->getOperand(2).getImm(); MachineInstr *DefInstr2 = MRI->getVRegDef(DefInstr->getOperand(1).getReg()); if (DefInstr2 && (DefInstr2->getOpcode() == PPC::LIS8 || @@ -685,7 +685,7 @@ bool PPCCTRLoops::convertToCTRLoop(MachineLoop *L) { const TargetRegisterClass *SrcRC = MF->getRegInfo().getRegClass(TripCount->getReg()); CountReg = MF->getRegInfo().createVirtualRegister(RC); - unsigned CopyOp = (isPPC64 && SrcRC == GPRC) ? + unsigned CopyOp = (isPPC64 && GPRC->hasSubClassEq(SrcRC)) ? (unsigned) PPC::EXTSW_32_64 : (unsigned) TargetOpcode::COPY; BuildMI(*Preheader, InsertPos, dl, diff --git a/lib/Target/PowerPC/PPCCallingConv.td b/lib/Target/PowerPC/PPCCallingConv.td index caeb179..c8a29a3 100644 --- a/lib/Target/PowerPC/PPCCallingConv.td +++ b/lib/Target/PowerPC/PPCCallingConv.td @@ -136,3 +136,9 @@ def CSR_SVR464 : CalleeSavedRegs<(add X14, X15, X16, X17, X18, X19, X20, VRSAV F27, F28, F29, F30, F31, CR2, CR3, CR4, V20, V21, V22, V23, V24, V25, V26, V27, V28, V29, V30, V31)>; + +def CSR_NoRegs : CalleeSavedRegs<(add VRSAVE)>; +def CSR_NoRegs_Darwin : CalleeSavedRegs<(add)>; + +def CSR_NoRegs_Altivec : CalleeSavedRegs<(add (sequence "V%u", 0, 31), VRSAVE)>; + diff --git a/lib/Target/PowerPC/PPCCodeEmitter.cpp b/lib/Target/PowerPC/PPCCodeEmitter.cpp index d68bfd1..6478718 100644 --- a/lib/Target/PowerPC/PPCCodeEmitter.cpp +++ b/lib/Target/PowerPC/PPCCodeEmitter.cpp @@ -142,7 +142,7 @@ unsigned PPCCodeEmitter::get_crbitm_encoding(const MachineInstr &MI, assert((MI.getOpcode() == PPC::MTCRF || MI.getOpcode() == PPC::MTCRF8 || MI.getOpcode() == PPC::MFOCRF) && (MO.getReg() >= PPC::CR0 && MO.getReg() <= PPC::CR7)); - return 0x80 >> getPPCRegisterNumbering(MO.getReg()); + return 0x80 >> TM.getRegisterInfo()->getEncodingValue(MO.getReg()); } MachineRelocation PPCCodeEmitter::GetRelocation(const MachineOperand &MO, @@ -260,7 +260,7 @@ unsigned PPCCodeEmitter::getMachineOpValue(const MachineInstr &MI, assert((MI.getOpcode() != PPC::MTCRF && MI.getOpcode() != PPC::MTCRF8 && MI.getOpcode() != PPC::MFOCRF) || MO.getReg() < PPC::CR0 || MO.getReg() > PPC::CR7); - return getPPCRegisterNumbering(MO.getReg()); + return TM.getRegisterInfo()->getEncodingValue(MO.getReg()); } assert(MO.isImm() && diff --git a/lib/Target/PowerPC/PPCFrameLowering.cpp b/lib/Target/PowerPC/PPCFrameLowering.cpp index 353560d..9ec10f6 100644 --- a/lib/Target/PowerPC/PPCFrameLowering.cpp +++ b/lib/Target/PowerPC/PPCFrameLowering.cpp @@ -103,6 +103,7 @@ static void RemoveVRSaveCode(MachineInstr *MI) { // transform this into the appropriate ORI instruction. static void HandleVRSaveUpdate(MachineInstr *MI, const TargetInstrInfo &TII) { MachineFunction *MF = MI->getParent()->getParent(); + const TargetRegisterInfo *TRI = MF->getTarget().getRegisterInfo(); DebugLoc dl = MI->getDebugLoc(); unsigned UsedRegMask = 0; @@ -115,7 +116,7 @@ static void HandleVRSaveUpdate(MachineInstr *MI, const TargetInstrInfo &TII) { for (MachineRegisterInfo::livein_iterator I = MF->getRegInfo().livein_begin(), E = MF->getRegInfo().livein_end(); I != E; ++I) { - unsigned RegNo = getPPCRegisterNumbering(I->first); + unsigned RegNo = TRI->getEncodingValue(I->first); if (VRRegNo[RegNo] == I->first) // If this really is a vector reg. UsedRegMask &= ~(1 << (31-RegNo)); // Doesn't need to be marked. } @@ -131,7 +132,7 @@ static void HandleVRSaveUpdate(MachineInstr *MI, const TargetInstrInfo &TII) { const MachineOperand &MO = Ret.getOperand(I); if (!MO.isReg() || !PPC::VRRCRegClass.contains(MO.getReg())) continue; - unsigned RegNo = getPPCRegisterNumbering(MO.getReg()); + unsigned RegNo = TRI->getEncodingValue(MO.getReg()); UsedRegMask &= ~(1 << (31-RegNo)); } } @@ -188,6 +189,11 @@ static bool spillsCR(const MachineFunction &MF) { return FuncInfo->isCRSpilled(); } +static bool spillsVRSAVE(const MachineFunction &MF) { + const PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>(); + return FuncInfo->isVRSAVESpilled(); +} + static bool hasSpills(const MachineFunction &MF) { const PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>(); return FuncInfo->hasSpills(); @@ -217,9 +223,7 @@ unsigned PPCFrameLowering::determineFrameLayout(MachineFunction &MF, // If we are a leaf function, and use up to 224 bytes of stack space, // don't have a frame pointer, calls, or dynamic alloca then we do not need - // to adjust the stack pointer (we fit in the Red Zone). For 64-bit - // SVR4, we also require a stack frame if we need to spill the CR, - // since this spill area is addressed relative to the stack pointer. + // to adjust the stack pointer (we fit in the Red Zone). // The 32-bit SVR4 ABI has no Red Zone. However, it can still generate // stackless code if all local vars are reg-allocated. bool DisableRedZone = MF.getFunction()->getAttributes(). @@ -231,9 +235,6 @@ unsigned PPCFrameLowering::determineFrameLayout(MachineFunction &MF, FrameSize <= 224 && // Fits in red zone. !MFI->hasVarSizedObjects() && // No dynamic alloca. !MFI->adjustsStack() && // No calls. - !(Subtarget.isPPC64() && // No 64-bit SVR4 CRsave. - Subtarget.isSVR4ABI() - && spillsCR(MF)) && (!ALIGN_STACK || MaxAlign <= TargetAlign)) { // No special alignment. // No need for frame if (UpdateMF) @@ -299,6 +300,31 @@ bool PPCFrameLowering::needsFP(const MachineFunction &MF) const { MF.getInfo<PPCFunctionInfo>()->hasFastCall()); } +void PPCFrameLowering::replaceFPWithRealFP(MachineFunction &MF) const { + bool is31 = needsFP(MF); + unsigned FPReg = is31 ? PPC::R31 : PPC::R1; + unsigned FP8Reg = is31 ? PPC::X31 : PPC::X1; + + for (MachineFunction::iterator BI = MF.begin(), BE = MF.end(); + BI != BE; ++BI) + for (MachineBasicBlock::iterator MBBI = BI->end(); MBBI != BI->begin(); ) { + --MBBI; + for (unsigned I = 0, E = MBBI->getNumOperands(); I != E; ++I) { + MachineOperand &MO = MBBI->getOperand(I); + if (!MO.isReg()) + continue; + + switch (MO.getReg()) { + case PPC::FP: + MO.setReg(FPReg); + break; + case PPC::FP8: + MO.setReg(FP8Reg); + break; + } + } + } +} void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { MachineBasicBlock &MBB = MF.front(); // Prolog goes in entry BB @@ -332,6 +358,9 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { unsigned FrameSize = determineFrameLayout(MF); int NegFrameSize = -FrameSize; + if (MFI->isFrameAddressTaken()) + replaceFPWithRealFP(MF); + // Get processor type. bool isPPC64 = Subtarget.isPPC64(); // Get operating system @@ -339,6 +368,7 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { // Check if the link register (LR) must be saved. PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>(); bool MustSaveLR = FI->mustSaveLR(); + const SmallVector<unsigned, 3> &MustSaveCRs = FI->getMustSaveCRs(); // Do we have a frame pointer for this function? bool HasFP = hasFP(MF); @@ -360,6 +390,13 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { if (MustSaveLR) BuildMI(MBB, MBBI, dl, TII.get(PPC::MFLR8), PPC::X0); + if (!MustSaveCRs.empty()) { + MachineInstrBuilder MIB = + BuildMI(MBB, MBBI, dl, TII.get(PPC::MFCR8), PPC::X12); + for (unsigned i = 0, e = MustSaveCRs.size(); i != e; ++i) + MIB.addReg(MustSaveCRs[i], RegState::ImplicitKill); + } + if (HasFP) BuildMI(MBB, MBBI, dl, TII.get(PPC::STD)) .addReg(PPC::X31) @@ -371,6 +408,12 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { .addReg(PPC::X0) .addImm(LROffset / 4) .addReg(PPC::X1); + + if (!MustSaveCRs.empty()) + BuildMI(MBB, MBBI, dl, TII.get(PPC::STW8)) + .addReg(PPC::X12, getKillRegState(true)) + .addImm(8) + .addReg(PPC::X1); } else { if (MustSaveLR) BuildMI(MBB, MBBI, dl, TII.get(PPC::MFLR), PPC::R0); @@ -383,6 +426,9 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { .addImm(FPOffset) .addReg(PPC::R1); + assert(MustSaveCRs.empty() && + "Prologue CR saving supported only in 64-bit mode"); + if (MustSaveLR) BuildMI(MBB, MBBI, dl, TII.get(PPC::STW)) .addReg(PPC::R0) @@ -546,7 +592,7 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { // spilled CRs. if (Subtarget.isSVR4ABI() && (PPC::CR2 <= Reg && Reg <= PPC::CR4) - && !spillsCR(MF)) + && MustSaveCRs.empty()) continue; // For 64-bit SVR4 when we have spilled CRs, the spill location @@ -602,6 +648,7 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF, // Check if the link register (LR) has been saved. PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>(); bool MustSaveLR = FI->mustSaveLR(); + const SmallVector<unsigned, 3> &MustSaveCRs = FI->getMustSaveCRs(); // Do we have a frame pointer for this function? bool HasFP = hasFP(MF); @@ -702,10 +749,19 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF, BuildMI(MBB, MBBI, dl, TII.get(PPC::LD), PPC::X0) .addImm(LROffset/4).addReg(PPC::X1); + if (!MustSaveCRs.empty()) + BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ8), PPC::X12) + .addImm(8).addReg(PPC::X1); + if (HasFP) BuildMI(MBB, MBBI, dl, TII.get(PPC::LD), PPC::X31) .addImm(FPOffset/4).addReg(PPC::X1); + if (!MustSaveCRs.empty()) + for (unsigned i = 0, e = MustSaveCRs.size(); i != e; ++i) + BuildMI(MBB, MBBI, dl, TII.get(PPC::MTCRF8), MustSaveCRs[i]) + .addReg(PPC::X12, getKillRegState(i == e-1)); + if (MustSaveLR) BuildMI(MBB, MBBI, dl, TII.get(PPC::MTLR8)).addReg(PPC::X0); } else { @@ -713,6 +769,9 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF, BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ), PPC::R0) .addImm(LROffset).addReg(PPC::R1); + assert(MustSaveCRs.empty() && + "Epilogue CR restoring supported only in 64-bit mode"); + if (HasFP) BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ), PPC::R31) .addImm(FPOffset).addReg(PPC::R1); @@ -917,6 +976,7 @@ void PPCFrameLowering::processFunctionBeforeFrameFinalized(MachineFunction &MF, } PPCFunctionInfo *PFI = MF.getInfo<PPCFunctionInfo>(); + const TargetRegisterInfo *TRI = MF.getTarget().getRegisterInfo(); int64_t LowerBound = 0; @@ -936,7 +996,7 @@ void PPCFrameLowering::processFunctionBeforeFrameFinalized(MachineFunction &MF, FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI)); } - LowerBound -= (31 - getPPCRegisterNumbering(MinFPR) + 1) * 8; + LowerBound -= (31 - TRI->getEncodingValue(MinFPR) + 1) * 8; } // Check whether the frame pointer register is allocated. If so, make sure it @@ -970,8 +1030,8 @@ void PPCFrameLowering::processFunctionBeforeFrameFinalized(MachineFunction &MF, } unsigned MinReg = - std::min<unsigned>(getPPCRegisterNumbering(MinGPR), - getPPCRegisterNumbering(MinG8R)); + std::min<unsigned>(TRI->getEncodingValue(MinGPR), + TRI->getEncodingValue(MinG8R)); if (Subtarget.isPPC64()) { LowerBound -= (31 - MinReg + 1) * 8; @@ -1053,14 +1113,21 @@ PPCFrameLowering::addScavengingSpillSlot(MachineFunction &MF, // needed alignment padding. unsigned StackSize = determineFrameLayout(MF, false, true); MachineFrameInfo *MFI = MF.getFrameInfo(); - if (MFI->hasVarSizedObjects() || spillsCR(MF) || hasNonRISpills(MF) || - (hasSpills(MF) && !isInt<16>(StackSize))) { + if (MFI->hasVarSizedObjects() || spillsCR(MF) || spillsVRSAVE(MF) || + hasNonRISpills(MF) || (hasSpills(MF) && !isInt<16>(StackSize))) { const TargetRegisterClass *GPRC = &PPC::GPRCRegClass; const TargetRegisterClass *G8RC = &PPC::G8RCRegClass; const TargetRegisterClass *RC = Subtarget.isPPC64() ? G8RC : GPRC; - RS->setScavengingFrameIndex(MFI->CreateStackObject(RC->getSize(), + RS->addScavengingFrameIndex(MFI->CreateStackObject(RC->getSize(), RC->getAlignment(), false)); + + // These kinds of spills might need two registers. + if (spillsCR(MF) || spillsVRSAVE(MF)) + RS->addScavengingFrameIndex(MFI->CreateStackObject(RC->getSize(), + RC->getAlignment(), + false)); + } } @@ -1080,44 +1147,41 @@ PPCFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB, *static_cast<const PPCInstrInfo*>(MF->getTarget().getInstrInfo()); DebugLoc DL; bool CRSpilled = false; + MachineInstrBuilder CRMIB; for (unsigned i = 0, e = CSI.size(); i != e; ++i) { unsigned Reg = CSI[i].getReg(); // CR2 through CR4 are the nonvolatile CR fields. bool IsCRField = PPC::CR2 <= Reg && Reg <= PPC::CR4; - if (CRSpilled && IsCRField) - continue; - // Add the callee-saved register as live-in; it's killed at the spill. MBB.addLiveIn(Reg); + if (CRSpilled && IsCRField) { + CRMIB.addReg(Reg, RegState::ImplicitKill); + continue; + } + // Insert the spill to the stack frame. if (IsCRField) { - CRSpilled = true; - // The first time we see a CR field, store the whole CR into the - // save slot via GPR12 (available in the prolog for 32- and 64-bit). + PPCFunctionInfo *FuncInfo = MF->getInfo<PPCFunctionInfo>(); if (Subtarget.isPPC64()) { - // 64-bit: SP+8 - MBB.insert(MI, BuildMI(*MF, DL, TII.get(PPC::MFCR), PPC::X12)); - MBB.insert(MI, BuildMI(*MF, DL, TII.get(PPC::STW)) - .addReg(PPC::X12, - getKillRegState(true)) - .addImm(8) - .addReg(PPC::X1)); + // The actual spill will happen at the start of the prologue. + FuncInfo->addMustSaveCR(Reg); } else { + CRSpilled = true; + // 32-bit: FP-relative. Note that we made sure CR2-CR4 all have // the same frame index in PPCRegisterInfo::hasReservedSpillSlot. - MBB.insert(MI, BuildMI(*MF, DL, TII.get(PPC::MFCR), PPC::R12)); + CRMIB = BuildMI(*MF, DL, TII.get(PPC::MFCR), PPC::R12) + .addReg(Reg, RegState::ImplicitKill); + + MBB.insert(MI, CRMIB); MBB.insert(MI, addFrameReference(BuildMI(*MF, DL, TII.get(PPC::STW)) .addReg(PPC::R12, getKillRegState(true)), CSI[i].getFrameIdx())); } - - // Record that we spill the CR in this function. - PPCFunctionInfo *FuncInfo = MF->getInfo<PPCFunctionInfo>(); - FuncInfo->setSpillsCR(); } else { const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); TII.storeRegToStackSlot(MBB, MI, Reg, true, @@ -1128,7 +1192,8 @@ PPCFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB, } static void -restoreCRs(bool isPPC64, bool CR2Spilled, bool CR3Spilled, bool CR4Spilled, +restoreCRs(bool isPPC64, bool is31, + bool CR2Spilled, bool CR3Spilled, bool CR4Spilled, MachineBasicBlock &MBB, MachineBasicBlock::iterator MI, const std::vector<CalleeSavedInfo> &CSI, unsigned CSIIndex) { @@ -1138,14 +1203,10 @@ restoreCRs(bool isPPC64, bool CR2Spilled, bool CR3Spilled, bool CR4Spilled, DebugLoc DL; unsigned RestoreOp, MoveReg; - if (isPPC64) { - // 64-bit: SP+8 - MBB.insert(MI, BuildMI(*MF, DL, TII.get(PPC::LWZ), PPC::X12) - .addImm(8) - .addReg(PPC::X1)); - RestoreOp = PPC::MTCRF8; - MoveReg = PPC::X12; - } else { + if (isPPC64) + // This is handled during epilogue generation. + return; + else { // 32-bit: FP-relative MBB.insert(MI, addFrameReference(BuildMI(*MF, DL, TII.get(PPC::LWZ), PPC::R12), @@ -1156,15 +1217,15 @@ restoreCRs(bool isPPC64, bool CR2Spilled, bool CR3Spilled, bool CR4Spilled, if (CR2Spilled) MBB.insert(MI, BuildMI(*MF, DL, TII.get(RestoreOp), PPC::CR2) - .addReg(MoveReg)); + .addReg(MoveReg, getKillRegState(!CR3Spilled && !CR4Spilled))); if (CR3Spilled) MBB.insert(MI, BuildMI(*MF, DL, TII.get(RestoreOp), PPC::CR3) - .addReg(MoveReg)); + .addReg(MoveReg, getKillRegState(!CR4Spilled))); if (CR4Spilled) MBB.insert(MI, BuildMI(*MF, DL, TII.get(RestoreOp), PPC::CR4) - .addReg(MoveReg)); + .addReg(MoveReg, getKillRegState(true))); } void PPCFrameLowering:: @@ -1255,7 +1316,9 @@ PPCFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB, // least one CR register, restore all spilled CRs together. if ((CR2Spilled || CR3Spilled || CR4Spilled) && !(PPC::CR2 <= Reg && Reg <= PPC::CR4)) { - restoreCRs(Subtarget.isPPC64(), CR2Spilled, CR3Spilled, CR4Spilled, + bool is31 = needsFP(*MF); + restoreCRs(Subtarget.isPPC64(), is31, + CR2Spilled, CR3Spilled, CR4Spilled, MBB, I, CSI, CSIIndex); CR2Spilled = CR3Spilled = CR4Spilled = false; } @@ -1278,9 +1341,11 @@ PPCFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB, } // If we haven't yet spilled the CRs, do so now. - if (CR2Spilled || CR3Spilled || CR4Spilled) - restoreCRs(Subtarget.isPPC64(), CR2Spilled, CR3Spilled, CR4Spilled, + if (CR2Spilled || CR3Spilled || CR4Spilled) { + bool is31 = needsFP(*MF); + restoreCRs(Subtarget.isPPC64(), is31, CR2Spilled, CR3Spilled, CR4Spilled, MBB, I, CSI, CSIIndex); + } return true; } diff --git a/lib/Target/PowerPC/PPCFrameLowering.h b/lib/Target/PowerPC/PPCFrameLowering.h index 53ee326..6f5f936 100644 --- a/lib/Target/PowerPC/PPCFrameLowering.h +++ b/lib/Target/PowerPC/PPCFrameLowering.h @@ -43,6 +43,7 @@ public: bool hasFP(const MachineFunction &MF) const; bool needsFP(const MachineFunction &MF) const; + void replaceFPWithRealFP(MachineFunction &MF) const; void processFunctionBeforeCalleeSavedScan(MachineFunction &MF, RegScavenger *RS = NULL) const; diff --git a/lib/Target/PowerPC/PPCHazardRecognizers.cpp b/lib/Target/PowerPC/PPCHazardRecognizers.cpp index 6ed1fb9..4bf1e33 100644 --- a/lib/Target/PowerPC/PPCHazardRecognizers.cpp +++ b/lib/Target/PowerPC/PPCHazardRecognizers.cpp @@ -179,7 +179,7 @@ getHazardType(SUnit *SU, int Stalls) { } // Do not allow MTCTR and BCTRL to be in the same dispatch group. - if (HasCTRSet && (Opcode == PPC::BCTRL_Darwin || Opcode == PPC::BCTRL_SVR4)) + if (HasCTRSet && Opcode == PPC::BCTRL) return NoopHazard; // If this is a load following a store, make sure it's not to the same or diff --git a/lib/Target/PowerPC/PPCISelDAGToDAG.cpp b/lib/Target/PowerPC/PPCISelDAGToDAG.cpp index 17bea8a..aed0fbb 100644 --- a/lib/Target/PowerPC/PPCISelDAGToDAG.cpp +++ b/lib/Target/PowerPC/PPCISelDAGToDAG.cpp @@ -120,10 +120,10 @@ namespace { } /// SelectAddrImmOffs - Return true if the operand is valid for a preinc - /// immediate field. Because preinc imms have already been validated, just - /// accept it. + /// immediate field. Note that the operand at this point is already the + /// result of a prior SelectAddressRegImm call. bool SelectAddrImmOffs(SDValue N, SDValue &Out) const { - if (isa<ConstantSDNode>(N) || N.getOpcode() == PPCISD::Lo || + if (N.getOpcode() == ISD::TargetConstant || N.getOpcode() == ISD::TargetGlobalAddress) { Out = N; return true; @@ -132,18 +132,6 @@ namespace { return false; } - /// SelectAddrIdxOffs - Return true if the operand is valid for a preinc - /// index field. Because preinc imms have already been validated, just - /// accept it. - bool SelectAddrIdxOffs(SDValue N, SDValue &Out) const { - if (isa<ConstantSDNode>(N) || N.getOpcode() == PPCISD::Lo || - N.getOpcode() == ISD::TargetGlobalAddress) - return false; - - Out = N; - return true; - } - /// SelectAddrIdx - Given the specified addressed, check to see if it can be /// represented as an indexed [r+r] operation. Returns false if it can /// be represented by [r+imm], which are preferred. @@ -164,6 +152,12 @@ namespace { return PPCLowering.SelectAddressRegImmShift(N, Disp, Base, *CurDAG); } + // Select an address into a single register. + bool SelectAddr(SDValue N, SDValue &Base) { + Base = N; + return true; + } + /// SelectInlineAsmMemoryOperand - Implement addressing mode selection for /// inline asm expressions. It is always correct to compute the value into /// a register. The case of adding a (possibly relocatable) constant to a @@ -463,7 +457,7 @@ SDNode *PPCDAGToDAGISel::SelectBitfieldInsert(SDNode *N) { SH &= 31; SDValue Ops[] = { Op0, Op1, getI32Imm(SH), getI32Imm(MB), getI32Imm(ME) }; - return CurDAG->getMachineNode(PPC::RLWIMI, dl, MVT::i32, Ops, 5); + return CurDAG->getMachineNode(PPC::RLWIMI, dl, MVT::i32, Ops); } } return 0; @@ -786,7 +780,7 @@ SDNode *PPCDAGToDAGISel::SelectSETCC(SDNode *N) { } case ISD::SETGT: { SDValue Ops[] = { Op, getI32Imm(1), getI32Imm(31), getI32Imm(31) }; - Op = SDValue(CurDAG->getMachineNode(PPC::RLWINM, dl, MVT::i32, Ops, 4), + Op = SDValue(CurDAG->getMachineNode(PPC::RLWINM, dl, MVT::i32, Ops), 0); return CurDAG->SelectNodeTo(N, PPC::XORI, MVT::i32, Op, getI32Imm(1)); @@ -879,7 +873,7 @@ SDNode *PPCDAGToDAGISel::SelectSETCC(SDNode *N) { // Get the specified bit. SDValue Tmp = - SDValue(CurDAG->getMachineNode(PPC::RLWINM, dl, MVT::i32, Ops, 4), 0); + SDValue(CurDAG->getMachineNode(PPC::RLWINM, dl, MVT::i32, Ops), 0); if (Inv) { assert(OtherCondIdx == -1 && "Can't have split plus negation"); return CurDAG->SelectNodeTo(N, PPC::XORI, MVT::i32, Tmp, getI32Imm(1)); @@ -891,7 +885,7 @@ SDNode *PPCDAGToDAGISel::SelectSETCC(SDNode *N) { // Get the other bit of the comparison. Ops[1] = getI32Imm((32-(3-OtherCondIdx)) & 31); SDValue OtherCond = - SDValue(CurDAG->getMachineNode(PPC::RLWINM, dl, MVT::i32, Ops, 4), 0); + SDValue(CurDAG->getMachineNode(PPC::RLWINM, dl, MVT::i32, Ops), 0); return CurDAG->SelectNodeTo(N, PPC::OR, MVT::i32, Tmp, OtherCond); } @@ -1050,7 +1044,7 @@ SDNode *PPCDAGToDAGISel::Select(SDNode *N) { break; SDValue Offset = LD->getOffset(); - if (isa<ConstantSDNode>(Offset) || + if (Offset.getOpcode() == ISD::TargetConstant || Offset.getOpcode() == ISD::TargetGlobalAddress) { unsigned Opcode; @@ -1085,7 +1079,7 @@ SDNode *PPCDAGToDAGISel::Select(SDNode *N) { SDValue Ops[] = { Offset, Base, Chain }; return CurDAG->getMachineNode(Opcode, dl, LD->getValueType(0), PPCLowering.getPointerTy(), - MVT::Other, Ops, 3); + MVT::Other, Ops); } else { unsigned Opcode; bool isSExt = LD->getExtensionType() == ISD::SEXTLOAD; @@ -1117,10 +1111,10 @@ SDNode *PPCDAGToDAGISel::Select(SDNode *N) { SDValue Chain = LD->getChain(); SDValue Base = LD->getBasePtr(); - SDValue Ops[] = { Offset, Base, Chain }; + SDValue Ops[] = { Base, Offset, Chain }; return CurDAG->getMachineNode(Opcode, dl, LD->getValueType(0), PPCLowering.getPointerTy(), - MVT::Other, Ops, 3); + MVT::Other, Ops); } } @@ -1169,7 +1163,7 @@ SDNode *PPCDAGToDAGISel::Select(SDNode *N) { SDValue Ops[] = { N->getOperand(0).getOperand(0), N->getOperand(0).getOperand(1), getI32Imm(0), getI32Imm(MB),getI32Imm(ME) }; - return CurDAG->getMachineNode(PPC::RLWIMI, dl, MVT::i32, Ops, 5); + return CurDAG->getMachineNode(PPC::RLWIMI, dl, MVT::i32, Ops); } } @@ -1483,8 +1477,7 @@ void PPCDAGToDAGISel::PostprocessISelDAG() { default: continue; case PPC::ADDI8: - case PPC::ADDI8L: - case PPC::ADDIL: + case PPC::ADDI: // In some cases (such as TLS) the relocation information // is already in place on the operand, so copying the operand // is sufficient. diff --git a/lib/Target/PowerPC/PPCISelLowering.cpp b/lib/Target/PowerPC/PPCISelLowering.cpp index 741e25e..3fcafdc 100644 --- a/lib/Target/PowerPC/PPCISelLowering.cpp +++ b/lib/Target/PowerPC/PPCISelLowering.cpp @@ -70,6 +70,8 @@ static TargetLoweringObjectFile *CreateTLOF(const PPCTargetMachine &TM) { PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM) : TargetLowering(TM, CreateTLOF(TM)), PPCSubTarget(*TM.getSubtargetImpl()) { const PPCSubtarget *Subtarget = &TM.getSubtarget<PPCSubtarget>(); + PPCRegInfo = TM.getRegisterInfo(); + PPCII = TM.getInstrInfo(); setPow2DivIsCheap(); @@ -115,6 +117,7 @@ PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM) setOperationAction(ISD::FTRUNC, MVT::ppcf128, Expand); setOperationAction(ISD::FRINT, MVT::ppcf128, Expand); setOperationAction(ISD::FNEARBYINT, MVT::ppcf128, Expand); + setOperationAction(ISD::FREM, MVT::ppcf128, Expand); // PowerPC has no SREM/UREM instructions setOperationAction(ISD::SREM, MVT::i32, Expand); @@ -149,26 +152,58 @@ PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM) setOperationAction(ISD::FLT_ROUNDS_, MVT::i32, Custom); // If we're enabling GP optimizations, use hardware square root - if (!Subtarget->hasFSQRT()) { + if (!Subtarget->hasFSQRT() && + !(TM.Options.UnsafeFPMath && + Subtarget->hasFRSQRTE() && Subtarget->hasFRE())) setOperationAction(ISD::FSQRT, MVT::f64, Expand); + + if (!Subtarget->hasFSQRT() && + !(TM.Options.UnsafeFPMath && + Subtarget->hasFRSQRTES() && Subtarget->hasFRES())) setOperationAction(ISD::FSQRT, MVT::f32, Expand); - } setOperationAction(ISD::FCOPYSIGN, MVT::f64, Expand); setOperationAction(ISD::FCOPYSIGN, MVT::f32, Expand); + if (Subtarget->hasFPRND()) { + setOperationAction(ISD::FFLOOR, MVT::f64, Legal); + setOperationAction(ISD::FCEIL, MVT::f64, Legal); + setOperationAction(ISD::FTRUNC, MVT::f64, Legal); + + setOperationAction(ISD::FFLOOR, MVT::f32, Legal); + setOperationAction(ISD::FCEIL, MVT::f32, Legal); + setOperationAction(ISD::FTRUNC, MVT::f32, Legal); + + // frin does not implement "ties to even." Thus, this is safe only in + // fast-math mode. + if (TM.Options.UnsafeFPMath) { + setOperationAction(ISD::FNEARBYINT, MVT::f64, Legal); + setOperationAction(ISD::FNEARBYINT, MVT::f32, Legal); + + // These need to set FE_INEXACT, and use a custom inserter. + setOperationAction(ISD::FRINT, MVT::f64, Legal); + setOperationAction(ISD::FRINT, MVT::f32, Legal); + } + } + // PowerPC does not have BSWAP, CTPOP or CTTZ setOperationAction(ISD::BSWAP, MVT::i32 , Expand); - setOperationAction(ISD::CTPOP, MVT::i32 , Expand); setOperationAction(ISD::CTTZ , MVT::i32 , Expand); setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i32, Expand); setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i32, Expand); setOperationAction(ISD::BSWAP, MVT::i64 , Expand); - setOperationAction(ISD::CTPOP, MVT::i64 , Expand); setOperationAction(ISD::CTTZ , MVT::i64 , Expand); setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i64, Expand); setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i64, Expand); + if (Subtarget->hasPOPCNTD()) { + setOperationAction(ISD::CTPOP, MVT::i32 , Legal); + setOperationAction(ISD::CTPOP, MVT::i64 , Legal); + } else { + setOperationAction(ISD::CTPOP, MVT::i32 , Expand); + setOperationAction(ISD::CTPOP, MVT::i64 , Expand); + } + // PowerPC does not have ROTR setOperationAction(ISD::ROTR, MVT::i32 , Expand); setOperationAction(ISD::ROTR, MVT::i64 , Expand); @@ -211,6 +246,14 @@ PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM) setOperationAction(ISD::EXCEPTIONADDR, MVT::i32, Expand); setOperationAction(ISD::EHSELECTION, MVT::i32, Expand); + // NOTE: EH_SJLJ_SETJMP/_LONGJMP supported here is NOT intended to support + // SjLj exception handling but a light-weight setjmp/longjmp replacement to + // support continuation, user-level threading, and etc.. As a result, no + // other SjLj exception interfaces are implemented and please don't build + // your own exception handling based on them. + // LLVM/Clang supports zero-cost DWARF exception handling. + setOperationAction(ISD::EH_SJLJ_SETJMP, MVT::i32, Custom); + setOperationAction(ISD::EH_SJLJ_LONGJMP, MVT::Other, Custom); // We want to legalize GlobalAddress and ConstantPool nodes into the // appropriate instructions to materialize the address. @@ -290,15 +333,28 @@ PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM) // We cannot do this with Promote because i64 is not a legal type. setOperationAction(ISD::FP_TO_UINT, MVT::i32, Custom); - // FIXME: disable this lowered code. This generates 64-bit register values, - // and we don't model the fact that the top part is clobbered by calls. We - // need to flag these together so that the value isn't live across a call. - //setOperationAction(ISD::SINT_TO_FP, MVT::i32, Custom); + if (PPCSubTarget.hasLFIWAX() || Subtarget->isPPC64()) + setOperationAction(ISD::SINT_TO_FP, MVT::i32, Custom); } else { // PowerPC does not have FP_TO_UINT on 32-bit implementations. setOperationAction(ISD::FP_TO_UINT, MVT::i32, Expand); } + // With the instructions enabled under FPCVT, we can do everything. + if (PPCSubTarget.hasFPCVT()) { + if (Subtarget->has64BitSupport()) { + setOperationAction(ISD::FP_TO_SINT, MVT::i64, Custom); + setOperationAction(ISD::FP_TO_UINT, MVT::i64, Custom); + setOperationAction(ISD::SINT_TO_FP, MVT::i64, Custom); + setOperationAction(ISD::UINT_TO_FP, MVT::i64, Custom); + } + + setOperationAction(ISD::FP_TO_SINT, MVT::i32, Custom); + setOperationAction(ISD::FP_TO_UINT, MVT::i32, Custom); + setOperationAction(ISD::SINT_TO_FP, MVT::i32, Custom); + setOperationAction(ISD::UINT_TO_FP, MVT::i32, Custom); + } + if (Subtarget->use64BitRegs()) { // 64-bit PowerPC implementations can support i64 types directly addRegisterClass(MVT::i64, &PPC::G8RCRegClass); @@ -420,6 +476,12 @@ PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM) setOperationAction(ISD::MUL, MVT::v4f32, Legal); setOperationAction(ISD::FMA, MVT::v4f32, Legal); + + if (TM.Options.UnsafeFPMath) { + setOperationAction(ISD::FDIV, MVT::v4f32, Legal); + setOperationAction(ISD::FSQRT, MVT::v4f32, Legal); + } + setOperationAction(ISD::MUL, MVT::v4i32, Custom); setOperationAction(ISD::MUL, MVT::v8i16, Custom); setOperationAction(ISD::MUL, MVT::v16i8, Custom); @@ -452,7 +514,8 @@ PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM) setOperationAction(ISD::ATOMIC_STORE, MVT::i64, Expand); setBooleanContents(ZeroOrOneBooleanContent); - setBooleanVectorContents(ZeroOrOneBooleanContent); // FIXME: Is this correct? + // Altivec instructions set fields to all zeros or all ones. + setBooleanVectorContents(ZeroOrNegativeOneBooleanContent); if (isPPC64) { setStackPointerRegisterToSaveRestore(PPC::X1); @@ -470,6 +533,12 @@ PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM) setTargetDAGCombine(ISD::BR_CC); setTargetDAGCombine(ISD::BSWAP); + // Use reciprocal estimates. + if (TM.Options.UnsafeFPMath) { + setTargetDAGCombine(ISD::FDIV); + setTargetDAGCombine(ISD::FSQRT); + } + // Darwin long double math library functions have $LDBL128 appended. if (Subtarget->isDarwin()) { setLibcallName(RTLIB::COS_PPCF128, "cosl$LDBL128"); @@ -511,7 +580,6 @@ PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM) MaxStoresPerMemmoveOptSize = 8; setPrefFunctionAlignment(4); - BenefitFromCodePlacementOpt = true; } } @@ -542,6 +610,8 @@ const char *PPCTargetLowering::getTargetNodeName(unsigned Opcode) const { case PPCISD::FCFID: return "PPCISD::FCFID"; case PPCISD::FCTIDZ: return "PPCISD::FCTIDZ"; case PPCISD::FCTIWZ: return "PPCISD::FCTIWZ"; + case PPCISD::FRE: return "PPCISD::FRE"; + case PPCISD::FRSQRTE: return "PPCISD::FRSQRTE"; case PPCISD::STFIWX: return "PPCISD::STFIWX"; case PPCISD::VMADDFP: return "PPCISD::VMADDFP"; case PPCISD::VNMSUBFP: return "PPCISD::VNMSUBFP"; @@ -557,16 +627,13 @@ const char *PPCTargetLowering::getTargetNodeName(unsigned Opcode) const { case PPCISD::SRL: return "PPCISD::SRL"; case PPCISD::SRA: return "PPCISD::SRA"; case PPCISD::SHL: return "PPCISD::SHL"; - case PPCISD::EXTSW_32: return "PPCISD::EXTSW_32"; - case PPCISD::STD_32: return "PPCISD::STD_32"; - case PPCISD::CALL_SVR4: return "PPCISD::CALL_SVR4"; - case PPCISD::CALL_NOP_SVR4: return "PPCISD::CALL_NOP_SVR4"; - case PPCISD::CALL_Darwin: return "PPCISD::CALL_Darwin"; - case PPCISD::NOP: return "PPCISD::NOP"; + case PPCISD::CALL: return "PPCISD::CALL"; + case PPCISD::CALL_NOP: return "PPCISD::CALL_NOP"; case PPCISD::MTCTR: return "PPCISD::MTCTR"; - case PPCISD::BCTRL_Darwin: return "PPCISD::BCTRL_Darwin"; - case PPCISD::BCTRL_SVR4: return "PPCISD::BCTRL_SVR4"; + case PPCISD::BCTRL: return "PPCISD::BCTRL"; case PPCISD::RET_FLAG: return "PPCISD::RET_FLAG"; + case PPCISD::EH_SJLJ_SETJMP: return "PPCISD::EH_SJLJ_SETJMP"; + case PPCISD::EH_SJLJ_LONGJMP: return "PPCISD::EH_SJLJ_LONGJMP"; case PPCISD::MFCR: return "PPCISD::MFCR"; case PPCISD::VCMP: return "PPCISD::VCMP"; case PPCISD::VCMPo: return "PPCISD::VCMPo"; @@ -576,10 +643,7 @@ const char *PPCTargetLowering::getTargetNodeName(unsigned Opcode) const { case PPCISD::STCX: return "PPCISD::STCX"; case PPCISD::COND_BRANCH: return "PPCISD::COND_BRANCH"; case PPCISD::MFFS: return "PPCISD::MFFS"; - case PPCISD::MTFSB0: return "PPCISD::MTFSB0"; - case PPCISD::MTFSB1: return "PPCISD::MTFSB1"; case PPCISD::FADDRTZ: return "PPCISD::FADDRTZ"; - case PPCISD::MTFSF: return "PPCISD::MTFSF"; case PPCISD::TC_RETURN: return "PPCISD::TC_RETURN"; case PPCISD::CR6SET: return "PPCISD::CR6SET"; case PPCISD::CR6UNSET: return "PPCISD::CR6UNSET"; @@ -1031,7 +1095,7 @@ bool PPCTargetLowering::SelectAddressRegImm(SDValue N, SDValue &Disp, short Imm; if (isIntS16Immediate(CN, Imm)) { Disp = DAG.getTargetConstant(Imm, CN->getValueType(0)); - Base = DAG.getRegister(PPCSubTarget.isPPC64() ? PPC::X0 : PPC::R0, + Base = DAG.getRegister(PPCSubTarget.isPPC64() ? PPC::ZERO8 : PPC::ZERO, CN->getValueType(0)); return true; } @@ -1080,7 +1144,7 @@ bool PPCTargetLowering::SelectAddressRegRegOnly(SDValue N, SDValue &Base, } // Otherwise, do it the hard way, using R0 as the base register. - Base = DAG.getRegister(PPCSubTarget.isPPC64() ? PPC::X0 : PPC::R0, + Base = DAG.getRegister(PPCSubTarget.isPPC64() ? PPC::ZERO8 : PPC::ZERO, N.getValueType()); Index = N; return true; @@ -1143,7 +1207,7 @@ bool PPCTargetLowering::SelectAddressRegImmShift(SDValue N, SDValue &Disp, short Imm; if (isIntS16Immediate(CN, Imm)) { Disp = DAG.getTargetConstant((unsigned short)Imm >> 2, getPointerTy()); - Base = DAG.getRegister(PPCSubTarget.isPPC64() ? PPC::X0 : PPC::R0, + Base = DAG.getRegister(PPCSubTarget.isPPC64() ? PPC::ZERO8 : PPC::ZERO, CN->getValueType(0)); return true; } @@ -1181,6 +1245,7 @@ bool PPCTargetLowering::getPreIndexedAddressParts(SDNode *N, SDValue &Base, SelectionDAG &DAG) const { if (DisablePPCPreinc) return false; + bool isLoad = true; SDValue Ptr; EVT VT; unsigned Alignment; @@ -1192,6 +1257,7 @@ bool PPCTargetLowering::getPreIndexedAddressParts(SDNode *N, SDValue &Base, Ptr = ST->getBasePtr(); VT = ST->getMemoryVT(); Alignment = ST->getAlignment(); + isLoad = false; } else return false; @@ -1199,7 +1265,25 @@ bool PPCTargetLowering::getPreIndexedAddressParts(SDNode *N, SDValue &Base, if (VT.isVector()) return false; - if (SelectAddressRegReg(Ptr, Offset, Base, DAG)) { + if (SelectAddressRegReg(Ptr, Base, Offset, DAG)) { + + // Common code will reject creating a pre-inc form if the base pointer + // is a frame index, or if N is a store and the base pointer is either + // the same as or a predecessor of the value being stored. Check for + // those situations here, and try with swapped Base/Offset instead. + bool Swap = false; + + if (isa<FrameIndexSDNode>(Base) || isa<RegisterSDNode>(Base)) + Swap = true; + else if (!isLoad) { + SDValue Val = cast<StoreSDNode>(N)->getValue(); + if (Val == Base || Base.getNode()->isPredecessorOf(Val.getNode())) + Swap = true; + } + + if (Swap) + std::swap(Base, Offset); + AM = ISD::PRE_INC; return true; } @@ -3105,7 +3189,7 @@ unsigned PrepareCall(SelectionDAG &DAG, SDValue &Callee, SDValue &InFlag, NodeTys.push_back(MVT::Other); // Returns a chain NodeTys.push_back(MVT::Glue); // Returns a flag for retval copy to use. - unsigned CallOpc = isSVR4ABI ? PPCISD::CALL_SVR4 : PPCISD::CALL_Darwin; + unsigned CallOpc = PPCISD::CALL; bool needIndirectCall = true; if (SDNode *Dest = isBLACompatibleAddress(Callee, DAG)) { @@ -3238,8 +3322,11 @@ unsigned PrepareCall(SelectionDAG &DAG, SDValue &Callee, SDValue &InFlag, NodeTys.push_back(MVT::Other); NodeTys.push_back(MVT::Glue); Ops.push_back(Chain); - CallOpc = isSVR4ABI ? PPCISD::BCTRL_SVR4 : PPCISD::BCTRL_Darwin; + CallOpc = PPCISD::BCTRL; Callee.setNode(0); + // Add use of X11 (holding environment pointer) + if (isSVR4ABI && isPPC64) + Ops.push_back(DAG.getRegister(PPC::X11, PtrVT)); // Add CTR register as callee so a bctr can be emitted later. if (isTailCall) Ops.push_back(DAG.getRegister(isPPC64 ? PPC::CTR8 : PPC::CTR, PtrVT)); @@ -3378,7 +3465,7 @@ PPCTargetLowering::FinishCall(CallingConv::ID CallConv, DebugLoc dl, bool needsTOCRestore = false; if (!isTailCall && PPCSubTarget.isSVR4ABI()&& PPCSubTarget.isPPC64()) { - if (CallOpc == PPCISD::BCTRL_SVR4) { + if (CallOpc == PPCISD::BCTRL) { // This is a call through a function pointer. // Restore the caller TOC from the save area into R2. // See PrepareCall() for more information about calls through function @@ -3389,9 +3476,9 @@ PPCTargetLowering::FinishCall(CallingConv::ID CallConv, DebugLoc dl, // from allocating it), resulting in an additional register being // allocated and an unnecessary move instruction being generated. needsTOCRestore = true; - } else if ((CallOpc == PPCISD::CALL_SVR4) && !isLocalCall(Callee)) { + } else if ((CallOpc == PPCISD::CALL) && !isLocalCall(Callee)) { // Otherwise insert NOP for non-local calls. - CallOpc = PPCISD::CALL_NOP_SVR4; + CallOpc = PPCISD::CALL_NOP; } } @@ -4564,6 +4651,21 @@ SDValue PPCTargetLowering::LowerDYNAMIC_STACKALLOC(SDValue Op, return DAG.getNode(PPCISD::DYNALLOC, dl, VTs, Ops, 3); } +SDValue PPCTargetLowering::lowerEH_SJLJ_SETJMP(SDValue Op, + SelectionDAG &DAG) const { + DebugLoc DL = Op.getDebugLoc(); + return DAG.getNode(PPCISD::EH_SJLJ_SETJMP, DL, + DAG.getVTList(MVT::i32, MVT::Other), + Op.getOperand(0), Op.getOperand(1)); +} + +SDValue PPCTargetLowering::lowerEH_SJLJ_LONGJMP(SDValue Op, + SelectionDAG &DAG) const { + DebugLoc DL = Op.getDebugLoc(); + return DAG.getNode(PPCISD::EH_SJLJ_LONGJMP, DL, MVT::Other, + Op.getOperand(0), Op.getOperand(1)); +} + /// LowerSELECT_CC - Lower floating point select_cc's into fsel instruction when /// possible. SDValue PPCTargetLowering::LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const { @@ -4572,10 +4674,14 @@ SDValue PPCTargetLowering::LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const { !Op.getOperand(2).getValueType().isFloatingPoint()) return Op; - ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(4))->get(); + // We might be able to do better than this under some circumstances, but in + // general, fsel-based lowering of select is a finite-math-only optimization. + // For more information, see section F.3 of the 2.06 ISA specification. + if (!DAG.getTarget().Options.NoInfsFPMath || + !DAG.getTarget().Options.NoNaNsFPMath) + return Op; - // Cannot handle SETEQ/SETNE. - if (CC == ISD::SETEQ || CC == ISD::SETNE) return Op; + ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(4))->get(); EVT ResVT = Op.getValueType(); EVT CmpVT = Op.getOperand(0).getValueType(); @@ -4585,9 +4691,20 @@ SDValue PPCTargetLowering::LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const { // If the RHS of the comparison is a 0.0, we don't need to do the // subtraction at all. + SDValue Sel1; if (isFloatingPointZero(RHS)) switch (CC) { default: break; // SETUO etc aren't handled by fsel. + case ISD::SETNE: + std::swap(TV, FV); + case ISD::SETEQ: + if (LHS.getValueType() == MVT::f32) // Comparison is always 64-bits + LHS = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, LHS); + Sel1 = DAG.getNode(PPCISD::FSEL, dl, ResVT, LHS, TV, FV); + if (Sel1.getValueType() == MVT::f32) // Comparison is always 64-bits + Sel1 = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Sel1); + return DAG.getNode(PPCISD::FSEL, dl, ResVT, + DAG.getNode(ISD::FNEG, dl, MVT::f64, LHS), Sel1, FV); case ISD::SETULT: case ISD::SETLT: std::swap(TV, FV); // fsel is natively setge, swap operands for setlt @@ -4610,30 +4727,41 @@ SDValue PPCTargetLowering::LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const { SDValue Cmp; switch (CC) { default: break; // SETUO etc aren't handled by fsel. + case ISD::SETNE: + std::swap(TV, FV); + case ISD::SETEQ: + Cmp = DAG.getNode(ISD::FSUB, dl, CmpVT, LHS, RHS); + if (Cmp.getValueType() == MVT::f32) // Comparison is always 64-bits + Cmp = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Cmp); + Sel1 = DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, TV, FV); + if (Sel1.getValueType() == MVT::f32) // Comparison is always 64-bits + Sel1 = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Sel1); + return DAG.getNode(PPCISD::FSEL, dl, ResVT, + DAG.getNode(ISD::FNEG, dl, MVT::f64, Cmp), Sel1, FV); case ISD::SETULT: case ISD::SETLT: Cmp = DAG.getNode(ISD::FSUB, dl, CmpVT, LHS, RHS); if (Cmp.getValueType() == MVT::f32) // Comparison is always 64-bits Cmp = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Cmp); - return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, FV, TV); + return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, FV, TV); case ISD::SETOGE: case ISD::SETGE: Cmp = DAG.getNode(ISD::FSUB, dl, CmpVT, LHS, RHS); if (Cmp.getValueType() == MVT::f32) // Comparison is always 64-bits Cmp = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Cmp); - return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, TV, FV); + return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, TV, FV); case ISD::SETUGT: case ISD::SETGT: Cmp = DAG.getNode(ISD::FSUB, dl, CmpVT, RHS, LHS); if (Cmp.getValueType() == MVT::f32) // Comparison is always 64-bits Cmp = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Cmp); - return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, FV, TV); + return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, FV, TV); case ISD::SETOLE: case ISD::SETLE: Cmp = DAG.getNode(ISD::FSUB, dl, CmpVT, RHS, LHS); if (Cmp.getValueType() == MVT::f32) // Comparison is always 64-bits Cmp = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Cmp); - return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, TV, FV); + return DAG.getNode(PPCISD::FSEL, dl, ResVT, Cmp, TV, FV); } return Op; } @@ -4651,37 +4779,72 @@ SDValue PPCTargetLowering::LowerFP_TO_INT(SDValue Op, SelectionDAG &DAG, default: llvm_unreachable("Unhandled FP_TO_INT type in custom expander!"); case MVT::i32: Tmp = DAG.getNode(Op.getOpcode()==ISD::FP_TO_SINT ? PPCISD::FCTIWZ : - PPCISD::FCTIDZ, + (PPCSubTarget.hasFPCVT() ? PPCISD::FCTIWUZ : + PPCISD::FCTIDZ), dl, MVT::f64, Src); break; case MVT::i64: - Tmp = DAG.getNode(PPCISD::FCTIDZ, dl, MVT::f64, Src); + assert((Op.getOpcode() == ISD::FP_TO_SINT || PPCSubTarget.hasFPCVT()) && + "i64 FP_TO_UINT is supported only with FPCVT"); + Tmp = DAG.getNode(Op.getOpcode()==ISD::FP_TO_SINT ? PPCISD::FCTIDZ : + PPCISD::FCTIDUZ, + dl, MVT::f64, Src); break; } // Convert the FP value to an int value through memory. - SDValue FIPtr = DAG.CreateStackTemporary(MVT::f64); + bool i32Stack = Op.getValueType() == MVT::i32 && PPCSubTarget.hasSTFIWX() && + (Op.getOpcode() == ISD::FP_TO_SINT || PPCSubTarget.hasFPCVT()); + SDValue FIPtr = DAG.CreateStackTemporary(i32Stack ? MVT::i32 : MVT::f64); + int FI = cast<FrameIndexSDNode>(FIPtr)->getIndex(); + MachinePointerInfo MPI = MachinePointerInfo::getFixedStack(FI); // Emit a store to the stack slot. - SDValue Chain = DAG.getStore(DAG.getEntryNode(), dl, Tmp, FIPtr, - MachinePointerInfo(), false, false, 0); + SDValue Chain; + if (i32Stack) { + MachineFunction &MF = DAG.getMachineFunction(); + MachineMemOperand *MMO = + MF.getMachineMemOperand(MPI, MachineMemOperand::MOStore, 4, 4); + SDValue Ops[] = { DAG.getEntryNode(), Tmp, FIPtr }; + Chain = DAG.getMemIntrinsicNode(PPCISD::STFIWX, dl, + DAG.getVTList(MVT::Other), Ops, array_lengthof(Ops), + MVT::i32, MMO); + } else + Chain = DAG.getStore(DAG.getEntryNode(), dl, Tmp, FIPtr, + MPI, false, false, 0); // Result is a load from the stack slot. If loading 4 bytes, make sure to // add in a bias. - if (Op.getValueType() == MVT::i32) + if (Op.getValueType() == MVT::i32 && !i32Stack) { FIPtr = DAG.getNode(ISD::ADD, dl, FIPtr.getValueType(), FIPtr, DAG.getConstant(4, FIPtr.getValueType())); - return DAG.getLoad(Op.getValueType(), dl, Chain, FIPtr, MachinePointerInfo(), + MPI = MachinePointerInfo(); + } + + return DAG.getLoad(Op.getValueType(), dl, Chain, FIPtr, MPI, false, false, false, 0); } -SDValue PPCTargetLowering::LowerSINT_TO_FP(SDValue Op, +SDValue PPCTargetLowering::LowerINT_TO_FP(SDValue Op, SelectionDAG &DAG) const { DebugLoc dl = Op.getDebugLoc(); // Don't handle ppc_fp128 here; let it be lowered to a libcall. if (Op.getValueType() != MVT::f32 && Op.getValueType() != MVT::f64) return SDValue(); + assert((Op.getOpcode() == ISD::SINT_TO_FP || PPCSubTarget.hasFPCVT()) && + "UINT_TO_FP is supported only with FPCVT"); + + // If we have FCFIDS, then use it when converting to single-precision. + // Otherwise, convert to double-precision and then round. + unsigned FCFOp = (PPCSubTarget.hasFPCVT() && Op.getValueType() == MVT::f32) ? + (Op.getOpcode() == ISD::UINT_TO_FP ? + PPCISD::FCFIDUS : PPCISD::FCFIDS) : + (Op.getOpcode() == ISD::UINT_TO_FP ? + PPCISD::FCFIDU : PPCISD::FCFID); + MVT FCFTy = (PPCSubTarget.hasFPCVT() && Op.getValueType() == MVT::f32) ? + MVT::f32 : MVT::f64; + if (Op.getOperand(0).getValueType() == MVT::i64) { SDValue SINT = Op.getOperand(0); // When converting to single-precision, we actually need to convert @@ -4695,6 +4858,7 @@ SDValue PPCTargetLowering::LowerSINT_TO_FP(SDValue Op, // However, if -enable-unsafe-fp-math is in effect, accept double // rounding to avoid the extra overhead. if (Op.getValueType() == MVT::f32 && + !PPCSubTarget.hasFPCVT() && !DAG.getTarget().Options.UnsafeFPMath) { // Twiddle input to make sure the low 11 bits are zero. (If this @@ -4728,44 +4892,69 @@ SDValue PPCTargetLowering::LowerSINT_TO_FP(SDValue Op, SINT = DAG.getNode(ISD::SELECT, dl, MVT::i64, Cond, Round, SINT); } + SDValue Bits = DAG.getNode(ISD::BITCAST, dl, MVT::f64, SINT); - SDValue FP = DAG.getNode(PPCISD::FCFID, dl, MVT::f64, Bits); - if (Op.getValueType() == MVT::f32) + SDValue FP = DAG.getNode(FCFOp, dl, FCFTy, Bits); + + if (Op.getValueType() == MVT::f32 && !PPCSubTarget.hasFPCVT()) FP = DAG.getNode(ISD::FP_ROUND, dl, MVT::f32, FP, DAG.getIntPtrConstant(0)); return FP; } assert(Op.getOperand(0).getValueType() == MVT::i32 && - "Unhandled SINT_TO_FP type in custom expander!"); + "Unhandled INT_TO_FP type in custom expander!"); // Since we only generate this in 64-bit mode, we can take advantage of // 64-bit registers. In particular, sign extend the input value into the // 64-bit register with extsw, store the WHOLE 64-bit value into the stack // then lfd it and fcfid it. MachineFunction &MF = DAG.getMachineFunction(); MachineFrameInfo *FrameInfo = MF.getFrameInfo(); - int FrameIdx = FrameInfo->CreateStackObject(8, 8, false); EVT PtrVT = DAG.getTargetLoweringInfo().getPointerTy(); - SDValue FIdx = DAG.getFrameIndex(FrameIdx, PtrVT); - SDValue Ext64 = DAG.getNode(PPCISD::EXTSW_32, dl, MVT::i32, + SDValue Ld; + if (PPCSubTarget.hasLFIWAX() || PPCSubTarget.hasFPCVT()) { + int FrameIdx = FrameInfo->CreateStackObject(4, 4, false); + SDValue FIdx = DAG.getFrameIndex(FrameIdx, PtrVT); + + SDValue Store = DAG.getStore(DAG.getEntryNode(), dl, Op.getOperand(0), FIdx, + MachinePointerInfo::getFixedStack(FrameIdx), + false, false, 0); + + assert(cast<StoreSDNode>(Store)->getMemoryVT() == MVT::i32 && + "Expected an i32 store"); + MachineMemOperand *MMO = + MF.getMachineMemOperand(MachinePointerInfo::getFixedStack(FrameIdx), + MachineMemOperand::MOLoad, 4, 4); + SDValue Ops[] = { Store, FIdx }; + Ld = DAG.getMemIntrinsicNode(Op.getOpcode() == ISD::UINT_TO_FP ? + PPCISD::LFIWZX : PPCISD::LFIWAX, + dl, DAG.getVTList(MVT::f64, MVT::Other), + Ops, 2, MVT::i32, MMO); + } else { + assert(PPCSubTarget.isPPC64() && + "i32->FP without LFIWAX supported only on PPC64"); + + int FrameIdx = FrameInfo->CreateStackObject(8, 8, false); + SDValue FIdx = DAG.getFrameIndex(FrameIdx, PtrVT); + + SDValue Ext64 = DAG.getNode(ISD::SIGN_EXTEND, dl, MVT::i64, Op.getOperand(0)); - // STD the extended value into the stack slot. - MachineMemOperand *MMO = - MF.getMachineMemOperand(MachinePointerInfo::getFixedStack(FrameIdx), - MachineMemOperand::MOStore, 8, 8); - SDValue Ops[] = { DAG.getEntryNode(), Ext64, FIdx }; - SDValue Store = - DAG.getMemIntrinsicNode(PPCISD::STD_32, dl, DAG.getVTList(MVT::Other), - Ops, 4, MVT::i64, MMO); - // Load the value as a double. - SDValue Ld = DAG.getLoad(MVT::f64, dl, Store, FIdx, MachinePointerInfo(), - false, false, false, 0); + // STD the extended value into the stack slot. + SDValue Store = DAG.getStore(DAG.getEntryNode(), dl, Ext64, FIdx, + MachinePointerInfo::getFixedStack(FrameIdx), + false, false, 0); + + // Load the value as a double. + Ld = DAG.getLoad(MVT::f64, dl, Store, FIdx, + MachinePointerInfo::getFixedStack(FrameIdx), + false, false, false, 0); + } // FCFID it and return it. - SDValue FP = DAG.getNode(PPCISD::FCFID, dl, MVT::f64, Ld); - if (Op.getValueType() == MVT::f32) + SDValue FP = DAG.getNode(FCFOp, dl, FCFTy, Ld); + if (Op.getValueType() == MVT::f32 && !PPCSubTarget.hasFPCVT()) FP = DAG.getNode(ISD::FP_ROUND, dl, MVT::f32, FP, DAG.getIntPtrConstant(0)); return FP; } @@ -5560,11 +5749,15 @@ SDValue PPCTargetLowering::LowerOperation(SDValue Op, SelectionDAG &DAG) const { case ISD::DYNAMIC_STACKALLOC: return LowerDYNAMIC_STACKALLOC(Op, DAG, PPCSubTarget); + case ISD::EH_SJLJ_SETJMP: return lowerEH_SJLJ_SETJMP(Op, DAG); + case ISD::EH_SJLJ_LONGJMP: return lowerEH_SJLJ_LONGJMP(Op, DAG); + case ISD::SELECT_CC: return LowerSELECT_CC(Op, DAG); case ISD::FP_TO_UINT: case ISD::FP_TO_SINT: return LowerFP_TO_INT(Op, DAG, Op.getDebugLoc()); - case ISD::SINT_TO_FP: return LowerSINT_TO_FP(Op, DAG); + case ISD::UINT_TO_FP: + case ISD::SINT_TO_FP: return LowerINT_TO_FP(Op, DAG); case ISD::FLT_ROUNDS_: return LowerFLT_ROUNDS_(Op, DAG); // Lower 64-bit shifts. @@ -5618,50 +5811,8 @@ void PPCTargetLowering::ReplaceNodeResults(SDNode *N, MVT::f64, N->getOperand(0), DAG.getIntPtrConstant(1)); - // This sequence changes FPSCR to do round-to-zero, adds the two halves - // of the long double, and puts FPSCR back the way it was. We do not - // actually model FPSCR. - std::vector<EVT> NodeTys; - SDValue Ops[4], Result, MFFSreg, InFlag, FPreg; - - NodeTys.push_back(MVT::f64); // Return register - NodeTys.push_back(MVT::Glue); // Returns a flag for later insns - Result = DAG.getNode(PPCISD::MFFS, dl, NodeTys, &InFlag, 0); - MFFSreg = Result.getValue(0); - InFlag = Result.getValue(1); - - NodeTys.clear(); - NodeTys.push_back(MVT::Glue); // Returns a flag - Ops[0] = DAG.getConstant(31, MVT::i32); - Ops[1] = InFlag; - Result = DAG.getNode(PPCISD::MTFSB1, dl, NodeTys, Ops, 2); - InFlag = Result.getValue(0); - - NodeTys.clear(); - NodeTys.push_back(MVT::Glue); // Returns a flag - Ops[0] = DAG.getConstant(30, MVT::i32); - Ops[1] = InFlag; - Result = DAG.getNode(PPCISD::MTFSB0, dl, NodeTys, Ops, 2); - InFlag = Result.getValue(0); - - NodeTys.clear(); - NodeTys.push_back(MVT::f64); // result of add - NodeTys.push_back(MVT::Glue); // Returns a flag - Ops[0] = Lo; - Ops[1] = Hi; - Ops[2] = InFlag; - Result = DAG.getNode(PPCISD::FADDRTZ, dl, NodeTys, Ops, 3); - FPreg = Result.getValue(0); - InFlag = Result.getValue(1); - - NodeTys.clear(); - NodeTys.push_back(MVT::f64); - Ops[0] = DAG.getConstant(1, MVT::i32); - Ops[1] = MFFSreg; - Ops[2] = FPreg; - Ops[3] = InFlag; - Result = DAG.getNode(PPCISD::MTFSF, dl, NodeTys, Ops, 4); - FPreg = Result.getValue(0); + // Add the two halves of the long double in round-to-zero mode. + SDValue FPreg = DAG.getNode(PPCISD::FADDRTZ, dl, MVT::f64, Lo, Hi); // We know the low half is about to be thrown away, so just use something // convenient. @@ -5753,7 +5904,7 @@ PPCTargetLowering::EmitPartwordAtomicBinary(MachineInstr *MI, // registers without caring whether they're 32 or 64, but here we're // doing actual arithmetic on the addresses. bool is64bit = PPCSubTarget.isPPC64(); - unsigned ZeroReg = is64bit ? PPC::X0 : PPC::R0; + unsigned ZeroReg = is64bit ? PPC::ZERO8 : PPC::ZERO; const BasicBlock *LLVM_BB = BB->getBasicBlock(); MachineFunction *F = BB->getParent(); @@ -5857,7 +6008,7 @@ PPCTargetLowering::EmitPartwordAtomicBinary(MachineInstr *MI, .addReg(TmpReg).addReg(MaskReg); BuildMI(BB, dl, TII->get(is64bit ? PPC::OR8 : PPC::OR), Tmp4Reg) .addReg(Tmp3Reg).addReg(Tmp2Reg); - BuildMI(BB, dl, TII->get(is64bit ? PPC::STDCX : PPC::STWCX)) + BuildMI(BB, dl, TII->get(PPC::STWCX)) .addReg(Tmp4Reg).addReg(ZeroReg).addReg(PtrReg); BuildMI(BB, dl, TII->get(PPC::BCC)) .addImm(PPC::PRED_NE).addReg(PPC::CR0).addMBB(loopMBB); @@ -5872,9 +6023,238 @@ PPCTargetLowering::EmitPartwordAtomicBinary(MachineInstr *MI, return BB; } +llvm::MachineBasicBlock* +PPCTargetLowering::emitEHSjLjSetJmp(MachineInstr *MI, + MachineBasicBlock *MBB) const { + DebugLoc DL = MI->getDebugLoc(); + const TargetInstrInfo *TII = getTargetMachine().getInstrInfo(); + + MachineFunction *MF = MBB->getParent(); + MachineRegisterInfo &MRI = MF->getRegInfo(); + + const BasicBlock *BB = MBB->getBasicBlock(); + MachineFunction::iterator I = MBB; + ++I; + + // Memory Reference + MachineInstr::mmo_iterator MMOBegin = MI->memoperands_begin(); + MachineInstr::mmo_iterator MMOEnd = MI->memoperands_end(); + + unsigned DstReg = MI->getOperand(0).getReg(); + const TargetRegisterClass *RC = MRI.getRegClass(DstReg); + assert(RC->hasType(MVT::i32) && "Invalid destination!"); + unsigned mainDstReg = MRI.createVirtualRegister(RC); + unsigned restoreDstReg = MRI.createVirtualRegister(RC); + + MVT PVT = getPointerTy(); + assert((PVT == MVT::i64 || PVT == MVT::i32) && + "Invalid Pointer Size!"); + // For v = setjmp(buf), we generate + // + // thisMBB: + // SjLjSetup mainMBB + // bl mainMBB + // v_restore = 1 + // b sinkMBB + // + // mainMBB: + // buf[LabelOffset] = LR + // v_main = 0 + // + // sinkMBB: + // v = phi(main, restore) + // + + MachineBasicBlock *thisMBB = MBB; + MachineBasicBlock *mainMBB = MF->CreateMachineBasicBlock(BB); + MachineBasicBlock *sinkMBB = MF->CreateMachineBasicBlock(BB); + MF->insert(I, mainMBB); + MF->insert(I, sinkMBB); + + MachineInstrBuilder MIB; + + // Transfer the remainder of BB and its successor edges to sinkMBB. + sinkMBB->splice(sinkMBB->begin(), MBB, + llvm::next(MachineBasicBlock::iterator(MI)), MBB->end()); + sinkMBB->transferSuccessorsAndUpdatePHIs(MBB); + + // Note that the structure of the jmp_buf used here is not compatible + // with that used by libc, and is not designed to be. Specifically, it + // stores only those 'reserved' registers that LLVM does not otherwise + // understand how to spill. Also, by convention, by the time this + // intrinsic is called, Clang has already stored the frame address in the + // first slot of the buffer and stack address in the third. Following the + // X86 target code, we'll store the jump address in the second slot. We also + // need to save the TOC pointer (R2) to handle jumps between shared + // libraries, and that will be stored in the fourth slot. The thread + // identifier (R13) is not affected. + + // thisMBB: + const int64_t LabelOffset = 1 * PVT.getStoreSize(); + const int64_t TOCOffset = 3 * PVT.getStoreSize(); + + // Prepare IP either in reg. + const TargetRegisterClass *PtrRC = getRegClassFor(PVT); + unsigned LabelReg = MRI.createVirtualRegister(PtrRC); + unsigned BufReg = MI->getOperand(1).getReg(); + + if (PPCSubTarget.isPPC64() && PPCSubTarget.isSVR4ABI()) { + MIB = BuildMI(*thisMBB, MI, DL, TII->get(PPC::STD)) + .addReg(PPC::X2) + .addImm(TOCOffset / 4) + .addReg(BufReg); + + MIB.setMemRefs(MMOBegin, MMOEnd); + } + + // Setup + MIB = BuildMI(*thisMBB, MI, DL, TII->get(PPC::BCLalways)).addMBB(mainMBB); + MIB.addRegMask(PPCRegInfo->getNoPreservedMask()); + + BuildMI(*thisMBB, MI, DL, TII->get(PPC::LI), restoreDstReg).addImm(1); + + MIB = BuildMI(*thisMBB, MI, DL, TII->get(PPC::EH_SjLj_Setup)) + .addMBB(mainMBB); + MIB = BuildMI(*thisMBB, MI, DL, TII->get(PPC::B)).addMBB(sinkMBB); + + thisMBB->addSuccessor(mainMBB, /* weight */ 0); + thisMBB->addSuccessor(sinkMBB, /* weight */ 1); + + // mainMBB: + // mainDstReg = 0 + MIB = BuildMI(mainMBB, DL, + TII->get(PPCSubTarget.isPPC64() ? PPC::MFLR8 : PPC::MFLR), LabelReg); + + // Store IP + if (PPCSubTarget.isPPC64()) { + MIB = BuildMI(mainMBB, DL, TII->get(PPC::STD)) + .addReg(LabelReg) + .addImm(LabelOffset / 4) + .addReg(BufReg); + } else { + MIB = BuildMI(mainMBB, DL, TII->get(PPC::STW)) + .addReg(LabelReg) + .addImm(LabelOffset) + .addReg(BufReg); + } + + MIB.setMemRefs(MMOBegin, MMOEnd); + + BuildMI(mainMBB, DL, TII->get(PPC::LI), mainDstReg).addImm(0); + mainMBB->addSuccessor(sinkMBB); + + // sinkMBB: + BuildMI(*sinkMBB, sinkMBB->begin(), DL, + TII->get(PPC::PHI), DstReg) + .addReg(mainDstReg).addMBB(mainMBB) + .addReg(restoreDstReg).addMBB(thisMBB); + + MI->eraseFromParent(); + return sinkMBB; +} + +MachineBasicBlock * +PPCTargetLowering::emitEHSjLjLongJmp(MachineInstr *MI, + MachineBasicBlock *MBB) const { + DebugLoc DL = MI->getDebugLoc(); + const TargetInstrInfo *TII = getTargetMachine().getInstrInfo(); + + MachineFunction *MF = MBB->getParent(); + MachineRegisterInfo &MRI = MF->getRegInfo(); + + // Memory Reference + MachineInstr::mmo_iterator MMOBegin = MI->memoperands_begin(); + MachineInstr::mmo_iterator MMOEnd = MI->memoperands_end(); + + MVT PVT = getPointerTy(); + assert((PVT == MVT::i64 || PVT == MVT::i32) && + "Invalid Pointer Size!"); + + const TargetRegisterClass *RC = + (PVT == MVT::i64) ? &PPC::G8RCRegClass : &PPC::GPRCRegClass; + unsigned Tmp = MRI.createVirtualRegister(RC); + // Since FP is only updated here but NOT referenced, it's treated as GPR. + unsigned FP = (PVT == MVT::i64) ? PPC::X31 : PPC::R31; + unsigned SP = (PVT == MVT::i64) ? PPC::X1 : PPC::R1; + + MachineInstrBuilder MIB; + + const int64_t LabelOffset = 1 * PVT.getStoreSize(); + const int64_t SPOffset = 2 * PVT.getStoreSize(); + const int64_t TOCOffset = 3 * PVT.getStoreSize(); + + unsigned BufReg = MI->getOperand(0).getReg(); + + // Reload FP (the jumped-to function may not have had a + // frame pointer, and if so, then its r31 will be restored + // as necessary). + if (PVT == MVT::i64) { + MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LD), FP) + .addImm(0) + .addReg(BufReg); + } else { + MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LWZ), FP) + .addImm(0) + .addReg(BufReg); + } + MIB.setMemRefs(MMOBegin, MMOEnd); + + // Reload IP + if (PVT == MVT::i64) { + MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LD), Tmp) + .addImm(LabelOffset / 4) + .addReg(BufReg); + } else { + MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LWZ), Tmp) + .addImm(LabelOffset) + .addReg(BufReg); + } + MIB.setMemRefs(MMOBegin, MMOEnd); + + // Reload SP + if (PVT == MVT::i64) { + MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LD), SP) + .addImm(SPOffset / 4) + .addReg(BufReg); + } else { + MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LWZ), SP) + .addImm(SPOffset) + .addReg(BufReg); + } + MIB.setMemRefs(MMOBegin, MMOEnd); + + // FIXME: When we also support base pointers, that register must also be + // restored here. + + // Reload TOC + if (PVT == MVT::i64 && PPCSubTarget.isSVR4ABI()) { + MIB = BuildMI(*MBB, MI, DL, TII->get(PPC::LD), PPC::X2) + .addImm(TOCOffset / 4) + .addReg(BufReg); + + MIB.setMemRefs(MMOBegin, MMOEnd); + } + + // Jump + BuildMI(*MBB, MI, DL, + TII->get(PVT == MVT::i64 ? PPC::MTCTR8 : PPC::MTCTR)).addReg(Tmp); + BuildMI(*MBB, MI, DL, TII->get(PVT == MVT::i64 ? PPC::BCTR8 : PPC::BCTR)); + + MI->eraseFromParent(); + return MBB; +} + MachineBasicBlock * PPCTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI, MachineBasicBlock *BB) const { + if (MI->getOpcode() == PPC::EH_SjLj_SetJmp32 || + MI->getOpcode() == PPC::EH_SjLj_SetJmp64) { + return emitEHSjLjSetJmp(MI, BB); + } else if (MI->getOpcode() == PPC::EH_SjLj_LongJmp32 || + MI->getOpcode() == PPC::EH_SjLj_LongJmp64) { + return emitEHSjLjLongJmp(MI, BB); + } + const TargetInstrInfo *TII = getTargetMachine().getInstrInfo(); // To "insert" these instructions we actually have to insert their @@ -5887,29 +6267,13 @@ PPCTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI, if (PPCSubTarget.hasISEL() && (MI->getOpcode() == PPC::SELECT_CC_I4 || MI->getOpcode() == PPC::SELECT_CC_I8)) { - unsigned OpCode = MI->getOpcode() == PPC::SELECT_CC_I8 ? - PPC::ISEL8 : PPC::ISEL; - unsigned SelectPred = MI->getOperand(4).getImm(); - DebugLoc dl = MI->getDebugLoc(); + SmallVector<MachineOperand, 2> Cond; + Cond.push_back(MI->getOperand(4)); + Cond.push_back(MI->getOperand(1)); - // The SelectPred is ((BI << 5) | BO) for a BCC - unsigned BO = SelectPred & 0xF; - assert((BO == 12 || BO == 4) && "invalid predicate BO field for isel"); - - unsigned TrueOpNo, FalseOpNo; - if (BO == 12) { - TrueOpNo = 2; - FalseOpNo = 3; - } else { - TrueOpNo = 3; - FalseOpNo = 2; - SelectPred = PPC::InvertPredicate((PPC::Predicate)SelectPred); - } - - BuildMI(*BB, MI, dl, TII->get(OpCode), MI->getOperand(0).getReg()) - .addReg(MI->getOperand(TrueOpNo).getReg()) - .addReg(MI->getOperand(FalseOpNo).getReg()) - .addImm(SelectPred).addReg(MI->getOperand(1).getReg()); + DebugLoc dl = MI->getDebugLoc(); + PPCII->insertSelect(*BB, MI, dl, MI->getOperand(0).getReg(), Cond, + MI->getOperand(2).getReg(), MI->getOperand(3).getReg()); } else if (MI->getOpcode() == PPC::SELECT_CC_I4 || MI->getOpcode() == PPC::SELECT_CC_I8 || MI->getOpcode() == PPC::SELECT_CC_F4 || @@ -6142,7 +6506,7 @@ PPCTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI, unsigned TmpDestReg = RegInfo.createVirtualRegister(RC); unsigned Ptr1Reg; unsigned TmpReg = RegInfo.createVirtualRegister(RC); - unsigned ZeroReg = is64bit ? PPC::X0 : PPC::R0; + unsigned ZeroReg = is64bit ? PPC::ZERO8 : PPC::ZERO; // thisMBB: // ... // fallthrough --> loopMBB @@ -6245,6 +6609,75 @@ PPCTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI, BB = exitMBB; BuildMI(*BB, BB->begin(), dl, TII->get(PPC::SRW),dest).addReg(TmpReg) .addReg(ShiftReg); + } else if (MI->getOpcode() == PPC::FADDrtz) { + // This pseudo performs an FADD with rounding mode temporarily forced + // to round-to-zero. We emit this via custom inserter since the FPSCR + // is not modeled at the SelectionDAG level. + unsigned Dest = MI->getOperand(0).getReg(); + unsigned Src1 = MI->getOperand(1).getReg(); + unsigned Src2 = MI->getOperand(2).getReg(); + DebugLoc dl = MI->getDebugLoc(); + + MachineRegisterInfo &RegInfo = F->getRegInfo(); + unsigned MFFSReg = RegInfo.createVirtualRegister(&PPC::F8RCRegClass); + + // Save FPSCR value. + BuildMI(*BB, MI, dl, TII->get(PPC::MFFS), MFFSReg); + + // Set rounding mode to round-to-zero. + BuildMI(*BB, MI, dl, TII->get(PPC::MTFSB1)).addImm(31); + BuildMI(*BB, MI, dl, TII->get(PPC::MTFSB0)).addImm(30); + + // Perform addition. + BuildMI(*BB, MI, dl, TII->get(PPC::FADD), Dest).addReg(Src1).addReg(Src2); + + // Restore FPSCR value. + BuildMI(*BB, MI, dl, TII->get(PPC::MTFSF)).addImm(1).addReg(MFFSReg); + } else if (MI->getOpcode() == PPC::FRINDrint || + MI->getOpcode() == PPC::FRINSrint) { + bool isf32 = MI->getOpcode() == PPC::FRINSrint; + unsigned Dest = MI->getOperand(0).getReg(); + unsigned Src = MI->getOperand(1).getReg(); + DebugLoc dl = MI->getDebugLoc(); + + MachineRegisterInfo &RegInfo = F->getRegInfo(); + unsigned CRReg = RegInfo.createVirtualRegister(&PPC::CRRCRegClass); + + // Perform the rounding. + BuildMI(*BB, MI, dl, TII->get(isf32 ? PPC::FRINS : PPC::FRIND), Dest) + .addReg(Src); + + // Compare the results. + BuildMI(*BB, MI, dl, TII->get(isf32 ? PPC::FCMPUS : PPC::FCMPUD), CRReg) + .addReg(Dest).addReg(Src); + + // If the results were not equal, then set the FPSCR XX bit. + MachineBasicBlock *midMBB = F->CreateMachineBasicBlock(LLVM_BB); + MachineBasicBlock *exitMBB = F->CreateMachineBasicBlock(LLVM_BB); + F->insert(It, midMBB); + F->insert(It, exitMBB); + exitMBB->splice(exitMBB->begin(), BB, + llvm::next(MachineBasicBlock::iterator(MI)), + BB->end()); + exitMBB->transferSuccessorsAndUpdatePHIs(BB); + + BuildMI(*BB, MI, dl, TII->get(PPC::BCC)) + .addImm(PPC::PRED_EQ).addReg(CRReg).addMBB(exitMBB); + + BB->addSuccessor(midMBB); + BB->addSuccessor(exitMBB); + + BB = midMBB; + + // Set the FPSCR XX bit (FE_INEXACT). Note that we cannot just set + // the FI bit here because that will not automatically set XX also, + // and XX is what libm interprets as the FE_INEXACT flag. + BuildMI(BB, dl, TII->get(PPC::MTFSB1)).addImm(/* 38 - 32 = */ 6); + BuildMI(BB, dl, TII->get(PPC::B)).addMBB(exitMBB); + + BB->addSuccessor(exitMBB); + + BB = exitMBB; } else { llvm_unreachable("Unexpected instr type to insert"); } @@ -6257,6 +6690,139 @@ PPCTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI, // Target Optimization Hooks //===----------------------------------------------------------------------===// +SDValue PPCTargetLowering::DAGCombineFastRecip(SDValue Op, + DAGCombinerInfo &DCI) const { + if (DCI.isAfterLegalizeVectorOps()) + return SDValue(); + + EVT VT = Op.getValueType(); + + if ((VT == MVT::f32 && PPCSubTarget.hasFRES()) || + (VT == MVT::f64 && PPCSubTarget.hasFRE()) || + (VT == MVT::v4f32 && PPCSubTarget.hasAltivec())) { + + // Newton iteration for a function: F(X) is X_{i+1} = X_i - F(X_i)/F'(X_i) + // For the reciprocal, we need to find the zero of the function: + // F(X) = A X - 1 [which has a zero at X = 1/A] + // => + // X_{i+1} = X_i (2 - A X_i) = X_i + X_i (1 - A X_i) [this second form + // does not require additional intermediate precision] + + // Convergence is quadratic, so we essentially double the number of digits + // correct after every iteration. The minimum architected relative + // accuracy is 2^-5. When hasRecipPrec(), this is 2^-14. IEEE float has + // 23 digits and double has 52 digits. + int Iterations = PPCSubTarget.hasRecipPrec() ? 1 : 3; + if (VT.getScalarType() == MVT::f64) + ++Iterations; + + SelectionDAG &DAG = DCI.DAG; + DebugLoc dl = Op.getDebugLoc(); + + SDValue FPOne = + DAG.getConstantFP(1.0, VT.getScalarType()); + if (VT.isVector()) { + assert(VT.getVectorNumElements() == 4 && + "Unknown vector type"); + FPOne = DAG.getNode(ISD::BUILD_VECTOR, dl, VT, + FPOne, FPOne, FPOne, FPOne); + } + + SDValue Est = DAG.getNode(PPCISD::FRE, dl, VT, Op); + DCI.AddToWorklist(Est.getNode()); + + // Newton iterations: Est = Est + Est (1 - Arg * Est) + for (int i = 0; i < Iterations; ++i) { + SDValue NewEst = DAG.getNode(ISD::FMUL, dl, VT, Op, Est); + DCI.AddToWorklist(NewEst.getNode()); + + NewEst = DAG.getNode(ISD::FSUB, dl, VT, FPOne, NewEst); + DCI.AddToWorklist(NewEst.getNode()); + + NewEst = DAG.getNode(ISD::FMUL, dl, VT, Est, NewEst); + DCI.AddToWorklist(NewEst.getNode()); + + Est = DAG.getNode(ISD::FADD, dl, VT, Est, NewEst); + DCI.AddToWorklist(Est.getNode()); + } + + return Est; + } + + return SDValue(); +} + +SDValue PPCTargetLowering::DAGCombineFastRecipFSQRT(SDValue Op, + DAGCombinerInfo &DCI) const { + if (DCI.isAfterLegalizeVectorOps()) + return SDValue(); + + EVT VT = Op.getValueType(); + + if ((VT == MVT::f32 && PPCSubTarget.hasFRSQRTES()) || + (VT == MVT::f64 && PPCSubTarget.hasFRSQRTE()) || + (VT == MVT::v4f32 && PPCSubTarget.hasAltivec())) { + + // Newton iteration for a function: F(X) is X_{i+1} = X_i - F(X_i)/F'(X_i) + // For the reciprocal sqrt, we need to find the zero of the function: + // F(X) = 1/X^2 - A [which has a zero at X = 1/sqrt(A)] + // => + // X_{i+1} = X_i (1.5 - A X_i^2 / 2) + // As a result, we precompute A/2 prior to the iteration loop. + + // Convergence is quadratic, so we essentially double the number of digits + // correct after every iteration. The minimum architected relative + // accuracy is 2^-5. When hasRecipPrec(), this is 2^-14. IEEE float has + // 23 digits and double has 52 digits. + int Iterations = PPCSubTarget.hasRecipPrec() ? 1 : 3; + if (VT.getScalarType() == MVT::f64) + ++Iterations; + + SelectionDAG &DAG = DCI.DAG; + DebugLoc dl = Op.getDebugLoc(); + + SDValue FPThreeHalves = + DAG.getConstantFP(1.5, VT.getScalarType()); + if (VT.isVector()) { + assert(VT.getVectorNumElements() == 4 && + "Unknown vector type"); + FPThreeHalves = DAG.getNode(ISD::BUILD_VECTOR, dl, VT, + FPThreeHalves, FPThreeHalves, + FPThreeHalves, FPThreeHalves); + } + + SDValue Est = DAG.getNode(PPCISD::FRSQRTE, dl, VT, Op); + DCI.AddToWorklist(Est.getNode()); + + // We now need 0.5*Arg which we can write as (1.5*Arg - Arg) so that + // this entire sequence requires only one FP constant. + SDValue HalfArg = DAG.getNode(ISD::FMUL, dl, VT, FPThreeHalves, Op); + DCI.AddToWorklist(HalfArg.getNode()); + + HalfArg = DAG.getNode(ISD::FSUB, dl, VT, HalfArg, Op); + DCI.AddToWorklist(HalfArg.getNode()); + + // Newton iterations: Est = Est * (1.5 - HalfArg * Est * Est) + for (int i = 0; i < Iterations; ++i) { + SDValue NewEst = DAG.getNode(ISD::FMUL, dl, VT, Est, Est); + DCI.AddToWorklist(NewEst.getNode()); + + NewEst = DAG.getNode(ISD::FMUL, dl, VT, HalfArg, NewEst); + DCI.AddToWorklist(NewEst.getNode()); + + NewEst = DAG.getNode(ISD::FSUB, dl, VT, FPThreeHalves, NewEst); + DCI.AddToWorklist(NewEst.getNode()); + + Est = DAG.getNode(ISD::FMUL, dl, VT, Est, NewEst); + DCI.AddToWorklist(Est.getNode()); + } + + return Est; + } + + return SDValue(); +} + SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const { const TargetMachine &TM = getTargetMachine(); @@ -6283,7 +6849,72 @@ SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N, return N->getOperand(0); } break; + case ISD::FDIV: { + assert(TM.Options.UnsafeFPMath && + "Reciprocal estimates require UnsafeFPMath"); + + if (N->getOperand(1).getOpcode() == ISD::FSQRT) { + SDValue RV = + DAGCombineFastRecipFSQRT(N->getOperand(1).getOperand(0), DCI); + if (RV.getNode() != 0) { + DCI.AddToWorklist(RV.getNode()); + return DAG.getNode(ISD::FMUL, dl, N->getValueType(0), + N->getOperand(0), RV); + } + } else if (N->getOperand(1).getOpcode() == ISD::FP_EXTEND && + N->getOperand(1).getOperand(0).getOpcode() == ISD::FSQRT) { + SDValue RV = + DAGCombineFastRecipFSQRT(N->getOperand(1).getOperand(0).getOperand(0), + DCI); + if (RV.getNode() != 0) { + DCI.AddToWorklist(RV.getNode()); + RV = DAG.getNode(ISD::FP_EXTEND, N->getOperand(1).getDebugLoc(), + N->getValueType(0), RV); + DCI.AddToWorklist(RV.getNode()); + return DAG.getNode(ISD::FMUL, dl, N->getValueType(0), + N->getOperand(0), RV); + } + } else if (N->getOperand(1).getOpcode() == ISD::FP_ROUND && + N->getOperand(1).getOperand(0).getOpcode() == ISD::FSQRT) { + SDValue RV = + DAGCombineFastRecipFSQRT(N->getOperand(1).getOperand(0).getOperand(0), + DCI); + if (RV.getNode() != 0) { + DCI.AddToWorklist(RV.getNode()); + RV = DAG.getNode(ISD::FP_ROUND, N->getOperand(1).getDebugLoc(), + N->getValueType(0), RV, + N->getOperand(1).getOperand(1)); + DCI.AddToWorklist(RV.getNode()); + return DAG.getNode(ISD::FMUL, dl, N->getValueType(0), + N->getOperand(0), RV); + } + } + SDValue RV = DAGCombineFastRecip(N->getOperand(1), DCI); + if (RV.getNode() != 0) { + DCI.AddToWorklist(RV.getNode()); + return DAG.getNode(ISD::FMUL, dl, N->getValueType(0), + N->getOperand(0), RV); + } + + } + break; + case ISD::FSQRT: { + assert(TM.Options.UnsafeFPMath && + "Reciprocal estimates require UnsafeFPMath"); + + // Compute this as 1/(1/sqrt(X)), which is the reciprocal of the + // reciprocal sqrt. + SDValue RV = DAGCombineFastRecipFSQRT(N->getOperand(0), DCI); + if (RV.getNode() != 0) { + DCI.AddToWorklist(RV.getNode()); + RV = DAGCombineFastRecip(RV, DCI); + if (RV.getNode() != 0) + return RV; + } + + } + break; case ISD::SINT_TO_FP: if (TM.getSubtarget<PPCSubtarget>().has64BitSupport()) { if (N->getOperand(0).getOpcode() == ISD::FP_TO_SINT) { @@ -6330,8 +6961,15 @@ SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N, Val = DAG.getNode(PPCISD::FCTIWZ, dl, MVT::f64, Val); DCI.AddToWorklist(Val.getNode()); - Val = DAG.getNode(PPCISD::STFIWX, dl, MVT::Other, N->getOperand(0), Val, - N->getOperand(2), N->getOperand(3)); + SDValue Ops[] = { + N->getOperand(0), Val, N->getOperand(2), + DAG.getValueType(N->getOperand(1).getValueType()) + }; + + Val = DAG.getMemIntrinsicNode(PPCISD::STFIWX, dl, + DAG.getVTList(MVT::Other), Ops, array_lengthof(Ops), + cast<StoreSDNode>(N)->getMemoryVT(), + cast<StoreSDNode>(N)->getMemOperand()); DCI.AddToWorklist(Val.getNode()); return Val; } @@ -6341,7 +6979,10 @@ SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N, N->getOperand(1).getOpcode() == ISD::BSWAP && N->getOperand(1).getNode()->hasOneUse() && (N->getOperand(1).getValueType() == MVT::i32 || - N->getOperand(1).getValueType() == MVT::i16)) { + N->getOperand(1).getValueType() == MVT::i16 || + (TM.getSubtarget<PPCSubtarget>().hasLDBRX() && + TM.getSubtarget<PPCSubtarget>().isPPC64() && + N->getOperand(1).getValueType() == MVT::i64))) { SDValue BSwapOp = N->getOperand(1).getOperand(0); // Do an any-extend to 32-bits if this is a half-word input. if (BSwapOp.getValueType() == MVT::i16) @@ -6362,7 +7003,10 @@ SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N, // Turn BSWAP (LOAD) -> lhbrx/lwbrx. if (ISD::isNON_EXTLoad(N->getOperand(0).getNode()) && N->getOperand(0).hasOneUse() && - (N->getValueType(0) == MVT::i32 || N->getValueType(0) == MVT::i16)) { + (N->getValueType(0) == MVT::i32 || N->getValueType(0) == MVT::i16 || + (TM.getSubtarget<PPCSubtarget>().hasLDBRX() && + TM.getSubtarget<PPCSubtarget>().isPPC64() && + N->getValueType(0) == MVT::i64))) { SDValue Load = N->getOperand(0); LoadSDNode *LD = cast<LoadSDNode>(Load); // Create the byte-swapping load. @@ -6373,8 +7017,9 @@ SDValue PPCTargetLowering::PerformDAGCombine(SDNode *N, }; SDValue BSLoad = DAG.getMemIntrinsicNode(PPCISD::LBRX, dl, - DAG.getVTList(MVT::i32, MVT::Other), Ops, 3, - LD->getMemoryVT(), LD->getMemOperand()); + DAG.getVTList(N->getValueType(0) == MVT::i64 ? + MVT::i64 : MVT::i32, MVT::Other), + Ops, 3, LD->getMemoryVT(), LD->getMemOperand()); // If this is an i16 load, insert the truncate. SDValue ResVal = BSLoad; @@ -6631,6 +7276,9 @@ PPCTargetLowering::getRegForInlineAsmConstraint(const std::string &Constraint, // GCC RS6000 Constraint Letters switch (Constraint[0]) { case 'b': // R1-R31 + if (VT == MVT::i64 && PPCSubTarget.isPPC64()) + return std::make_pair(0U, &PPC::G8RC_NOX0RegClass); + return std::make_pair(0U, &PPC::GPRC_NOR0RegClass); case 'r': // R0-R31 if (VT == MVT::i64 && PPCSubTarget.isPPC64()) return std::make_pair(0U, &PPC::G8RCRegClass); @@ -6815,13 +7463,16 @@ SDValue PPCTargetLowering::LowerFRAMEADDR(SDValue Op, MachineFunction &MF = DAG.getMachineFunction(); MachineFrameInfo *MFI = MF.getFrameInfo(); MFI->setFrameAddressIsTaken(true); - bool is31 = (getTargetMachine().Options.DisableFramePointerElim(MF) || - MFI->hasVarSizedObjects()) && - MFI->getStackSize() && - !MF.getFunction()->getAttributes(). - hasAttribute(AttributeSet::FunctionIndex, Attribute::Naked); - unsigned FrameReg = isPPC64 ? (is31 ? PPC::X31 : PPC::X1) : - (is31 ? PPC::R31 : PPC::R1); + + // Naked functions never have a frame pointer, and so we use r1. For all + // other functions, this decision must be delayed until during PEI. + unsigned FrameReg; + if (MF.getFunction()->getAttributes().hasAttribute( + AttributeSet::FunctionIndex, Attribute::Naked)) + FrameReg = isPPC64 ? PPC::X1 : PPC::R1; + else + FrameReg = isPPC64 ? PPC::FP8 : PPC::FP; + SDValue FrameAddr = DAG.getCopyFromReg(DAG.getEntryNode(), dl, FrameReg, PtrVT); while (Depth--) diff --git a/lib/Target/PowerPC/PPCISelLowering.h b/lib/Target/PowerPC/PPCISelLowering.h index 8d44d9f..423e983 100644 --- a/lib/Target/PowerPC/PPCISelLowering.h +++ b/lib/Target/PowerPC/PPCISelLowering.h @@ -16,6 +16,8 @@ #define LLVM_TARGET_POWERPC_PPC32ISELLOWERING_H #include "PPC.h" +#include "PPCInstrInfo.h" +#include "PPCRegisterInfo.h" #include "PPCSubtarget.h" #include "llvm/CodeGen/SelectionDAG.h" #include "llvm/Target/TargetLowering.h" @@ -35,14 +37,21 @@ namespace llvm { /// was temporarily in the f64 operand. FCFID, + /// Newer FCFID[US] integer-to-floating-point conversion instructions for + /// unsigned integers and single-precision outputs. + FCFIDU, FCFIDS, FCFIDUS, + /// FCTI[D,W]Z - The FCTIDZ and FCTIWZ instructions, taking an f32 or f64 /// operand, producing an f64 value containing the integer representation /// of that FP value. FCTIDZ, FCTIWZ, - /// STFIWX - The STFIWX instruction. The first operand is an input token - /// chain, then an f64 value to store, then an address to store it to. - STFIWX, + /// Newer FCTI[D,W]UZ floating-point-to-integer conversion instructions for + /// unsigned integers. + FCTIDUZ, FCTIWUZ, + + /// Reciprocal estimate instructions (unary FP ops). + FRE, FRSQRTE, // VMADDFP, VNMSUBFP - The VMADDFP and VNMSUBFP instructions, taking // three v4f32 operands and producing a v4f32 result. @@ -90,17 +99,10 @@ namespace llvm { /// code. SRL, SRA, SHL, - /// EXTSW_32 - This is the EXTSW instruction for use with "32-bit" - /// registers. - EXTSW_32, - /// CALL - A direct function call. - /// CALL_NOP_SVR4 is a call with the special NOP which follows 64-bit + /// CALL_NOP is a call with the special NOP which follows 64-bit /// SVR4 calls. - CALL_Darwin, CALL_SVR4, CALL_NOP_SVR4, - - /// NOP - Special NOP which follows 64-bit SVR4 calls. - NOP, + CALL, CALL_NOP, /// CHAIN,FLAG = MTCTR(VAL, CHAIN[, INFLAG]) - Directly corresponds to a /// MTCTR instruction. @@ -108,7 +110,7 @@ namespace llvm { /// CHAIN,FLAG = BCTRL(CHAIN, INFLAG) - Directly corresponds to a /// BCTRL instruction. - BCTRL_Darwin, BCTRL_SVR4, + BCTRL, /// Return with a flag operand, matched by 'blr' RET_FLAG, @@ -119,6 +121,12 @@ namespace llvm { /// are undefined. MFCR, + // EH_SJLJ_SETJMP - SjLj exception handling setjmp. + EH_SJLJ_SETJMP, + + // EH_SJLJ_LONGJMP - SjLj exception handling longjmp. + EH_SJLJ_LONGJMP, + /// RESVEC = VCMP(LHS, RHS, OPC) - Represents one of the altivec VCMP* /// instructions. For lack of better number, we use the opcode number /// encoding for the OPC field to identify the compare. For example, 838 @@ -138,26 +146,13 @@ namespace llvm { /// an optional input flag argument. COND_BRANCH, - // The following 5 instructions are used only as part of the - // long double-to-int conversion sequence. - - /// OUTFLAG = MFFS F8RC - This moves the FPSCR (not modelled) into the - /// register. - MFFS, - - /// OUTFLAG = MTFSB0 INFLAG - This clears a bit in the FPSCR. - MTFSB0, - - /// OUTFLAG = MTFSB1 INFLAG - This sets a bit in the FPSCR. - MTFSB1, - - /// F8RC, OUTFLAG = FADDRTZ F8RC, F8RC, INFLAG - This is an FADD done with - /// rounding towards zero. It has flags added so it won't move past the - /// FPSCR-setting instructions. + /// F8RC = FADDRTZ F8RC, F8RC - This is an FADD done with rounding + /// towards zero. Used only as part of the long double-to-int + /// conversion sequence. FADDRTZ, - /// MTFSF = F8RC, INFLAG - This moves the register into the FPSCR. - MTFSF, + /// F8RC = MFFS - This moves the FPSCR (not modeled) into the register. + MFFS, /// LARX = This corresponds to PPC l{w|d}arx instrcution: load and /// reserve indexed. This is used to implement atomic operations. @@ -243,14 +238,11 @@ namespace llvm { /// optimizations due to constant folding. VADD_SPLAT, - /// STD_32 - This is the STD instruction for use with "32-bit" registers. - STD_32 = ISD::FIRST_TARGET_MEMORY_OPCODE, - /// CHAIN = STBRX CHAIN, GPRC, Ptr, Type - This is a /// byte-swapping store instruction. It byte-swaps the low "Type" bits of /// the GPRC input, then stores it through Ptr. Type can be either i16 or /// i32. - STBRX, + STBRX = ISD::FIRST_TARGET_MEMORY_OPCODE, /// GPRC, CHAIN = LBRX CHAIN, Ptr, Type - This is a /// byte-swapping load instruction. It loads "Type" bits, byte swaps it, @@ -258,6 +250,20 @@ namespace llvm { /// or i32. LBRX, + /// STFIWX - The STFIWX instruction. The first operand is an input token + /// chain, then an f64 value to store, then an address to store it to. + STFIWX, + + /// GPRC, CHAIN = LFIWAX CHAIN, Ptr - This is a floating-point + /// load which sign-extends from a 32-bit integer value into the + /// destination 64-bit register. + LFIWAX, + + /// GPRC, CHAIN = LFIWZX CHAIN, Ptr - This is a floating-point + /// load which zero-extends from a 32-bit integer value into the + /// destination 64-bit register. + LFIWZX, + /// G8RC = ADDIS_TOC_HA %X2, Symbol - For medium and large code model, /// produces an ADDIS8 instruction that adds the TOC base register to /// sym@toc@ha. @@ -321,6 +327,8 @@ namespace llvm { class PPCTargetLowering : public TargetLowering { const PPCSubtarget &PPCSubTarget; + const PPCRegisterInfo *PPCRegInfo; + const PPCInstrInfo *PPCII; public: explicit PPCTargetLowering(PPCTargetMachine &TM); @@ -395,6 +403,12 @@ namespace llvm { MachineBasicBlock *MBB, bool is8bit, unsigned Opcode) const; + MachineBasicBlock *emitEHSjLjSetJmp(MachineInstr *MI, + MachineBasicBlock *MBB) const; + + MachineBasicBlock *emitEHSjLjLongJmp(MachineInstr *MI, + MachineBasicBlock *MBB) const; + ConstraintType getConstraintType(const std::string &Constraint) const; /// Examine constraint string and operand type and determine a weight value. @@ -498,7 +512,7 @@ namespace llvm { const PPCSubtarget &Subtarget) const; SDValue LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const; SDValue LowerFP_TO_INT(SDValue Op, SelectionDAG &DAG, DebugLoc dl) const; - SDValue LowerSINT_TO_FP(SDValue Op, SelectionDAG &DAG) const; + SDValue LowerINT_TO_FP(SDValue Op, SelectionDAG &DAG) const; SDValue LowerFLT_ROUNDS_(SDValue Op, SelectionDAG &DAG) const; SDValue LowerSHL_PARTS(SDValue Op, SelectionDAG &DAG) const; SDValue LowerSRL_PARTS(SDValue Op, SelectionDAG &DAG) const; @@ -608,6 +622,12 @@ namespace llvm { const SmallVectorImpl<ISD::InputArg> &Ins, DebugLoc dl, SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals) const; + + SDValue lowerEH_SJLJ_SETJMP(SDValue Op, SelectionDAG &DAG) const; + SDValue lowerEH_SJLJ_LONGJMP(SDValue Op, SelectionDAG &DAG) const; + + SDValue DAGCombineFastRecip(SDValue Op, DAGCombinerInfo &DCI) const; + SDValue DAGCombineFastRecipFSQRT(SDValue Op, DAGCombinerInfo &DCI) const; }; } diff --git a/lib/Target/PowerPC/PPCInstr64Bit.td b/lib/Target/PowerPC/PPCInstr64Bit.td index bca1bd5..e5d0b91 100644 --- a/lib/Target/PowerPC/PPCInstr64Bit.td +++ b/lib/Target/PowerPC/PPCInstr64Bit.td @@ -30,12 +30,7 @@ def symbolLo64 : Operand<i64> { let EncoderMethod = "getLO16Encoding"; } def tocentry : Operand<iPTR> { - let MIOperandInfo = (ops i32imm:$imm); -} -def memrs : Operand<iPTR> { // memri where the immediate is a symbolLo64 - let PrintMethod = "printMemRegImm"; - let EncoderMethod = "getMemRIXEncoding"; - let MIOperandInfo = (ops symbolLo64:$off, ptr_rc:$reg); + let MIOperandInfo = (ops i64imm:$imm); } def tlsreg : Operand<i64> { let EncoderMethod = "getTLSRegEncoding"; @@ -71,135 +66,136 @@ def HI48_64 : SDNodeXForm<imm, [{ // Calls. // +let Interpretation64Bit = 1 in { +let isTerminator = 1, isBarrier = 1, PPC970_Unit = 7 in { + let isBranch = 1, isIndirectBranch = 1, Uses = [CTR8] in { + def BCTR8 : XLForm_2_ext<19, 528, 20, 0, 0, (outs), (ins), "bctr", BrB, []>, + Requires<[In64BitMode]>; + + let isCodeGenOnly = 1 in + def BCCTR8 : XLForm_2_br<19, 528, 0, (outs), (ins pred:$cond), + "b${cond:cc}ctr ${cond:reg}", BrB, []>, + Requires<[In64BitMode]>; + } +} + let Defs = [LR8] in def MovePCtoLR8 : Pseudo<(outs), (ins), "#MovePCtoLR8", []>, PPC970_Unit_BRU; -// Darwin ABI Calls. -let isCall = 1, PPC970_Unit = 7, Defs = [LR8] in { - // Convenient aliases for call instructions - let Uses = [RM] in { - def BL8_Darwin : IForm<18, 0, 1, - (outs), (ins calltarget:$func), - "bl $func", BrB, []>; // See Pat patterns below. - def BLA8_Darwin : IForm<18, 1, 1, - (outs), (ins aaddr:$func), - "bla $func", BrB, [(PPCcall_Darwin (i64 imm:$func))]>; +let isBranch = 1, isTerminator = 1, hasCtrlDep = 1, PPC970_Unit = 7 in { + let Defs = [CTR8], Uses = [CTR8] in { + def BDZ8 : BForm_1<16, 18, 0, 0, (outs), (ins condbrtarget:$dst), + "bdz $dst">; + def BDNZ8 : BForm_1<16, 16, 0, 0, (outs), (ins condbrtarget:$dst), + "bdnz $dst">; } - let Uses = [CTR8, RM] in { - def BCTRL8_Darwin : XLForm_2_ext<19, 528, 20, 0, 1, - (outs), (ins), - "bctrl", BrB, - [(PPCbctrl_Darwin)]>, Requires<[In64BitMode]>; + + let isReturn = 1, Defs = [CTR8], Uses = [CTR8, LR8, RM] in { + def BDZLR8 : XLForm_2_ext<19, 16, 18, 0, 0, (outs), (ins), + "bdzlr", BrB, []>; + def BDNZLR8 : XLForm_2_ext<19, 16, 16, 0, 0, (outs), (ins), + "bdnzlr", BrB, []>; } } -// ELF 64 ABI Calls = Darwin ABI Calls -// Used to define BL8_ELF and BLA8_ELF + + let isCall = 1, PPC970_Unit = 7, Defs = [LR8] in { // Convenient aliases for call instructions let Uses = [RM] in { - def BL8_ELF : IForm<18, 0, 1, - (outs), (ins calltarget:$func), - "bl $func", BrB, []>; // See Pat patterns below. + def BL8 : IForm<18, 0, 1, (outs), (ins calltarget:$func), + "bl $func", BrB, []>; // See Pat patterns below. - let isCodeGenOnly = 1 in - def BL8_NOP_ELF : IForm_and_DForm_4_zero<18, 0, 1, 24, + def BLA8 : IForm<18, 1, 1, (outs), (ins aaddr:$func), + "bla $func", BrB, [(PPCcall (i64 imm:$func))]>; + } + let Uses = [RM], isCodeGenOnly = 1 in { + def BL8_NOP : IForm_and_DForm_4_zero<18, 0, 1, 24, (outs), (ins calltarget:$func), "bl $func\n\tnop", BrB, []>; - let isCodeGenOnly = 1 in - def BL8_NOP_ELF_TLSGD : IForm_and_DForm_4_zero<18, 0, 1, 24, + def BL8_NOP_TLSGD : IForm_and_DForm_4_zero<18, 0, 1, 24, (outs), (ins calltarget:$func, tlsgd:$sym), "bl $func($sym)\n\tnop", BrB, []>; - let isCodeGenOnly = 1 in - def BL8_NOP_ELF_TLSLD : IForm_and_DForm_4_zero<18, 0, 1, 24, + def BL8_NOP_TLSLD : IForm_and_DForm_4_zero<18, 0, 1, 24, (outs), (ins calltarget:$func, tlsgd:$sym), "bl $func($sym)\n\tnop", BrB, []>; - def BLA8_ELF : IForm<18, 1, 1, - (outs), (ins aaddr:$func), - "bla $func", BrB, [(PPCcall_SVR4 (i64 imm:$func))]>; - - let isCodeGenOnly = 1 in - def BLA8_NOP_ELF : IForm_and_DForm_4_zero<18, 1, 1, 24, + def BLA8_NOP : IForm_and_DForm_4_zero<18, 1, 1, 24, (outs), (ins aaddr:$func), "bla $func\n\tnop", BrB, - [(PPCcall_nop_SVR4 (i64 imm:$func))]>; + [(PPCcall_nop (i64 imm:$func))]>; } - let Uses = [X11, CTR8, RM] in { - def BCTRL8_ELF : XLForm_2_ext<19, 528, 20, 0, 1, - (outs), (ins), - "bctrl", BrB, - [(PPCbctrl_SVR4)]>, Requires<[In64BitMode]>; + let Uses = [CTR8, RM] in { + def BCTRL8 : XLForm_2_ext<19, 528, 20, 0, 1, (outs), (ins), + "bctrl", BrB, [(PPCbctrl)]>, + Requires<[In64BitMode]>; + + let isCodeGenOnly = 1 in + def BCCTRL8 : XLForm_2_br<19, 528, 1, (outs), (ins pred:$cond), + "b${cond:cc}ctrl ${cond:reg}", BrB, []>, + Requires<[In64BitMode]>; } } - +} // Interpretation64Bit // Calls -def : Pat<(PPCcall_Darwin (i64 tglobaladdr:$dst)), - (BL8_Darwin tglobaladdr:$dst)>; -def : Pat<(PPCcall_Darwin (i64 texternalsym:$dst)), - (BL8_Darwin texternalsym:$dst)>; - -def : Pat<(PPCcall_SVR4 (i64 tglobaladdr:$dst)), - (BL8_ELF tglobaladdr:$dst)>; -def : Pat<(PPCcall_nop_SVR4 (i64 tglobaladdr:$dst)), - (BL8_NOP_ELF tglobaladdr:$dst)>; - -def : Pat<(PPCcall_SVR4 (i64 texternalsym:$dst)), - (BL8_ELF texternalsym:$dst)>; -def : Pat<(PPCcall_nop_SVR4 (i64 texternalsym:$dst)), - (BL8_NOP_ELF texternalsym:$dst)>; +def : Pat<(PPCcall (i64 tglobaladdr:$dst)), + (BL8 tglobaladdr:$dst)>; +def : Pat<(PPCcall_nop (i64 tglobaladdr:$dst)), + (BL8_NOP tglobaladdr:$dst)>; -def : Pat<(PPCnop), - (NOP)>; +def : Pat<(PPCcall (i64 texternalsym:$dst)), + (BL8 texternalsym:$dst)>; +def : Pat<(PPCcall_nop (i64 texternalsym:$dst)), + (BL8_NOP texternalsym:$dst)>; // Atomic operations let usesCustomInserter = 1 in { let Defs = [CR0] in { def ATOMIC_LOAD_ADD_I64 : Pseudo< - (outs G8RC:$dst), (ins memrr:$ptr, G8RC:$incr), "#ATOMIC_LOAD_ADD_I64", - [(set G8RC:$dst, (atomic_load_add_64 xoaddr:$ptr, G8RC:$incr))]>; + (outs g8rc:$dst), (ins memrr:$ptr, g8rc:$incr), "#ATOMIC_LOAD_ADD_I64", + [(set i64:$dst, (atomic_load_add_64 xoaddr:$ptr, i64:$incr))]>; def ATOMIC_LOAD_SUB_I64 : Pseudo< - (outs G8RC:$dst), (ins memrr:$ptr, G8RC:$incr), "#ATOMIC_LOAD_SUB_I64", - [(set G8RC:$dst, (atomic_load_sub_64 xoaddr:$ptr, G8RC:$incr))]>; + (outs g8rc:$dst), (ins memrr:$ptr, g8rc:$incr), "#ATOMIC_LOAD_SUB_I64", + [(set i64:$dst, (atomic_load_sub_64 xoaddr:$ptr, i64:$incr))]>; def ATOMIC_LOAD_OR_I64 : Pseudo< - (outs G8RC:$dst), (ins memrr:$ptr, G8RC:$incr), "#ATOMIC_LOAD_OR_I64", - [(set G8RC:$dst, (atomic_load_or_64 xoaddr:$ptr, G8RC:$incr))]>; + (outs g8rc:$dst), (ins memrr:$ptr, g8rc:$incr), "#ATOMIC_LOAD_OR_I64", + [(set i64:$dst, (atomic_load_or_64 xoaddr:$ptr, i64:$incr))]>; def ATOMIC_LOAD_XOR_I64 : Pseudo< - (outs G8RC:$dst), (ins memrr:$ptr, G8RC:$incr), "#ATOMIC_LOAD_XOR_I64", - [(set G8RC:$dst, (atomic_load_xor_64 xoaddr:$ptr, G8RC:$incr))]>; + (outs g8rc:$dst), (ins memrr:$ptr, g8rc:$incr), "#ATOMIC_LOAD_XOR_I64", + [(set i64:$dst, (atomic_load_xor_64 xoaddr:$ptr, i64:$incr))]>; def ATOMIC_LOAD_AND_I64 : Pseudo< - (outs G8RC:$dst), (ins memrr:$ptr, G8RC:$incr), "#ATOMIC_LOAD_AND_i64", - [(set G8RC:$dst, (atomic_load_and_64 xoaddr:$ptr, G8RC:$incr))]>; + (outs g8rc:$dst), (ins memrr:$ptr, g8rc:$incr), "#ATOMIC_LOAD_AND_i64", + [(set i64:$dst, (atomic_load_and_64 xoaddr:$ptr, i64:$incr))]>; def ATOMIC_LOAD_NAND_I64 : Pseudo< - (outs G8RC:$dst), (ins memrr:$ptr, G8RC:$incr), "#ATOMIC_LOAD_NAND_I64", - [(set G8RC:$dst, (atomic_load_nand_64 xoaddr:$ptr, G8RC:$incr))]>; + (outs g8rc:$dst), (ins memrr:$ptr, g8rc:$incr), "#ATOMIC_LOAD_NAND_I64", + [(set i64:$dst, (atomic_load_nand_64 xoaddr:$ptr, i64:$incr))]>; def ATOMIC_CMP_SWAP_I64 : Pseudo< - (outs G8RC:$dst), (ins memrr:$ptr, G8RC:$old, G8RC:$new), "#ATOMIC_CMP_SWAP_I64", - [(set G8RC:$dst, - (atomic_cmp_swap_64 xoaddr:$ptr, G8RC:$old, G8RC:$new))]>; + (outs g8rc:$dst), (ins memrr:$ptr, g8rc:$old, g8rc:$new), "#ATOMIC_CMP_SWAP_I64", + [(set i64:$dst, (atomic_cmp_swap_64 xoaddr:$ptr, i64:$old, i64:$new))]>; def ATOMIC_SWAP_I64 : Pseudo< - (outs G8RC:$dst), (ins memrr:$ptr, G8RC:$new), "#ATOMIC_SWAP_I64", - [(set G8RC:$dst, (atomic_swap_64 xoaddr:$ptr, G8RC:$new))]>; + (outs g8rc:$dst), (ins memrr:$ptr, g8rc:$new), "#ATOMIC_SWAP_I64", + [(set i64:$dst, (atomic_swap_64 xoaddr:$ptr, i64:$new))]>; } } // Instructions to support atomic operations -def LDARX : XForm_1<31, 84, (outs G8RC:$rD), (ins memrr:$ptr), +def LDARX : XForm_1<31, 84, (outs g8rc:$rD), (ins memrr:$ptr), "ldarx $rD, $ptr", LdStLDARX, - [(set G8RC:$rD, (PPClarx xoaddr:$ptr))]>; + [(set i64:$rD, (PPClarx xoaddr:$ptr))]>; let Defs = [CR0] in -def STDCX : XForm_1<31, 214, (outs), (ins G8RC:$rS, memrr:$dst), +def STDCX : XForm_1<31, 214, (outs), (ins g8rc:$rS, memrr:$dst), "stdcx. $rS, $dst", LdStSTDCX, - [(PPCstcx G8RC:$rS, xoaddr:$dst)]>, + [(PPCstcx i64:$rS, xoaddr:$dst)]>, isDOT; +let Interpretation64Bit = 1 in { let isCall = 1, isTerminator = 1, isReturn = 1, isBarrier = 1, Uses = [RM] in def TCRETURNdi8 :Pseudo< (outs), (ins calltarget:$dst, i32imm:$offset), @@ -216,17 +212,12 @@ def TCRETURNri8 : Pseudo<(outs), (ins CTRRC8:$dst, i32imm:$offset), "#TC_RETURNr8 $dst $offset", []>; +let isCodeGenOnly = 1 in { let isTerminator = 1, isBarrier = 1, PPC970_Unit = 7, isBranch = 1, - isIndirectBranch = 1, isCall = 1, Uses = [CTR8, RM] in { - let isReturn = 1 in { - def TAILBCTR8 : XLForm_2_ext<19, 528, 20, 0, 0, (outs), (ins), "bctr", BrB, []>, - Requires<[In64BitMode]>; - } - - def BCTR8 : XLForm_2_ext<19, 528, 20, 0, 0, (outs), (ins), "bctr", BrB, []>, - Requires<[In64BitMode]>; -} + isIndirectBranch = 1, isCall = 1, isReturn = 1, Uses = [CTR8, RM] in +def TAILBCTR8 : XLForm_2_ext<19, 528, 20, 0, 0, (outs), (ins), "bctr", BrB, []>, + Requires<[In64BitMode]>; let isBranch = 1, isTerminator = 1, hasCtrlDep = 1, PPC970_Unit = 7, @@ -242,6 +233,9 @@ def TAILBA8 : IForm<18, 0, 0, (outs), (ins aaddr:$dst), "ba $dst", BrB, []>; +} +} // Interpretation64Bit + def : Pat<(PPCtc_return (i64 tglobaladdr:$dst), imm:$imm), (TCRETURNdi8 tglobaladdr:$dst, imm:$imm)>; @@ -251,44 +245,53 @@ def : Pat<(PPCtc_return (i64 texternalsym:$dst), imm:$imm), def : Pat<(PPCtc_return CTRRC8:$dst, imm:$imm), (TCRETURNri8 CTRRC8:$dst, imm:$imm)>; -let isBranch = 1, isTerminator = 1, hasCtrlDep = 1, PPC970_Unit = 7 in { - let Defs = [CTR8], Uses = [CTR8] in { - def BDZ8 : BForm_1<16, 18, 0, 0, (outs), (ins condbrtarget:$dst), - "bdz $dst">; - def BDNZ8 : BForm_1<16, 16, 0, 0, (outs), (ins condbrtarget:$dst), - "bdnz $dst">; - } -} -// 64-but CR instructions -def MTCRF8 : XFXForm_5<31, 144, (outs crbitm:$FXM), (ins G8RC:$rS), +// 64-bit CR instructions +let Interpretation64Bit = 1 in { +let neverHasSideEffects = 1 in { +def MTCRF8 : XFXForm_5<31, 144, (outs crbitm:$FXM), (ins g8rc:$rS), "mtcrf $FXM, $rS", BrMCRX>, PPC970_MicroCode, PPC970_Unit_CRU; -def MFCR8pseud: XFXForm_3<31, 19, (outs G8RC:$rT), (ins crbitm:$FXM), +let isCodeGenOnly = 1 in +def MFCR8pseud: XFXForm_3<31, 19, (outs g8rc:$rT), (ins crbitm:$FXM), "#MFCR8pseud", SprMFCR>, PPC970_MicroCode, PPC970_Unit_CRU; - -def MFCR8 : XFXForm_3<31, 19, (outs G8RC:$rT), (ins), +} // neverHasSideEffects = 1 + +let neverHasSideEffects = 1 in +def MFCR8 : XFXForm_3<31, 19, (outs g8rc:$rT), (ins), "mfcr $rT", SprMFCR>, PPC970_MicroCode, PPC970_Unit_CRU; +let hasSideEffects = 1, isBarrier = 1, usesCustomInserter = 1 in { + def EH_SjLj_SetJmp64 : Pseudo<(outs gprc:$dst), (ins memr:$buf), + "#EH_SJLJ_SETJMP64", + [(set i32:$dst, (PPCeh_sjlj_setjmp addr:$buf))]>, + Requires<[In64BitMode]>; + let isTerminator = 1 in + def EH_SjLj_LongJmp64 : Pseudo<(outs), (ins memr:$buf), + "#EH_SJLJ_LONGJMP64", + [(PPCeh_sjlj_longjmp addr:$buf)]>, + Requires<[In64BitMode]>; +} + //===----------------------------------------------------------------------===// // 64-bit SPR manipulation instrs. let Uses = [CTR8] in { -def MFCTR8 : XFXForm_1_ext<31, 339, 9, (outs G8RC:$rT), (ins), +def MFCTR8 : XFXForm_1_ext<31, 339, 9, (outs g8rc:$rT), (ins), "mfctr $rT", SprMFSPR>, PPC970_DGroup_First, PPC970_Unit_FXU; } -let Pattern = [(PPCmtctr G8RC:$rS)], Defs = [CTR8] in { -def MTCTR8 : XFXForm_7_ext<31, 467, 9, (outs), (ins G8RC:$rS), +let Pattern = [(PPCmtctr i64:$rS)], Defs = [CTR8] in { +def MTCTR8 : XFXForm_7_ext<31, 467, 9, (outs), (ins g8rc:$rS), "mtctr $rS", SprMTSPR>, PPC970_DGroup_First, PPC970_Unit_FXU; } -let Pattern = [(set G8RC:$rT, readcyclecounter)] in -def MFTB8 : XFXForm_1_ext<31, 339, 268, (outs G8RC:$rT), (ins), +let Pattern = [(set i64:$rT, readcyclecounter)] in +def MFTB8 : XFXForm_1_ext<31, 339, 268, (outs g8rc:$rT), (ins), "mfspr $rT, 268", SprMFTB>, PPC970_DGroup_First, PPC970_Unit_FXU; // Note that encoding mftb using mfspr is now the preferred form, @@ -297,248 +300,265 @@ def MFTB8 : XFXForm_1_ext<31, 339, 268, (outs G8RC:$rT), (ins), // the POWER3. let Defs = [X1], Uses = [X1] in -def DYNALLOC8 : Pseudo<(outs G8RC:$result), (ins G8RC:$negsize, memri:$fpsi),"#DYNALLOC8", - [(set G8RC:$result, - (PPCdynalloc G8RC:$negsize, iaddr:$fpsi))]>; +def DYNALLOC8 : Pseudo<(outs g8rc:$result), (ins g8rc:$negsize, memri:$fpsi),"#DYNALLOC8", + [(set i64:$result, + (PPCdynalloc i64:$negsize, iaddr:$fpsi))]>; let Defs = [LR8] in { -def MTLR8 : XFXForm_7_ext<31, 467, 8, (outs), (ins G8RC:$rS), +def MTLR8 : XFXForm_7_ext<31, 467, 8, (outs), (ins g8rc:$rS), "mtlr $rS", SprMTSPR>, PPC970_DGroup_First, PPC970_Unit_FXU; } let Uses = [LR8] in { -def MFLR8 : XFXForm_1_ext<31, 339, 8, (outs G8RC:$rT), (ins), +def MFLR8 : XFXForm_1_ext<31, 339, 8, (outs g8rc:$rT), (ins), "mflr $rT", SprMFSPR>, PPC970_DGroup_First, PPC970_Unit_FXU; } +} // Interpretation64Bit //===----------------------------------------------------------------------===// // Fixed point instructions. // let PPC970_Unit = 1 in { // FXU Operations. +let Interpretation64Bit = 1 in { +let neverHasSideEffects = 1 in { let isReMaterializable = 1, isAsCheapAsAMove = 1, isMoveImm = 1 in { -def LI8 : DForm_2_r0<14, (outs G8RC:$rD), (ins symbolLo64:$imm), +def LI8 : DForm_2_r0<14, (outs g8rc:$rD), (ins symbolLo64:$imm), "li $rD, $imm", IntSimple, - [(set G8RC:$rD, immSExt16:$imm)]>; -def LIS8 : DForm_2_r0<15, (outs G8RC:$rD), (ins symbolHi64:$imm), + [(set i64:$rD, immSExt16:$imm)]>; +def LIS8 : DForm_2_r0<15, (outs g8rc:$rD), (ins symbolHi64:$imm), "lis $rD, $imm", IntSimple, - [(set G8RC:$rD, imm16ShiftedSExt:$imm)]>; + [(set i64:$rD, imm16ShiftedSExt:$imm)]>; } // Logical ops. -def NAND8: XForm_6<31, 476, (outs G8RC:$rA), (ins G8RC:$rS, G8RC:$rB), - "nand $rA, $rS, $rB", IntSimple, - [(set G8RC:$rA, (not (and G8RC:$rS, G8RC:$rB)))]>; -def AND8 : XForm_6<31, 28, (outs G8RC:$rA), (ins G8RC:$rS, G8RC:$rB), - "and $rA, $rS, $rB", IntSimple, - [(set G8RC:$rA, (and G8RC:$rS, G8RC:$rB))]>; -def ANDC8: XForm_6<31, 60, (outs G8RC:$rA), (ins G8RC:$rS, G8RC:$rB), - "andc $rA, $rS, $rB", IntSimple, - [(set G8RC:$rA, (and G8RC:$rS, (not G8RC:$rB)))]>; -def OR8 : XForm_6<31, 444, (outs G8RC:$rA), (ins G8RC:$rS, G8RC:$rB), - "or $rA, $rS, $rB", IntSimple, - [(set G8RC:$rA, (or G8RC:$rS, G8RC:$rB))]>; -def NOR8 : XForm_6<31, 124, (outs G8RC:$rA), (ins G8RC:$rS, G8RC:$rB), - "nor $rA, $rS, $rB", IntSimple, - [(set G8RC:$rA, (not (or G8RC:$rS, G8RC:$rB)))]>; -def ORC8 : XForm_6<31, 412, (outs G8RC:$rA), (ins G8RC:$rS, G8RC:$rB), - "orc $rA, $rS, $rB", IntSimple, - [(set G8RC:$rA, (or G8RC:$rS, (not G8RC:$rB)))]>; -def EQV8 : XForm_6<31, 284, (outs G8RC:$rA), (ins G8RC:$rS, G8RC:$rB), - "eqv $rA, $rS, $rB", IntSimple, - [(set G8RC:$rA, (not (xor G8RC:$rS, G8RC:$rB)))]>; -def XOR8 : XForm_6<31, 316, (outs G8RC:$rA), (ins G8RC:$rS, G8RC:$rB), - "xor $rA, $rS, $rB", IntSimple, - [(set G8RC:$rA, (xor G8RC:$rS, G8RC:$rB))]>; +defm NAND8: XForm_6r<31, 476, (outs g8rc:$rA), (ins g8rc:$rS, g8rc:$rB), + "nand", "$rA, $rS, $rB", IntSimple, + [(set i64:$rA, (not (and i64:$rS, i64:$rB)))]>; +defm AND8 : XForm_6r<31, 28, (outs g8rc:$rA), (ins g8rc:$rS, g8rc:$rB), + "and", "$rA, $rS, $rB", IntSimple, + [(set i64:$rA, (and i64:$rS, i64:$rB))]>; +defm ANDC8: XForm_6r<31, 60, (outs g8rc:$rA), (ins g8rc:$rS, g8rc:$rB), + "andc", "$rA, $rS, $rB", IntSimple, + [(set i64:$rA, (and i64:$rS, (not i64:$rB)))]>; +defm OR8 : XForm_6r<31, 444, (outs g8rc:$rA), (ins g8rc:$rS, g8rc:$rB), + "or", "$rA, $rS, $rB", IntSimple, + [(set i64:$rA, (or i64:$rS, i64:$rB))]>; +defm NOR8 : XForm_6r<31, 124, (outs g8rc:$rA), (ins g8rc:$rS, g8rc:$rB), + "nor", "$rA, $rS, $rB", IntSimple, + [(set i64:$rA, (not (or i64:$rS, i64:$rB)))]>; +defm ORC8 : XForm_6r<31, 412, (outs g8rc:$rA), (ins g8rc:$rS, g8rc:$rB), + "orc", "$rA, $rS, $rB", IntSimple, + [(set i64:$rA, (or i64:$rS, (not i64:$rB)))]>; +defm EQV8 : XForm_6r<31, 284, (outs g8rc:$rA), (ins g8rc:$rS, g8rc:$rB), + "eqv", "$rA, $rS, $rB", IntSimple, + [(set i64:$rA, (not (xor i64:$rS, i64:$rB)))]>; +defm XOR8 : XForm_6r<31, 316, (outs g8rc:$rA), (ins g8rc:$rS, g8rc:$rB), + "xor", "$rA, $rS, $rB", IntSimple, + [(set i64:$rA, (xor i64:$rS, i64:$rB))]>; // Logical ops with immediate. -def ANDIo8 : DForm_4<28, (outs G8RC:$dst), (ins G8RC:$src1, u16imm:$src2), +let Defs = [CR0] in { +def ANDIo8 : DForm_4<28, (outs g8rc:$dst), (ins g8rc:$src1, u16imm:$src2), "andi. $dst, $src1, $src2", IntGeneral, - [(set G8RC:$dst, (and G8RC:$src1, immZExt16:$src2))]>, + [(set i64:$dst, (and i64:$src1, immZExt16:$src2))]>, isDOT; -def ANDISo8 : DForm_4<29, (outs G8RC:$dst), (ins G8RC:$src1, u16imm:$src2), +def ANDISo8 : DForm_4<29, (outs g8rc:$dst), (ins g8rc:$src1, u16imm:$src2), "andis. $dst, $src1, $src2", IntGeneral, - [(set G8RC:$dst, (and G8RC:$src1,imm16ShiftedZExt:$src2))]>, + [(set i64:$dst, (and i64:$src1, imm16ShiftedZExt:$src2))]>, isDOT; -def ORI8 : DForm_4<24, (outs G8RC:$dst), (ins G8RC:$src1, u16imm:$src2), +} +def ORI8 : DForm_4<24, (outs g8rc:$dst), (ins g8rc:$src1, u16imm:$src2), "ori $dst, $src1, $src2", IntSimple, - [(set G8RC:$dst, (or G8RC:$src1, immZExt16:$src2))]>; -def ORIS8 : DForm_4<25, (outs G8RC:$dst), (ins G8RC:$src1, u16imm:$src2), + [(set i64:$dst, (or i64:$src1, immZExt16:$src2))]>; +def ORIS8 : DForm_4<25, (outs g8rc:$dst), (ins g8rc:$src1, u16imm:$src2), "oris $dst, $src1, $src2", IntSimple, - [(set G8RC:$dst, (or G8RC:$src1, imm16ShiftedZExt:$src2))]>; -def XORI8 : DForm_4<26, (outs G8RC:$dst), (ins G8RC:$src1, u16imm:$src2), + [(set i64:$dst, (or i64:$src1, imm16ShiftedZExt:$src2))]>; +def XORI8 : DForm_4<26, (outs g8rc:$dst), (ins g8rc:$src1, u16imm:$src2), "xori $dst, $src1, $src2", IntSimple, - [(set G8RC:$dst, (xor G8RC:$src1, immZExt16:$src2))]>; -def XORIS8 : DForm_4<27, (outs G8RC:$dst), (ins G8RC:$src1, u16imm:$src2), + [(set i64:$dst, (xor i64:$src1, immZExt16:$src2))]>; +def XORIS8 : DForm_4<27, (outs g8rc:$dst), (ins g8rc:$src1, u16imm:$src2), "xoris $dst, $src1, $src2", IntSimple, - [(set G8RC:$dst, (xor G8RC:$src1, imm16ShiftedZExt:$src2))]>; + [(set i64:$dst, (xor i64:$src1, imm16ShiftedZExt:$src2))]>; -def ADD8 : XOForm_1<31, 266, 0, (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB), - "add $rT, $rA, $rB", IntSimple, - [(set G8RC:$rT, (add G8RC:$rA, G8RC:$rB))]>; +defm ADD8 : XOForm_1r<31, 266, 0, (outs g8rc:$rT), (ins g8rc:$rA, g8rc:$rB), + "add", "$rT, $rA, $rB", IntSimple, + [(set i64:$rT, (add i64:$rA, i64:$rB))]>; // ADD8 has a special form: reg = ADD8(reg, sym@tls) for use by the // initial-exec thread-local storage model. -def ADD8TLS : XOForm_1<31, 266, 0, (outs G8RC:$rT), (ins G8RC:$rA, tlsreg:$rB), +let isCodeGenOnly = 1 in +def ADD8TLS : XOForm_1<31, 266, 0, (outs g8rc:$rT), (ins g8rc:$rA, tlsreg:$rB), "add $rT, $rA, $rB@tls", IntSimple, - [(set G8RC:$rT, (add G8RC:$rA, tglobaltlsaddr:$rB))]>; + [(set i64:$rT, (add i64:$rA, tglobaltlsaddr:$rB))]>; -let Defs = [CARRY] in { -def ADDC8 : XOForm_1<31, 10, 0, (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB), - "addc $rT, $rA, $rB", IntGeneral, - [(set G8RC:$rT, (addc G8RC:$rA, G8RC:$rB))]>, - PPC970_DGroup_Cracked; -def ADDIC8 : DForm_2<12, (outs G8RC:$rD), (ins G8RC:$rA, s16imm64:$imm), +defm ADDC8 : XOForm_1rc<31, 10, 0, (outs g8rc:$rT), (ins g8rc:$rA, g8rc:$rB), + "addc", "$rT, $rA, $rB", IntGeneral, + [(set i64:$rT, (addc i64:$rA, i64:$rB))]>, + PPC970_DGroup_Cracked; +let Defs = [CARRY] in +def ADDIC8 : DForm_2<12, (outs g8rc:$rD), (ins g8rc:$rA, s16imm64:$imm), "addic $rD, $rA, $imm", IntGeneral, - [(set G8RC:$rD, (addc G8RC:$rA, immSExt16:$imm))]>; -} -def ADDI8 : DForm_2<14, (outs G8RC:$rD), (ins G8RC:$rA, s16imm64:$imm), + [(set i64:$rD, (addc i64:$rA, immSExt16:$imm))]>; +def ADDI8 : DForm_2<14, (outs g8rc:$rD), (ins g8rc_nox0:$rA, symbolLo64:$imm), "addi $rD, $rA, $imm", IntSimple, - [(set G8RC:$rD, (add G8RC:$rA, immSExt16:$imm))]>; -def ADDI8L : DForm_2<14, (outs G8RC:$rD), (ins G8RC:$rA, symbolLo64:$imm), - "addi $rD, $rA, $imm", IntSimple, - [(set G8RC:$rD, (add G8RC:$rA, immSExt16:$imm))]>; -def ADDIS8 : DForm_2<15, (outs G8RC:$rD), (ins G8RC:$rA, symbolHi64:$imm), + [(set i64:$rD, (add i64:$rA, immSExt16:$imm))]>; +def ADDIS8 : DForm_2<15, (outs g8rc:$rD), (ins g8rc_nox0:$rA, symbolHi64:$imm), "addis $rD, $rA, $imm", IntSimple, - [(set G8RC:$rD, (add G8RC:$rA, imm16ShiftedSExt:$imm))]>; + [(set i64:$rD, (add i64:$rA, imm16ShiftedSExt:$imm))]>; let Defs = [CARRY] in { -def SUBFIC8: DForm_2< 8, (outs G8RC:$rD), (ins G8RC:$rA, s16imm64:$imm), +def SUBFIC8: DForm_2< 8, (outs g8rc:$rD), (ins g8rc:$rA, s16imm64:$imm), "subfic $rD, $rA, $imm", IntGeneral, - [(set G8RC:$rD, (subc immSExt16:$imm, G8RC:$rA))]>; -def SUBFC8 : XOForm_1<31, 8, 0, (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB), - "subfc $rT, $rA, $rB", IntGeneral, - [(set G8RC:$rT, (subc G8RC:$rB, G8RC:$rA))]>, - PPC970_DGroup_Cracked; -} -def SUBF8 : XOForm_1<31, 40, 0, (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB), - "subf $rT, $rA, $rB", IntGeneral, - [(set G8RC:$rT, (sub G8RC:$rB, G8RC:$rA))]>; -def NEG8 : XOForm_3<31, 104, 0, (outs G8RC:$rT), (ins G8RC:$rA), - "neg $rT, $rA", IntSimple, - [(set G8RC:$rT, (ineg G8RC:$rA))]>; -let Uses = [CARRY], Defs = [CARRY] in { -def ADDE8 : XOForm_1<31, 138, 0, (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB), - "adde $rT, $rA, $rB", IntGeneral, - [(set G8RC:$rT, (adde G8RC:$rA, G8RC:$rB))]>; -def ADDME8 : XOForm_3<31, 234, 0, (outs G8RC:$rT), (ins G8RC:$rA), - "addme $rT, $rA", IntGeneral, - [(set G8RC:$rT, (adde G8RC:$rA, -1))]>; -def ADDZE8 : XOForm_3<31, 202, 0, (outs G8RC:$rT), (ins G8RC:$rA), - "addze $rT, $rA", IntGeneral, - [(set G8RC:$rT, (adde G8RC:$rA, 0))]>; -def SUBFE8 : XOForm_1<31, 136, 0, (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB), - "subfe $rT, $rA, $rB", IntGeneral, - [(set G8RC:$rT, (sube G8RC:$rB, G8RC:$rA))]>; -def SUBFME8 : XOForm_3<31, 232, 0, (outs G8RC:$rT), (ins G8RC:$rA), - "subfme $rT, $rA", IntGeneral, - [(set G8RC:$rT, (sube -1, G8RC:$rA))]>; -def SUBFZE8 : XOForm_3<31, 200, 0, (outs G8RC:$rT), (ins G8RC:$rA), - "subfze $rT, $rA", IntGeneral, - [(set G8RC:$rT, (sube 0, G8RC:$rA))]>; -} - - -def MULHD : XOForm_1<31, 73, 0, (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB), - "mulhd $rT, $rA, $rB", IntMulHW, - [(set G8RC:$rT, (mulhs G8RC:$rA, G8RC:$rB))]>; -def MULHDU : XOForm_1<31, 9, 0, (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB), - "mulhdu $rT, $rA, $rB", IntMulHWU, - [(set G8RC:$rT, (mulhu G8RC:$rA, G8RC:$rB))]>; - -def CMPD : XForm_16_ext<31, 0, (outs CRRC:$crD), (ins G8RC:$rA, G8RC:$rB), - "cmpd $crD, $rA, $rB", IntCompare>, isPPC64; -def CMPLD : XForm_16_ext<31, 32, (outs CRRC:$crD), (ins G8RC:$rA, G8RC:$rB), - "cmpld $crD, $rA, $rB", IntCompare>, isPPC64; -def CMPDI : DForm_5_ext<11, (outs CRRC:$crD), (ins G8RC:$rA, s16imm:$imm), - "cmpdi $crD, $rA, $imm", IntCompare>, isPPC64; -def CMPLDI : DForm_6_ext<10, (outs CRRC:$dst), (ins G8RC:$src1, u16imm:$src2), - "cmpldi $dst, $src1, $src2", IntCompare>, isPPC64; - -def SLD : XForm_6<31, 27, (outs G8RC:$rA), (ins G8RC:$rS, GPRC:$rB), - "sld $rA, $rS, $rB", IntRotateD, - [(set G8RC:$rA, (PPCshl G8RC:$rS, GPRC:$rB))]>, isPPC64; -def SRD : XForm_6<31, 539, (outs G8RC:$rA), (ins G8RC:$rS, GPRC:$rB), - "srd $rA, $rS, $rB", IntRotateD, - [(set G8RC:$rA, (PPCsrl G8RC:$rS, GPRC:$rB))]>, isPPC64; -let Defs = [CARRY] in { -def SRAD : XForm_6<31, 794, (outs G8RC:$rA), (ins G8RC:$rS, GPRC:$rB), - "srad $rA, $rS, $rB", IntRotateD, - [(set G8RC:$rA, (PPCsra G8RC:$rS, GPRC:$rB))]>, isPPC64; + [(set i64:$rD, (subc immSExt16:$imm, i64:$rA))]>; +defm SUBFC8 : XOForm_1r<31, 8, 0, (outs g8rc:$rT), (ins g8rc:$rA, g8rc:$rB), + "subfc", "$rT, $rA, $rB", IntGeneral, + [(set i64:$rT, (subc i64:$rB, i64:$rA))]>, + PPC970_DGroup_Cracked; +} +defm SUBF8 : XOForm_1r<31, 40, 0, (outs g8rc:$rT), (ins g8rc:$rA, g8rc:$rB), + "subf", "$rT, $rA, $rB", IntGeneral, + [(set i64:$rT, (sub i64:$rB, i64:$rA))]>; +defm NEG8 : XOForm_3r<31, 104, 0, (outs g8rc:$rT), (ins g8rc:$rA), + "neg", "$rT, $rA", IntSimple, + [(set i64:$rT, (ineg i64:$rA))]>; +let Uses = [CARRY] in { +defm ADDE8 : XOForm_1rc<31, 138, 0, (outs g8rc:$rT), (ins g8rc:$rA, g8rc:$rB), + "adde", "$rT, $rA, $rB", IntGeneral, + [(set i64:$rT, (adde i64:$rA, i64:$rB))]>; +defm ADDME8 : XOForm_3rc<31, 234, 0, (outs g8rc:$rT), (ins g8rc:$rA), + "addme", "$rT, $rA", IntGeneral, + [(set i64:$rT, (adde i64:$rA, -1))]>; +defm ADDZE8 : XOForm_3rc<31, 202, 0, (outs g8rc:$rT), (ins g8rc:$rA), + "addze", "$rT, $rA", IntGeneral, + [(set i64:$rT, (adde i64:$rA, 0))]>; +defm SUBFE8 : XOForm_1rc<31, 136, 0, (outs g8rc:$rT), (ins g8rc:$rA, g8rc:$rB), + "subfe", "$rT, $rA, $rB", IntGeneral, + [(set i64:$rT, (sube i64:$rB, i64:$rA))]>; +defm SUBFME8 : XOForm_3rc<31, 232, 0, (outs g8rc:$rT), (ins g8rc:$rA), + "subfme", "$rT, $rA", IntGeneral, + [(set i64:$rT, (sube -1, i64:$rA))]>; +defm SUBFZE8 : XOForm_3rc<31, 200, 0, (outs g8rc:$rT), (ins g8rc:$rA), + "subfze", "$rT, $rA", IntGeneral, + [(set i64:$rT, (sube 0, i64:$rA))]>; } - -def EXTSB8 : XForm_11<31, 954, (outs G8RC:$rA), (ins G8RC:$rS), - "extsb $rA, $rS", IntSimple, - [(set G8RC:$rA, (sext_inreg G8RC:$rS, i8))]>; -def EXTSH8 : XForm_11<31, 922, (outs G8RC:$rA), (ins G8RC:$rS), - "extsh $rA, $rS", IntSimple, - [(set G8RC:$rA, (sext_inreg G8RC:$rS, i16))]>; - -def EXTSW : XForm_11<31, 986, (outs G8RC:$rA), (ins G8RC:$rS), - "extsw $rA, $rS", IntSimple, - [(set G8RC:$rA, (sext_inreg G8RC:$rS, i32))]>, isPPC64; -/// EXTSW_32 - Just like EXTSW, but works on '32-bit' registers. -def EXTSW_32 : XForm_11<31, 986, (outs GPRC:$rA), (ins GPRC:$rS), - "extsw $rA, $rS", IntSimple, - [(set GPRC:$rA, (PPCextsw_32 GPRC:$rS))]>, isPPC64; -def EXTSW_32_64 : XForm_11<31, 986, (outs G8RC:$rA), (ins GPRC:$rS), - "extsw $rA, $rS", IntSimple, - [(set G8RC:$rA, (sext GPRC:$rS))]>, isPPC64; -let Defs = [CARRY] in { -def SRADI : XSForm_1<31, 413, (outs G8RC:$rA), (ins G8RC:$rS, u6imm:$SH), - "sradi $rA, $rS, $SH", IntRotateDI, - [(set G8RC:$rA, (sra G8RC:$rS, (i32 imm:$SH)))]>, isPPC64; -} -def CNTLZD : XForm_11<31, 58, (outs G8RC:$rA), (ins G8RC:$rS), - "cntlzd $rA, $rS", IntGeneral, - [(set G8RC:$rA, (ctlz G8RC:$rS))]>; - -def DIVD : XOForm_1<31, 489, 0, (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB), - "divd $rT, $rA, $rB", IntDivD, - [(set G8RC:$rT, (sdiv G8RC:$rA, G8RC:$rB))]>, isPPC64, - PPC970_DGroup_First, PPC970_DGroup_Cracked; -def DIVDU : XOForm_1<31, 457, 0, (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB), - "divdu $rT, $rA, $rB", IntDivD, - [(set G8RC:$rT, (udiv G8RC:$rA, G8RC:$rB))]>, isPPC64, - PPC970_DGroup_First, PPC970_DGroup_Cracked; -def MULLD : XOForm_1<31, 233, 0, (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB), - "mulld $rT, $rA, $rB", IntMulHD, - [(set G8RC:$rT, (mul G8RC:$rA, G8RC:$rB))]>, isPPC64; +defm MULHD : XOForm_1r<31, 73, 0, (outs g8rc:$rT), (ins g8rc:$rA, g8rc:$rB), + "mulhd", "$rT, $rA, $rB", IntMulHW, + [(set i64:$rT, (mulhs i64:$rA, i64:$rB))]>; +defm MULHDU : XOForm_1r<31, 9, 0, (outs g8rc:$rT), (ins g8rc:$rA, g8rc:$rB), + "mulhdu", "$rT, $rA, $rB", IntMulHWU, + [(set i64:$rT, (mulhu i64:$rA, i64:$rB))]>; +} +} // Interpretation64Bit + +let isCompare = 1, neverHasSideEffects = 1 in { + def CMPD : XForm_16_ext<31, 0, (outs crrc:$crD), (ins g8rc:$rA, g8rc:$rB), + "cmpd $crD, $rA, $rB", IntCompare>, isPPC64; + def CMPLD : XForm_16_ext<31, 32, (outs crrc:$crD), (ins g8rc:$rA, g8rc:$rB), + "cmpld $crD, $rA, $rB", IntCompare>, isPPC64; + def CMPDI : DForm_5_ext<11, (outs crrc:$crD), (ins g8rc:$rA, s16imm:$imm), + "cmpdi $crD, $rA, $imm", IntCompare>, isPPC64; + def CMPLDI : DForm_6_ext<10, (outs crrc:$dst), (ins g8rc:$src1, u16imm:$src2), + "cmpldi $dst, $src1, $src2", IntCompare>, isPPC64; +} +let neverHasSideEffects = 1 in { +defm SLD : XForm_6r<31, 27, (outs g8rc:$rA), (ins g8rc:$rS, gprc:$rB), + "sld", "$rA, $rS, $rB", IntRotateD, + [(set i64:$rA, (PPCshl i64:$rS, i32:$rB))]>, isPPC64; +defm SRD : XForm_6r<31, 539, (outs g8rc:$rA), (ins g8rc:$rS, gprc:$rB), + "srd", "$rA, $rS, $rB", IntRotateD, + [(set i64:$rA, (PPCsrl i64:$rS, i32:$rB))]>, isPPC64; +defm SRAD : XForm_6rc<31, 794, (outs g8rc:$rA), (ins g8rc:$rS, gprc:$rB), + "srad", "$rA, $rS, $rB", IntRotateD, + [(set i64:$rA, (PPCsra i64:$rS, i32:$rB))]>, isPPC64; + +let Interpretation64Bit = 1 in { +defm EXTSB8 : XForm_11r<31, 954, (outs g8rc:$rA), (ins g8rc:$rS), + "extsb", "$rA, $rS", IntSimple, + [(set i64:$rA, (sext_inreg i64:$rS, i8))]>; +defm EXTSH8 : XForm_11r<31, 922, (outs g8rc:$rA), (ins g8rc:$rS), + "extsh", "$rA, $rS", IntSimple, + [(set i64:$rA, (sext_inreg i64:$rS, i16))]>; +} // Interpretation64Bit + +defm EXTSW : XForm_11r<31, 986, (outs g8rc:$rA), (ins g8rc:$rS), + "extsw", "$rA, $rS", IntSimple, + [(set i64:$rA, (sext_inreg i64:$rS, i32))]>, isPPC64; +let Interpretation64Bit = 1 in +defm EXTSW_32_64 : XForm_11r<31, 986, (outs g8rc:$rA), (ins gprc:$rS), + "extsw", "$rA, $rS", IntSimple, + [(set i64:$rA, (sext i32:$rS))]>, isPPC64; + +defm SRADI : XSForm_1rc<31, 413, (outs g8rc:$rA), (ins g8rc:$rS, u6imm:$SH), + "sradi", "$rA, $rS, $SH", IntRotateDI, + [(set i64:$rA, (sra i64:$rS, (i32 imm:$SH)))]>, isPPC64; +defm CNTLZD : XForm_11r<31, 58, (outs g8rc:$rA), (ins g8rc:$rS), + "cntlzd", "$rA, $rS", IntGeneral, + [(set i64:$rA, (ctlz i64:$rS))]>; +defm POPCNTD : XForm_11r<31, 506, (outs g8rc:$rA), (ins g8rc:$rS), + "popcntd", "$rA, $rS", IntGeneral, + [(set i64:$rA, (ctpop i64:$rS))]>; + +// popcntw also does a population count on the high 32 bits (storing the +// results in the high 32-bits of the output). We'll ignore that here (which is +// safe because we never separately use the high part of the 64-bit registers). +defm POPCNTW : XForm_11r<31, 378, (outs gprc:$rA), (ins gprc:$rS), + "popcntw", "$rA, $rS", IntGeneral, + [(set i32:$rA, (ctpop i32:$rS))]>; + +defm DIVD : XOForm_1r<31, 489, 0, (outs g8rc:$rT), (ins g8rc:$rA, g8rc:$rB), + "divd", "$rT, $rA, $rB", IntDivD, + [(set i64:$rT, (sdiv i64:$rA, i64:$rB))]>, isPPC64, + PPC970_DGroup_First, PPC970_DGroup_Cracked; +defm DIVDU : XOForm_1r<31, 457, 0, (outs g8rc:$rT), (ins g8rc:$rA, g8rc:$rB), + "divdu", "$rT, $rA, $rB", IntDivD, + [(set i64:$rT, (udiv i64:$rA, i64:$rB))]>, isPPC64, + PPC970_DGroup_First, PPC970_DGroup_Cracked; +defm MULLD : XOForm_1r<31, 233, 0, (outs g8rc:$rT), (ins g8rc:$rA, g8rc:$rB), + "mulld", "$rT, $rA, $rB", IntMulHD, + [(set i64:$rT, (mul i64:$rA, i64:$rB))]>, isPPC64; +} + +let neverHasSideEffects = 1 in { let isCommutable = 1 in { -def RLDIMI : MDForm_1<30, 3, - (outs G8RC:$rA), (ins G8RC:$rSi, G8RC:$rS, u6imm:$SH, u6imm:$MB), - "rldimi $rA, $rS, $SH, $MB", IntRotateDI, - []>, isPPC64, RegConstraint<"$rSi = $rA">, - NoEncode<"$rSi">; +defm RLDIMI : MDForm_1r<30, 3, (outs g8rc:$rA), + (ins g8rc:$rSi, g8rc:$rS, u6imm:$SH, u6imm:$MBE), + "rldimi", "$rA, $rS, $SH, $MBE", IntRotateDI, + []>, isPPC64, RegConstraint<"$rSi = $rA">, + NoEncode<"$rSi">; } // Rotate instructions. -def RLDCL : MDForm_1<30, 0, - (outs G8RC:$rA), (ins G8RC:$rS, GPRC:$rB, u6imm:$MBE), - "rldcl $rA, $rS, $rB, $MBE", IntRotateD, - []>, isPPC64; -def RLDICL : MDForm_1<30, 0, - (outs G8RC:$rA), (ins G8RC:$rS, u6imm:$SH, u6imm:$MBE), - "rldicl $rA, $rS, $SH, $MBE", IntRotateDI, - []>, isPPC64; -def RLDICR : MDForm_1<30, 1, - (outs G8RC:$rA), (ins G8RC:$rS, u6imm:$SH, u6imm:$MBE), - "rldicr $rA, $rS, $SH, $MBE", IntRotateDI, - []>, isPPC64; - -def RLWINM8 : MForm_2<21, - (outs G8RC:$rA), (ins G8RC:$rS, u5imm:$SH, u5imm:$MB, u5imm:$ME), - "rlwinm $rA, $rS, $SH, $MB, $ME", IntGeneral, - []>; - +defm RLDCL : MDSForm_1r<30, 8, + (outs g8rc:$rA), (ins g8rc:$rS, gprc:$rB, u6imm:$MBE), + "rldcl", "$rA, $rS, $rB, $MBE", IntRotateD, + []>, isPPC64; +defm RLDICL : MDForm_1r<30, 0, + (outs g8rc:$rA), (ins g8rc:$rS, u6imm:$SH, u6imm:$MBE), + "rldicl", "$rA, $rS, $SH, $MBE", IntRotateDI, + []>, isPPC64; +defm RLDICR : MDForm_1r<30, 1, + (outs g8rc:$rA), (ins g8rc:$rS, u6imm:$SH, u6imm:$MBE), + "rldicr", "$rA, $rS, $SH, $MBE", IntRotateDI, + []>, isPPC64; + +let Interpretation64Bit = 1 in { +defm RLWINM8 : MForm_2r<21, (outs g8rc:$rA), + (ins g8rc:$rS, u5imm:$SH, u5imm:$MB, u5imm:$ME), + "rlwinm", "$rA, $rS, $SH, $MB, $ME", IntGeneral, + []>; + +let isSelect = 1 in def ISEL8 : AForm_4<31, 15, - (outs G8RC:$rT), (ins G8RC:$rA, G8RC:$rB, pred:$cond), + (outs g8rc:$rT), (ins g8rc_nox0:$rA, g8rc:$rB, crbitrc:$cond), "isel $rT, $rA, $rB, $cond", IntGeneral, []>; +} // Interpretation64Bit +} // neverHasSideEffects = 1 } // End FXU Operations. @@ -549,157 +569,161 @@ def ISEL8 : AForm_4<31, 15, // Sign extending loads. let canFoldAsLoad = 1, PPC970_Unit = 2 in { -def LHA8: DForm_1<42, (outs G8RC:$rD), (ins memri:$src), +let Interpretation64Bit = 1 in +def LHA8: DForm_1<42, (outs g8rc:$rD), (ins memri:$src), "lha $rD, $src", LdStLHA, - [(set G8RC:$rD, (sextloadi16 iaddr:$src))]>, + [(set i64:$rD, (sextloadi16 iaddr:$src))]>, PPC970_DGroup_Cracked; -def LWA : DSForm_1<58, 2, (outs G8RC:$rD), (ins memrix:$src), +def LWA : DSForm_1<58, 2, (outs g8rc:$rD), (ins memrix:$src), "lwa $rD, $src", LdStLWA, - [(set G8RC:$rD, + [(set i64:$rD, (aligned4sextloadi32 ixaddr:$src))]>, isPPC64, PPC970_DGroup_Cracked; -def LHAX8: XForm_1<31, 343, (outs G8RC:$rD), (ins memrr:$src), +let Interpretation64Bit = 1 in +def LHAX8: XForm_1<31, 343, (outs g8rc:$rD), (ins memrr:$src), "lhax $rD, $src", LdStLHA, - [(set G8RC:$rD, (sextloadi16 xaddr:$src))]>, + [(set i64:$rD, (sextloadi16 xaddr:$src))]>, PPC970_DGroup_Cracked; -def LWAX : XForm_1<31, 341, (outs G8RC:$rD), (ins memrr:$src), +def LWAX : XForm_1<31, 341, (outs g8rc:$rD), (ins memrr:$src), "lwax $rD, $src", LdStLHA, - [(set G8RC:$rD, (sextloadi32 xaddr:$src))]>, isPPC64, + [(set i64:$rD, (sextloadi32 xaddr:$src))]>, isPPC64, PPC970_DGroup_Cracked; // Update forms. -let mayLoad = 1 in -def LHAU8 : DForm_1a<43, (outs G8RC:$rD, ptr_rc:$ea_result), (ins symbolLo:$disp, - ptr_rc:$rA), - "lhau $rD, $disp($rA)", LdStLHAU, - []>, RegConstraint<"$rA = $ea_result">, +let mayLoad = 1, neverHasSideEffects = 1 in { +let Interpretation64Bit = 1 in +def LHAU8 : DForm_1<43, (outs g8rc:$rD, ptr_rc_nor0:$ea_result), + (ins memri:$addr), + "lhau $rD, $addr", LdStLHAU, + []>, RegConstraint<"$addr.reg = $ea_result">, NoEncode<"$ea_result">; // NO LWAU! -def LHAUX8 : XForm_1<31, 375, (outs G8RC:$rD, ptr_rc:$ea_result), +let Interpretation64Bit = 1 in +def LHAUX8 : XForm_1<31, 375, (outs g8rc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "lhaux $rD, $addr", LdStLHAU, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">; -def LWAUX : XForm_1<31, 373, (outs G8RC:$rD, ptr_rc:$ea_result), +def LWAUX : XForm_1<31, 373, (outs g8rc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "lwaux $rD, $addr", LdStLHAU, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">, isPPC64; } +} +let Interpretation64Bit = 1 in { // Zero extending loads. let canFoldAsLoad = 1, PPC970_Unit = 2 in { -def LBZ8 : DForm_1<34, (outs G8RC:$rD), (ins memri:$src), +def LBZ8 : DForm_1<34, (outs g8rc:$rD), (ins memri:$src), "lbz $rD, $src", LdStLoad, - [(set G8RC:$rD, (zextloadi8 iaddr:$src))]>; -def LHZ8 : DForm_1<40, (outs G8RC:$rD), (ins memri:$src), + [(set i64:$rD, (zextloadi8 iaddr:$src))]>; +def LHZ8 : DForm_1<40, (outs g8rc:$rD), (ins memri:$src), "lhz $rD, $src", LdStLoad, - [(set G8RC:$rD, (zextloadi16 iaddr:$src))]>; -def LWZ8 : DForm_1<32, (outs G8RC:$rD), (ins memri:$src), + [(set i64:$rD, (zextloadi16 iaddr:$src))]>; +def LWZ8 : DForm_1<32, (outs g8rc:$rD), (ins memri:$src), "lwz $rD, $src", LdStLoad, - [(set G8RC:$rD, (zextloadi32 iaddr:$src))]>, isPPC64; + [(set i64:$rD, (zextloadi32 iaddr:$src))]>, isPPC64; -def LBZX8 : XForm_1<31, 87, (outs G8RC:$rD), (ins memrr:$src), +def LBZX8 : XForm_1<31, 87, (outs g8rc:$rD), (ins memrr:$src), "lbzx $rD, $src", LdStLoad, - [(set G8RC:$rD, (zextloadi8 xaddr:$src))]>; -def LHZX8 : XForm_1<31, 279, (outs G8RC:$rD), (ins memrr:$src), + [(set i64:$rD, (zextloadi8 xaddr:$src))]>; +def LHZX8 : XForm_1<31, 279, (outs g8rc:$rD), (ins memrr:$src), "lhzx $rD, $src", LdStLoad, - [(set G8RC:$rD, (zextloadi16 xaddr:$src))]>; -def LWZX8 : XForm_1<31, 23, (outs G8RC:$rD), (ins memrr:$src), + [(set i64:$rD, (zextloadi16 xaddr:$src))]>; +def LWZX8 : XForm_1<31, 23, (outs g8rc:$rD), (ins memrr:$src), "lwzx $rD, $src", LdStLoad, - [(set G8RC:$rD, (zextloadi32 xaddr:$src))]>; + [(set i64:$rD, (zextloadi32 xaddr:$src))]>; // Update forms. -let mayLoad = 1 in { -def LBZU8 : DForm_1<35, (outs G8RC:$rD, ptr_rc:$ea_result), (ins memri:$addr), +let mayLoad = 1, neverHasSideEffects = 1 in { +def LBZU8 : DForm_1<35, (outs g8rc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr), "lbzu $rD, $addr", LdStLoadUpd, []>, RegConstraint<"$addr.reg = $ea_result">, NoEncode<"$ea_result">; -def LHZU8 : DForm_1<41, (outs G8RC:$rD, ptr_rc:$ea_result), (ins memri:$addr), +def LHZU8 : DForm_1<41, (outs g8rc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr), "lhzu $rD, $addr", LdStLoadUpd, []>, RegConstraint<"$addr.reg = $ea_result">, NoEncode<"$ea_result">; -def LWZU8 : DForm_1<33, (outs G8RC:$rD, ptr_rc:$ea_result), (ins memri:$addr), +def LWZU8 : DForm_1<33, (outs g8rc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr), "lwzu $rD, $addr", LdStLoadUpd, []>, RegConstraint<"$addr.reg = $ea_result">, NoEncode<"$ea_result">; -def LBZUX8 : XForm_1<31, 119, (outs G8RC:$rD, ptr_rc:$ea_result), +def LBZUX8 : XForm_1<31, 119, (outs g8rc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "lbzux $rD, $addr", LdStLoadUpd, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">; -def LHZUX8 : XForm_1<31, 311, (outs G8RC:$rD, ptr_rc:$ea_result), +def LHZUX8 : XForm_1<31, 311, (outs g8rc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "lhzux $rD, $addr", LdStLoadUpd, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">; -def LWZUX8 : XForm_1<31, 55, (outs G8RC:$rD, ptr_rc:$ea_result), +def LWZUX8 : XForm_1<31, 55, (outs g8rc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "lwzux $rD, $addr", LdStLoadUpd, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">; } } +} // Interpretation64Bit // Full 8-byte loads. let canFoldAsLoad = 1, PPC970_Unit = 2 in { -def LD : DSForm_1<58, 0, (outs G8RC:$rD), (ins memrix:$src), +def LD : DSForm_1<58, 0, (outs g8rc:$rD), (ins memrix:$src), "ld $rD, $src", LdStLD, - [(set G8RC:$rD, (aligned4load ixaddr:$src))]>, isPPC64; -def LDrs : DSForm_1<58, 0, (outs G8RC:$rD), (ins memrs:$src), - "ld $rD, $src", LdStLD, - []>, isPPC64; + [(set i64:$rD, (aligned4load ixaddr:$src))]>, isPPC64; // The following three definitions are selected for small code model only. // Otherwise, we need to create two instructions to form a 32-bit offset, // so we have a custom matcher for TOC_ENTRY in PPCDAGToDAGIsel::Select(). -def LDtoc: Pseudo<(outs G8RC:$rD), (ins tocentry:$disp, G8RC:$reg), +def LDtoc: Pseudo<(outs g8rc:$rD), (ins tocentry:$disp, g8rc:$reg), "#LDtoc", - [(set G8RC:$rD, - (PPCtoc_entry tglobaladdr:$disp, G8RC:$reg))]>, isPPC64; -def LDtocJTI: Pseudo<(outs G8RC:$rD), (ins tocentry:$disp, G8RC:$reg), + [(set i64:$rD, + (PPCtoc_entry tglobaladdr:$disp, i64:$reg))]>, isPPC64; +def LDtocJTI: Pseudo<(outs g8rc:$rD), (ins tocentry:$disp, g8rc:$reg), "#LDtocJTI", - [(set G8RC:$rD, - (PPCtoc_entry tjumptable:$disp, G8RC:$reg))]>, isPPC64; -def LDtocCPT: Pseudo<(outs G8RC:$rD), (ins tocentry:$disp, G8RC:$reg), + [(set i64:$rD, + (PPCtoc_entry tjumptable:$disp, i64:$reg))]>, isPPC64; +def LDtocCPT: Pseudo<(outs g8rc:$rD), (ins tocentry:$disp, g8rc:$reg), "#LDtocCPT", - [(set G8RC:$rD, - (PPCtoc_entry tconstpool:$disp, G8RC:$reg))]>, isPPC64; + [(set i64:$rD, + (PPCtoc_entry tconstpool:$disp, i64:$reg))]>, isPPC64; -let hasSideEffects = 1 in { +let hasSideEffects = 1, isCodeGenOnly = 1 in { let RST = 2, DS = 2 in -def LDinto_toc: DSForm_1a<58, 0, (outs), (ins G8RC:$reg), +def LDinto_toc: DSForm_1a<58, 0, (outs), (ins g8rc:$reg), "ld 2, 8($reg)", LdStLD, - [(PPCload_toc G8RC:$reg)]>, isPPC64; + [(PPCload_toc i64:$reg)]>, isPPC64; let RST = 2, DS = 10, RA = 1 in def LDtoc_restore : DSForm_1a<58, 0, (outs), (ins), "ld 2, 40(1)", LdStLD, [(PPCtoc_restore)]>, isPPC64; } -def LDX : XForm_1<31, 21, (outs G8RC:$rD), (ins memrr:$src), +def LDX : XForm_1<31, 21, (outs g8rc:$rD), (ins memrr:$src), "ldx $rD, $src", LdStLD, - [(set G8RC:$rD, (load xaddr:$src))]>, isPPC64; -let isCodeGenOnly = 1 in -def LDXu : XForm_1<31, 21, (outs G8RC:$rD), (ins memrr:$src), - "ldx $rD, $src", LdStLD, - [(set G8RC:$rD, (load xaddr:$src))]>, isPPC64; - -let mayLoad = 1 in -def LDU : DSForm_1<58, 1, (outs G8RC:$rD, ptr_rc:$ea_result), (ins memrix:$addr), + [(set i64:$rD, (load xaddr:$src))]>, isPPC64; +def LDBRX : XForm_1<31, 532, (outs g8rc:$rD), (ins memrr:$src), + "ldbrx $rD, $src", LdStLoad, + [(set i64:$rD, (PPClbrx xoaddr:$src, i64))]>, isPPC64; + +let mayLoad = 1, neverHasSideEffects = 1 in { +def LDU : DSForm_1<58, 1, (outs g8rc:$rD, ptr_rc_nor0:$ea_result), (ins memrix:$addr), "ldu $rD, $addr", LdStLDU, []>, RegConstraint<"$addr.reg = $ea_result">, isPPC64, NoEncode<"$ea_result">; -def LDUX : XForm_1<31, 53, (outs G8RC:$rD, ptr_rc:$ea_result), +def LDUX : XForm_1<31, 53, (outs g8rc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "ldux $rD, $addr", LdStLDU, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">, isPPC64; } +} def : Pat<(PPCload ixaddr:$src), (LD ixaddr:$src)>; @@ -707,189 +731,173 @@ def : Pat<(PPCload xaddr:$src), (LDX xaddr:$src)>; // Support for medium and large code model. -def ADDIStocHA: Pseudo<(outs G8RC:$rD), (ins G8RC:$reg, tocentry:$disp), +def ADDIStocHA: Pseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, tocentry:$disp), "#ADDIStocHA", - [(set G8RC:$rD, - (PPCaddisTocHA G8RC:$reg, tglobaladdr:$disp))]>, + [(set i64:$rD, + (PPCaddisTocHA i64:$reg, tglobaladdr:$disp))]>, isPPC64; -def LDtocL: Pseudo<(outs G8RC:$rD), (ins tocentry:$disp, G8RC:$reg), +def LDtocL: Pseudo<(outs g8rc:$rD), (ins tocentry:$disp, g8rc_nox0:$reg), "#LDtocL", - [(set G8RC:$rD, - (PPCldTocL tglobaladdr:$disp, G8RC:$reg))]>, isPPC64; -def ADDItocL: Pseudo<(outs G8RC:$rD), (ins G8RC:$reg, tocentry:$disp), + [(set i64:$rD, + (PPCldTocL tglobaladdr:$disp, i64:$reg))]>, isPPC64; +def ADDItocL: Pseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, tocentry:$disp), "#ADDItocL", - [(set G8RC:$rD, - (PPCaddiTocL G8RC:$reg, tglobaladdr:$disp))]>, isPPC64; + [(set i64:$rD, + (PPCaddiTocL i64:$reg, tglobaladdr:$disp))]>, isPPC64; // Support for thread-local storage. -def ADDISgotTprelHA: Pseudo<(outs G8RC:$rD), (ins G8RC:$reg, symbolHi64:$disp), +def ADDISgotTprelHA: Pseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, symbolHi64:$disp), "#ADDISgotTprelHA", - [(set G8RC:$rD, - (PPCaddisGotTprelHA G8RC:$reg, + [(set i64:$rD, + (PPCaddisGotTprelHA i64:$reg, tglobaltlsaddr:$disp))]>, isPPC64; -def LDgotTprelL: Pseudo<(outs G8RC:$rD), (ins symbolLo64:$disp, G8RC:$reg), +def LDgotTprelL: Pseudo<(outs g8rc:$rD), (ins symbolLo64:$disp, g8rc_nox0:$reg), "#LDgotTprelL", - [(set G8RC:$rD, - (PPCldGotTprelL tglobaltlsaddr:$disp, G8RC:$reg))]>, + [(set i64:$rD, + (PPCldGotTprelL tglobaltlsaddr:$disp, i64:$reg))]>, isPPC64; -def : Pat<(PPCaddTls G8RC:$in, tglobaltlsaddr:$g), - (ADD8TLS G8RC:$in, tglobaltlsaddr:$g)>; -def ADDIStlsgdHA: Pseudo<(outs G8RC:$rD), (ins G8RC:$reg, symbolHi64:$disp), +def : Pat<(PPCaddTls i64:$in, tglobaltlsaddr:$g), + (ADD8TLS $in, tglobaltlsaddr:$g)>; +def ADDIStlsgdHA: Pseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, symbolHi64:$disp), "#ADDIStlsgdHA", - [(set G8RC:$rD, - (PPCaddisTlsgdHA G8RC:$reg, tglobaltlsaddr:$disp))]>, + [(set i64:$rD, + (PPCaddisTlsgdHA i64:$reg, tglobaltlsaddr:$disp))]>, isPPC64; -def ADDItlsgdL : Pseudo<(outs G8RC:$rD), (ins G8RC:$reg, symbolLo64:$disp), +def ADDItlsgdL : Pseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, symbolLo64:$disp), "#ADDItlsgdL", - [(set G8RC:$rD, - (PPCaddiTlsgdL G8RC:$reg, tglobaltlsaddr:$disp))]>, + [(set i64:$rD, + (PPCaddiTlsgdL i64:$reg, tglobaltlsaddr:$disp))]>, isPPC64; -def GETtlsADDR : Pseudo<(outs G8RC:$rD), (ins G8RC:$reg, tlsgd:$sym), +def GETtlsADDR : Pseudo<(outs g8rc:$rD), (ins g8rc:$reg, tlsgd:$sym), "#GETtlsADDR", - [(set G8RC:$rD, - (PPCgetTlsAddr G8RC:$reg, tglobaltlsaddr:$sym))]>, + [(set i64:$rD, + (PPCgetTlsAddr i64:$reg, tglobaltlsaddr:$sym))]>, isPPC64; -def ADDIStlsldHA: Pseudo<(outs G8RC:$rD), (ins G8RC:$reg, symbolHi64:$disp), +def ADDIStlsldHA: Pseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, symbolHi64:$disp), "#ADDIStlsldHA", - [(set G8RC:$rD, - (PPCaddisTlsldHA G8RC:$reg, tglobaltlsaddr:$disp))]>, + [(set i64:$rD, + (PPCaddisTlsldHA i64:$reg, tglobaltlsaddr:$disp))]>, isPPC64; -def ADDItlsldL : Pseudo<(outs G8RC:$rD), (ins G8RC:$reg, symbolLo64:$disp), +def ADDItlsldL : Pseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, symbolLo64:$disp), "#ADDItlsldL", - [(set G8RC:$rD, - (PPCaddiTlsldL G8RC:$reg, tglobaltlsaddr:$disp))]>, + [(set i64:$rD, + (PPCaddiTlsldL i64:$reg, tglobaltlsaddr:$disp))]>, isPPC64; -def GETtlsldADDR : Pseudo<(outs G8RC:$rD), (ins G8RC:$reg, tlsgd:$sym), +def GETtlsldADDR : Pseudo<(outs g8rc:$rD), (ins g8rc:$reg, tlsgd:$sym), "#GETtlsldADDR", - [(set G8RC:$rD, - (PPCgetTlsldAddr G8RC:$reg, tglobaltlsaddr:$sym))]>, + [(set i64:$rD, + (PPCgetTlsldAddr i64:$reg, tglobaltlsaddr:$sym))]>, isPPC64; -def ADDISdtprelHA: Pseudo<(outs G8RC:$rD), (ins G8RC:$reg, symbolHi64:$disp), +def ADDISdtprelHA: Pseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, symbolHi64:$disp), "#ADDISdtprelHA", - [(set G8RC:$rD, - (PPCaddisDtprelHA G8RC:$reg, + [(set i64:$rD, + (PPCaddisDtprelHA i64:$reg, tglobaltlsaddr:$disp))]>, isPPC64; -def ADDIdtprelL : Pseudo<(outs G8RC:$rD), (ins G8RC:$reg, symbolLo64:$disp), +def ADDIdtprelL : Pseudo<(outs g8rc:$rD), (ins g8rc_nox0:$reg, symbolLo64:$disp), "#ADDIdtprelL", - [(set G8RC:$rD, - (PPCaddiDtprelL G8RC:$reg, tglobaltlsaddr:$disp))]>, + [(set i64:$rD, + (PPCaddiDtprelL i64:$reg, tglobaltlsaddr:$disp))]>, isPPC64; let PPC970_Unit = 2 in { +let Interpretation64Bit = 1 in { // Truncating stores. -def STB8 : DForm_1<38, (outs), (ins G8RC:$rS, memri:$src), +def STB8 : DForm_1<38, (outs), (ins g8rc:$rS, memri:$src), "stb $rS, $src", LdStStore, - [(truncstorei8 G8RC:$rS, iaddr:$src)]>; -def STH8 : DForm_1<44, (outs), (ins G8RC:$rS, memri:$src), + [(truncstorei8 i64:$rS, iaddr:$src)]>; +def STH8 : DForm_1<44, (outs), (ins g8rc:$rS, memri:$src), "sth $rS, $src", LdStStore, - [(truncstorei16 G8RC:$rS, iaddr:$src)]>; -def STW8 : DForm_1<36, (outs), (ins G8RC:$rS, memri:$src), + [(truncstorei16 i64:$rS, iaddr:$src)]>; +def STW8 : DForm_1<36, (outs), (ins g8rc:$rS, memri:$src), "stw $rS, $src", LdStStore, - [(truncstorei32 G8RC:$rS, iaddr:$src)]>; -def STBX8 : XForm_8<31, 215, (outs), (ins G8RC:$rS, memrr:$dst), + [(truncstorei32 i64:$rS, iaddr:$src)]>; +def STBX8 : XForm_8<31, 215, (outs), (ins g8rc:$rS, memrr:$dst), "stbx $rS, $dst", LdStStore, - [(truncstorei8 G8RC:$rS, xaddr:$dst)]>, + [(truncstorei8 i64:$rS, xaddr:$dst)]>, PPC970_DGroup_Cracked; -def STHX8 : XForm_8<31, 407, (outs), (ins G8RC:$rS, memrr:$dst), +def STHX8 : XForm_8<31, 407, (outs), (ins g8rc:$rS, memrr:$dst), "sthx $rS, $dst", LdStStore, - [(truncstorei16 G8RC:$rS, xaddr:$dst)]>, + [(truncstorei16 i64:$rS, xaddr:$dst)]>, PPC970_DGroup_Cracked; -def STWX8 : XForm_8<31, 151, (outs), (ins G8RC:$rS, memrr:$dst), +def STWX8 : XForm_8<31, 151, (outs), (ins g8rc:$rS, memrr:$dst), "stwx $rS, $dst", LdStStore, - [(truncstorei32 G8RC:$rS, xaddr:$dst)]>, + [(truncstorei32 i64:$rS, xaddr:$dst)]>, PPC970_DGroup_Cracked; +} // Interpretation64Bit + // Normal 8-byte stores. -def STD : DSForm_1<62, 0, (outs), (ins G8RC:$rS, memrix:$dst), +def STD : DSForm_1<62, 0, (outs), (ins g8rc:$rS, memrix:$dst), "std $rS, $dst", LdStSTD, - [(aligned4store G8RC:$rS, ixaddr:$dst)]>, isPPC64; -def STDX : XForm_8<31, 149, (outs), (ins G8RC:$rS, memrr:$dst), + [(aligned4store i64:$rS, ixaddr:$dst)]>, isPPC64; +def STDX : XForm_8<31, 149, (outs), (ins g8rc:$rS, memrr:$dst), "stdx $rS, $dst", LdStSTD, - [(store G8RC:$rS, xaddr:$dst)]>, isPPC64, + [(store i64:$rS, xaddr:$dst)]>, isPPC64, + PPC970_DGroup_Cracked; +def STDBRX: XForm_8<31, 660, (outs), (ins g8rc:$rS, memrr:$dst), + "stdbrx $rS, $dst", LdStStore, + [(PPCstbrx i64:$rS, xoaddr:$dst, i64)]>, isPPC64, PPC970_DGroup_Cracked; } -let PPC970_Unit = 2 in { +// Stores with Update (pre-inc). +let PPC970_Unit = 2, mayStore = 1 in { +let Interpretation64Bit = 1 in { +def STBU8 : DForm_1<39, (outs ptr_rc_nor0:$ea_res), (ins g8rc:$rS, memri:$dst), + "stbu $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">; +def STHU8 : DForm_1<45, (outs ptr_rc_nor0:$ea_res), (ins g8rc:$rS, memri:$dst), + "sthu $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">; +def STWU8 : DForm_1<37, (outs ptr_rc_nor0:$ea_res), (ins g8rc:$rS, memri:$dst), + "stwu $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">; +def STDU : DSForm_1<62, 1, (outs ptr_rc_nor0:$ea_res), (ins g8rc:$rS, memrix:$dst), + "stdu $rS, $dst", LdStSTDU, []>, + RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">, + isPPC64; -def STBU8 : DForm_1a<39, (outs ptr_rc:$ea_res), (ins G8RC:$rS, - symbolLo:$ptroff, ptr_rc:$ptrreg), - "stbu $rS, $ptroff($ptrreg)", LdStStoreUpd, - [(set ptr_rc:$ea_res, - (pre_truncsti8 G8RC:$rS, ptr_rc:$ptrreg, - iaddroff:$ptroff))]>, - RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">; -def STHU8 : DForm_1a<45, (outs ptr_rc:$ea_res), (ins G8RC:$rS, - symbolLo:$ptroff, ptr_rc:$ptrreg), - "sthu $rS, $ptroff($ptrreg)", LdStStoreUpd, - [(set ptr_rc:$ea_res, - (pre_truncsti16 G8RC:$rS, ptr_rc:$ptrreg, - iaddroff:$ptroff))]>, - RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">; - -def STWU8 : DForm_1a<37, (outs ptr_rc:$ea_res), (ins G8RC:$rS, - symbolLo:$ptroff, ptr_rc:$ptrreg), - "stwu $rS, $ptroff($ptrreg)", LdStStoreUpd, - [(set ptr_rc:$ea_res, - (pre_truncsti32 G8RC:$rS, ptr_rc:$ptrreg, - iaddroff:$ptroff))]>, - RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">; - -def STDU : DSForm_1a<62, 1, (outs ptr_rc:$ea_res), (ins G8RC:$rS, - s16immX4:$ptroff, ptr_rc:$ptrreg), - "stdu $rS, $ptroff($ptrreg)", LdStSTDU, - [(set ptr_rc:$ea_res, - (aligned4pre_store G8RC:$rS, ptr_rc:$ptrreg, - iaddroff:$ptroff))]>, - RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">, - isPPC64; - - -def STBUX8 : XForm_8<31, 247, (outs ptr_rc:$ea_res), - (ins G8RC:$rS, ptr_rc:$ptroff, ptr_rc:$ptrreg), - "stbux $rS, $ptroff, $ptrreg", LdStStoreUpd, - [(set ptr_rc:$ea_res, - (pre_truncsti8 G8RC:$rS, - ptr_rc:$ptrreg, xaddroff:$ptroff))]>, - RegConstraint<"$ptroff = $ea_res">, NoEncode<"$ea_res">, +def STBUX8: XForm_8<31, 247, (outs ptr_rc_nor0:$ea_res), (ins g8rc:$rS, memrr:$dst), + "stbux $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">, PPC970_DGroup_Cracked; - -def STHUX8 : XForm_8<31, 439, (outs ptr_rc:$ea_res), - (ins G8RC:$rS, ptr_rc:$ptroff, ptr_rc:$ptrreg), - "sthux $rS, $ptroff, $ptrreg", LdStStoreUpd, - [(set ptr_rc:$ea_res, - (pre_truncsti16 G8RC:$rS, - ptr_rc:$ptrreg, xaddroff:$ptroff))]>, - RegConstraint<"$ptroff = $ea_res">, NoEncode<"$ea_res">, +def STHUX8: XForm_8<31, 439, (outs ptr_rc_nor0:$ea_res), (ins g8rc:$rS, memrr:$dst), + "sthux $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">, PPC970_DGroup_Cracked; - -def STWUX8 : XForm_8<31, 183, (outs ptr_rc:$ea_res), - (ins G8RC:$rS, ptr_rc:$ptroff, ptr_rc:$ptrreg), - "stwux $rS, $ptroff, $ptrreg", LdStStoreUpd, - [(set ptr_rc:$ea_res, - (pre_truncsti32 G8RC:$rS, - ptr_rc:$ptrreg, xaddroff:$ptroff))]>, - RegConstraint<"$ptroff = $ea_res">, NoEncode<"$ea_res">, +def STWUX8: XForm_8<31, 183, (outs ptr_rc_nor0:$ea_res), (ins g8rc:$rS, memrr:$dst), + "stwux $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">, PPC970_DGroup_Cracked; +} // Interpretation64Bit -def STDUX : XForm_8<31, 181, (outs ptr_rc:$ea_res), - (ins G8RC:$rS, ptr_rc:$ptroff, ptr_rc:$ptrreg), - "stdux $rS, $ptroff, $ptrreg", LdStSTDU, - [(set ptr_rc:$ea_res, - (pre_store G8RC:$rS, ptr_rc:$ptrreg, xaddroff:$ptroff))]>, - RegConstraint<"$ptroff = $ea_res">, NoEncode<"$ea_res">, +def STDUX : XForm_8<31, 181, (outs ptr_rc_nor0:$ea_res), (ins g8rc:$rS, memrr:$dst), + "stdux $rS, $dst", LdStSTDU, []>, + RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">, PPC970_DGroup_Cracked, isPPC64; - -// STD_32/STDX_32 - Just like STD/STDX, but uses a '32-bit' input register. -def STD_32 : DSForm_1<62, 0, (outs), (ins GPRC:$rT, memrix:$dst), - "std $rT, $dst", LdStSTD, - [(PPCstd_32 GPRC:$rT, ixaddr:$dst)]>, isPPC64; -def STDX_32 : XForm_8<31, 149, (outs), (ins GPRC:$rT, memrr:$dst), - "stdx $rT, $dst", LdStSTD, - [(PPCstd_32 GPRC:$rT, xaddr:$dst)]>, isPPC64, - PPC970_DGroup_Cracked; } +// Patterns to match the pre-inc stores. We can't put the patterns on +// the instruction definitions directly as ISel wants the address base +// and offset to be separate operands, not a single complex operand. +def : Pat<(pre_truncsti8 i64:$rS, iPTR:$ptrreg, iaddroff:$ptroff), + (STBU8 $rS, iaddroff:$ptroff, $ptrreg)>; +def : Pat<(pre_truncsti16 i64:$rS, iPTR:$ptrreg, iaddroff:$ptroff), + (STHU8 $rS, iaddroff:$ptroff, $ptrreg)>; +def : Pat<(pre_truncsti32 i64:$rS, iPTR:$ptrreg, iaddroff:$ptroff), + (STWU8 $rS, iaddroff:$ptroff, $ptrreg)>; +def : Pat<(aligned4pre_store i64:$rS, iPTR:$ptrreg, iaddroff:$ptroff), + (STDU $rS, iaddroff:$ptroff, $ptrreg)>; + +def : Pat<(pre_truncsti8 i64:$rS, iPTR:$ptrreg, iPTR:$ptroff), + (STBUX8 $rS, $ptrreg, $ptroff)>; +def : Pat<(pre_truncsti16 i64:$rS, iPTR:$ptrreg, iPTR:$ptroff), + (STHUX8 $rS, $ptrreg, $ptroff)>; +def : Pat<(pre_truncsti32 i64:$rS, iPTR:$ptrreg, iPTR:$ptroff), + (STWUX8 $rS, $ptrreg, $ptroff)>; +def : Pat<(pre_store i64:$rS, iPTR:$ptrreg, iPTR:$ptroff), + (STDUX $rS, $ptrreg, $ptroff)>; //===----------------------------------------------------------------------===// @@ -897,13 +905,30 @@ def STDX_32 : XForm_8<31, 149, (outs), (ins GPRC:$rT, memrr:$dst), // -let PPC970_Unit = 3, Uses = [RM] in { // FPU Operations. -def FCFID : XForm_26<63, 846, (outs F8RC:$frD), (ins F8RC:$frB), - "fcfid $frD, $frB", FPGeneral, - [(set F8RC:$frD, (PPCfcfid F8RC:$frB))]>, isPPC64; -def FCTIDZ : XForm_26<63, 815, (outs F8RC:$frD), (ins F8RC:$frB), - "fctidz $frD, $frB", FPGeneral, - [(set F8RC:$frD, (PPCfctidz F8RC:$frB))]>, isPPC64; +let PPC970_Unit = 3, neverHasSideEffects = 1, + Uses = [RM] in { // FPU Operations. +defm FCFID : XForm_26r<63, 846, (outs f8rc:$frD), (ins f8rc:$frB), + "fcfid", "$frD, $frB", FPGeneral, + [(set f64:$frD, (PPCfcfid f64:$frB))]>, isPPC64; +defm FCTIDZ : XForm_26r<63, 815, (outs f8rc:$frD), (ins f8rc:$frB), + "fctidz", "$frD, $frB", FPGeneral, + [(set f64:$frD, (PPCfctidz f64:$frB))]>, isPPC64; + +defm FCFIDU : XForm_26r<63, 974, (outs f8rc:$frD), (ins f8rc:$frB), + "fcfidu", "$frD, $frB", FPGeneral, + [(set f64:$frD, (PPCfcfidu f64:$frB))]>, isPPC64; +defm FCFIDS : XForm_26r<59, 846, (outs f4rc:$frD), (ins f8rc:$frB), + "fcfids", "$frD, $frB", FPGeneral, + [(set f32:$frD, (PPCfcfids f64:$frB))]>, isPPC64; +defm FCFIDUS : XForm_26r<59, 974, (outs f4rc:$frD), (ins f8rc:$frB), + "fcfidus", "$frD, $frB", FPGeneral, + [(set f32:$frD, (PPCfcfidus f64:$frB))]>, isPPC64; +defm FCTIDUZ : XForm_26r<63, 943, (outs f8rc:$frD), (ins f8rc:$frB), + "fctiduz", "$frD, $frB", FPGeneral, + [(set f64:$frD, (PPCfctiduz f64:$frB))]>, isPPC64; +defm FCTIWUZ : XForm_26r<63, 143, (outs f8rc:$frD), (ins f8rc:$frB), + "fctiwuz", "$frD, $frB", FPGeneral, + [(set f64:$frD, (PPCfctiwuz f64:$frB))]>, isPPC64; } @@ -912,13 +937,13 @@ def FCTIDZ : XForm_26<63, 815, (outs F8RC:$frD), (ins F8RC:$frB), // // Extensions and truncates to/from 32-bit regs. -def : Pat<(i64 (zext GPRC:$in)), - (RLDICL (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GPRC:$in, sub_32), +def : Pat<(i64 (zext i32:$in)), + (RLDICL (INSERT_SUBREG (i64 (IMPLICIT_DEF)), $in, sub_32), 0, 32)>; -def : Pat<(i64 (anyext GPRC:$in)), - (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GPRC:$in, sub_32)>; -def : Pat<(i32 (trunc G8RC:$in)), - (EXTRACT_SUBREG G8RC:$in, sub_32)>; +def : Pat<(i64 (anyext i32:$in)), + (INSERT_SUBREG (i64 (IMPLICIT_DEF)), $in, sub_32)>; +def : Pat<(i32 (trunc i64:$in)), + (EXTRACT_SUBREG $in, sub_32)>; // Extending loads with i64 targets. def : Pat<(zextloadi1 iaddr:$src), @@ -945,24 +970,24 @@ def : Pat<(extloadi32 xaddr:$src), // Standard shifts. These are represented separately from the real shifts above // so that we can distinguish between shifts that allow 6-bit and 7-bit shift // amounts. -def : Pat<(sra G8RC:$rS, GPRC:$rB), - (SRAD G8RC:$rS, GPRC:$rB)>; -def : Pat<(srl G8RC:$rS, GPRC:$rB), - (SRD G8RC:$rS, GPRC:$rB)>; -def : Pat<(shl G8RC:$rS, GPRC:$rB), - (SLD G8RC:$rS, GPRC:$rB)>; +def : Pat<(sra i64:$rS, i32:$rB), + (SRAD $rS, $rB)>; +def : Pat<(srl i64:$rS, i32:$rB), + (SRD $rS, $rB)>; +def : Pat<(shl i64:$rS, i32:$rB), + (SLD $rS, $rB)>; // SHL/SRL -def : Pat<(shl G8RC:$in, (i32 imm:$imm)), - (RLDICR G8RC:$in, imm:$imm, (SHL64 imm:$imm))>; -def : Pat<(srl G8RC:$in, (i32 imm:$imm)), - (RLDICL G8RC:$in, (SRL64 imm:$imm), imm:$imm)>; +def : Pat<(shl i64:$in, (i32 imm:$imm)), + (RLDICR $in, imm:$imm, (SHL64 imm:$imm))>; +def : Pat<(srl i64:$in, (i32 imm:$imm)), + (RLDICL $in, (SRL64 imm:$imm), imm:$imm)>; // ROTL -def : Pat<(rotl G8RC:$in, GPRC:$sh), - (RLDCL G8RC:$in, GPRC:$sh, 0)>; -def : Pat<(rotl G8RC:$in, (i32 imm:$imm)), - (RLDICL G8RC:$in, imm:$imm, 0)>; +def : Pat<(rotl i64:$in, i32:$sh), + (RLDCL $in, $sh, 0)>; +def : Pat<(rotl i64:$in, (i32 imm:$imm)), + (RLDICL $in, imm:$imm, 0)>; // Hi and Lo for Darwin Global Addresses. def : Pat<(PPChi tglobaladdr:$in, 0), (LIS8 tglobaladdr:$in)>; @@ -973,18 +998,18 @@ def : Pat<(PPChi tjumptable:$in , 0), (LIS8 tjumptable:$in)>; def : Pat<(PPClo tjumptable:$in , 0), (LI8 tjumptable:$in)>; def : Pat<(PPChi tblockaddress:$in, 0), (LIS8 tblockaddress:$in)>; def : Pat<(PPClo tblockaddress:$in, 0), (LI8 tblockaddress:$in)>; -def : Pat<(PPChi tglobaltlsaddr:$g, G8RC:$in), - (ADDIS8 G8RC:$in, tglobaltlsaddr:$g)>; -def : Pat<(PPClo tglobaltlsaddr:$g, G8RC:$in), - (ADDI8L G8RC:$in, tglobaltlsaddr:$g)>; -def : Pat<(add G8RC:$in, (PPChi tglobaladdr:$g, 0)), - (ADDIS8 G8RC:$in, tglobaladdr:$g)>; -def : Pat<(add G8RC:$in, (PPChi tconstpool:$g, 0)), - (ADDIS8 G8RC:$in, tconstpool:$g)>; -def : Pat<(add G8RC:$in, (PPChi tjumptable:$g, 0)), - (ADDIS8 G8RC:$in, tjumptable:$g)>; -def : Pat<(add G8RC:$in, (PPChi tblockaddress:$g, 0)), - (ADDIS8 G8RC:$in, tblockaddress:$g)>; +def : Pat<(PPChi tglobaltlsaddr:$g, i64:$in), + (ADDIS8 $in, tglobaltlsaddr:$g)>; +def : Pat<(PPClo tglobaltlsaddr:$g, i64:$in), + (ADDI8 $in, tglobaltlsaddr:$g)>; +def : Pat<(add i64:$in, (PPChi tglobaladdr:$g, 0)), + (ADDIS8 $in, tglobaladdr:$g)>; +def : Pat<(add i64:$in, (PPChi tconstpool:$g, 0)), + (ADDIS8 $in, tconstpool:$g)>; +def : Pat<(add i64:$in, (PPChi tjumptable:$g, 0)), + (ADDIS8 $in, tjumptable:$g)>; +def : Pat<(add i64:$in, (PPChi tblockaddress:$g, 0)), + (ADDIS8 $in, tblockaddress:$g)>; // Patterns to match r+r indexed loads and stores for // addresses without at least 4-byte alignment. @@ -992,6 +1017,6 @@ def : Pat<(i64 (unaligned4sextloadi32 xoaddr:$src)), (LWAX xoaddr:$src)>; def : Pat<(i64 (unaligned4load xoaddr:$src)), (LDX xoaddr:$src)>; -def : Pat<(unaligned4store G8RC:$rS, xoaddr:$dst), - (STDX G8RC:$rS, xoaddr:$dst)>; +def : Pat<(unaligned4store i64:$rS, xoaddr:$dst), + (STDX $rS, xoaddr:$dst)>; diff --git a/lib/Target/PowerPC/PPCInstrAltivec.td b/lib/Target/PowerPC/PPCInstrAltivec.td index 0ed7ff2..cc9cf0a 100644 --- a/lib/Target/PowerPC/PPCInstrAltivec.td +++ b/lib/Target/PowerPC/PPCInstrAltivec.td @@ -161,23 +161,64 @@ def vecspltisw : PatLeaf<(build_vector), [{ //===----------------------------------------------------------------------===// // Helpers for defining instructions that directly correspond to intrinsics. -// VA1a_Int - A VAForm_1a intrinsic definition. -class VA1a_Int<bits<6> xo, string opc, Intrinsic IntID> - : VAForm_1a<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB, VRRC:$vC), +// VA1a_Int_Ty - A VAForm_1a intrinsic definition of specific type. +class VA1a_Int_Ty<bits<6> xo, string opc, Intrinsic IntID, ValueType Ty> + : VAForm_1a<xo, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB, vrrc:$vC), !strconcat(opc, " $vD, $vA, $vB, $vC"), VecFP, - [(set VRRC:$vD, (IntID VRRC:$vA, VRRC:$vB, VRRC:$vC))]>; + [(set Ty:$vD, (IntID Ty:$vA, Ty:$vB, Ty:$vC))]>; -// VX1_Int - A VXForm_1 intrinsic definition. -class VX1_Int<bits<11> xo, string opc, Intrinsic IntID> - : VXForm_1<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), +// VA1a_Int_Ty2 - A VAForm_1a intrinsic definition where the type of the +// inputs doesn't match the type of the output. +class VA1a_Int_Ty2<bits<6> xo, string opc, Intrinsic IntID, ValueType OutTy, + ValueType InTy> + : VAForm_1a<xo, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB, vrrc:$vC), + !strconcat(opc, " $vD, $vA, $vB, $vC"), VecFP, + [(set OutTy:$vD, (IntID InTy:$vA, InTy:$vB, InTy:$vC))]>; + +// VA1a_Int_Ty3 - A VAForm_1a intrinsic definition where there are two +// input types and an output type. +class VA1a_Int_Ty3<bits<6> xo, string opc, Intrinsic IntID, ValueType OutTy, + ValueType In1Ty, ValueType In2Ty> + : VAForm_1a<xo, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB, vrrc:$vC), + !strconcat(opc, " $vD, $vA, $vB, $vC"), VecFP, + [(set OutTy:$vD, + (IntID In1Ty:$vA, In1Ty:$vB, In2Ty:$vC))]>; + +// VX1_Int_Ty - A VXForm_1 intrinsic definition of specific type. +class VX1_Int_Ty<bits<11> xo, string opc, Intrinsic IntID, ValueType Ty> + : VXForm_1<xo, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + !strconcat(opc, " $vD, $vA, $vB"), VecFP, + [(set Ty:$vD, (IntID Ty:$vA, Ty:$vB))]>; + +// VX1_Int_Ty2 - A VXForm_1 intrinsic definition where the type of the +// inputs doesn't match the type of the output. +class VX1_Int_Ty2<bits<11> xo, string opc, Intrinsic IntID, ValueType OutTy, + ValueType InTy> + : VXForm_1<xo, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), + !strconcat(opc, " $vD, $vA, $vB"), VecFP, + [(set OutTy:$vD, (IntID InTy:$vA, InTy:$vB))]>; + +// VX1_Int_Ty3 - A VXForm_1 intrinsic definition where there are two +// input types and an output type. +class VX1_Int_Ty3<bits<11> xo, string opc, Intrinsic IntID, ValueType OutTy, + ValueType In1Ty, ValueType In2Ty> + : VXForm_1<xo, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), !strconcat(opc, " $vD, $vA, $vB"), VecFP, - [(set VRRC:$vD, (IntID VRRC:$vA, VRRC:$vB))]>; + [(set OutTy:$vD, (IntID In1Ty:$vA, In2Ty:$vB))]>; -// VX2_Int - A VXForm_2 intrinsic definition. -class VX2_Int<bits<11> xo, string opc, Intrinsic IntID> - : VXForm_2<xo, (outs VRRC:$vD), (ins VRRC:$vB), +// VX2_Int_SP - A VXForm_2 intrinsic definition of vector single-precision type. +class VX2_Int_SP<bits<11> xo, string opc, Intrinsic IntID> + : VXForm_2<xo, (outs vrrc:$vD), (ins vrrc:$vB), !strconcat(opc, " $vD, $vB"), VecFP, - [(set VRRC:$vD, (IntID VRRC:$vB))]>; + [(set v4f32:$vD, (IntID v4f32:$vB))]>; + +// VX2_Int_Ty2 - A VXForm_2 intrinsic definition where the type of the +// inputs doesn't match the type of the output. +class VX2_Int_Ty2<bits<11> xo, string opc, Intrinsic IntID, ValueType OutTy, + ValueType InTy> + : VXForm_2<xo, (outs vrrc:$vD), (ins vrrc:$vB), + !strconcat(opc, " $vD, $vB"), VecFP, + [(set OutTy:$vD, (IntID InTy:$vB))]>; //===----------------------------------------------------------------------===// // Instruction Definitions. @@ -185,6 +226,7 @@ class VX2_Int<bits<11> xo, string opc, Intrinsic IntID> def HasAltivec : Predicate<"PPCSubTarget.hasAltivec()">; let Predicates = [HasAltivec] in { +let isCodeGenOnly = 1 in { def DSS : DSS_Form<822, (outs), (ins u5imm:$ZERO0, u5imm:$STRM,u5imm:$ZERO1,u5imm:$ZERO2), "dss $STRM", LdStLoad /*FIXME*/, []>; @@ -192,357 +234,398 @@ def DSSALL : DSS_Form<822, (outs), (ins u5imm:$ONE, u5imm:$ZERO0,u5imm:$ZERO1,u5imm:$ZERO2), "dssall", LdStLoad /*FIXME*/, []>; def DST : DSS_Form<342, (outs), - (ins u5imm:$ZERO, u5imm:$STRM, GPRC:$rA, GPRC:$rB), + (ins u5imm:$ZERO, u5imm:$STRM, gprc:$rA, gprc:$rB), "dst $rA, $rB, $STRM", LdStLoad /*FIXME*/, []>; def DSTT : DSS_Form<342, (outs), - (ins u5imm:$ONE, u5imm:$STRM, GPRC:$rA, GPRC:$rB), + (ins u5imm:$ONE, u5imm:$STRM, gprc:$rA, gprc:$rB), "dstt $rA, $rB, $STRM", LdStLoad /*FIXME*/, []>; def DSTST : DSS_Form<374, (outs), - (ins u5imm:$ZERO, u5imm:$STRM, GPRC:$rA, GPRC:$rB), + (ins u5imm:$ZERO, u5imm:$STRM, gprc:$rA, gprc:$rB), "dstst $rA, $rB, $STRM", LdStLoad /*FIXME*/, []>; def DSTSTT : DSS_Form<374, (outs), - (ins u5imm:$ONE, u5imm:$STRM, GPRC:$rA, GPRC:$rB), + (ins u5imm:$ONE, u5imm:$STRM, gprc:$rA, gprc:$rB), "dststt $rA, $rB, $STRM", LdStLoad /*FIXME*/, []>; def DST64 : DSS_Form<342, (outs), - (ins u5imm:$ZERO, u5imm:$STRM, G8RC:$rA, GPRC:$rB), + (ins u5imm:$ZERO, u5imm:$STRM, g8rc:$rA, gprc:$rB), "dst $rA, $rB, $STRM", LdStLoad /*FIXME*/, []>; def DSTT64 : DSS_Form<342, (outs), - (ins u5imm:$ONE, u5imm:$STRM, G8RC:$rA, GPRC:$rB), + (ins u5imm:$ONE, u5imm:$STRM, g8rc:$rA, gprc:$rB), "dstt $rA, $rB, $STRM", LdStLoad /*FIXME*/, []>; def DSTST64 : DSS_Form<374, (outs), - (ins u5imm:$ZERO, u5imm:$STRM, G8RC:$rA, GPRC:$rB), + (ins u5imm:$ZERO, u5imm:$STRM, g8rc:$rA, gprc:$rB), "dstst $rA, $rB, $STRM", LdStLoad /*FIXME*/, []>; def DSTSTT64 : DSS_Form<374, (outs), - (ins u5imm:$ONE, u5imm:$STRM, G8RC:$rA, GPRC:$rB), + (ins u5imm:$ONE, u5imm:$STRM, g8rc:$rA, gprc:$rB), "dststt $rA, $rB, $STRM", LdStLoad /*FIXME*/, []>; +} -def MFVSCR : VXForm_4<1540, (outs VRRC:$vD), (ins), +def MFVSCR : VXForm_4<1540, (outs vrrc:$vD), (ins), "mfvscr $vD", LdStStore, - [(set VRRC:$vD, (int_ppc_altivec_mfvscr))]>; -def MTVSCR : VXForm_5<1604, (outs), (ins VRRC:$vB), + [(set v8i16:$vD, (int_ppc_altivec_mfvscr))]>; +def MTVSCR : VXForm_5<1604, (outs), (ins vrrc:$vB), "mtvscr $vB", LdStLoad, - [(int_ppc_altivec_mtvscr VRRC:$vB)]>; + [(int_ppc_altivec_mtvscr v4i32:$vB)]>; let canFoldAsLoad = 1, PPC970_Unit = 2 in { // Loads. -def LVEBX: XForm_1<31, 7, (outs VRRC:$vD), (ins memrr:$src), +def LVEBX: XForm_1<31, 7, (outs vrrc:$vD), (ins memrr:$src), "lvebx $vD, $src", LdStLoad, - [(set VRRC:$vD, (int_ppc_altivec_lvebx xoaddr:$src))]>; -def LVEHX: XForm_1<31, 39, (outs VRRC:$vD), (ins memrr:$src), + [(set v16i8:$vD, (int_ppc_altivec_lvebx xoaddr:$src))]>; +def LVEHX: XForm_1<31, 39, (outs vrrc:$vD), (ins memrr:$src), "lvehx $vD, $src", LdStLoad, - [(set VRRC:$vD, (int_ppc_altivec_lvehx xoaddr:$src))]>; -def LVEWX: XForm_1<31, 71, (outs VRRC:$vD), (ins memrr:$src), + [(set v8i16:$vD, (int_ppc_altivec_lvehx xoaddr:$src))]>; +def LVEWX: XForm_1<31, 71, (outs vrrc:$vD), (ins memrr:$src), "lvewx $vD, $src", LdStLoad, - [(set VRRC:$vD, (int_ppc_altivec_lvewx xoaddr:$src))]>; -def LVX : XForm_1<31, 103, (outs VRRC:$vD), (ins memrr:$src), + [(set v4i32:$vD, (int_ppc_altivec_lvewx xoaddr:$src))]>; +def LVX : XForm_1<31, 103, (outs vrrc:$vD), (ins memrr:$src), "lvx $vD, $src", LdStLoad, - [(set VRRC:$vD, (int_ppc_altivec_lvx xoaddr:$src))]>; -def LVXL : XForm_1<31, 359, (outs VRRC:$vD), (ins memrr:$src), + [(set v4i32:$vD, (int_ppc_altivec_lvx xoaddr:$src))]>; +def LVXL : XForm_1<31, 359, (outs vrrc:$vD), (ins memrr:$src), "lvxl $vD, $src", LdStLoad, - [(set VRRC:$vD, (int_ppc_altivec_lvxl xoaddr:$src))]>; + [(set v4i32:$vD, (int_ppc_altivec_lvxl xoaddr:$src))]>; } -def LVSL : XForm_1<31, 6, (outs VRRC:$vD), (ins memrr:$src), +def LVSL : XForm_1<31, 6, (outs vrrc:$vD), (ins memrr:$src), "lvsl $vD, $src", LdStLoad, - [(set VRRC:$vD, (int_ppc_altivec_lvsl xoaddr:$src))]>, + [(set v16i8:$vD, (int_ppc_altivec_lvsl xoaddr:$src))]>, PPC970_Unit_LSU; -def LVSR : XForm_1<31, 38, (outs VRRC:$vD), (ins memrr:$src), +def LVSR : XForm_1<31, 38, (outs vrrc:$vD), (ins memrr:$src), "lvsr $vD, $src", LdStLoad, - [(set VRRC:$vD, (int_ppc_altivec_lvsr xoaddr:$src))]>, + [(set v16i8:$vD, (int_ppc_altivec_lvsr xoaddr:$src))]>, PPC970_Unit_LSU; let PPC970_Unit = 2 in { // Stores. -def STVEBX: XForm_8<31, 135, (outs), (ins VRRC:$rS, memrr:$dst), +def STVEBX: XForm_8<31, 135, (outs), (ins vrrc:$rS, memrr:$dst), "stvebx $rS, $dst", LdStStore, - [(int_ppc_altivec_stvebx VRRC:$rS, xoaddr:$dst)]>; -def STVEHX: XForm_8<31, 167, (outs), (ins VRRC:$rS, memrr:$dst), + [(int_ppc_altivec_stvebx v16i8:$rS, xoaddr:$dst)]>; +def STVEHX: XForm_8<31, 167, (outs), (ins vrrc:$rS, memrr:$dst), "stvehx $rS, $dst", LdStStore, - [(int_ppc_altivec_stvehx VRRC:$rS, xoaddr:$dst)]>; -def STVEWX: XForm_8<31, 199, (outs), (ins VRRC:$rS, memrr:$dst), + [(int_ppc_altivec_stvehx v8i16:$rS, xoaddr:$dst)]>; +def STVEWX: XForm_8<31, 199, (outs), (ins vrrc:$rS, memrr:$dst), "stvewx $rS, $dst", LdStStore, - [(int_ppc_altivec_stvewx VRRC:$rS, xoaddr:$dst)]>; -def STVX : XForm_8<31, 231, (outs), (ins VRRC:$rS, memrr:$dst), + [(int_ppc_altivec_stvewx v4i32:$rS, xoaddr:$dst)]>; +def STVX : XForm_8<31, 231, (outs), (ins vrrc:$rS, memrr:$dst), "stvx $rS, $dst", LdStStore, - [(int_ppc_altivec_stvx VRRC:$rS, xoaddr:$dst)]>; -def STVXL : XForm_8<31, 487, (outs), (ins VRRC:$rS, memrr:$dst), + [(int_ppc_altivec_stvx v4i32:$rS, xoaddr:$dst)]>; +def STVXL : XForm_8<31, 487, (outs), (ins vrrc:$rS, memrr:$dst), "stvxl $rS, $dst", LdStStore, - [(int_ppc_altivec_stvxl VRRC:$rS, xoaddr:$dst)]>; + [(int_ppc_altivec_stvxl v4i32:$rS, xoaddr:$dst)]>; } let PPC970_Unit = 5 in { // VALU Operations. // VA-Form instructions. 3-input AltiVec ops. -def VMADDFP : VAForm_1<46, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vC, VRRC:$vB), +def VMADDFP : VAForm_1<46, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vC, vrrc:$vB), "vmaddfp $vD, $vA, $vC, $vB", VecFP, - [(set VRRC:$vD, (fma VRRC:$vA, VRRC:$vC, VRRC:$vB))]>; -def VNMSUBFP: VAForm_1<47, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vC, VRRC:$vB), + [(set v4f32:$vD, + (fma v4f32:$vA, v4f32:$vC, v4f32:$vB))]>; + +// FIXME: The fma+fneg pattern won't match because fneg is not legal. +def VNMSUBFP: VAForm_1<47, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vC, vrrc:$vB), "vnmsubfp $vD, $vA, $vC, $vB", VecFP, - [(set VRRC:$vD, (fneg (fma VRRC:$vA, VRRC:$vC, - (fneg VRRC:$vB))))]>; + [(set v4f32:$vD, (fneg (fma v4f32:$vA, v4f32:$vC, + (fneg v4f32:$vB))))]>; -def VMHADDSHS : VA1a_Int<32, "vmhaddshs", int_ppc_altivec_vmhaddshs>; -def VMHRADDSHS : VA1a_Int<33, "vmhraddshs", int_ppc_altivec_vmhraddshs>; -def VMLADDUHM : VA1a_Int<34, "vmladduhm", int_ppc_altivec_vmladduhm>; -def VPERM : VA1a_Int<43, "vperm", int_ppc_altivec_vperm>; -def VSEL : VA1a_Int<42, "vsel", int_ppc_altivec_vsel>; +def VMHADDSHS : VA1a_Int_Ty<32, "vmhaddshs", int_ppc_altivec_vmhaddshs, v8i16>; +def VMHRADDSHS : VA1a_Int_Ty<33, "vmhraddshs", int_ppc_altivec_vmhraddshs, + v8i16>; +def VMLADDUHM : VA1a_Int_Ty<34, "vmladduhm", int_ppc_altivec_vmladduhm, v8i16>; + +def VPERM : VA1a_Int_Ty3<43, "vperm", int_ppc_altivec_vperm, + v4i32, v4i32, v16i8>; +def VSEL : VA1a_Int_Ty<42, "vsel", int_ppc_altivec_vsel, v4i32>; // Shuffles. -def VSLDOI : VAForm_2<44, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB, u5imm:$SH), +def VSLDOI : VAForm_2<44, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB, u5imm:$SH), "vsldoi $vD, $vA, $vB, $SH", VecFP, - [(set VRRC:$vD, - (vsldoi_shuffle:$SH (v16i8 VRRC:$vA), VRRC:$vB))]>; + [(set v16i8:$vD, + (vsldoi_shuffle:$SH v16i8:$vA, v16i8:$vB))]>; // VX-Form instructions. AltiVec arithmetic ops. -def VADDFP : VXForm_1<10, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), +def VADDFP : VXForm_1<10, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vaddfp $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (fadd VRRC:$vA, VRRC:$vB))]>; + [(set v4f32:$vD, (fadd v4f32:$vA, v4f32:$vB))]>; -def VADDUBM : VXForm_1<0, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), +def VADDUBM : VXForm_1<0, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vaddubm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (add (v16i8 VRRC:$vA), VRRC:$vB))]>; -def VADDUHM : VXForm_1<64, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v16i8:$vD, (add v16i8:$vA, v16i8:$vB))]>; +def VADDUHM : VXForm_1<64, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vadduhm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (add (v8i16 VRRC:$vA), VRRC:$vB))]>; -def VADDUWM : VXForm_1<128, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v8i16:$vD, (add v8i16:$vA, v8i16:$vB))]>; +def VADDUWM : VXForm_1<128, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vadduwm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (add (v4i32 VRRC:$vA), VRRC:$vB))]>; + [(set v4i32:$vD, (add v4i32:$vA, v4i32:$vB))]>; -def VADDCUW : VX1_Int<384, "vaddcuw", int_ppc_altivec_vaddcuw>; -def VADDSBS : VX1_Int<768, "vaddsbs", int_ppc_altivec_vaddsbs>; -def VADDSHS : VX1_Int<832, "vaddshs", int_ppc_altivec_vaddshs>; -def VADDSWS : VX1_Int<896, "vaddsws", int_ppc_altivec_vaddsws>; -def VADDUBS : VX1_Int<512, "vaddubs", int_ppc_altivec_vaddubs>; -def VADDUHS : VX1_Int<576, "vadduhs", int_ppc_altivec_vadduhs>; -def VADDUWS : VX1_Int<640, "vadduws", int_ppc_altivec_vadduws>; +def VADDCUW : VX1_Int_Ty<384, "vaddcuw", int_ppc_altivec_vaddcuw, v4i32>; +def VADDSBS : VX1_Int_Ty<768, "vaddsbs", int_ppc_altivec_vaddsbs, v16i8>; +def VADDSHS : VX1_Int_Ty<832, "vaddshs", int_ppc_altivec_vaddshs, v8i16>; +def VADDSWS : VX1_Int_Ty<896, "vaddsws", int_ppc_altivec_vaddsws, v4i32>; +def VADDUBS : VX1_Int_Ty<512, "vaddubs", int_ppc_altivec_vaddubs, v16i8>; +def VADDUHS : VX1_Int_Ty<576, "vadduhs", int_ppc_altivec_vadduhs, v8i16>; +def VADDUWS : VX1_Int_Ty<640, "vadduws", int_ppc_altivec_vadduws, v4i32>; -def VAND : VXForm_1<1028, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), +def VAND : VXForm_1<1028, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vand $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (and (v4i32 VRRC:$vA), VRRC:$vB))]>; -def VANDC : VXForm_1<1092, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v4i32:$vD, (and v4i32:$vA, v4i32:$vB))]>; +def VANDC : VXForm_1<1092, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vandc $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (and (v4i32 VRRC:$vA), - (vnot_ppc VRRC:$vB)))]>; + [(set v4i32:$vD, (and v4i32:$vA, + (vnot_ppc v4i32:$vB)))]>; -def VCFSX : VXForm_1<842, (outs VRRC:$vD), (ins u5imm:$UIMM, VRRC:$vB), +def VCFSX : VXForm_1<842, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), "vcfsx $vD, $vB, $UIMM", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vcfsx VRRC:$vB, imm:$UIMM))]>; -def VCFUX : VXForm_1<778, (outs VRRC:$vD), (ins u5imm:$UIMM, VRRC:$vB), + [(set v4f32:$vD, + (int_ppc_altivec_vcfsx v4i32:$vB, imm:$UIMM))]>; +def VCFUX : VXForm_1<778, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), "vcfux $vD, $vB, $UIMM", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vcfux VRRC:$vB, imm:$UIMM))]>; -def VCTSXS : VXForm_1<970, (outs VRRC:$vD), (ins u5imm:$UIMM, VRRC:$vB), + [(set v4f32:$vD, + (int_ppc_altivec_vcfux v4i32:$vB, imm:$UIMM))]>; +def VCTSXS : VXForm_1<970, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), "vctsxs $vD, $vB, $UIMM", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vctsxs VRRC:$vB, imm:$UIMM))]>; -def VCTUXS : VXForm_1<906, (outs VRRC:$vD), (ins u5imm:$UIMM, VRRC:$vB), + [(set v4i32:$vD, + (int_ppc_altivec_vctsxs v4f32:$vB, imm:$UIMM))]>; +def VCTUXS : VXForm_1<906, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), "vctuxs $vD, $vB, $UIMM", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vctuxs VRRC:$vB, imm:$UIMM))]>; + [(set v4i32:$vD, + (int_ppc_altivec_vctuxs v4f32:$vB, imm:$UIMM))]>; // Defines with the UIM field set to 0 for floating-point // to integer (fp_to_sint/fp_to_uint) conversions and integer // to floating-point (sint_to_fp/uint_to_fp) conversions. let VA = 0 in { -def VCFSX_0 : VXForm_1<842, (outs VRRC:$vD), (ins VRRC:$vB), +def VCFSX_0 : VXForm_1<842, (outs vrrc:$vD), (ins vrrc:$vB), "vcfsx $vD, $vB, 0", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vcfsx VRRC:$vB, 0))]>; -def VCTUXS_0 : VXForm_1<906, (outs VRRC:$vD), (ins VRRC:$vB), + [(set v4f32:$vD, + (int_ppc_altivec_vcfsx v4i32:$vB, 0))]>; +def VCTUXS_0 : VXForm_1<906, (outs vrrc:$vD), (ins vrrc:$vB), "vctuxs $vD, $vB, 0", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vctuxs VRRC:$vB, 0))]>; -def VCFUX_0 : VXForm_1<778, (outs VRRC:$vD), (ins VRRC:$vB), + [(set v4i32:$vD, + (int_ppc_altivec_vctuxs v4f32:$vB, 0))]>; +def VCFUX_0 : VXForm_1<778, (outs vrrc:$vD), (ins vrrc:$vB), "vcfux $vD, $vB, 0", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vcfux VRRC:$vB, 0))]>; -def VCTSXS_0 : VXForm_1<970, (outs VRRC:$vD), (ins VRRC:$vB), + [(set v4f32:$vD, + (int_ppc_altivec_vcfux v4i32:$vB, 0))]>; +def VCTSXS_0 : VXForm_1<970, (outs vrrc:$vD), (ins vrrc:$vB), "vctsxs $vD, $vB, 0", VecFP, - [(set VRRC:$vD, - (int_ppc_altivec_vctsxs VRRC:$vB, 0))]>; + [(set v4i32:$vD, + (int_ppc_altivec_vctsxs v4f32:$vB, 0))]>; } -def VEXPTEFP : VX2_Int<394, "vexptefp", int_ppc_altivec_vexptefp>; -def VLOGEFP : VX2_Int<458, "vlogefp", int_ppc_altivec_vlogefp>; - -def VAVGSB : VX1_Int<1282, "vavgsb", int_ppc_altivec_vavgsb>; -def VAVGSH : VX1_Int<1346, "vavgsh", int_ppc_altivec_vavgsh>; -def VAVGSW : VX1_Int<1410, "vavgsw", int_ppc_altivec_vavgsw>; -def VAVGUB : VX1_Int<1026, "vavgub", int_ppc_altivec_vavgub>; -def VAVGUH : VX1_Int<1090, "vavguh", int_ppc_altivec_vavguh>; -def VAVGUW : VX1_Int<1154, "vavguw", int_ppc_altivec_vavguw>; - -def VMAXFP : VX1_Int<1034, "vmaxfp", int_ppc_altivec_vmaxfp>; -def VMAXSB : VX1_Int< 258, "vmaxsb", int_ppc_altivec_vmaxsb>; -def VMAXSH : VX1_Int< 322, "vmaxsh", int_ppc_altivec_vmaxsh>; -def VMAXSW : VX1_Int< 386, "vmaxsw", int_ppc_altivec_vmaxsw>; -def VMAXUB : VX1_Int< 2, "vmaxub", int_ppc_altivec_vmaxub>; -def VMAXUH : VX1_Int< 66, "vmaxuh", int_ppc_altivec_vmaxuh>; -def VMAXUW : VX1_Int< 130, "vmaxuw", int_ppc_altivec_vmaxuw>; -def VMINFP : VX1_Int<1098, "vminfp", int_ppc_altivec_vminfp>; -def VMINSB : VX1_Int< 770, "vminsb", int_ppc_altivec_vminsb>; -def VMINSH : VX1_Int< 834, "vminsh", int_ppc_altivec_vminsh>; -def VMINSW : VX1_Int< 898, "vminsw", int_ppc_altivec_vminsw>; -def VMINUB : VX1_Int< 514, "vminub", int_ppc_altivec_vminub>; -def VMINUH : VX1_Int< 578, "vminuh", int_ppc_altivec_vminuh>; -def VMINUW : VX1_Int< 642, "vminuw", int_ppc_altivec_vminuw>; - -def VMRGHB : VXForm_1< 12, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), +def VEXPTEFP : VX2_Int_SP<394, "vexptefp", int_ppc_altivec_vexptefp>; +def VLOGEFP : VX2_Int_SP<458, "vlogefp", int_ppc_altivec_vlogefp>; + +def VAVGSB : VX1_Int_Ty<1282, "vavgsb", int_ppc_altivec_vavgsb, v16i8>; +def VAVGSH : VX1_Int_Ty<1346, "vavgsh", int_ppc_altivec_vavgsh, v8i16>; +def VAVGSW : VX1_Int_Ty<1410, "vavgsw", int_ppc_altivec_vavgsw, v4i32>; +def VAVGUB : VX1_Int_Ty<1026, "vavgub", int_ppc_altivec_vavgub, v16i8>; +def VAVGUH : VX1_Int_Ty<1090, "vavguh", int_ppc_altivec_vavguh, v8i16>; +def VAVGUW : VX1_Int_Ty<1154, "vavguw", int_ppc_altivec_vavguw, v4i32>; + +def VMAXFP : VX1_Int_Ty<1034, "vmaxfp", int_ppc_altivec_vmaxfp, v4f32>; +def VMAXSB : VX1_Int_Ty< 258, "vmaxsb", int_ppc_altivec_vmaxsb, v16i8>; +def VMAXSH : VX1_Int_Ty< 322, "vmaxsh", int_ppc_altivec_vmaxsh, v8i16>; +def VMAXSW : VX1_Int_Ty< 386, "vmaxsw", int_ppc_altivec_vmaxsw, v4i32>; +def VMAXUB : VX1_Int_Ty< 2, "vmaxub", int_ppc_altivec_vmaxub, v16i8>; +def VMAXUH : VX1_Int_Ty< 66, "vmaxuh", int_ppc_altivec_vmaxuh, v8i16>; +def VMAXUW : VX1_Int_Ty< 130, "vmaxuw", int_ppc_altivec_vmaxuw, v4i32>; +def VMINFP : VX1_Int_Ty<1098, "vminfp", int_ppc_altivec_vminfp, v4f32>; +def VMINSB : VX1_Int_Ty< 770, "vminsb", int_ppc_altivec_vminsb, v16i8>; +def VMINSH : VX1_Int_Ty< 834, "vminsh", int_ppc_altivec_vminsh, v8i16>; +def VMINSW : VX1_Int_Ty< 898, "vminsw", int_ppc_altivec_vminsw, v4i32>; +def VMINUB : VX1_Int_Ty< 514, "vminub", int_ppc_altivec_vminub, v16i8>; +def VMINUH : VX1_Int_Ty< 578, "vminuh", int_ppc_altivec_vminuh, v8i16>; +def VMINUW : VX1_Int_Ty< 642, "vminuw", int_ppc_altivec_vminuw, v4i32>; + +def VMRGHB : VXForm_1< 12, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vmrghb $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (vmrghb_shuffle VRRC:$vA, VRRC:$vB))]>; -def VMRGHH : VXForm_1< 76, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v16i8:$vD, (vmrghb_shuffle v16i8:$vA, v16i8:$vB))]>; +def VMRGHH : VXForm_1< 76, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vmrghh $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (vmrghh_shuffle VRRC:$vA, VRRC:$vB))]>; -def VMRGHW : VXForm_1<140, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v16i8:$vD, (vmrghh_shuffle v16i8:$vA, v16i8:$vB))]>; +def VMRGHW : VXForm_1<140, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vmrghw $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (vmrghw_shuffle VRRC:$vA, VRRC:$vB))]>; -def VMRGLB : VXForm_1<268, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v16i8:$vD, (vmrghw_shuffle v16i8:$vA, v16i8:$vB))]>; +def VMRGLB : VXForm_1<268, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vmrglb $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (vmrglb_shuffle VRRC:$vA, VRRC:$vB))]>; -def VMRGLH : VXForm_1<332, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v16i8:$vD, (vmrglb_shuffle v16i8:$vA, v16i8:$vB))]>; +def VMRGLH : VXForm_1<332, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vmrglh $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (vmrglh_shuffle VRRC:$vA, VRRC:$vB))]>; -def VMRGLW : VXForm_1<396, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v16i8:$vD, (vmrglh_shuffle v16i8:$vA, v16i8:$vB))]>; +def VMRGLW : VXForm_1<396, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vmrglw $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (vmrglw_shuffle VRRC:$vA, VRRC:$vB))]>; - -def VMSUMMBM : VA1a_Int<37, "vmsummbm", int_ppc_altivec_vmsummbm>; -def VMSUMSHM : VA1a_Int<40, "vmsumshm", int_ppc_altivec_vmsumshm>; -def VMSUMSHS : VA1a_Int<41, "vmsumshs", int_ppc_altivec_vmsumshs>; -def VMSUMUBM : VA1a_Int<36, "vmsumubm", int_ppc_altivec_vmsumubm>; -def VMSUMUHM : VA1a_Int<38, "vmsumuhm", int_ppc_altivec_vmsumuhm>; -def VMSUMUHS : VA1a_Int<39, "vmsumuhs", int_ppc_altivec_vmsumuhs>; - -def VMULESB : VX1_Int<776, "vmulesb", int_ppc_altivec_vmulesb>; -def VMULESH : VX1_Int<840, "vmulesh", int_ppc_altivec_vmulesh>; -def VMULEUB : VX1_Int<520, "vmuleub", int_ppc_altivec_vmuleub>; -def VMULEUH : VX1_Int<584, "vmuleuh", int_ppc_altivec_vmuleuh>; -def VMULOSB : VX1_Int<264, "vmulosb", int_ppc_altivec_vmulosb>; -def VMULOSH : VX1_Int<328, "vmulosh", int_ppc_altivec_vmulosh>; -def VMULOUB : VX1_Int< 8, "vmuloub", int_ppc_altivec_vmuloub>; -def VMULOUH : VX1_Int< 72, "vmulouh", int_ppc_altivec_vmulouh>; + [(set v16i8:$vD, (vmrglw_shuffle v16i8:$vA, v16i8:$vB))]>; + +def VMSUMMBM : VA1a_Int_Ty3<37, "vmsummbm", int_ppc_altivec_vmsummbm, + v4i32, v16i8, v4i32>; +def VMSUMSHM : VA1a_Int_Ty3<40, "vmsumshm", int_ppc_altivec_vmsumshm, + v4i32, v8i16, v4i32>; +def VMSUMSHS : VA1a_Int_Ty3<41, "vmsumshs", int_ppc_altivec_vmsumshs, + v4i32, v8i16, v4i32>; +def VMSUMUBM : VA1a_Int_Ty3<36, "vmsumubm", int_ppc_altivec_vmsumubm, + v4i32, v16i8, v4i32>; +def VMSUMUHM : VA1a_Int_Ty3<38, "vmsumuhm", int_ppc_altivec_vmsumuhm, + v4i32, v8i16, v4i32>; +def VMSUMUHS : VA1a_Int_Ty3<39, "vmsumuhs", int_ppc_altivec_vmsumuhs, + v4i32, v8i16, v4i32>; + +def VMULESB : VX1_Int_Ty2<776, "vmulesb", int_ppc_altivec_vmulesb, + v8i16, v16i8>; +def VMULESH : VX1_Int_Ty2<840, "vmulesh", int_ppc_altivec_vmulesh, + v4i32, v8i16>; +def VMULEUB : VX1_Int_Ty2<520, "vmuleub", int_ppc_altivec_vmuleub, + v8i16, v16i8>; +def VMULEUH : VX1_Int_Ty2<584, "vmuleuh", int_ppc_altivec_vmuleuh, + v4i32, v8i16>; +def VMULOSB : VX1_Int_Ty2<264, "vmulosb", int_ppc_altivec_vmulosb, + v8i16, v16i8>; +def VMULOSH : VX1_Int_Ty2<328, "vmulosh", int_ppc_altivec_vmulosh, + v4i32, v8i16>; +def VMULOUB : VX1_Int_Ty2< 8, "vmuloub", int_ppc_altivec_vmuloub, + v8i16, v16i8>; +def VMULOUH : VX1_Int_Ty2< 72, "vmulouh", int_ppc_altivec_vmulouh, + v4i32, v8i16>; -def VREFP : VX2_Int<266, "vrefp", int_ppc_altivec_vrefp>; -def VRFIM : VX2_Int<714, "vrfim", int_ppc_altivec_vrfim>; -def VRFIN : VX2_Int<522, "vrfin", int_ppc_altivec_vrfin>; -def VRFIP : VX2_Int<650, "vrfip", int_ppc_altivec_vrfip>; -def VRFIZ : VX2_Int<586, "vrfiz", int_ppc_altivec_vrfiz>; -def VRSQRTEFP : VX2_Int<330, "vrsqrtefp", int_ppc_altivec_vrsqrtefp>; +def VREFP : VX2_Int_SP<266, "vrefp", int_ppc_altivec_vrefp>; +def VRFIM : VX2_Int_SP<714, "vrfim", int_ppc_altivec_vrfim>; +def VRFIN : VX2_Int_SP<522, "vrfin", int_ppc_altivec_vrfin>; +def VRFIP : VX2_Int_SP<650, "vrfip", int_ppc_altivec_vrfip>; +def VRFIZ : VX2_Int_SP<586, "vrfiz", int_ppc_altivec_vrfiz>; +def VRSQRTEFP : VX2_Int_SP<330, "vrsqrtefp", int_ppc_altivec_vrsqrtefp>; -def VSUBCUW : VX1_Int<74, "vsubcuw", int_ppc_altivec_vsubcuw>; +def VSUBCUW : VX1_Int_Ty<1408, "vsubcuw", int_ppc_altivec_vsubcuw, v4i32>; -def VSUBFP : VXForm_1<74, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), +def VSUBFP : VXForm_1<74, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vsubfp $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (fsub VRRC:$vA, VRRC:$vB))]>; -def VSUBUBM : VXForm_1<1024, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v4f32:$vD, (fsub v4f32:$vA, v4f32:$vB))]>; +def VSUBUBM : VXForm_1<1024, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vsububm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (sub (v16i8 VRRC:$vA), VRRC:$vB))]>; -def VSUBUHM : VXForm_1<1088, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v16i8:$vD, (sub v16i8:$vA, v16i8:$vB))]>; +def VSUBUHM : VXForm_1<1088, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vsubuhm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (sub (v8i16 VRRC:$vA), VRRC:$vB))]>; -def VSUBUWM : VXForm_1<1152, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v8i16:$vD, (sub v8i16:$vA, v8i16:$vB))]>; +def VSUBUWM : VXForm_1<1152, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vsubuwm $vD, $vA, $vB", VecGeneral, - [(set VRRC:$vD, (sub (v4i32 VRRC:$vA), VRRC:$vB))]>; + [(set v4i32:$vD, (sub v4i32:$vA, v4i32:$vB))]>; -def VSUBSBS : VX1_Int<1792, "vsubsbs" , int_ppc_altivec_vsubsbs>; -def VSUBSHS : VX1_Int<1856, "vsubshs" , int_ppc_altivec_vsubshs>; -def VSUBSWS : VX1_Int<1920, "vsubsws" , int_ppc_altivec_vsubsws>; -def VSUBUBS : VX1_Int<1536, "vsububs" , int_ppc_altivec_vsububs>; -def VSUBUHS : VX1_Int<1600, "vsubuhs" , int_ppc_altivec_vsubuhs>; -def VSUBUWS : VX1_Int<1664, "vsubuws" , int_ppc_altivec_vsubuws>; -def VSUMSWS : VX1_Int<1928, "vsumsws" , int_ppc_altivec_vsumsws>; -def VSUM2SWS: VX1_Int<1672, "vsum2sws", int_ppc_altivec_vsum2sws>; -def VSUM4SBS: VX1_Int<1672, "vsum4sbs", int_ppc_altivec_vsum4sbs>; -def VSUM4SHS: VX1_Int<1608, "vsum4shs", int_ppc_altivec_vsum4shs>; -def VSUM4UBS: VX1_Int<1544, "vsum4ubs", int_ppc_altivec_vsum4ubs>; - -def VNOR : VXForm_1<1284, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), +def VSUBSBS : VX1_Int_Ty<1792, "vsubsbs" , int_ppc_altivec_vsubsbs, v16i8>; +def VSUBSHS : VX1_Int_Ty<1856, "vsubshs" , int_ppc_altivec_vsubshs, v8i16>; +def VSUBSWS : VX1_Int_Ty<1920, "vsubsws" , int_ppc_altivec_vsubsws, v4i32>; +def VSUBUBS : VX1_Int_Ty<1536, "vsububs" , int_ppc_altivec_vsububs, v16i8>; +def VSUBUHS : VX1_Int_Ty<1600, "vsubuhs" , int_ppc_altivec_vsubuhs, v8i16>; +def VSUBUWS : VX1_Int_Ty<1664, "vsubuws" , int_ppc_altivec_vsubuws, v4i32>; + +def VSUMSWS : VX1_Int_Ty<1928, "vsumsws" , int_ppc_altivec_vsumsws, v4i32>; +def VSUM2SWS: VX1_Int_Ty<1672, "vsum2sws", int_ppc_altivec_vsum2sws, v4i32>; + +def VSUM4SBS: VX1_Int_Ty3<1800, "vsum4sbs", int_ppc_altivec_vsum4sbs, + v4i32, v16i8, v4i32>; +def VSUM4SHS: VX1_Int_Ty3<1608, "vsum4shs", int_ppc_altivec_vsum4shs, + v4i32, v8i16, v4i32>; +def VSUM4UBS: VX1_Int_Ty3<1544, "vsum4ubs", int_ppc_altivec_vsum4ubs, + v4i32, v16i8, v4i32>; + +def VNOR : VXForm_1<1284, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vnor $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (vnot_ppc (or (v4i32 VRRC:$vA), - VRRC:$vB)))]>; -def VOR : VXForm_1<1156, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v4i32:$vD, (vnot_ppc (or v4i32:$vA, + v4i32:$vB)))]>; +def VOR : VXForm_1<1156, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vor $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (or (v4i32 VRRC:$vA), VRRC:$vB))]>; -def VXOR : VXForm_1<1220, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v4i32:$vD, (or v4i32:$vA, v4i32:$vB))]>; +def VXOR : VXForm_1<1220, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vxor $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, (xor (v4i32 VRRC:$vA), VRRC:$vB))]>; + [(set v4i32:$vD, (xor v4i32:$vA, v4i32:$vB))]>; + +def VRLB : VX1_Int_Ty< 4, "vrlb", int_ppc_altivec_vrlb, v16i8>; +def VRLH : VX1_Int_Ty< 68, "vrlh", int_ppc_altivec_vrlh, v8i16>; +def VRLW : VX1_Int_Ty< 132, "vrlw", int_ppc_altivec_vrlw, v4i32>; -def VRLB : VX1_Int< 4, "vrlb", int_ppc_altivec_vrlb>; -def VRLH : VX1_Int< 68, "vrlh", int_ppc_altivec_vrlh>; -def VRLW : VX1_Int< 132, "vrlw", int_ppc_altivec_vrlw>; +def VSL : VX1_Int_Ty< 452, "vsl" , int_ppc_altivec_vsl, v4i32 >; +def VSLO : VX1_Int_Ty<1036, "vslo", int_ppc_altivec_vslo, v4i32>; -def VSL : VX1_Int< 452, "vsl" , int_ppc_altivec_vsl >; -def VSLO : VX1_Int<1036, "vslo", int_ppc_altivec_vslo>; -def VSLB : VX1_Int< 260, "vslb", int_ppc_altivec_vslb>; -def VSLH : VX1_Int< 324, "vslh", int_ppc_altivec_vslh>; -def VSLW : VX1_Int< 388, "vslw", int_ppc_altivec_vslw>; +def VSLB : VX1_Int_Ty< 260, "vslb", int_ppc_altivec_vslb, v16i8>; +def VSLH : VX1_Int_Ty< 324, "vslh", int_ppc_altivec_vslh, v8i16>; +def VSLW : VX1_Int_Ty< 388, "vslw", int_ppc_altivec_vslw, v4i32>; -def VSPLTB : VXForm_1<524, (outs VRRC:$vD), (ins u5imm:$UIMM, VRRC:$vB), +def VSPLTB : VXForm_1<524, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), "vspltb $vD, $vB, $UIMM", VecPerm, - [(set VRRC:$vD, - (vspltb_shuffle:$UIMM (v16i8 VRRC:$vB), (undef)))]>; -def VSPLTH : VXForm_1<588, (outs VRRC:$vD), (ins u5imm:$UIMM, VRRC:$vB), + [(set v16i8:$vD, + (vspltb_shuffle:$UIMM v16i8:$vB, (undef)))]>; +def VSPLTH : VXForm_1<588, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), "vsplth $vD, $vB, $UIMM", VecPerm, - [(set VRRC:$vD, - (vsplth_shuffle:$UIMM (v16i8 VRRC:$vB), (undef)))]>; -def VSPLTW : VXForm_1<652, (outs VRRC:$vD), (ins u5imm:$UIMM, VRRC:$vB), + [(set v16i8:$vD, + (vsplth_shuffle:$UIMM v16i8:$vB, (undef)))]>; +def VSPLTW : VXForm_1<652, (outs vrrc:$vD), (ins u5imm:$UIMM, vrrc:$vB), "vspltw $vD, $vB, $UIMM", VecPerm, - [(set VRRC:$vD, - (vspltw_shuffle:$UIMM (v16i8 VRRC:$vB), (undef)))]>; + [(set v16i8:$vD, + (vspltw_shuffle:$UIMM v16i8:$vB, (undef)))]>; -def VSR : VX1_Int< 708, "vsr" , int_ppc_altivec_vsr>; -def VSRO : VX1_Int<1100, "vsro" , int_ppc_altivec_vsro>; -def VSRAB : VX1_Int< 772, "vsrab", int_ppc_altivec_vsrab>; -def VSRAH : VX1_Int< 836, "vsrah", int_ppc_altivec_vsrah>; -def VSRAW : VX1_Int< 900, "vsraw", int_ppc_altivec_vsraw>; -def VSRB : VX1_Int< 516, "vsrb" , int_ppc_altivec_vsrb>; -def VSRH : VX1_Int< 580, "vsrh" , int_ppc_altivec_vsrh>; -def VSRW : VX1_Int< 644, "vsrw" , int_ppc_altivec_vsrw>; +def VSR : VX1_Int_Ty< 708, "vsr" , int_ppc_altivec_vsr, v4i32>; +def VSRO : VX1_Int_Ty<1100, "vsro" , int_ppc_altivec_vsro, v4i32>; +def VSRAB : VX1_Int_Ty< 772, "vsrab", int_ppc_altivec_vsrab, v16i8>; +def VSRAH : VX1_Int_Ty< 836, "vsrah", int_ppc_altivec_vsrah, v8i16>; +def VSRAW : VX1_Int_Ty< 900, "vsraw", int_ppc_altivec_vsraw, v4i32>; +def VSRB : VX1_Int_Ty< 516, "vsrb" , int_ppc_altivec_vsrb , v16i8>; +def VSRH : VX1_Int_Ty< 580, "vsrh" , int_ppc_altivec_vsrh , v8i16>; +def VSRW : VX1_Int_Ty< 644, "vsrw" , int_ppc_altivec_vsrw , v4i32>; -def VSPLTISB : VXForm_3<780, (outs VRRC:$vD), (ins s5imm:$SIMM), + +def VSPLTISB : VXForm_3<780, (outs vrrc:$vD), (ins s5imm:$SIMM), "vspltisb $vD, $SIMM", VecPerm, - [(set VRRC:$vD, (v16i8 vecspltisb:$SIMM))]>; -def VSPLTISH : VXForm_3<844, (outs VRRC:$vD), (ins s5imm:$SIMM), + [(set v16i8:$vD, (v16i8 vecspltisb:$SIMM))]>; +def VSPLTISH : VXForm_3<844, (outs vrrc:$vD), (ins s5imm:$SIMM), "vspltish $vD, $SIMM", VecPerm, - [(set VRRC:$vD, (v8i16 vecspltish:$SIMM))]>; -def VSPLTISW : VXForm_3<908, (outs VRRC:$vD), (ins s5imm:$SIMM), + [(set v8i16:$vD, (v8i16 vecspltish:$SIMM))]>; +def VSPLTISW : VXForm_3<908, (outs vrrc:$vD), (ins s5imm:$SIMM), "vspltisw $vD, $SIMM", VecPerm, - [(set VRRC:$vD, (v4i32 vecspltisw:$SIMM))]>; + [(set v4i32:$vD, (v4i32 vecspltisw:$SIMM))]>; // Vector Pack. -def VPKPX : VX1_Int<782, "vpkpx", int_ppc_altivec_vpkpx>; -def VPKSHSS : VX1_Int<398, "vpkshss", int_ppc_altivec_vpkshss>; -def VPKSHUS : VX1_Int<270, "vpkshus", int_ppc_altivec_vpkshus>; -def VPKSWSS : VX1_Int<462, "vpkswss", int_ppc_altivec_vpkswss>; -def VPKSWUS : VX1_Int<334, "vpkswus", int_ppc_altivec_vpkswus>; -def VPKUHUM : VXForm_1<14, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), +def VPKPX : VX1_Int_Ty2<782, "vpkpx", int_ppc_altivec_vpkpx, + v8i16, v4i32>; +def VPKSHSS : VX1_Int_Ty2<398, "vpkshss", int_ppc_altivec_vpkshss, + v16i8, v8i16>; +def VPKSHUS : VX1_Int_Ty2<270, "vpkshus", int_ppc_altivec_vpkshus, + v16i8, v8i16>; +def VPKSWSS : VX1_Int_Ty2<462, "vpkswss", int_ppc_altivec_vpkswss, + v16i8, v4i32>; +def VPKSWUS : VX1_Int_Ty2<334, "vpkswus", int_ppc_altivec_vpkswus, + v8i16, v4i32>; +def VPKUHUM : VXForm_1<14, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vpkuhum $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (vpkuhum_shuffle (v16i8 VRRC:$vA), VRRC:$vB))]>; -def VPKUHUS : VX1_Int<142, "vpkuhus", int_ppc_altivec_vpkuhus>; -def VPKUWUM : VXForm_1<78, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB), + [(set v16i8:$vD, + (vpkuhum_shuffle v16i8:$vA, v16i8:$vB))]>; +def VPKUHUS : VX1_Int_Ty2<142, "vpkuhus", int_ppc_altivec_vpkuhus, + v16i8, v8i16>; +def VPKUWUM : VXForm_1<78, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB), "vpkuwum $vD, $vA, $vB", VecFP, - [(set VRRC:$vD, - (vpkuwum_shuffle (v16i8 VRRC:$vA), VRRC:$vB))]>; -def VPKUWUS : VX1_Int<206, "vpkuwus", int_ppc_altivec_vpkuwus>; + [(set v16i8:$vD, + (vpkuwum_shuffle v16i8:$vA, v16i8:$vB))]>; +def VPKUWUS : VX1_Int_Ty2<206, "vpkuwus", int_ppc_altivec_vpkuwus, + v8i16, v4i32>; // Vector Unpack. -def VUPKHPX : VX2_Int<846, "vupkhpx", int_ppc_altivec_vupkhpx>; -def VUPKHSB : VX2_Int<526, "vupkhsb", int_ppc_altivec_vupkhsb>; -def VUPKHSH : VX2_Int<590, "vupkhsh", int_ppc_altivec_vupkhsh>; -def VUPKLPX : VX2_Int<974, "vupklpx", int_ppc_altivec_vupklpx>; -def VUPKLSB : VX2_Int<654, "vupklsb", int_ppc_altivec_vupklsb>; -def VUPKLSH : VX2_Int<718, "vupklsh", int_ppc_altivec_vupklsh>; +def VUPKHPX : VX2_Int_Ty2<846, "vupkhpx", int_ppc_altivec_vupkhpx, + v4i32, v8i16>; +def VUPKHSB : VX2_Int_Ty2<526, "vupkhsb", int_ppc_altivec_vupkhsb, + v8i16, v16i8>; +def VUPKHSH : VX2_Int_Ty2<590, "vupkhsh", int_ppc_altivec_vupkhsh, + v4i32, v8i16>; +def VUPKLPX : VX2_Int_Ty2<974, "vupklpx", int_ppc_altivec_vupklpx, + v4i32, v8i16>; +def VUPKLSB : VX2_Int_Ty2<654, "vupklsb", int_ppc_altivec_vupklsb, + v8i16, v16i8>; +def VUPKLSH : VX2_Int_Ty2<718, "vupklsh", int_ppc_altivec_vupklsh, + v4i32, v8i16>; // Altivec Comparisons. class VCMP<bits<10> xo, string asmstr, ValueType Ty> - : VXRForm_1<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),asmstr,VecFPCompare, - [(set VRRC:$vD, (Ty (PPCvcmp VRRC:$vA, VRRC:$vB, xo)))]>; + : VXRForm_1<xo, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),asmstr,VecFPCompare, + [(set Ty:$vD, (Ty (PPCvcmp Ty:$vA, Ty:$vB, xo)))]>; class VCMPo<bits<10> xo, string asmstr, ValueType Ty> - : VXRForm_1<xo, (outs VRRC:$vD), (ins VRRC:$vA, VRRC:$vB),asmstr,VecFPCompare, - [(set VRRC:$vD, (Ty (PPCvcmp_o VRRC:$vA, VRRC:$vB, xo)))]> { + : VXRForm_1<xo, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),asmstr,VecFPCompare, + [(set Ty:$vD, (Ty (PPCvcmp_o Ty:$vA, Ty:$vB, xo)))]> { let Defs = [CR6]; let RC = 1; } @@ -581,13 +664,14 @@ def VCMPGTSWo : VCMPo<902, "vcmpgtsw. $vD, $vA, $vB", v4i32>; def VCMPGTUW : VCMP <646, "vcmpgtuw $vD, $vA, $vB" , v4i32>; def VCMPGTUWo : VCMPo<646, "vcmpgtuw. $vD, $vA, $vB", v4i32>; -def V_SET0 : VXForm_setzero<1220, (outs VRRC:$vD), (ins), +let isCodeGenOnly = 1 in +def V_SET0 : VXForm_setzero<1220, (outs vrrc:$vD), (ins), "vxor $vD, $vD, $vD", VecFP, - [(set VRRC:$vD, (v4i32 immAllZerosV))]>; + [(set v4i32:$vD, (v4i32 immAllZerosV))]>; let IMM=-1 in { -def V_SETALLONES : VXForm_3<908, (outs VRRC:$vD), (ins), +def V_SETALLONES : VXForm_3<908, (outs vrrc:$vD), (ins), "vspltisw $vD, -1", VecFP, - [(set VRRC:$vD, (v4i32 immAllOnesV))]>; + [(set v4i32:$vD, (v4i32 immAllOnesV))]>; } } // VALU Operations. @@ -600,31 +684,31 @@ def : Pat<(int_ppc_altivec_dssall), (DSSALL 1, 0, 0, 0)>; def : Pat<(int_ppc_altivec_dss imm:$STRM), (DSS 0, imm:$STRM, 0, 0)>; // * 32-bit -def : Pat<(int_ppc_altivec_dst GPRC:$rA, GPRC:$rB, imm:$STRM), - (DST 0, imm:$STRM, GPRC:$rA, GPRC:$rB)>; -def : Pat<(int_ppc_altivec_dstt GPRC:$rA, GPRC:$rB, imm:$STRM), - (DSTT 1, imm:$STRM, GPRC:$rA, GPRC:$rB)>; -def : Pat<(int_ppc_altivec_dstst GPRC:$rA, GPRC:$rB, imm:$STRM), - (DSTST 0, imm:$STRM, GPRC:$rA, GPRC:$rB)>; -def : Pat<(int_ppc_altivec_dststt GPRC:$rA, GPRC:$rB, imm:$STRM), - (DSTSTT 1, imm:$STRM, GPRC:$rA, GPRC:$rB)>; +def : Pat<(int_ppc_altivec_dst i32:$rA, i32:$rB, imm:$STRM), + (DST 0, imm:$STRM, $rA, $rB)>; +def : Pat<(int_ppc_altivec_dstt i32:$rA, i32:$rB, imm:$STRM), + (DSTT 1, imm:$STRM, $rA, $rB)>; +def : Pat<(int_ppc_altivec_dstst i32:$rA, i32:$rB, imm:$STRM), + (DSTST 0, imm:$STRM, $rA, $rB)>; +def : Pat<(int_ppc_altivec_dststt i32:$rA, i32:$rB, imm:$STRM), + (DSTSTT 1, imm:$STRM, $rA, $rB)>; // * 64-bit -def : Pat<(int_ppc_altivec_dst G8RC:$rA, GPRC:$rB, imm:$STRM), - (DST64 0, imm:$STRM, (i64 G8RC:$rA), GPRC:$rB)>; -def : Pat<(int_ppc_altivec_dstt G8RC:$rA, GPRC:$rB, imm:$STRM), - (DSTT64 1, imm:$STRM, (i64 G8RC:$rA), GPRC:$rB)>; -def : Pat<(int_ppc_altivec_dstst G8RC:$rA, GPRC:$rB, imm:$STRM), - (DSTST64 0, imm:$STRM, (i64 G8RC:$rA), GPRC:$rB)>; -def : Pat<(int_ppc_altivec_dststt G8RC:$rA, GPRC:$rB, imm:$STRM), - (DSTSTT64 1, imm:$STRM, (i64 G8RC:$rA), GPRC:$rB)>; +def : Pat<(int_ppc_altivec_dst i64:$rA, i32:$rB, imm:$STRM), + (DST64 0, imm:$STRM, $rA, $rB)>; +def : Pat<(int_ppc_altivec_dstt i64:$rA, i32:$rB, imm:$STRM), + (DSTT64 1, imm:$STRM, $rA, $rB)>; +def : Pat<(int_ppc_altivec_dstst i64:$rA, i32:$rB, imm:$STRM), + (DSTST64 0, imm:$STRM, $rA, $rB)>; +def : Pat<(int_ppc_altivec_dststt i64:$rA, i32:$rB, imm:$STRM), + (DSTSTT64 1, imm:$STRM, $rA, $rB)>; // Loads. def : Pat<(v4i32 (load xoaddr:$src)), (LVX xoaddr:$src)>; // Stores. -def : Pat<(store (v4i32 VRRC:$rS), xoaddr:$dst), - (STVX (v4i32 VRRC:$rS), xoaddr:$dst)>; +def : Pat<(store v4i32:$rS, xoaddr:$dst), + (STVX $rS, xoaddr:$dst)>; // Bit conversions. def : Pat<(v16i8 (bitconvert (v8i16 VRRC:$src))), (v16i8 VRRC:$src)>; @@ -646,96 +730,99 @@ def : Pat<(v4f32 (bitconvert (v4i32 VRRC:$src))), (v4f32 VRRC:$src)>; // Shuffles. // Match vsldoi(x,x), vpkuwum(x,x), vpkuhum(x,x) -def:Pat<(vsldoi_unary_shuffle:$in (v16i8 VRRC:$vA), undef), - (VSLDOI VRRC:$vA, VRRC:$vA, (VSLDOI_unary_get_imm VRRC:$in))>; -def:Pat<(vpkuwum_unary_shuffle (v16i8 VRRC:$vA), undef), - (VPKUWUM VRRC:$vA, VRRC:$vA)>; -def:Pat<(vpkuhum_unary_shuffle (v16i8 VRRC:$vA), undef), - (VPKUHUM VRRC:$vA, VRRC:$vA)>; +def:Pat<(vsldoi_unary_shuffle:$in v16i8:$vA, undef), + (VSLDOI $vA, $vA, (VSLDOI_unary_get_imm $in))>; +def:Pat<(vpkuwum_unary_shuffle v16i8:$vA, undef), + (VPKUWUM $vA, $vA)>; +def:Pat<(vpkuhum_unary_shuffle v16i8:$vA, undef), + (VPKUHUM $vA, $vA)>; // Match vmrg*(x,x) -def:Pat<(vmrglb_unary_shuffle (v16i8 VRRC:$vA), undef), - (VMRGLB VRRC:$vA, VRRC:$vA)>; -def:Pat<(vmrglh_unary_shuffle (v16i8 VRRC:$vA), undef), - (VMRGLH VRRC:$vA, VRRC:$vA)>; -def:Pat<(vmrglw_unary_shuffle (v16i8 VRRC:$vA), undef), - (VMRGLW VRRC:$vA, VRRC:$vA)>; -def:Pat<(vmrghb_unary_shuffle (v16i8 VRRC:$vA), undef), - (VMRGHB VRRC:$vA, VRRC:$vA)>; -def:Pat<(vmrghh_unary_shuffle (v16i8 VRRC:$vA), undef), - (VMRGHH VRRC:$vA, VRRC:$vA)>; -def:Pat<(vmrghw_unary_shuffle (v16i8 VRRC:$vA), undef), - (VMRGHW VRRC:$vA, VRRC:$vA)>; +def:Pat<(vmrglb_unary_shuffle v16i8:$vA, undef), + (VMRGLB $vA, $vA)>; +def:Pat<(vmrglh_unary_shuffle v16i8:$vA, undef), + (VMRGLH $vA, $vA)>; +def:Pat<(vmrglw_unary_shuffle v16i8:$vA, undef), + (VMRGLW $vA, $vA)>; +def:Pat<(vmrghb_unary_shuffle v16i8:$vA, undef), + (VMRGHB $vA, $vA)>; +def:Pat<(vmrghh_unary_shuffle v16i8:$vA, undef), + (VMRGHH $vA, $vA)>; +def:Pat<(vmrghw_unary_shuffle v16i8:$vA, undef), + (VMRGHW $vA, $vA)>; // Logical Operations -def : Pat<(v4i32 (vnot_ppc VRRC:$vA)), (VNOR VRRC:$vA, VRRC:$vA)>; +def : Pat<(vnot_ppc v4i32:$vA), (VNOR $vA, $vA)>; -def : Pat<(v4i32 (vnot_ppc (or VRRC:$A, VRRC:$B))), - (VNOR VRRC:$A, VRRC:$B)>; -def : Pat<(v4i32 (and VRRC:$A, (vnot_ppc VRRC:$B))), - (VANDC VRRC:$A, VRRC:$B)>; +def : Pat<(vnot_ppc (or v4i32:$A, v4i32:$B)), + (VNOR $A, $B)>; +def : Pat<(and v4i32:$A, (vnot_ppc v4i32:$B)), + (VANDC $A, $B)>; -def : Pat<(fmul VRRC:$vA, VRRC:$vB), - (VMADDFP VRRC:$vA, VRRC:$vB, +def : Pat<(fmul v4f32:$vA, v4f32:$vB), + (VMADDFP $vA, $vB, (v4i32 (VSLW (V_SETALLONES), (V_SETALLONES))))>; // Fused multiply add and multiply sub for packed float. These are represented // separately from the real instructions above, for operations that must have // the additional precision, such as Newton-Rhapson (used by divide, sqrt) -def : Pat<(PPCvmaddfp VRRC:$A, VRRC:$B, VRRC:$C), - (VMADDFP VRRC:$A, VRRC:$B, VRRC:$C)>; -def : Pat<(PPCvnmsubfp VRRC:$A, VRRC:$B, VRRC:$C), - (VNMSUBFP VRRC:$A, VRRC:$B, VRRC:$C)>; +def : Pat<(PPCvmaddfp v4f32:$A, v4f32:$B, v4f32:$C), + (VMADDFP $A, $B, $C)>; +def : Pat<(PPCvnmsubfp v4f32:$A, v4f32:$B, v4f32:$C), + (VNMSUBFP $A, $B, $C)>; + +def : Pat<(int_ppc_altivec_vmaddfp v4f32:$A, v4f32:$B, v4f32:$C), + (VMADDFP $A, $B, $C)>; +def : Pat<(int_ppc_altivec_vnmsubfp v4f32:$A, v4f32:$B, v4f32:$C), + (VNMSUBFP $A, $B, $C)>; -def : Pat<(int_ppc_altivec_vmaddfp VRRC:$A, VRRC:$B, VRRC:$C), - (VMADDFP VRRC:$A, VRRC:$B, VRRC:$C)>; -def : Pat<(int_ppc_altivec_vnmsubfp VRRC:$A, VRRC:$B, VRRC:$C), - (VNMSUBFP VRRC:$A, VRRC:$B, VRRC:$C)>; +def : Pat<(PPCvperm v16i8:$vA, v16i8:$vB, v16i8:$vC), + (VPERM $vA, $vB, $vC)>; -def : Pat<(PPCvperm (v16i8 VRRC:$vA), VRRC:$vB, VRRC:$vC), - (VPERM VRRC:$vA, VRRC:$vB, VRRC:$vC)>; +def : Pat<(PPCfre v4f32:$A), (VREFP $A)>; +def : Pat<(PPCfrsqrte v4f32:$A), (VRSQRTEFP $A)>; // Vector shifts -def : Pat<(v16i8 (shl (v16i8 VRRC:$vA), (v16i8 VRRC:$vB))), - (v16i8 (VSLB VRRC:$vA, VRRC:$vB))>; -def : Pat<(v8i16 (shl (v8i16 VRRC:$vA), (v8i16 VRRC:$vB))), - (v8i16 (VSLH VRRC:$vA, VRRC:$vB))>; -def : Pat<(v4i32 (shl (v4i32 VRRC:$vA), (v4i32 VRRC:$vB))), - (v4i32 (VSLW VRRC:$vA, VRRC:$vB))>; - -def : Pat<(v16i8 (srl (v16i8 VRRC:$vA), (v16i8 VRRC:$vB))), - (v16i8 (VSRB VRRC:$vA, VRRC:$vB))>; -def : Pat<(v8i16 (srl (v8i16 VRRC:$vA), (v8i16 VRRC:$vB))), - (v8i16 (VSRH VRRC:$vA, VRRC:$vB))>; -def : Pat<(v4i32 (srl (v4i32 VRRC:$vA), (v4i32 VRRC:$vB))), - (v4i32 (VSRW VRRC:$vA, VRRC:$vB))>; - -def : Pat<(v16i8 (sra (v16i8 VRRC:$vA), (v16i8 VRRC:$vB))), - (v16i8 (VSRAB VRRC:$vA, VRRC:$vB))>; -def : Pat<(v8i16 (sra (v8i16 VRRC:$vA), (v8i16 VRRC:$vB))), - (v8i16 (VSRAH VRRC:$vA, VRRC:$vB))>; -def : Pat<(v4i32 (sra (v4i32 VRRC:$vA), (v4i32 VRRC:$vB))), - (v4i32 (VSRAW VRRC:$vA, VRRC:$vB))>; +def : Pat<(v16i8 (shl v16i8:$vA, v16i8:$vB)), + (v16i8 (VSLB $vA, $vB))>; +def : Pat<(v8i16 (shl v8i16:$vA, v8i16:$vB)), + (v8i16 (VSLH $vA, $vB))>; +def : Pat<(v4i32 (shl v4i32:$vA, v4i32:$vB)), + (v4i32 (VSLW $vA, $vB))>; + +def : Pat<(v16i8 (srl v16i8:$vA, v16i8:$vB)), + (v16i8 (VSRB $vA, $vB))>; +def : Pat<(v8i16 (srl v8i16:$vA, v8i16:$vB)), + (v8i16 (VSRH $vA, $vB))>; +def : Pat<(v4i32 (srl v4i32:$vA, v4i32:$vB)), + (v4i32 (VSRW $vA, $vB))>; + +def : Pat<(v16i8 (sra v16i8:$vA, v16i8:$vB)), + (v16i8 (VSRAB $vA, $vB))>; +def : Pat<(v8i16 (sra v8i16:$vA, v8i16:$vB)), + (v8i16 (VSRAH $vA, $vB))>; +def : Pat<(v4i32 (sra v4i32:$vA, v4i32:$vB)), + (v4i32 (VSRAW $vA, $vB))>; // Float to integer and integer to float conversions -def : Pat<(v4i32 (fp_to_sint (v4f32 VRRC:$vA))), - (VCTSXS_0 VRRC:$vA)>; -def : Pat<(v4i32 (fp_to_uint (v4f32 VRRC:$vA))), - (VCTUXS_0 VRRC:$vA)>; -def : Pat<(v4f32 (sint_to_fp (v4i32 VRRC:$vA))), - (VCFSX_0 VRRC:$vA)>; -def : Pat<(v4f32 (uint_to_fp (v4i32 VRRC:$vA))), - (VCFUX_0 VRRC:$vA)>; +def : Pat<(v4i32 (fp_to_sint v4f32:$vA)), + (VCTSXS_0 $vA)>; +def : Pat<(v4i32 (fp_to_uint v4f32:$vA)), + (VCTUXS_0 $vA)>; +def : Pat<(v4f32 (sint_to_fp v4i32:$vA)), + (VCFSX_0 $vA)>; +def : Pat<(v4f32 (uint_to_fp v4i32:$vA)), + (VCFUX_0 $vA)>; // Floating-point rounding -def : Pat<(v4f32 (ffloor (v4f32 VRRC:$vA))), - (VRFIM VRRC:$vA)>; -def : Pat<(v4f32 (fceil (v4f32 VRRC:$vA))), - (VRFIP VRRC:$vA)>; -def : Pat<(v4f32 (ftrunc (v4f32 VRRC:$vA))), - (VRFIZ VRRC:$vA)>; -def : Pat<(v4f32 (fnearbyint (v4f32 VRRC:$vA))), - (VRFIN VRRC:$vA)>; +def : Pat<(v4f32 (ffloor v4f32:$vA)), + (VRFIM $vA)>; +def : Pat<(v4f32 (fceil v4f32:$vA)), + (VRFIP $vA)>; +def : Pat<(v4f32 (ftrunc v4f32:$vA)), + (VRFIZ $vA)>; +def : Pat<(v4f32 (fnearbyint v4f32:$vA)), + (VRFIN $vA)>; } // end HasAltivec diff --git a/lib/Target/PowerPC/PPCInstrFormats.td b/lib/Target/PowerPC/PPCInstrFormats.td index c3c171c..41b4e01 100644 --- a/lib/Target/PowerPC/PPCInstrFormats.td +++ b/lib/Target/PowerPC/PPCInstrFormats.td @@ -35,6 +35,15 @@ class I<bits<6> opcode, dag OOL, dag IOL, string asmstr, InstrItinClass itin> let TSFlags{1} = PPC970_Single; let TSFlags{2} = PPC970_Cracked; let TSFlags{5-3} = PPC970_Unit; + + // Fields used for relation models. + string BaseName = ""; + + // For cases where multiple instruction definitions really represent the + // same underlying instruction but with one definition for 64-bit arguments + // and one for 32-bit arguments, this bit breaks the degeneracy between + // the two forms and allows TableGen to generate mapping tables. + bit Interpretation64Bit = 0; } class PPC970_DGroup_First { bits<1> PPC970_First = 1; } @@ -80,6 +89,10 @@ class I2<bits<6> opcode1, bits<6> opcode2, dag OOL, dag IOL, string asmstr, let TSFlags{1} = PPC970_Single; let TSFlags{2} = PPC970_Cracked; let TSFlags{5-3} = PPC970_Unit; + + // Fields used for relation models. + string BaseName = ""; + bit Interpretation64Bit = 0; } // 1.7.1 I-Form @@ -120,6 +133,18 @@ class BForm_1<bits<6> opcode, bits<5> bo, bit aa, bit lk, dag OOL, dag IOL, let CR = 0; } +class BForm_2<bits<6> opcode, bits<5> bo, bits<5> bi, bit aa, bit lk, + dag OOL, dag IOL, string asmstr> + : I<opcode, OOL, IOL, asmstr, BrB> { + bits<14> BD; + + let Inst{6-10} = bo; + let Inst{11-15} = bi; + let Inst{16-29} = BD; + let Inst{30} = aa; + let Inst{31} = lk; +} + // 1.7.4 D-Form class DForm_base<bits<6> opcode, dag OOL, dag IOL, string asmstr, InstrItinClass itin, list<dag> pattern> @@ -165,7 +190,12 @@ class DForm_1a<bits<6> opcode, dag OOL, dag IOL, string asmstr, class DForm_2<bits<6> opcode, dag OOL, dag IOL, string asmstr, InstrItinClass itin, list<dag> pattern> - : DForm_base<opcode, OOL, IOL, asmstr, itin, pattern>; + : DForm_base<opcode, OOL, IOL, asmstr, itin, pattern> { + + // Even though ADDICo does not really have an RC bit, provide + // the declaration of one here so that isDOT has something to set. + bit RC = 0; +} class DForm_2_r0<bits<6> opcode, dag OOL, dag IOL, string asmstr, InstrItinClass itin, list<dag> pattern> @@ -553,9 +583,9 @@ class XLForm_2_br<bits<6> opcode, bits<10> xo, bit lk, bits<7> BIBO; // 2 bits of BI and 5 bits of BO. bits<3> CR; - let BO = BIBO{2-6}; - let BI{0-1} = BIBO{0-1}; - let BI{2-4} = CR; + let BO = BIBO{4-0}; + let BI{0-1} = BIBO{5-6}; + let BI{2-4} = CR{0-2}; let BH = 0; } @@ -664,14 +694,13 @@ class XFXForm_7_ext<bits<6> opcode, bits<10> xo, bits<10> spr, // This is probably 1.7.9, but I don't have the reference that uses this // numbering scheme... class XFLForm<bits<6> opcode, bits<10> xo, dag OOL, dag IOL, string asmstr, - string cstr, InstrItinClass itin, list<dag>pattern> + InstrItinClass itin, list<dag>pattern> : I<opcode, OOL, IOL, asmstr, itin> { bits<8> FM; bits<5> rT; bit RC = 0; // set by isDOT let Pattern = pattern; - let Constraints = cstr; let Inst{6} = 0; let Inst{7-14} = FM; @@ -765,16 +794,14 @@ class AForm_4<bits<6> opcode, bits<5> xo, dag OOL, dag IOL, string asmstr, bits<5> RT; bits<5> RA; bits<5> RB; - bits<7> BIBO; // 2 bits of BI and 5 bits of BO (must be 12). - bits<3> CR; + bits<5> COND; let Pattern = pattern; let Inst{6-10} = RT; let Inst{11-15} = RA; let Inst{16-20} = RB; - let Inst{21-23} = CR; - let Inst{24-25} = BIBO{6-5}; + let Inst{21-25} = COND; let Inst{26-30} = xo; let Inst{31} = 0; } @@ -828,6 +855,25 @@ class MDForm_1<bits<6> opcode, bits<3> xo, dag OOL, dag IOL, string asmstr, let Inst{31} = RC; } +class MDSForm_1<bits<6> opcode, bits<4> xo, dag OOL, dag IOL, string asmstr, + InstrItinClass itin, list<dag> pattern> + : I<opcode, OOL, IOL, asmstr, itin> { + bits<5> RA; + bits<5> RS; + bits<5> RB; + bits<6> MBE; + + let Pattern = pattern; + + bit RC = 0; // set by isDOT + + let Inst{6-10} = RS; + let Inst{11-15} = RA; + let Inst{16-20} = RB; + let Inst{21-26} = MBE{4,3,2,1,0,5}; + let Inst{27-30} = xo; + let Inst{31} = RC; +} // E-1 VA-Form @@ -987,6 +1033,7 @@ class VXRForm_1<bits<10> xo, dag OOL, dag IOL, string asmstr, //===----------------------------------------------------------------------===// class Pseudo<dag OOL, dag IOL, string asmstr, list<dag> pattern> : I<0, OOL, IOL, asmstr, NoItinerary> { + let isCodeGenOnly = 1; let PPC64 = 0; let Pattern = pattern; let Inst{31-0} = 0; diff --git a/lib/Target/PowerPC/PPCInstrInfo.cpp b/lib/Target/PowerPC/PPCInstrInfo.cpp index 7fe7880..847bd22 100644 --- a/lib/Target/PowerPC/PPCInstrInfo.cpp +++ b/lib/Target/PowerPC/PPCInstrInfo.cpp @@ -18,8 +18,10 @@ #include "PPCInstrBuilder.h" #include "PPCMachineFunctionInfo.h" #include "PPCTargetMachine.h" +#include "llvm/ADT/Statistic.h" #include "llvm/ADT/STLExtras.h" #include "llvm/CodeGen/MachineFrameInfo.h" +#include "llvm/CodeGen/MachineFunctionPass.h" #include "llvm/CodeGen/MachineInstrBuilder.h" #include "llvm/CodeGen/MachineMemOperand.h" #include "llvm/CodeGen/MachineRegisterInfo.h" @@ -30,6 +32,7 @@ #include "llvm/Support/TargetRegistry.h" #include "llvm/Support/raw_ostream.h" +#define GET_INSTRMAP_INFO #define GET_INSTRINFO_CTOR #include "PPCGenInstrInfo.inc" @@ -39,6 +42,9 @@ static cl:: opt<bool> DisableCTRLoopAnal("disable-ppc-ctrloop-analysis", cl::Hidden, cl::desc("Disable analysis for CTR loops")); +static cl::opt<bool> DisableCmpOpt("disable-ppc-cmp-opt", +cl::desc("Disable compare instruction optimization"), cl::Hidden); + PPCInstrInfo::PPCInstrInfo(PPCTargetMachine &tm) : PPCGenInstrInfo(PPC::ADJCALLSTACKDOWN, PPC::ADJCALLSTACKUP), TM(tm), RI(*TM.getSubtargetImpl(), *this) {} @@ -94,12 +100,18 @@ bool PPCInstrInfo::isCoalescableExtInstr(const MachineInstr &MI, unsigned PPCInstrInfo::isLoadFromStackSlot(const MachineInstr *MI, int &FrameIndex) const { + // Note: This list must be kept consistent with LoadRegFromStackSlot. switch (MI->getOpcode()) { default: break; case PPC::LD: case PPC::LWZ: case PPC::LFS: case PPC::LFD: + case PPC::RESTORE_CR: + case PPC::LVX: + case PPC::RESTORE_VRSAVE: + // Check for the operands added by addFrameReference (the immediate is the + // offset which defaults to 0). if (MI->getOperand(1).isImm() && !MI->getOperand(1).getImm() && MI->getOperand(2).isFI()) { FrameIndex = MI->getOperand(2).getIndex(); @@ -112,12 +124,18 @@ unsigned PPCInstrInfo::isLoadFromStackSlot(const MachineInstr *MI, unsigned PPCInstrInfo::isStoreToStackSlot(const MachineInstr *MI, int &FrameIndex) const { + // Note: This list must be kept consistent with StoreRegToStackSlot. switch (MI->getOpcode()) { default: break; case PPC::STD: case PPC::STW: case PPC::STFS: case PPC::STFD: + case PPC::SPILL_CR: + case PPC::STVX: + case PPC::SPILL_VRSAVE: + // Check for the operands added by addFrameReference (the immediate is the + // offset which defaults to 0). if (MI->getOperand(1).isImm() && !MI->getOperand(1).getImm() && MI->getOperand(2).isFI()) { FrameIndex = MI->getOperand(2).getIndex(); @@ -135,7 +153,8 @@ PPCInstrInfo::commuteInstruction(MachineInstr *MI, bool NewMI) const { MachineFunction &MF = *MI->getParent()->getParent(); // Normal instructions can be commuted the obvious way. - if (MI->getOpcode() != PPC::RLWIMI) + if (MI->getOpcode() != PPC::RLWIMI && + MI->getOpcode() != PPC::RLWIMIo) return TargetInstrInfo::commuteInstruction(MI, NewMI); // Cannot commute if it has a non-zero rotate count. @@ -405,6 +424,105 @@ PPCInstrInfo::InsertBranch(MachineBasicBlock &MBB, MachineBasicBlock *TBB, return 2; } +// Select analysis. +bool PPCInstrInfo::canInsertSelect(const MachineBasicBlock &MBB, + const SmallVectorImpl<MachineOperand> &Cond, + unsigned TrueReg, unsigned FalseReg, + int &CondCycles, int &TrueCycles, int &FalseCycles) const { + if (!TM.getSubtargetImpl()->hasISEL()) + return false; + + if (Cond.size() != 2) + return false; + + // If this is really a bdnz-like condition, then it cannot be turned into a + // select. + if (Cond[1].getReg() == PPC::CTR || Cond[1].getReg() == PPC::CTR8) + return false; + + // Check register classes. + const MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo(); + const TargetRegisterClass *RC = + RI.getCommonSubClass(MRI.getRegClass(TrueReg), MRI.getRegClass(FalseReg)); + if (!RC) + return false; + + // isel is for regular integer GPRs only. + if (!PPC::GPRCRegClass.hasSubClassEq(RC) && + !PPC::G8RCRegClass.hasSubClassEq(RC)) + return false; + + // FIXME: These numbers are for the A2, how well they work for other cores is + // an open question. On the A2, the isel instruction has a 2-cycle latency + // but single-cycle throughput. These numbers are used in combination with + // the MispredictPenalty setting from the active SchedMachineModel. + CondCycles = 1; + TrueCycles = 1; + FalseCycles = 1; + + return true; +} + +void PPCInstrInfo::insertSelect(MachineBasicBlock &MBB, + MachineBasicBlock::iterator MI, DebugLoc dl, + unsigned DestReg, + const SmallVectorImpl<MachineOperand> &Cond, + unsigned TrueReg, unsigned FalseReg) const { + assert(Cond.size() == 2 && + "PPC branch conditions have two components!"); + + assert(TM.getSubtargetImpl()->hasISEL() && + "Cannot insert select on target without ISEL support"); + + // Get the register classes. + MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo(); + const TargetRegisterClass *RC = + RI.getCommonSubClass(MRI.getRegClass(TrueReg), MRI.getRegClass(FalseReg)); + assert(RC && "TrueReg and FalseReg must have overlapping register classes"); + assert((PPC::GPRCRegClass.hasSubClassEq(RC) || + PPC::G8RCRegClass.hasSubClassEq(RC)) && + "isel is for regular integer GPRs only"); + + unsigned OpCode = + PPC::GPRCRegClass.hasSubClassEq(RC) ? PPC::ISEL : PPC::ISEL8; + unsigned SelectPred = Cond[0].getImm(); + + unsigned SubIdx; + bool SwapOps; + switch (SelectPred) { + default: llvm_unreachable("invalid predicate for isel"); + case PPC::PRED_EQ: SubIdx = PPC::sub_eq; SwapOps = false; break; + case PPC::PRED_NE: SubIdx = PPC::sub_eq; SwapOps = true; break; + case PPC::PRED_LT: SubIdx = PPC::sub_lt; SwapOps = false; break; + case PPC::PRED_GE: SubIdx = PPC::sub_lt; SwapOps = true; break; + case PPC::PRED_GT: SubIdx = PPC::sub_gt; SwapOps = false; break; + case PPC::PRED_LE: SubIdx = PPC::sub_gt; SwapOps = true; break; + case PPC::PRED_UN: SubIdx = PPC::sub_un; SwapOps = false; break; + case PPC::PRED_NU: SubIdx = PPC::sub_un; SwapOps = true; break; + } + + unsigned FirstReg = SwapOps ? FalseReg : TrueReg, + SecondReg = SwapOps ? TrueReg : FalseReg; + + // The first input register of isel cannot be r0. If it is a member + // of a register class that can be r0, then copy it first (the + // register allocator should eliminate the copy). + if (MRI.getRegClass(FirstReg)->contains(PPC::R0) || + MRI.getRegClass(FirstReg)->contains(PPC::X0)) { + const TargetRegisterClass *FirstRC = + MRI.getRegClass(FirstReg)->contains(PPC::X0) ? + &PPC::G8RC_NOX0RegClass : &PPC::GPRC_NOR0RegClass; + unsigned OldFirstReg = FirstReg; + FirstReg = MRI.createVirtualRegister(FirstRC); + BuildMI(MBB, MI, dl, get(TargetOpcode::COPY), FirstReg) + .addReg(OldFirstReg); + } + + BuildMI(MBB, MI, dl, get(OpCode), DestReg) + .addReg(FirstReg).addReg(SecondReg) + .addReg(Cond[1].getReg(), 0, SubIdx); +} + void PPCInstrInfo::copyPhysReg(MachineBasicBlock &MBB, MachineBasicBlock::iterator I, DebugLoc DL, unsigned DestReg, unsigned SrcReg, @@ -440,40 +558,21 @@ PPCInstrInfo::StoreRegToStackSlot(MachineFunction &MF, int FrameIdx, const TargetRegisterClass *RC, SmallVectorImpl<MachineInstr*> &NewMIs, - bool &NonRI) const{ + bool &NonRI, bool &SpillsVRS) const{ + // Note: If additional store instructions are added here, + // update isStoreToStackSlot. + DebugLoc DL; if (PPC::GPRCRegClass.hasSubClassEq(RC)) { - if (SrcReg != PPC::LR) { - NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STW)) - .addReg(SrcReg, - getKillRegState(isKill)), - FrameIdx)); - } else { - // FIXME: this spills LR immediately to memory in one step. To do this, - // we use R11, which we know cannot be used in the prolog/epilog. This is - // a hack. - NewMIs.push_back(BuildMI(MF, DL, get(PPC::MFLR), PPC::R11)); - NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STW)) - .addReg(PPC::R11, - getKillRegState(isKill)), - FrameIdx)); - } + NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STW)) + .addReg(SrcReg, + getKillRegState(isKill)), + FrameIdx)); } else if (PPC::G8RCRegClass.hasSubClassEq(RC)) { - if (SrcReg != PPC::LR8) { - NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STD)) - .addReg(SrcReg, - getKillRegState(isKill)), - FrameIdx)); - } else { - // FIXME: this spills LR immediately to memory in one step. To do this, - // we use X11, which we know cannot be used in the prolog/epilog. This is - // a hack. - NewMIs.push_back(BuildMI(MF, DL, get(PPC::MFLR8), PPC::X11)); - NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STD)) - .addReg(PPC::X11, - getKillRegState(isKill)), - FrameIdx)); - } + NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STD)) + .addReg(SrcReg, + getKillRegState(isKill)), + FrameIdx)); } else if (PPC::F8RCRegClass.hasSubClassEq(RC)) { NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STFD)) .addReg(SrcReg, @@ -522,7 +621,7 @@ PPCInstrInfo::StoreRegToStackSlot(MachineFunction &MF, Reg = PPC::CR7; return StoreRegToStackSlot(MF, Reg, isKill, FrameIdx, - &PPC::CRRCRegClass, NewMIs, NonRI); + &PPC::CRRCRegClass, NewMIs, NonRI, SpillsVRS); } else if (PPC::VRRCRegClass.hasSubClassEq(RC)) { NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::STVX)) @@ -530,6 +629,14 @@ PPCInstrInfo::StoreRegToStackSlot(MachineFunction &MF, getKillRegState(isKill)), FrameIdx)); NonRI = true; + } else if (PPC::VRSAVERCRegClass.hasSubClassEq(RC)) { + assert(TM.getSubtargetImpl()->isDarwin() && + "VRSAVE only needs spill/restore on Darwin"); + NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::SPILL_VRSAVE)) + .addReg(SrcReg, + getKillRegState(isKill)), + FrameIdx)); + SpillsVRS = true; } else { llvm_unreachable("Unknown regclass!"); } @@ -549,10 +656,14 @@ PPCInstrInfo::storeRegToStackSlot(MachineBasicBlock &MBB, PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>(); FuncInfo->setHasSpills(); - bool NonRI = false; - if (StoreRegToStackSlot(MF, SrcReg, isKill, FrameIdx, RC, NewMIs, NonRI)) + bool NonRI = false, SpillsVRS = false; + if (StoreRegToStackSlot(MF, SrcReg, isKill, FrameIdx, RC, NewMIs, + NonRI, SpillsVRS)) FuncInfo->setSpillsCR(); + if (SpillsVRS) + FuncInfo->setSpillsVRSAVE(); + if (NonRI) FuncInfo->setHasNonRISpills(); @@ -573,25 +684,16 @@ PPCInstrInfo::LoadRegFromStackSlot(MachineFunction &MF, DebugLoc DL, unsigned DestReg, int FrameIdx, const TargetRegisterClass *RC, SmallVectorImpl<MachineInstr*> &NewMIs, - bool &NonRI) const{ + bool &NonRI, bool &SpillsVRS) const{ + // Note: If additional load instructions are added here, + // update isLoadFromStackSlot. + if (PPC::GPRCRegClass.hasSubClassEq(RC)) { - if (DestReg != PPC::LR) { - NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LWZ), - DestReg), FrameIdx)); - } else { - NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LWZ), - PPC::R11), FrameIdx)); - NewMIs.push_back(BuildMI(MF, DL, get(PPC::MTLR)).addReg(PPC::R11)); - } + NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LWZ), + DestReg), FrameIdx)); } else if (PPC::G8RCRegClass.hasSubClassEq(RC)) { - if (DestReg != PPC::LR8) { - NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LD), DestReg), - FrameIdx)); - } else { - NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LD), - PPC::X11), FrameIdx)); - NewMIs.push_back(BuildMI(MF, DL, get(PPC::MTLR8)).addReg(PPC::X11)); - } + NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LD), DestReg), + FrameIdx)); } else if (PPC::F8RCRegClass.hasSubClassEq(RC)) { NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LFD), DestReg), FrameIdx)); @@ -632,12 +734,20 @@ PPCInstrInfo::LoadRegFromStackSlot(MachineFunction &MF, DebugLoc DL, Reg = PPC::CR7; return LoadRegFromStackSlot(MF, DL, Reg, FrameIdx, - &PPC::CRRCRegClass, NewMIs, NonRI); + &PPC::CRRCRegClass, NewMIs, NonRI, SpillsVRS); } else if (PPC::VRRCRegClass.hasSubClassEq(RC)) { NewMIs.push_back(addFrameReference(BuildMI(MF, DL, get(PPC::LVX), DestReg), FrameIdx)); NonRI = true; + } else if (PPC::VRSAVERCRegClass.hasSubClassEq(RC)) { + assert(TM.getSubtargetImpl()->isDarwin() && + "VRSAVE only needs spill/restore on Darwin"); + NewMIs.push_back(addFrameReference(BuildMI(MF, DL, + get(PPC::RESTORE_VRSAVE), + DestReg), + FrameIdx)); + SpillsVRS = true; } else { llvm_unreachable("Unknown regclass!"); } @@ -659,10 +769,14 @@ PPCInstrInfo::loadRegFromStackSlot(MachineBasicBlock &MBB, PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>(); FuncInfo->setHasSpills(); - bool NonRI = false; - if (LoadRegFromStackSlot(MF, DL, DestReg, FrameIdx, RC, NewMIs, NonRI)) + bool NonRI = false, SpillsVRS = false; + if (LoadRegFromStackSlot(MF, DL, DestReg, FrameIdx, RC, NewMIs, + NonRI, SpillsVRS)) FuncInfo->setSpillsCR(); + if (SpillsVRS) + FuncInfo->setSpillsVRSAVE(); + if (NonRI) FuncInfo->setHasNonRISpills(); @@ -699,6 +813,562 @@ ReverseBranchCondition(SmallVectorImpl<MachineOperand> &Cond) const { return false; } +bool PPCInstrInfo::FoldImmediate(MachineInstr *UseMI, MachineInstr *DefMI, + unsigned Reg, MachineRegisterInfo *MRI) const { + // For some instructions, it is legal to fold ZERO into the RA register field. + // A zero immediate should always be loaded with a single li. + unsigned DefOpc = DefMI->getOpcode(); + if (DefOpc != PPC::LI && DefOpc != PPC::LI8) + return false; + if (!DefMI->getOperand(1).isImm()) + return false; + if (DefMI->getOperand(1).getImm() != 0) + return false; + + // Note that we cannot here invert the arguments of an isel in order to fold + // a ZERO into what is presented as the second argument. All we have here + // is the condition bit, and that might come from a CR-logical bit operation. + + const MCInstrDesc &UseMCID = UseMI->getDesc(); + + // Only fold into real machine instructions. + if (UseMCID.isPseudo()) + return false; + + unsigned UseIdx; + for (UseIdx = 0; UseIdx < UseMI->getNumOperands(); ++UseIdx) + if (UseMI->getOperand(UseIdx).isReg() && + UseMI->getOperand(UseIdx).getReg() == Reg) + break; + + assert(UseIdx < UseMI->getNumOperands() && "Cannot find Reg in UseMI"); + assert(UseIdx < UseMCID.getNumOperands() && "No operand description for Reg"); + + const MCOperandInfo *UseInfo = &UseMCID.OpInfo[UseIdx]; + + // We can fold the zero if this register requires a GPRC_NOR0/G8RC_NOX0 + // register (which might also be specified as a pointer class kind). + if (UseInfo->isLookupPtrRegClass()) { + if (UseInfo->RegClass /* Kind */ != 1) + return false; + } else { + if (UseInfo->RegClass != PPC::GPRC_NOR0RegClassID && + UseInfo->RegClass != PPC::G8RC_NOX0RegClassID) + return false; + } + + // Make sure this is not tied to an output register (or otherwise + // constrained). This is true for ST?UX registers, for example, which + // are tied to their output registers. + if (UseInfo->Constraints != 0) + return false; + + unsigned ZeroReg; + if (UseInfo->isLookupPtrRegClass()) { + bool isPPC64 = TM.getSubtargetImpl()->isPPC64(); + ZeroReg = isPPC64 ? PPC::ZERO8 : PPC::ZERO; + } else { + ZeroReg = UseInfo->RegClass == PPC::G8RC_NOX0RegClassID ? + PPC::ZERO8 : PPC::ZERO; + } + + bool DeleteDef = MRI->hasOneNonDBGUse(Reg); + UseMI->getOperand(UseIdx).setReg(ZeroReg); + + if (DeleteDef) + DefMI->eraseFromParent(); + + return true; +} + +static bool MBBDefinesCTR(MachineBasicBlock &MBB) { + for (MachineBasicBlock::iterator I = MBB.begin(), IE = MBB.end(); + I != IE; ++I) + if (I->definesRegister(PPC::CTR) || I->definesRegister(PPC::CTR8)) + return true; + return false; +} + +// We should make sure that, if we're going to predicate both sides of a +// condition (a diamond), that both sides don't define the counter register. We +// can predicate counter-decrement-based branches, but while that predicates +// the branching, it does not predicate the counter decrement. If we tried to +// merge the triangle into one predicated block, we'd decrement the counter +// twice. +bool PPCInstrInfo::isProfitableToIfCvt(MachineBasicBlock &TMBB, + unsigned NumT, unsigned ExtraT, + MachineBasicBlock &FMBB, + unsigned NumF, unsigned ExtraF, + const BranchProbability &Probability) const { + return !(MBBDefinesCTR(TMBB) && MBBDefinesCTR(FMBB)); +} + + +bool PPCInstrInfo::isPredicated(const MachineInstr *MI) const { + // The predicated branches are identified by their type, not really by the + // explicit presence of a predicate. Furthermore, some of them can be + // predicated more than once. Because if conversion won't try to predicate + // any instruction which already claims to be predicated (by returning true + // here), always return false. In doing so, we let isPredicable() be the + // final word on whether not the instruction can be (further) predicated. + + return false; +} + +bool PPCInstrInfo::isUnpredicatedTerminator(const MachineInstr *MI) const { + if (!MI->isTerminator()) + return false; + + // Conditional branch is a special case. + if (MI->isBranch() && !MI->isBarrier()) + return true; + + return !isPredicated(MI); +} + +bool PPCInstrInfo::PredicateInstruction( + MachineInstr *MI, + const SmallVectorImpl<MachineOperand> &Pred) const { + unsigned OpC = MI->getOpcode(); + if (OpC == PPC::BLR) { + if (Pred[1].getReg() == PPC::CTR8 || Pred[1].getReg() == PPC::CTR) { + bool isPPC64 = TM.getSubtargetImpl()->isPPC64(); + MI->setDesc(get(Pred[0].getImm() ? + (isPPC64 ? PPC::BDNZLR8 : PPC::BDNZLR) : + (isPPC64 ? PPC::BDZLR8 : PPC::BDZLR))); + } else { + MI->setDesc(get(PPC::BCLR)); + MachineInstrBuilder(*MI->getParent()->getParent(), MI) + .addImm(Pred[0].getImm()) + .addReg(Pred[1].getReg()); + } + + return true; + } else if (OpC == PPC::B) { + if (Pred[1].getReg() == PPC::CTR8 || Pred[1].getReg() == PPC::CTR) { + bool isPPC64 = TM.getSubtargetImpl()->isPPC64(); + MI->setDesc(get(Pred[0].getImm() ? + (isPPC64 ? PPC::BDNZ8 : PPC::BDNZ) : + (isPPC64 ? PPC::BDZ8 : PPC::BDZ))); + } else { + MachineBasicBlock *MBB = MI->getOperand(0).getMBB(); + MI->RemoveOperand(0); + + MI->setDesc(get(PPC::BCC)); + MachineInstrBuilder(*MI->getParent()->getParent(), MI) + .addImm(Pred[0].getImm()) + .addReg(Pred[1].getReg()) + .addMBB(MBB); + } + + return true; + } else if (OpC == PPC::BCTR || OpC == PPC::BCTR8 || + OpC == PPC::BCTRL || OpC == PPC::BCTRL8) { + if (Pred[1].getReg() == PPC::CTR8 || Pred[1].getReg() == PPC::CTR) + llvm_unreachable("Cannot predicate bctr[l] on the ctr register"); + + bool setLR = OpC == PPC::BCTRL || OpC == PPC::BCTRL8; + bool isPPC64 = TM.getSubtargetImpl()->isPPC64(); + MI->setDesc(get(isPPC64 ? (setLR ? PPC::BCCTRL8 : PPC::BCCTR8) : + (setLR ? PPC::BCCTRL : PPC::BCCTR))); + MachineInstrBuilder(*MI->getParent()->getParent(), MI) + .addImm(Pred[0].getImm()) + .addReg(Pred[1].getReg()); + return true; + } + + return false; +} + +bool PPCInstrInfo::SubsumesPredicate( + const SmallVectorImpl<MachineOperand> &Pred1, + const SmallVectorImpl<MachineOperand> &Pred2) const { + assert(Pred1.size() == 2 && "Invalid PPC first predicate"); + assert(Pred2.size() == 2 && "Invalid PPC second predicate"); + + if (Pred1[1].getReg() == PPC::CTR8 || Pred1[1].getReg() == PPC::CTR) + return false; + if (Pred2[1].getReg() == PPC::CTR8 || Pred2[1].getReg() == PPC::CTR) + return false; + + PPC::Predicate P1 = (PPC::Predicate) Pred1[0].getImm(); + PPC::Predicate P2 = (PPC::Predicate) Pred2[0].getImm(); + + if (P1 == P2) + return true; + + // Does P1 subsume P2, e.g. GE subsumes GT. + if (P1 == PPC::PRED_LE && + (P2 == PPC::PRED_LT || P2 == PPC::PRED_EQ)) + return true; + if (P1 == PPC::PRED_GE && + (P2 == PPC::PRED_GT || P2 == PPC::PRED_EQ)) + return true; + + return false; +} + +bool PPCInstrInfo::DefinesPredicate(MachineInstr *MI, + std::vector<MachineOperand> &Pred) const { + // Note: At the present time, the contents of Pred from this function is + // unused by IfConversion. This implementation follows ARM by pushing the + // CR-defining operand. Because the 'DZ' and 'DNZ' count as types of + // predicate, instructions defining CTR or CTR8 are also included as + // predicate-defining instructions. + + const TargetRegisterClass *RCs[] = + { &PPC::CRRCRegClass, &PPC::CRBITRCRegClass, + &PPC::CTRRCRegClass, &PPC::CTRRC8RegClass }; + + bool Found = false; + for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) { + const MachineOperand &MO = MI->getOperand(i); + for (unsigned c = 0; c < array_lengthof(RCs) && !Found; ++c) { + const TargetRegisterClass *RC = RCs[c]; + if (MO.isReg()) { + if (MO.isDef() && RC->contains(MO.getReg())) { + Pred.push_back(MO); + Found = true; + } + } else if (MO.isRegMask()) { + for (TargetRegisterClass::iterator I = RC->begin(), + IE = RC->end(); I != IE; ++I) + if (MO.clobbersPhysReg(*I)) { + Pred.push_back(MO); + Found = true; + } + } + } + } + + return Found; +} + +bool PPCInstrInfo::isPredicable(MachineInstr *MI) const { + unsigned OpC = MI->getOpcode(); + switch (OpC) { + default: + return false; + case PPC::B: + case PPC::BLR: + case PPC::BCTR: + case PPC::BCTR8: + case PPC::BCTRL: + case PPC::BCTRL8: + return true; + } +} + +bool PPCInstrInfo::analyzeCompare(const MachineInstr *MI, + unsigned &SrcReg, unsigned &SrcReg2, + int &Mask, int &Value) const { + unsigned Opc = MI->getOpcode(); + + switch (Opc) { + default: return false; + case PPC::CMPWI: + case PPC::CMPLWI: + case PPC::CMPDI: + case PPC::CMPLDI: + SrcReg = MI->getOperand(1).getReg(); + SrcReg2 = 0; + Value = MI->getOperand(2).getImm(); + Mask = 0xFFFF; + return true; + case PPC::CMPW: + case PPC::CMPLW: + case PPC::CMPD: + case PPC::CMPLD: + case PPC::FCMPUS: + case PPC::FCMPUD: + SrcReg = MI->getOperand(1).getReg(); + SrcReg2 = MI->getOperand(2).getReg(); + return true; + } +} + +bool PPCInstrInfo::optimizeCompareInstr(MachineInstr *CmpInstr, + unsigned SrcReg, unsigned SrcReg2, + int Mask, int Value, + const MachineRegisterInfo *MRI) const { + if (DisableCmpOpt) + return false; + + int OpC = CmpInstr->getOpcode(); + unsigned CRReg = CmpInstr->getOperand(0).getReg(); + bool isFP = OpC == PPC::FCMPUS || OpC == PPC::FCMPUD; + unsigned CRRecReg = isFP ? PPC::CR1 : PPC::CR0; + + // The record forms set the condition register based on a signed comparison + // with zero (so says the ISA manual). This is not as straightforward as it + // seems, however, because this is always a 64-bit comparison on PPC64, even + // for instructions that are 32-bit in nature (like slw for example). + // So, on PPC32, for unsigned comparisons, we can use the record forms only + // for equality checks (as those don't depend on the sign). On PPC64, + // we are restricted to equality for unsigned 64-bit comparisons and for + // signed 32-bit comparisons the applicability is more restricted. + bool isPPC64 = TM.getSubtargetImpl()->isPPC64(); + bool is32BitSignedCompare = OpC == PPC::CMPWI || OpC == PPC::CMPW; + bool is32BitUnsignedCompare = OpC == PPC::CMPLWI || OpC == PPC::CMPLW; + bool is64BitUnsignedCompare = OpC == PPC::CMPLDI || OpC == PPC::CMPLD; + + // Get the unique definition of SrcReg. + MachineInstr *MI = MRI->getUniqueVRegDef(SrcReg); + if (!MI) return false; + int MIOpC = MI->getOpcode(); + + bool equalityOnly = false; + bool noSub = false; + if (isPPC64) { + if (is32BitSignedCompare) { + // We can perform this optimization only if MI is sign-extending. + if (MIOpC == PPC::SRAW || MIOpC == PPC::SRAWo || + MIOpC == PPC::SRAWI || MIOpC == PPC::SRAWIo || + MIOpC == PPC::EXTSB || MIOpC == PPC::EXTSBo || + MIOpC == PPC::EXTSH || MIOpC == PPC::EXTSHo || + MIOpC == PPC::EXTSW || MIOpC == PPC::EXTSWo) { + noSub = true; + } else + return false; + } else if (is32BitUnsignedCompare) { + // We can perform this optimization, equality only, if MI is + // zero-extending. + if (MIOpC == PPC::CNTLZW || MIOpC == PPC::CNTLZWo || + MIOpC == PPC::SLW || MIOpC == PPC::SLWo || + MIOpC == PPC::SRW || MIOpC == PPC::SRWo) { + noSub = true; + equalityOnly = true; + } else + return false; + } else if (!isFP) + equalityOnly = is64BitUnsignedCompare; + } else if (!isFP) + equalityOnly = is32BitUnsignedCompare; + + if (equalityOnly) { + // We need to check the uses of the condition register in order to reject + // non-equality comparisons. + for (MachineRegisterInfo::use_iterator I = MRI->use_begin(CRReg), + IE = MRI->use_end(); I != IE; ++I) { + MachineInstr *UseMI = &*I; + if (UseMI->getOpcode() == PPC::BCC) { + unsigned Pred = UseMI->getOperand(0).getImm(); + if (Pred == PPC::PRED_EQ || Pred == PPC::PRED_NE) + continue; + + return false; + } else if (UseMI->getOpcode() == PPC::ISEL || + UseMI->getOpcode() == PPC::ISEL8) { + unsigned SubIdx = UseMI->getOperand(3).getSubReg(); + if (SubIdx == PPC::sub_eq) + continue; + + return false; + } else + return false; + } + } + + // Get ready to iterate backward from CmpInstr. + MachineBasicBlock::iterator I = CmpInstr, E = MI, + B = CmpInstr->getParent()->begin(); + + // Scan forward to find the first use of the compare. + for (MachineBasicBlock::iterator EL = CmpInstr->getParent()->end(); + I != EL; ++I) { + bool FoundUse = false; + for (MachineRegisterInfo::use_iterator J = MRI->use_begin(CRReg), + JE = MRI->use_end(); J != JE; ++J) + if (&*J == &*I) { + FoundUse = true; + break; + } + + if (FoundUse) + break; + } + + // Early exit if we're at the beginning of the BB. + if (I == B) return false; + + // There are two possible candidates which can be changed to set CR[01]. + // One is MI, the other is a SUB instruction. + // For CMPrr(r1,r2), we are looking for SUB(r1,r2) or SUB(r2,r1). + MachineInstr *Sub = NULL; + if (SrcReg2 != 0) + // MI is not a candidate for CMPrr. + MI = NULL; + // FIXME: Conservatively refuse to convert an instruction which isn't in the + // same BB as the comparison. This is to allow the check below to avoid calls + // (and other explicit clobbers); instead we should really check for these + // more explicitly (in at least a few predecessors). + else if (MI->getParent() != CmpInstr->getParent() || Value != 0) { + // PPC does not have a record-form SUBri. + return false; + } + + // Search for Sub. + const TargetRegisterInfo *TRI = &getRegisterInfo(); + --I; + for (; I != E && !noSub; --I) { + const MachineInstr &Instr = *I; + unsigned IOpC = Instr.getOpcode(); + + if (&*I != CmpInstr && ( + Instr.modifiesRegister(CRRecReg, TRI) || + Instr.readsRegister(CRRecReg, TRI))) + // This instruction modifies or uses the record condition register after + // the one we want to change. While we could do this transformation, it + // would likely not be profitable. This transformation removes one + // instruction, and so even forcing RA to generate one move probably + // makes it unprofitable. + return false; + + // Check whether CmpInstr can be made redundant by the current instruction. + if ((OpC == PPC::CMPW || OpC == PPC::CMPLW || + OpC == PPC::CMPD || OpC == PPC::CMPLD) && + (IOpC == PPC::SUBF || IOpC == PPC::SUBF8) && + ((Instr.getOperand(1).getReg() == SrcReg && + Instr.getOperand(2).getReg() == SrcReg2) || + (Instr.getOperand(1).getReg() == SrcReg2 && + Instr.getOperand(2).getReg() == SrcReg))) { + Sub = &*I; + break; + } + + if (isFP && (IOpC == PPC::FSUB || IOpC == PPC::FSUBS) && + ((Instr.getOperand(1).getReg() == SrcReg && + Instr.getOperand(2).getReg() == SrcReg2) || + (Instr.getOperand(1).getReg() == SrcReg2 && + Instr.getOperand(2).getReg() == SrcReg))) { + Sub = &*I; + break; + } + + if (I == B) + // The 'and' is below the comparison instruction. + return false; + } + + // Return false if no candidates exist. + if (!MI && !Sub) + return false; + + // The single candidate is called MI. + if (!MI) MI = Sub; + + int NewOpC = -1; + MIOpC = MI->getOpcode(); + if (MIOpC == PPC::ANDIo || MIOpC == PPC::ANDIo8) + NewOpC = MIOpC; + else { + NewOpC = PPC::getRecordFormOpcode(MIOpC); + if (NewOpC == -1 && PPC::getNonRecordFormOpcode(MIOpC) != -1) + NewOpC = MIOpC; + } + + // FIXME: On the non-embedded POWER architectures, only some of the record + // forms are fast, and we should use only the fast ones. + + // The defining instruction has a record form (or is already a record + // form). It is possible, however, that we'll need to reverse the condition + // code of the users. + if (NewOpC == -1) + return false; + + SmallVector<std::pair<MachineOperand*, PPC::Predicate>, 4> PredsToUpdate; + SmallVector<std::pair<MachineOperand*, unsigned>, 4> SubRegsToUpdate; + + // If we have SUB(r1, r2) and CMP(r2, r1), the condition code based on CMP + // needs to be updated to be based on SUB. Push the condition code + // operands to OperandsToUpdate. If it is safe to remove CmpInstr, the + // condition code of these operands will be modified. + bool ShouldSwap = false; + if (Sub) { + ShouldSwap = SrcReg2 != 0 && Sub->getOperand(1).getReg() == SrcReg2 && + Sub->getOperand(2).getReg() == SrcReg; + + // The operands to subf are the opposite of sub, so only in the fixed-point + // case, invert the order. + if (!isFP) + ShouldSwap = !ShouldSwap; + } + + if (ShouldSwap) + for (MachineRegisterInfo::use_iterator I = MRI->use_begin(CRReg), + IE = MRI->use_end(); I != IE; ++I) { + MachineInstr *UseMI = &*I; + if (UseMI->getOpcode() == PPC::BCC) { + PPC::Predicate Pred = (PPC::Predicate) UseMI->getOperand(0).getImm(); + assert((!equalityOnly || + Pred == PPC::PRED_EQ || Pred == PPC::PRED_NE) && + "Invalid predicate for equality-only optimization"); + PredsToUpdate.push_back(std::make_pair(&((*I).getOperand(0)), + PPC::getSwappedPredicate(Pred))); + } else if (UseMI->getOpcode() == PPC::ISEL || + UseMI->getOpcode() == PPC::ISEL8) { + unsigned NewSubReg = UseMI->getOperand(3).getSubReg(); + assert((!equalityOnly || NewSubReg == PPC::sub_eq) && + "Invalid CR bit for equality-only optimization"); + + if (NewSubReg == PPC::sub_lt) + NewSubReg = PPC::sub_gt; + else if (NewSubReg == PPC::sub_gt) + NewSubReg = PPC::sub_lt; + + SubRegsToUpdate.push_back(std::make_pair(&((*I).getOperand(3)), + NewSubReg)); + } else // We need to abort on a user we don't understand. + return false; + } + + // Create a new virtual register to hold the value of the CR set by the + // record-form instruction. If the instruction was not previously in + // record form, then set the kill flag on the CR. + CmpInstr->eraseFromParent(); + + MachineBasicBlock::iterator MII = MI; + BuildMI(*MI->getParent(), llvm::next(MII), MI->getDebugLoc(), + get(TargetOpcode::COPY), CRReg) + .addReg(CRRecReg, MIOpC != NewOpC ? RegState::Kill : 0); + + if (MIOpC != NewOpC) { + // We need to be careful here: we're replacing one instruction with + // another, and we need to make sure that we get all of the right + // implicit uses and defs. On the other hand, the caller may be holding + // an iterator to this instruction, and so we can't delete it (this is + // specifically the case if this is the instruction directly after the + // compare). + + const MCInstrDesc &NewDesc = get(NewOpC); + MI->setDesc(NewDesc); + + if (NewDesc.ImplicitDefs) + for (const uint16_t *ImpDefs = NewDesc.getImplicitDefs(); + *ImpDefs; ++ImpDefs) + if (!MI->definesRegister(*ImpDefs)) + MI->addOperand(*MI->getParent()->getParent(), + MachineOperand::CreateReg(*ImpDefs, true, true)); + if (NewDesc.ImplicitUses) + for (const uint16_t *ImpUses = NewDesc.getImplicitUses(); + *ImpUses; ++ImpUses) + if (!MI->readsRegister(*ImpUses)) + MI->addOperand(*MI->getParent()->getParent(), + MachineOperand::CreateReg(*ImpUses, false, true)); + } + + // Modify the condition code of operands in OperandsToUpdate. + // Since we have SUB(r1, r2) and CMP(r2, r1), the condition code needs to + // be changed from r2 > r1 to r1 < r2, from r2 < r1 to r1 > r2, etc. + for (unsigned i = 0, e = PredsToUpdate.size(); i < e; i++) + PredsToUpdate[i].first->setImm(PredsToUpdate[i].second); + + for (unsigned i = 0, e = SubRegsToUpdate.size(); i < e; i++) + SubRegsToUpdate[i].first->setSubReg(SubRegsToUpdate[i].second); + + return true; +} + /// GetInstSize - Return the number of bytes of code the specified /// instruction may be. This returns the maximum number of bytes. /// @@ -714,10 +1384,159 @@ unsigned PPCInstrInfo::GetInstSizeInBytes(const MachineInstr *MI) const { case PPC::GC_LABEL: case PPC::DBG_VALUE: return 0; - case PPC::BL8_NOP_ELF: - case PPC::BLA8_NOP_ELF: + case PPC::BL8_NOP: + case PPC::BLA8_NOP: return 8; default: return 4; // PowerPC instructions are all 4 bytes } } + +#undef DEBUG_TYPE +#define DEBUG_TYPE "ppc-early-ret" +STATISTIC(NumBCLR, "Number of early conditional returns"); +STATISTIC(NumBLR, "Number of early returns"); + +namespace llvm { + void initializePPCEarlyReturnPass(PassRegistry&); +} + +namespace { + // PPCEarlyReturn pass - For simple functions without epilogue code, move + // returns up, and create conditional returns, to avoid unnecessary + // branch-to-blr sequences. + struct PPCEarlyReturn : public MachineFunctionPass { + static char ID; + PPCEarlyReturn() : MachineFunctionPass(ID) { + initializePPCEarlyReturnPass(*PassRegistry::getPassRegistry()); + } + + const PPCTargetMachine *TM; + const PPCInstrInfo *TII; + +protected: + bool processBlock(MachineBasicBlock &ReturnMBB) { + bool Changed = false; + + MachineBasicBlock::iterator I = ReturnMBB.begin(); + I = ReturnMBB.SkipPHIsAndLabels(I); + + // The block must be essentially empty except for the blr. + if (I == ReturnMBB.end() || I->getOpcode() != PPC::BLR || + I != ReturnMBB.getLastNonDebugInstr()) + return Changed; + + SmallVector<MachineBasicBlock*, 8> PredToRemove; + for (MachineBasicBlock::pred_iterator PI = ReturnMBB.pred_begin(), + PIE = ReturnMBB.pred_end(); PI != PIE; ++PI) { + bool OtherReference = false, BlockChanged = false; + for (MachineBasicBlock::iterator J = (*PI)->getLastNonDebugInstr();;) { + if (J->getOpcode() == PPC::B) { + if (J->getOperand(0).getMBB() == &ReturnMBB) { + // This is an unconditional branch to the return. Replace the + // branch with a blr. + BuildMI(**PI, J, J->getDebugLoc(), TII->get(PPC::BLR)); + MachineBasicBlock::iterator K = J--; + K->eraseFromParent(); + BlockChanged = true; + ++NumBLR; + continue; + } + } else if (J->getOpcode() == PPC::BCC) { + if (J->getOperand(2).getMBB() == &ReturnMBB) { + // This is a conditional branch to the return. Replace the branch + // with a bclr. + BuildMI(**PI, J, J->getDebugLoc(), TII->get(PPC::BCLR)) + .addImm(J->getOperand(0).getImm()) + .addReg(J->getOperand(1).getReg()); + MachineBasicBlock::iterator K = J--; + K->eraseFromParent(); + BlockChanged = true; + ++NumBCLR; + continue; + } + } else if (J->isBranch()) { + if (J->isIndirectBranch()) { + if (ReturnMBB.hasAddressTaken()) + OtherReference = true; + } else + for (unsigned i = 0; i < J->getNumOperands(); ++i) + if (J->getOperand(i).isMBB() && + J->getOperand(i).getMBB() == &ReturnMBB) + OtherReference = true; + } else if (!J->isTerminator() && !J->isDebugValue()) + break; + + if (J == (*PI)->begin()) + break; + + --J; + } + + if ((*PI)->canFallThrough() && (*PI)->isLayoutSuccessor(&ReturnMBB)) + OtherReference = true; + + // Predecessors are stored in a vector and can't be removed here. + if (!OtherReference && BlockChanged) { + PredToRemove.push_back(*PI); + } + + if (BlockChanged) + Changed = true; + } + + for (unsigned i = 0, ie = PredToRemove.size(); i != ie; ++i) + PredToRemove[i]->removeSuccessor(&ReturnMBB); + + if (Changed && !ReturnMBB.hasAddressTaken()) { + // We now might be able to merge this blr-only block into its + // by-layout predecessor. + if (ReturnMBB.pred_size() == 1 && + (*ReturnMBB.pred_begin())->isLayoutSuccessor(&ReturnMBB)) { + // Move the blr into the preceding block. + MachineBasicBlock &PrevMBB = **ReturnMBB.pred_begin(); + PrevMBB.splice(PrevMBB.end(), &ReturnMBB, I); + PrevMBB.removeSuccessor(&ReturnMBB); + } + + if (ReturnMBB.pred_empty()) + ReturnMBB.eraseFromParent(); + } + + return Changed; + } + +public: + virtual bool runOnMachineFunction(MachineFunction &MF) { + TM = static_cast<const PPCTargetMachine *>(&MF.getTarget()); + TII = TM->getInstrInfo(); + + bool Changed = false; + + // If the function does not have at least two blocks, then there is + // nothing to do. + if (MF.size() < 2) + return Changed; + + for (MachineFunction::iterator I = MF.begin(); I != MF.end();) { + MachineBasicBlock &B = *I++; + if (processBlock(B)) + Changed = true; + } + + return Changed; + } + + virtual void getAnalysisUsage(AnalysisUsage &AU) const { + MachineFunctionPass::getAnalysisUsage(AU); + } + }; +} + +INITIALIZE_PASS(PPCEarlyReturn, DEBUG_TYPE, + "PowerPC Early-Return Creation", false, false) + +char PPCEarlyReturn::ID = 0; +FunctionPass* +llvm::createPPCEarlyReturnPass() { return new PPCEarlyReturn(); } + diff --git a/lib/Target/PowerPC/PPCInstrInfo.h b/lib/Target/PowerPC/PPCInstrInfo.h index 5d4ae91..34a1a73 100644 --- a/lib/Target/PowerPC/PPCInstrInfo.h +++ b/lib/Target/PowerPC/PPCInstrInfo.h @@ -72,12 +72,12 @@ class PPCInstrInfo : public PPCGenInstrInfo { unsigned SrcReg, bool isKill, int FrameIdx, const TargetRegisterClass *RC, SmallVectorImpl<MachineInstr*> &NewMIs, - bool &NonRI) const; + bool &NonRI, bool &SpillsVRS) const; bool LoadRegFromStackSlot(MachineFunction &MF, DebugLoc DL, unsigned DestReg, int FrameIdx, const TargetRegisterClass *RC, SmallVectorImpl<MachineInstr*> &NewMIs, - bool &NonRI) const; + bool &NonRI, bool &SpillsVRS) const; public: explicit PPCInstrInfo(PPCTargetMachine &TM); @@ -120,6 +120,17 @@ public: MachineBasicBlock *FBB, const SmallVectorImpl<MachineOperand> &Cond, DebugLoc DL) const; + + // Select analysis. + virtual bool canInsertSelect(const MachineBasicBlock&, + const SmallVectorImpl<MachineOperand> &Cond, + unsigned, unsigned, int&, int&, int&) const; + virtual void insertSelect(MachineBasicBlock &MBB, + MachineBasicBlock::iterator MI, DebugLoc DL, + unsigned DstReg, + const SmallVectorImpl<MachineOperand> &Cond, + unsigned TrueReg, unsigned FalseReg) const; + virtual void copyPhysReg(MachineBasicBlock &MBB, MachineBasicBlock::iterator I, DebugLoc DL, unsigned DestReg, unsigned SrcReg, @@ -146,6 +157,66 @@ public: virtual bool ReverseBranchCondition(SmallVectorImpl<MachineOperand> &Cond) const; + virtual bool FoldImmediate(MachineInstr *UseMI, MachineInstr *DefMI, + unsigned Reg, MachineRegisterInfo *MRI) const; + + // If conversion by predication (only supported by some branch instructions). + // All of the profitability checks always return true; it is always + // profitable to use the predicated branches. + virtual bool isProfitableToIfCvt(MachineBasicBlock &MBB, + unsigned NumCycles, unsigned ExtraPredCycles, + const BranchProbability &Probability) const { + return true; + } + + virtual bool isProfitableToIfCvt(MachineBasicBlock &TMBB, + unsigned NumT, unsigned ExtraT, + MachineBasicBlock &FMBB, + unsigned NumF, unsigned ExtraF, + const BranchProbability &Probability) const; + + virtual bool isProfitableToDupForIfCvt(MachineBasicBlock &MBB, + unsigned NumCycles, + const BranchProbability + &Probability) const { + return true; + } + + virtual bool isProfitableToUnpredicate(MachineBasicBlock &TMBB, + MachineBasicBlock &FMBB) const { + return false; + } + + // Predication support. + bool isPredicated(const MachineInstr *MI) const; + + virtual bool isUnpredicatedTerminator(const MachineInstr *MI) const; + + virtual + bool PredicateInstruction(MachineInstr *MI, + const SmallVectorImpl<MachineOperand> &Pred) const; + + virtual + bool SubsumesPredicate(const SmallVectorImpl<MachineOperand> &Pred1, + const SmallVectorImpl<MachineOperand> &Pred2) const; + + virtual bool DefinesPredicate(MachineInstr *MI, + std::vector<MachineOperand> &Pred) const; + + virtual bool isPredicable(MachineInstr *MI) const; + + // Comparison optimization. + + + virtual bool analyzeCompare(const MachineInstr *MI, + unsigned &SrcReg, unsigned &SrcReg2, + int &Mask, int &Value) const; + + virtual bool optimizeCompareInstr(MachineInstr *CmpInstr, + unsigned SrcReg, unsigned SrcReg2, + int Mask, int Value, + const MachineRegisterInfo *MRI) const; + /// GetInstSize - Return the number of bytes of code the specified /// instruction may be. This returns the maximum number of bytes. /// diff --git a/lib/Target/PowerPC/PPCInstrInfo.td b/lib/Target/PowerPC/PPCInstrInfo.td index 3f181aa..7d3540e 100644 --- a/lib/Target/PowerPC/PPCInstrInfo.td +++ b/lib/Target/PowerPC/PPCInstrInfo.td @@ -20,6 +20,10 @@ include "PPCInstrFormats.td" def SDT_PPCstfiwx : SDTypeProfile<0, 2, [ // stfiwx SDTCisVT<0, f64>, SDTCisPtrTy<1> ]>; +def SDT_PPClfiwx : SDTypeProfile<1, 1, [ // lfiw[az]x + SDTCisVT<0, f64>, SDTCisPtrTy<1> +]>; + def SDT_PPCCallSeqStart : SDCallSeqStart<[ SDTCisVT<0, i32> ]>; def SDT_PPCCallSeqEnd : SDCallSeqEnd<[ SDTCisVT<0, i32>, SDTCisVT<1, i32> ]>; @@ -36,10 +40,10 @@ def SDT_PPCcondbr : SDTypeProfile<0, 3, [ ]>; def SDT_PPClbrx : SDTypeProfile<1, 2, [ - SDTCisVT<0, i32>, SDTCisPtrTy<1>, SDTCisVT<2, OtherVT> + SDTCisInt<0>, SDTCisPtrTy<1>, SDTCisVT<2, OtherVT> ]>; def SDT_PPCstbrx : SDTypeProfile<0, 3, [ - SDTCisVT<0, i32>, SDTCisPtrTy<1>, SDTCisVT<2, OtherVT> + SDTCisInt<0>, SDTCisPtrTy<1>, SDTCisVT<2, OtherVT> ]>; def SDT_PPClarx : SDTypeProfile<1, 1, [ @@ -53,32 +57,36 @@ def SDT_PPCTC_ret : SDTypeProfile<0, 2, [ SDTCisPtrTy<0>, SDTCisVT<1, i32> ]>; -def SDT_PPCnop : SDTypeProfile<0, 0, []>; //===----------------------------------------------------------------------===// // PowerPC specific DAG Nodes. // -def PPCfcfid : SDNode<"PPCISD::FCFID" , SDTFPUnaryOp, []>; +def PPCfre : SDNode<"PPCISD::FRE", SDTFPUnaryOp, []>; +def PPCfrsqrte: SDNode<"PPCISD::FRSQRTE", SDTFPUnaryOp, []>; + +def PPCfcfid : SDNode<"PPCISD::FCFID", SDTFPUnaryOp, []>; +def PPCfcfidu : SDNode<"PPCISD::FCFIDU", SDTFPUnaryOp, []>; +def PPCfcfids : SDNode<"PPCISD::FCFIDS", SDTFPRoundOp, []>; +def PPCfcfidus: SDNode<"PPCISD::FCFIDUS", SDTFPRoundOp, []>; def PPCfctidz : SDNode<"PPCISD::FCTIDZ", SDTFPUnaryOp, []>; def PPCfctiwz : SDNode<"PPCISD::FCTIWZ", SDTFPUnaryOp, []>; +def PPCfctiduz: SDNode<"PPCISD::FCTIDUZ",SDTFPUnaryOp, []>; +def PPCfctiwuz: SDNode<"PPCISD::FCTIWUZ",SDTFPUnaryOp, []>; def PPCstfiwx : SDNode<"PPCISD::STFIWX", SDT_PPCstfiwx, [SDNPHasChain, SDNPMayStore]>; +def PPClfiwax : SDNode<"PPCISD::LFIWAX", SDT_PPClfiwx, + [SDNPHasChain, SDNPMayLoad]>; +def PPClfiwzx : SDNode<"PPCISD::LFIWZX", SDT_PPClfiwx, + [SDNPHasChain, SDNPMayLoad]>; + +// Extract FPSCR (not modeled at the DAG level). +def PPCmffs : SDNode<"PPCISD::MFFS", + SDTypeProfile<1, 0, [SDTCisVT<0, f64>]>, []>; + +// Perform FADD in round-to-zero mode. +def PPCfaddrtz: SDNode<"PPCISD::FADDRTZ", SDTFPBinOp, []>; -// This sequence is used for long double->int conversions. It changes the -// bits in the FPSCR which is not modelled. -def PPCmffs : SDNode<"PPCISD::MFFS", SDTypeProfile<1, 0, [SDTCisVT<0, f64>]>, - [SDNPOutGlue]>; -def PPCmtfsb0 : SDNode<"PPCISD::MTFSB0", SDTypeProfile<0, 1, [SDTCisInt<0>]>, - [SDNPInGlue, SDNPOutGlue]>; -def PPCmtfsb1 : SDNode<"PPCISD::MTFSB1", SDTypeProfile<0, 1, [SDTCisInt<0>]>, - [SDNPInGlue, SDNPOutGlue]>; -def PPCfaddrtz: SDNode<"PPCISD::FADDRTZ", SDTFPBinOp, - [SDNPInGlue, SDNPOutGlue]>; -def PPCmtfsf : SDNode<"PPCISD::MTFSF", SDTypeProfile<1, 3, - [SDTCisVT<0, f64>, SDTCisInt<1>, SDTCisVT<2, f64>, - SDTCisVT<3, f64>]>, - [SDNPInGlue]>; def PPCfsel : SDNode<"PPCISD::FSEL", // Type constraint for fsel. @@ -113,10 +121,6 @@ def PPCsrl : SDNode<"PPCISD::SRL" , SDTIntShiftOp>; def PPCsra : SDNode<"PPCISD::SRA" , SDTIntShiftOp>; def PPCshl : SDNode<"PPCISD::SHL" , SDTIntShiftOp>; -def PPCextsw_32 : SDNode<"PPCISD::EXTSW_32" , SDTIntUnaryOp>; -def PPCstd_32 : SDNode<"PPCISD::STD_32" , SDTStore, - [SDNPHasChain, SDNPMayStore]>; - // These are target-independent nodes, but have target-specific formats. def callseq_start : SDNode<"ISD::CALLSEQ_START", SDT_PPCCallSeqStart, [SDNPHasChain, SDNPOutGlue]>; @@ -124,16 +128,12 @@ def callseq_end : SDNode<"ISD::CALLSEQ_END", SDT_PPCCallSeqEnd, [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue]>; def SDT_PPCCall : SDTypeProfile<0, -1, [SDTCisInt<0>]>; -def PPCcall_Darwin : SDNode<"PPCISD::CALL_Darwin", SDT_PPCCall, - [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue, - SDNPVariadic]>; -def PPCcall_SVR4 : SDNode<"PPCISD::CALL_SVR4", SDT_PPCCall, - [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue, - SDNPVariadic]>; -def PPCcall_nop_SVR4 : SDNode<"PPCISD::CALL_NOP_SVR4", SDT_PPCCall, - [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue, - SDNPVariadic]>; -def PPCnop : SDNode<"PPCISD::NOP", SDT_PPCnop, [SDNPInGlue, SDNPOutGlue]>; +def PPCcall : SDNode<"PPCISD::CALL", SDT_PPCCall, + [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue, + SDNPVariadic]>; +def PPCcall_nop : SDNode<"PPCISD::CALL_NOP", SDT_PPCCall, + [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue, + SDNPVariadic]>; def PPCload : SDNode<"PPCISD::LOAD", SDTypeProfile<1, 1, []>, [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue]>; def PPCload_toc : SDNode<"PPCISD::LOAD_TOC", SDTypeProfile<0, 1, []>, @@ -144,13 +144,9 @@ def PPCtoc_restore : SDNode<"PPCISD::TOC_RESTORE", SDTypeProfile<0, 0, []>, SDNPInGlue, SDNPOutGlue]>; def PPCmtctr : SDNode<"PPCISD::MTCTR", SDT_PPCCall, [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue]>; -def PPCbctrl_Darwin : SDNode<"PPCISD::BCTRL_Darwin", SDTNone, - [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue, - SDNPVariadic]>; - -def PPCbctrl_SVR4 : SDNode<"PPCISD::BCTRL_SVR4", SDTNone, - [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue, - SDNPVariadic]>; +def PPCbctrl : SDNode<"PPCISD::BCTRL", SDTNone, + [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue, + SDNPVariadic]>; def retflag : SDNode<"PPCISD::RET_FLAG", SDTNone, [SDNPHasChain, SDNPOptInGlue, SDNPVariadic]>; @@ -158,6 +154,14 @@ def retflag : SDNode<"PPCISD::RET_FLAG", SDTNone, def PPCtc_return : SDNode<"PPCISD::TC_RETURN", SDT_PPCTC_ret, [SDNPHasChain, SDNPOptInGlue, SDNPVariadic]>; +def PPCeh_sjlj_setjmp : SDNode<"PPCISD::EH_SJLJ_SETJMP", + SDTypeProfile<1, 1, [SDTCisInt<0>, + SDTCisPtrTy<1>]>, + [SDNPHasChain, SDNPSideEffect]>; +def PPCeh_sjlj_longjmp : SDNode<"PPCISD::EH_SJLJ_LONGJMP", + SDTypeProfile<0, 1, [SDTCisPtrTy<0>]>, + [SDNPHasChain, SDNPSideEffect]>; + def PPCvcmp : SDNode<"PPCISD::VCMP" , SDT_PPCvcmp, []>; def PPCvcmp_o : SDNode<"PPCISD::VCMPo", SDT_PPCvcmp, [SDNPOutGlue]>; @@ -315,10 +319,7 @@ def unaligned4sextloadi32 : PatFrag<(ops node:$ptr), (sextloadi32 node:$ptr), [{ // PowerPC Flag Definitions. class isPPC64 { bit PPC64 = 1; } -class isDOT { - list<Register> Defs = [CR0]; - bit RC = 1; -} +class isDOT { bit RC = 1; } class RegConstraint<string C> { string Constraints = C; @@ -331,6 +332,26 @@ class NoEncode<string E> { //===----------------------------------------------------------------------===// // PowerPC Operand Definitions. +// In the default PowerPC assembler syntax, registers are specified simply +// by number, so they cannot be distinguished from immediate values (without +// looking at the opcode). This means that the default operand matching logic +// for the asm parser does not work, and we need to specify custom matchers. +// Since those can only be specified with RegisterOperand classes and not +// directly on the RegisterClass, all instructions patterns used by the asm +// parser need to use a RegisterOperand (instead of a RegisterClass) for +// all their register operands. +// For this purpose, we define one RegisterOperand for each RegisterClass, +// using the same name as the class, just in lower case. +def gprc : RegisterOperand<GPRC>; +def g8rc : RegisterOperand<G8RC>; +def gprc_nor0 : RegisterOperand<GPRC_NOR0>; +def g8rc_nox0 : RegisterOperand<G8RC_NOX0>; +def f8rc : RegisterOperand<F8RC>; +def f4rc : RegisterOperand<F4RC>; +def vrrc : RegisterOperand<VRRC>; +def crbitrc : RegisterOperand<CRBITRC>; +def crrc : RegisterOperand<CRRC>; + def s5imm : Operand<i32> { let PrintMethod = "printS5ImmOperand"; } @@ -346,9 +367,6 @@ def s16imm : Operand<i32> { def u16imm : Operand<i32> { let PrintMethod = "printU16ImmOperand"; } -def s16immX4 : Operand<i32> { // Multiply imm by 4 before printing. - let PrintMethod = "printS16X4ImmOperand"; -} def directbrtarget : Operand<OtherVT> { let PrintMethod = "printBranchOperand"; let EncoderMethod = "getDirectBrEncoding"; @@ -376,26 +394,37 @@ def crbitm: Operand<i8> { let EncoderMethod = "get_crbitm_encoding"; } // Address operands +// A version of ptr_rc which excludes R0 (or X0 in 64-bit mode). +def ptr_rc_nor0 : PointerLikeRegClass<1>; + +def dispRI : Operand<iPTR>; +def dispRIX : Operand<iPTR>; + def memri : Operand<iPTR> { let PrintMethod = "printMemRegImm"; - let MIOperandInfo = (ops symbolLo:$imm, ptr_rc:$reg); + let MIOperandInfo = (ops dispRI:$imm, ptr_rc_nor0:$reg); let EncoderMethod = "getMemRIEncoding"; } def memrr : Operand<iPTR> { let PrintMethod = "printMemRegReg"; - let MIOperandInfo = (ops ptr_rc:$offreg, ptr_rc:$ptrreg); + let MIOperandInfo = (ops ptr_rc_nor0:$ptrreg, ptr_rc:$offreg); } def memrix : Operand<iPTR> { // memri where the imm is shifted 2 bits. let PrintMethod = "printMemRegImmShifted"; - let MIOperandInfo = (ops symbolLo:$imm, ptr_rc:$reg); + let MIOperandInfo = (ops dispRIX:$imm, ptr_rc_nor0:$reg); let EncoderMethod = "getMemRIXEncoding"; } -// PowerPC Predicate operand. 20 = (0<<5)|20 = always, CR0 is a dummy reg -// that doesn't matter. -def pred : PredicateOperand<OtherVT, (ops imm, CRRC), - (ops (i32 20), (i32 zero_reg))> { +// A single-register address. This is used with the SjLj +// pseudo-instructions. +def memr : Operand<iPTR> { + let MIOperandInfo = (ops ptr_rc:$ptrreg); +} + +// PowerPC Predicate operand. +def pred : Operand<OtherVT> { let PrintMethod = "printPredicateOperand"; + let MIOperandInfo = (ops i32imm:$bibo, crrc:$reg); } // Define PowerPC specific addressing mode. @@ -404,9 +433,12 @@ def xaddr : ComplexPattern<iPTR, 2, "SelectAddrIdx", [], []>; def xoaddr : ComplexPattern<iPTR, 2, "SelectAddrIdxOnly",[], []>; def ixaddr : ComplexPattern<iPTR, 2, "SelectAddrImmShift", [], []>; // "std" +// The address in a single register. This is used with the SjLj +// pseudo-instructions. +def addr : ComplexPattern<iPTR, 1, "SelectAddr",[], []>; + /// This is just the offset part of iaddr, used for preinc. def iaddroff : ComplexPattern<iPTR, 1, "SelectAddrImmOffs", [], []>; -def xaddroff : ComplexPattern<iPTR, 1, "SelectAddrIdxOffs", [], []>; //===----------------------------------------------------------------------===// // PowerPC Instruction Predicate Definitions. @@ -415,6 +447,252 @@ def In64BitMode : Predicate<"PPCSubTarget.isPPC64()">; def IsBookE : Predicate<"PPCSubTarget.isBookE()">; //===----------------------------------------------------------------------===// +// PowerPC Multiclass Definitions. + +multiclass XForm_6r<bits<6> opcode, bits<10> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : XForm_6<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR0] in + def o : XForm_6<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass XForm_6rc<bits<6> opcode, bits<10> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + let Defs = [CARRY] in + def NAME : XForm_6<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CARRY, CR0] in + def o : XForm_6<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass XForm_10r<bits<6> opcode, bits<10> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : XForm_10<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR0] in + def o : XForm_10<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass XForm_10rc<bits<6> opcode, bits<10> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + let Defs = [CARRY] in + def NAME : XForm_10<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CARRY, CR0] in + def o : XForm_10<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass XForm_11r<bits<6> opcode, bits<10> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : XForm_11<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR0] in + def o : XForm_11<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass XOForm_1r<bits<6> opcode, bits<9> xo, bit oe, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : XOForm_1<opcode, xo, oe, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR0] in + def o : XOForm_1<opcode, xo, oe, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass XOForm_1rc<bits<6> opcode, bits<9> xo, bit oe, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + let Defs = [CARRY] in + def NAME : XOForm_1<opcode, xo, oe, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CARRY, CR0] in + def o : XOForm_1<opcode, xo, oe, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass XOForm_3r<bits<6> opcode, bits<9> xo, bit oe, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : XOForm_3<opcode, xo, oe, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR0] in + def o : XOForm_3<opcode, xo, oe, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass XOForm_3rc<bits<6> opcode, bits<9> xo, bit oe, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + let Defs = [CARRY] in + def NAME : XOForm_3<opcode, xo, oe, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CARRY, CR0] in + def o : XOForm_3<opcode, xo, oe, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass MForm_2r<bits<6> opcode, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : MForm_2<opcode, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR0] in + def o : MForm_2<opcode, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass MDForm_1r<bits<6> opcode, bits<3> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : MDForm_1<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR0] in + def o : MDForm_1<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass MDSForm_1r<bits<6> opcode, bits<4> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : MDSForm_1<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR0] in + def o : MDSForm_1<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass XSForm_1rc<bits<6> opcode, bits<9> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + let Defs = [CARRY] in + def NAME : XSForm_1<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CARRY, CR0] in + def o : XSForm_1<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass XForm_26r<bits<6> opcode, bits<10> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : XForm_26<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR1] in + def o : XForm_26<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass AForm_1r<bits<6> opcode, bits<5> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : AForm_1<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR1] in + def o : AForm_1<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass AForm_2r<bits<6> opcode, bits<5> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : AForm_2<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR1] in + def o : AForm_2<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +multiclass AForm_3r<bits<6> opcode, bits<5> xo, dag OOL, dag IOL, + string asmbase, string asmstr, InstrItinClass itin, + list<dag> pattern> { + let BaseName = asmbase in { + def NAME : AForm_3<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(" ", asmstr)), itin, + pattern>, RecFormRel; + let Defs = [CR1] in + def o : AForm_3<opcode, xo, OOL, IOL, + !strconcat(asmbase, !strconcat(". ", asmstr)), itin, + []>, isDOT, RecFormRel; + } +} + +//===----------------------------------------------------------------------===// // PowerPC Instruction Definitions. // Pseudo-instructions: @@ -427,32 +705,37 @@ def ADJCALLSTACKUP : Pseudo<(outs), (ins u16imm:$amt1, u16imm:$amt2), "#ADJCAL [(callseq_end timm:$amt1, timm:$amt2)]>; } -def UPDATE_VRSAVE : Pseudo<(outs GPRC:$rD), (ins GPRC:$rS), +def UPDATE_VRSAVE : Pseudo<(outs gprc:$rD), (ins gprc:$rS), "UPDATE_VRSAVE $rD, $rS", []>; } let Defs = [R1], Uses = [R1] in -def DYNALLOC : Pseudo<(outs GPRC:$result), (ins GPRC:$negsize, memri:$fpsi), "#DYNALLOC", - [(set GPRC:$result, - (PPCdynalloc GPRC:$negsize, iaddr:$fpsi))]>; +def DYNALLOC : Pseudo<(outs gprc:$result), (ins gprc:$negsize, memri:$fpsi), "#DYNALLOC", + [(set i32:$result, + (PPCdynalloc i32:$negsize, iaddr:$fpsi))]>; // SELECT_CC_* - Used to implement the SELECT_CC DAG operation. Expanded after // instruction selection into a branch sequence. let usesCustomInserter = 1, // Expanded after instruction selection. PPC970_Single = 1 in { - def SELECT_CC_I4 : Pseudo<(outs GPRC:$dst), (ins CRRC:$cond, GPRC:$T, GPRC:$F, + // Note that SELECT_CC_I4 and SELECT_CC_I8 use the no-r0 register classes + // because either operand might become the first operand in an isel, and + // that operand cannot be r0. + def SELECT_CC_I4 : Pseudo<(outs gprc:$dst), (ins crrc:$cond, + gprc_nor0:$T, gprc_nor0:$F, i32imm:$BROPC), "#SELECT_CC_I4", []>; - def SELECT_CC_I8 : Pseudo<(outs G8RC:$dst), (ins CRRC:$cond, G8RC:$T, G8RC:$F, + def SELECT_CC_I8 : Pseudo<(outs g8rc:$dst), (ins crrc:$cond, + g8rc_nox0:$T, g8rc_nox0:$F, i32imm:$BROPC), "#SELECT_CC_I8", []>; - def SELECT_CC_F4 : Pseudo<(outs F4RC:$dst), (ins CRRC:$cond, F4RC:$T, F4RC:$F, + def SELECT_CC_F4 : Pseudo<(outs f4rc:$dst), (ins crrc:$cond, f4rc:$T, f4rc:$F, i32imm:$BROPC), "#SELECT_CC_F4", []>; - def SELECT_CC_F8 : Pseudo<(outs F8RC:$dst), (ins CRRC:$cond, F8RC:$T, F8RC:$F, + def SELECT_CC_F8 : Pseudo<(outs f8rc:$dst), (ins crrc:$cond, f8rc:$T, f8rc:$F, i32imm:$BROPC), "#SELECT_CC_F8", []>; - def SELECT_CC_VRRC: Pseudo<(outs VRRC:$dst), (ins CRRC:$cond, VRRC:$T, VRRC:$F, + def SELECT_CC_VRRC: Pseudo<(outs vrrc:$dst), (ins crrc:$cond, vrrc:$T, vrrc:$F, i32imm:$BROPC), "#SELECT_CC_VRRC", []>; } @@ -460,22 +743,26 @@ let usesCustomInserter = 1, // Expanded after instruction selection. // SPILL_CR - Indicate that we're dumping the CR register, so we'll need to // scavenge a register for it. let mayStore = 1 in -def SPILL_CR : Pseudo<(outs), (ins CRRC:$cond, memri:$F), +def SPILL_CR : Pseudo<(outs), (ins crrc:$cond, memri:$F), "#SPILL_CR", []>; // RESTORE_CR - Indicate that we're restoring the CR register (previously // spilled), so we'll need to scavenge a register for it. let mayLoad = 1 in -def RESTORE_CR : Pseudo<(outs CRRC:$cond), (ins memri:$F), +def RESTORE_CR : Pseudo<(outs crrc:$cond), (ins memri:$F), "#RESTORE_CR", []>; let isTerminator = 1, isBarrier = 1, PPC970_Unit = 7 in { - let isCodeGenOnly = 1, isReturn = 1, Uses = [LR, RM] in - def BLR : XLForm_2_br<19, 16, 0, (outs), (ins pred:$p), - "b${p:cc}lr ${p:reg}", BrB, - [(retflag)]>; - let isBranch = 1, isIndirectBranch = 1, Uses = [CTR] in + let isReturn = 1, Uses = [LR, RM] in + def BLR : XLForm_2_ext<19, 16, 20, 0, 0, (outs), (ins), "blr", BrB, + [(retflag)]>; + let isBranch = 1, isIndirectBranch = 1, Uses = [CTR] in { def BCTR : XLForm_2_ext<19, 528, 20, 0, 0, (outs), (ins), "bctr", BrB, []>; + + let isCodeGenOnly = 1 in + def BCCTR : XLForm_2_br<19, 528, 0, (outs), (ins pred:$cond), + "b${cond:cc}ctr ${cond:reg}", BrB, []>; + } } let Defs = [LR] in @@ -492,10 +779,21 @@ let isBranch = 1, isTerminator = 1, hasCtrlDep = 1, PPC970_Unit = 7 in { // BCC represents an arbitrary conditional branch on a predicate. // FIXME: should be able to write a pattern for PPCcondbranch, but can't use // a two-value operand where a dag node expects two operands. :( - let isCodeGenOnly = 1 in + let isCodeGenOnly = 1 in { def BCC : BForm<16, 0, 0, (outs), (ins pred:$cond, condbrtarget:$dst), "b${cond:cc} ${cond:reg}, $dst" - /*[(PPCcondbranch CRRC:$crS, imm:$opc, bb:$dst)]*/>; + /*[(PPCcondbranch crrc:$crS, imm:$opc, bb:$dst)]*/>; + let isReturn = 1, Uses = [LR, RM] in + def BCLR : XLForm_2_br<19, 16, 0, (outs), (ins pred:$cond), + "b${cond:cc}lr ${cond:reg}", BrB, []>; + + let isReturn = 1, Defs = [CTR], Uses = [CTR, LR, RM] in { + def BDZLR : XLForm_2_ext<19, 16, 18, 0, 0, (outs), (ins), + "bdzlr", BrB, []>; + def BDNZLR : XLForm_2_ext<19, 16, 16, 0, 0, (outs), (ins), + "bdnzlr", BrB, []>; + } + } let Defs = [CTR], Uses = [CTR] in { def BDZ : BForm_1<16, 18, 0, 0, (outs), (ins condbrtarget:$dst), @@ -505,46 +803,33 @@ let isBranch = 1, isTerminator = 1, hasCtrlDep = 1, PPC970_Unit = 7 in { } } -// Darwin ABI Calls. -let isCall = 1, PPC970_Unit = 7, Defs = [LR] in { - // Convenient aliases for call instructions - let Uses = [RM] in { - def BL_Darwin : IForm<18, 0, 1, - (outs), (ins calltarget:$func), - "bl $func", BrB, []>; // See Pat patterns below. - def BLA_Darwin : IForm<18, 1, 1, - (outs), (ins aaddr:$func), - "bla $func", BrB, [(PPCcall_Darwin (i32 imm:$func))]>; - } - let Uses = [CTR, RM] in { - def BCTRL_Darwin : XLForm_2_ext<19, 528, 20, 0, 1, - (outs), (ins), - "bctrl", BrB, - [(PPCbctrl_Darwin)]>, Requires<[In32BitMode]>; +// The unconditional BCL used by the SjLj setjmp code. +let isCall = 1, hasCtrlDep = 1, isCodeGenOnly = 1, PPC970_Unit = 7 in { + let Defs = [LR], Uses = [RM] in { + def BCLalways : BForm_2<16, 20, 31, 0, 1, (outs), (ins condbrtarget:$dst), + "bcl 20, 31, $dst">; } } -// SVR4 ABI Calls. let isCall = 1, PPC970_Unit = 7, Defs = [LR] in { // Convenient aliases for call instructions let Uses = [RM] in { - def BL_SVR4 : IForm<18, 0, 1, - (outs), (ins calltarget:$func), - "bl $func", BrB, []>; // See Pat patterns below. - def BLA_SVR4 : IForm<18, 1, 1, - (outs), (ins aaddr:$func), - "bla $func", BrB, - [(PPCcall_SVR4 (i32 imm:$func))]>; + def BL : IForm<18, 0, 1, (outs), (ins calltarget:$func), + "bl $func", BrB, []>; // See Pat patterns below. + def BLA : IForm<18, 1, 1, (outs), (ins aaddr:$func), + "bla $func", BrB, [(PPCcall (i32 imm:$func))]>; } let Uses = [CTR, RM] in { - def BCTRL_SVR4 : XLForm_2_ext<19, 528, 20, 0, 1, - (outs), (ins), - "bctrl", BrB, - [(PPCbctrl_SVR4)]>, Requires<[In32BitMode]>; + def BCTRL : XLForm_2_ext<19, 528, 20, 0, 1, (outs), (ins), + "bctrl", BrB, [(PPCbctrl)]>, + Requires<[In32BitMode]>; + + let isCodeGenOnly = 1 in + def BCCTRL : XLForm_2_br<19, 528, 1, (outs), (ins pred:$cond), + "b${cond:cc}ctrl ${cond:reg}", BrB, []>; } } - let isCall = 1, isTerminator = 1, isReturn = 1, isBarrier = 1, Uses = [RM] in def TCRETURNdi :Pseudo< (outs), (ins calltarget:$dst, i32imm:$offset), @@ -563,6 +848,8 @@ def TCRETURNri : Pseudo<(outs), (ins CTRRC:$dst, i32imm:$offset), []>; +let isCodeGenOnly = 1 in { + let isTerminator = 1, isBarrier = 1, PPC970_Unit = 7, isBranch = 1, isIndirectBranch = 1, isCall = 1, isReturn = 1, Uses = [CTR, RM] in def TAILBCTR : XLForm_2_ext<19, 528, 20, 0, 0, (outs), (ins), "bctr", BrB, []>, @@ -576,6 +863,7 @@ def TAILB : IForm<18, 0, 0, (outs), (ins calltarget:$dst), "b $dst", BrB, []>; +} let isBranch = 1, isTerminator = 1, hasCtrlDep = 1, PPC970_Unit = 7, isBarrier = 1, isCall = 1, isReturn = 1, Uses = [RM] in @@ -583,6 +871,22 @@ def TAILBA : IForm<18, 0, 0, (outs), (ins aaddr:$dst), "ba $dst", BrB, []>; +let hasSideEffects = 1, isBarrier = 1, usesCustomInserter = 1 in { + def EH_SjLj_SetJmp32 : Pseudo<(outs gprc:$dst), (ins memr:$buf), + "#EH_SJLJ_SETJMP32", + [(set i32:$dst, (PPCeh_sjlj_setjmp addr:$buf))]>, + Requires<[In32BitMode]>; + let isTerminator = 1 in + def EH_SjLj_LongJmp32 : Pseudo<(outs), (ins memr:$buf), + "#EH_SJLJ_LONGJMP32", + [(PPCeh_sjlj_longjmp addr:$buf)]>, + Requires<[In32BitMode]>; +} + +let isBranch = 1, isTerminator = 1 in { + def EH_SjLj_Setup : Pseudo<(outs), (ins directbrtarget:$dst), + "#EH_SjLj_Setup\t$dst", []>; +} // DCB* instructions. def DCBA : DCB_Form<758, 0, (outs), (ins memrr:$dst), @@ -617,94 +921,91 @@ def : Pat<(prefetch xoaddr:$dst, (i32 0), imm, (i32 1)), let usesCustomInserter = 1 in { let Defs = [CR0] in { def ATOMIC_LOAD_ADD_I8 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_ADD_I8", - [(set GPRC:$dst, (atomic_load_add_8 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_ADD_I8", + [(set i32:$dst, (atomic_load_add_8 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_SUB_I8 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_SUB_I8", - [(set GPRC:$dst, (atomic_load_sub_8 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_SUB_I8", + [(set i32:$dst, (atomic_load_sub_8 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_AND_I8 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_AND_I8", - [(set GPRC:$dst, (atomic_load_and_8 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_AND_I8", + [(set i32:$dst, (atomic_load_and_8 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_OR_I8 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_OR_I8", - [(set GPRC:$dst, (atomic_load_or_8 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_OR_I8", + [(set i32:$dst, (atomic_load_or_8 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_XOR_I8 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "ATOMIC_LOAD_XOR_I8", - [(set GPRC:$dst, (atomic_load_xor_8 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "ATOMIC_LOAD_XOR_I8", + [(set i32:$dst, (atomic_load_xor_8 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_NAND_I8 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_NAND_I8", - [(set GPRC:$dst, (atomic_load_nand_8 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_NAND_I8", + [(set i32:$dst, (atomic_load_nand_8 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_ADD_I16 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_ADD_I16", - [(set GPRC:$dst, (atomic_load_add_16 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_ADD_I16", + [(set i32:$dst, (atomic_load_add_16 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_SUB_I16 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_SUB_I16", - [(set GPRC:$dst, (atomic_load_sub_16 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_SUB_I16", + [(set i32:$dst, (atomic_load_sub_16 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_AND_I16 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_AND_I16", - [(set GPRC:$dst, (atomic_load_and_16 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_AND_I16", + [(set i32:$dst, (atomic_load_and_16 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_OR_I16 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_OR_I16", - [(set GPRC:$dst, (atomic_load_or_16 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_OR_I16", + [(set i32:$dst, (atomic_load_or_16 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_XOR_I16 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_XOR_I16", - [(set GPRC:$dst, (atomic_load_xor_16 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_XOR_I16", + [(set i32:$dst, (atomic_load_xor_16 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_NAND_I16 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_NAND_I16", - [(set GPRC:$dst, (atomic_load_nand_16 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_NAND_I16", + [(set i32:$dst, (atomic_load_nand_16 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_ADD_I32 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_ADD_I32", - [(set GPRC:$dst, (atomic_load_add_32 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_ADD_I32", + [(set i32:$dst, (atomic_load_add_32 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_SUB_I32 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_SUB_I32", - [(set GPRC:$dst, (atomic_load_sub_32 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_SUB_I32", + [(set i32:$dst, (atomic_load_sub_32 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_AND_I32 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_AND_I32", - [(set GPRC:$dst, (atomic_load_and_32 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_AND_I32", + [(set i32:$dst, (atomic_load_and_32 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_OR_I32 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_OR_I32", - [(set GPRC:$dst, (atomic_load_or_32 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_OR_I32", + [(set i32:$dst, (atomic_load_or_32 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_XOR_I32 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_XOR_I32", - [(set GPRC:$dst, (atomic_load_xor_32 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_XOR_I32", + [(set i32:$dst, (atomic_load_xor_32 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_LOAD_NAND_I32 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$incr), "#ATOMIC_LOAD_NAND_I32", - [(set GPRC:$dst, (atomic_load_nand_32 xoaddr:$ptr, GPRC:$incr))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$incr), "#ATOMIC_LOAD_NAND_I32", + [(set i32:$dst, (atomic_load_nand_32 xoaddr:$ptr, i32:$incr))]>; def ATOMIC_CMP_SWAP_I8 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$old, GPRC:$new), "#ATOMIC_CMP_SWAP_I8", - [(set GPRC:$dst, - (atomic_cmp_swap_8 xoaddr:$ptr, GPRC:$old, GPRC:$new))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$old, gprc:$new), "#ATOMIC_CMP_SWAP_I8", + [(set i32:$dst, (atomic_cmp_swap_8 xoaddr:$ptr, i32:$old, i32:$new))]>; def ATOMIC_CMP_SWAP_I16 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$old, GPRC:$new), "#ATOMIC_CMP_SWAP_I16 $dst $ptr $old $new", - [(set GPRC:$dst, - (atomic_cmp_swap_16 xoaddr:$ptr, GPRC:$old, GPRC:$new))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$old, gprc:$new), "#ATOMIC_CMP_SWAP_I16 $dst $ptr $old $new", + [(set i32:$dst, (atomic_cmp_swap_16 xoaddr:$ptr, i32:$old, i32:$new))]>; def ATOMIC_CMP_SWAP_I32 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$old, GPRC:$new), "#ATOMIC_CMP_SWAP_I32 $dst $ptr $old $new", - [(set GPRC:$dst, - (atomic_cmp_swap_32 xoaddr:$ptr, GPRC:$old, GPRC:$new))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$old, gprc:$new), "#ATOMIC_CMP_SWAP_I32 $dst $ptr $old $new", + [(set i32:$dst, (atomic_cmp_swap_32 xoaddr:$ptr, i32:$old, i32:$new))]>; def ATOMIC_SWAP_I8 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$new), "#ATOMIC_SWAP_i8", - [(set GPRC:$dst, (atomic_swap_8 xoaddr:$ptr, GPRC:$new))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$new), "#ATOMIC_SWAP_i8", + [(set i32:$dst, (atomic_swap_8 xoaddr:$ptr, i32:$new))]>; def ATOMIC_SWAP_I16 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$new), "#ATOMIC_SWAP_I16", - [(set GPRC:$dst, (atomic_swap_16 xoaddr:$ptr, GPRC:$new))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$new), "#ATOMIC_SWAP_I16", + [(set i32:$dst, (atomic_swap_16 xoaddr:$ptr, i32:$new))]>; def ATOMIC_SWAP_I32 : Pseudo< - (outs GPRC:$dst), (ins memrr:$ptr, GPRC:$new), "#ATOMIC_SWAP_I32", - [(set GPRC:$dst, (atomic_swap_32 xoaddr:$ptr, GPRC:$new))]>; + (outs gprc:$dst), (ins memrr:$ptr, gprc:$new), "#ATOMIC_SWAP_I32", + [(set i32:$dst, (atomic_swap_32 xoaddr:$ptr, i32:$new))]>; } } // Instructions to support atomic operations -def LWARX : XForm_1<31, 20, (outs GPRC:$rD), (ins memrr:$src), +def LWARX : XForm_1<31, 20, (outs gprc:$rD), (ins memrr:$src), "lwarx $rD, $src", LdStLWARX, - [(set GPRC:$rD, (PPClarx xoaddr:$src))]>; + [(set i32:$rD, (PPClarx xoaddr:$src))]>; let Defs = [CR0] in -def STWCX : XForm_1<31, 150, (outs), (ins GPRC:$rS, memrr:$dst), +def STWCX : XForm_1<31, 150, (outs), (ins gprc:$rS, memrr:$dst), "stwcx. $rS, $dst", LdStSTWCX, - [(PPCstcx GPRC:$rS, xoaddr:$dst)]>, + [(PPCstcx i32:$rS, xoaddr:$dst)]>, isDOT; let isTerminator = 1, isBarrier = 1, hasCtrlDep = 1 in @@ -716,96 +1017,96 @@ def TRAP : XForm_24<31, 4, (outs), (ins), "trap", LdStLoad, [(trap)]>; // Unindexed (r+i) Loads. let canFoldAsLoad = 1, PPC970_Unit = 2 in { -def LBZ : DForm_1<34, (outs GPRC:$rD), (ins memri:$src), +def LBZ : DForm_1<34, (outs gprc:$rD), (ins memri:$src), "lbz $rD, $src", LdStLoad, - [(set GPRC:$rD, (zextloadi8 iaddr:$src))]>; -def LHA : DForm_1<42, (outs GPRC:$rD), (ins memri:$src), + [(set i32:$rD, (zextloadi8 iaddr:$src))]>; +def LHA : DForm_1<42, (outs gprc:$rD), (ins memri:$src), "lha $rD, $src", LdStLHA, - [(set GPRC:$rD, (sextloadi16 iaddr:$src))]>, + [(set i32:$rD, (sextloadi16 iaddr:$src))]>, PPC970_DGroup_Cracked; -def LHZ : DForm_1<40, (outs GPRC:$rD), (ins memri:$src), +def LHZ : DForm_1<40, (outs gprc:$rD), (ins memri:$src), "lhz $rD, $src", LdStLoad, - [(set GPRC:$rD, (zextloadi16 iaddr:$src))]>; -def LWZ : DForm_1<32, (outs GPRC:$rD), (ins memri:$src), + [(set i32:$rD, (zextloadi16 iaddr:$src))]>; +def LWZ : DForm_1<32, (outs gprc:$rD), (ins memri:$src), "lwz $rD, $src", LdStLoad, - [(set GPRC:$rD, (load iaddr:$src))]>; + [(set i32:$rD, (load iaddr:$src))]>; -def LFS : DForm_1<48, (outs F4RC:$rD), (ins memri:$src), +def LFS : DForm_1<48, (outs f4rc:$rD), (ins memri:$src), "lfs $rD, $src", LdStLFD, - [(set F4RC:$rD, (load iaddr:$src))]>; -def LFD : DForm_1<50, (outs F8RC:$rD), (ins memri:$src), + [(set f32:$rD, (load iaddr:$src))]>; +def LFD : DForm_1<50, (outs f8rc:$rD), (ins memri:$src), "lfd $rD, $src", LdStLFD, - [(set F8RC:$rD, (load iaddr:$src))]>; + [(set f64:$rD, (load iaddr:$src))]>; // Unindexed (r+i) Loads with Update (preinc). -let mayLoad = 1 in { -def LBZU : DForm_1<35, (outs GPRC:$rD, ptr_rc:$ea_result), (ins memri:$addr), +let mayLoad = 1, neverHasSideEffects = 1 in { +def LBZU : DForm_1<35, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr), "lbzu $rD, $addr", LdStLoadUpd, []>, RegConstraint<"$addr.reg = $ea_result">, NoEncode<"$ea_result">; -def LHAU : DForm_1<43, (outs GPRC:$rD, ptr_rc:$ea_result), (ins memri:$addr), +def LHAU : DForm_1<43, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr), "lhau $rD, $addr", LdStLHAU, []>, RegConstraint<"$addr.reg = $ea_result">, NoEncode<"$ea_result">; -def LHZU : DForm_1<41, (outs GPRC:$rD, ptr_rc:$ea_result), (ins memri:$addr), +def LHZU : DForm_1<41, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr), "lhzu $rD, $addr", LdStLoadUpd, []>, RegConstraint<"$addr.reg = $ea_result">, NoEncode<"$ea_result">; -def LWZU : DForm_1<33, (outs GPRC:$rD, ptr_rc:$ea_result), (ins memri:$addr), +def LWZU : DForm_1<33, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr), "lwzu $rD, $addr", LdStLoadUpd, []>, RegConstraint<"$addr.reg = $ea_result">, NoEncode<"$ea_result">; -def LFSU : DForm_1<49, (outs F4RC:$rD, ptr_rc:$ea_result), (ins memri:$addr), +def LFSU : DForm_1<49, (outs f4rc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr), "lfsu $rD, $addr", LdStLFDU, []>, RegConstraint<"$addr.reg = $ea_result">, NoEncode<"$ea_result">; -def LFDU : DForm_1<51, (outs F8RC:$rD, ptr_rc:$ea_result), (ins memri:$addr), +def LFDU : DForm_1<51, (outs f8rc:$rD, ptr_rc_nor0:$ea_result), (ins memri:$addr), "lfdu $rD, $addr", LdStLFDU, []>, RegConstraint<"$addr.reg = $ea_result">, NoEncode<"$ea_result">; // Indexed (r+r) Loads with Update (preinc). -def LBZUX : XForm_1<31, 119, (outs GPRC:$rD, ptr_rc:$ea_result), +def LBZUX : XForm_1<31, 119, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "lbzux $rD, $addr", LdStLoadUpd, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">; -def LHAUX : XForm_1<31, 375, (outs GPRC:$rD, ptr_rc:$ea_result), +def LHAUX : XForm_1<31, 375, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "lhaux $rD, $addr", LdStLHAU, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">; -def LHZUX : XForm_1<31, 311, (outs GPRC:$rD, ptr_rc:$ea_result), +def LHZUX : XForm_1<31, 311, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "lhzux $rD, $addr", LdStLoadUpd, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">; -def LWZUX : XForm_1<31, 55, (outs GPRC:$rD, ptr_rc:$ea_result), +def LWZUX : XForm_1<31, 55, (outs gprc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "lwzux $rD, $addr", LdStLoadUpd, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">; -def LFSUX : XForm_1<31, 567, (outs F4RC:$rD, ptr_rc:$ea_result), +def LFSUX : XForm_1<31, 567, (outs f4rc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "lfsux $rD, $addr", LdStLFDU, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">; -def LFDUX : XForm_1<31, 631, (outs F8RC:$rD, ptr_rc:$ea_result), +def LFDUX : XForm_1<31, 631, (outs f8rc:$rD, ptr_rc_nor0:$ea_result), (ins memrr:$addr), "lfdux $rD, $addr", LdStLFDU, - []>, RegConstraint<"$addr.offreg = $ea_result">, + []>, RegConstraint<"$addr.ptrreg = $ea_result">, NoEncode<"$ea_result">; } } @@ -813,34 +1114,41 @@ def LFDUX : XForm_1<31, 631, (outs F8RC:$rD, ptr_rc:$ea_result), // Indexed (r+r) Loads. // let canFoldAsLoad = 1, PPC970_Unit = 2 in { -def LBZX : XForm_1<31, 87, (outs GPRC:$rD), (ins memrr:$src), +def LBZX : XForm_1<31, 87, (outs gprc:$rD), (ins memrr:$src), "lbzx $rD, $src", LdStLoad, - [(set GPRC:$rD, (zextloadi8 xaddr:$src))]>; -def LHAX : XForm_1<31, 343, (outs GPRC:$rD), (ins memrr:$src), + [(set i32:$rD, (zextloadi8 xaddr:$src))]>; +def LHAX : XForm_1<31, 343, (outs gprc:$rD), (ins memrr:$src), "lhax $rD, $src", LdStLHA, - [(set GPRC:$rD, (sextloadi16 xaddr:$src))]>, + [(set i32:$rD, (sextloadi16 xaddr:$src))]>, PPC970_DGroup_Cracked; -def LHZX : XForm_1<31, 279, (outs GPRC:$rD), (ins memrr:$src), +def LHZX : XForm_1<31, 279, (outs gprc:$rD), (ins memrr:$src), "lhzx $rD, $src", LdStLoad, - [(set GPRC:$rD, (zextloadi16 xaddr:$src))]>; -def LWZX : XForm_1<31, 23, (outs GPRC:$rD), (ins memrr:$src), + [(set i32:$rD, (zextloadi16 xaddr:$src))]>; +def LWZX : XForm_1<31, 23, (outs gprc:$rD), (ins memrr:$src), "lwzx $rD, $src", LdStLoad, - [(set GPRC:$rD, (load xaddr:$src))]>; + [(set i32:$rD, (load xaddr:$src))]>; -def LHBRX : XForm_1<31, 790, (outs GPRC:$rD), (ins memrr:$src), +def LHBRX : XForm_1<31, 790, (outs gprc:$rD), (ins memrr:$src), "lhbrx $rD, $src", LdStLoad, - [(set GPRC:$rD, (PPClbrx xoaddr:$src, i16))]>; -def LWBRX : XForm_1<31, 534, (outs GPRC:$rD), (ins memrr:$src), + [(set i32:$rD, (PPClbrx xoaddr:$src, i16))]>; +def LWBRX : XForm_1<31, 534, (outs gprc:$rD), (ins memrr:$src), "lwbrx $rD, $src", LdStLoad, - [(set GPRC:$rD, (PPClbrx xoaddr:$src, i32))]>; + [(set i32:$rD, (PPClbrx xoaddr:$src, i32))]>; -def LFSX : XForm_25<31, 535, (outs F4RC:$frD), (ins memrr:$src), +def LFSX : XForm_25<31, 535, (outs f4rc:$frD), (ins memrr:$src), "lfsx $frD, $src", LdStLFD, - [(set F4RC:$frD, (load xaddr:$src))]>; -def LFDX : XForm_25<31, 599, (outs F8RC:$frD), (ins memrr:$src), + [(set f32:$frD, (load xaddr:$src))]>; +def LFDX : XForm_25<31, 599, (outs f8rc:$frD), (ins memrr:$src), "lfdx $frD, $src", LdStLFD, - [(set F8RC:$frD, (load xaddr:$src))]>; + [(set f64:$frD, (load xaddr:$src))]>; + +def LFIWAX : XForm_25<31, 855, (outs f8rc:$frD), (ins memrr:$src), + "lfiwax $frD, $src", LdStLFD, + [(set f64:$frD, (PPClfiwax xoaddr:$src))]>; +def LFIWZX : XForm_25<31, 887, (outs f8rc:$frD), (ins memrr:$src), + "lfiwzx $frD, $src", LdStLFD, + [(set f64:$frD, (PPClfiwzx xoaddr:$src))]>; } //===----------------------------------------------------------------------===// @@ -849,139 +1157,130 @@ def LFDX : XForm_25<31, 599, (outs F8RC:$frD), (ins memrr:$src), // Unindexed (r+i) Stores. let PPC970_Unit = 2 in { -def STB : DForm_1<38, (outs), (ins GPRC:$rS, memri:$src), +def STB : DForm_1<38, (outs), (ins gprc:$rS, memri:$src), "stb $rS, $src", LdStStore, - [(truncstorei8 GPRC:$rS, iaddr:$src)]>; -def STH : DForm_1<44, (outs), (ins GPRC:$rS, memri:$src), + [(truncstorei8 i32:$rS, iaddr:$src)]>; +def STH : DForm_1<44, (outs), (ins gprc:$rS, memri:$src), "sth $rS, $src", LdStStore, - [(truncstorei16 GPRC:$rS, iaddr:$src)]>; -def STW : DForm_1<36, (outs), (ins GPRC:$rS, memri:$src), + [(truncstorei16 i32:$rS, iaddr:$src)]>; +def STW : DForm_1<36, (outs), (ins gprc:$rS, memri:$src), "stw $rS, $src", LdStStore, - [(store GPRC:$rS, iaddr:$src)]>; -def STFS : DForm_1<52, (outs), (ins F4RC:$rS, memri:$dst), + [(store i32:$rS, iaddr:$src)]>; +def STFS : DForm_1<52, (outs), (ins f4rc:$rS, memri:$dst), "stfs $rS, $dst", LdStSTFD, - [(store F4RC:$rS, iaddr:$dst)]>; -def STFD : DForm_1<54, (outs), (ins F8RC:$rS, memri:$dst), + [(store f32:$rS, iaddr:$dst)]>; +def STFD : DForm_1<54, (outs), (ins f8rc:$rS, memri:$dst), "stfd $rS, $dst", LdStSTFD, - [(store F8RC:$rS, iaddr:$dst)]>; + [(store f64:$rS, iaddr:$dst)]>; } // Unindexed (r+i) Stores with Update (preinc). -let PPC970_Unit = 2 in { -def STBU : DForm_1a<39, (outs ptr_rc:$ea_res), (ins GPRC:$rS, - symbolLo:$ptroff, ptr_rc:$ptrreg), - "stbu $rS, $ptroff($ptrreg)", LdStStoreUpd, - [(set ptr_rc:$ea_res, - (pre_truncsti8 GPRC:$rS, ptr_rc:$ptrreg, - iaddroff:$ptroff))]>, - RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">; -def STHU : DForm_1a<45, (outs ptr_rc:$ea_res), (ins GPRC:$rS, - symbolLo:$ptroff, ptr_rc:$ptrreg), - "sthu $rS, $ptroff($ptrreg)", LdStStoreUpd, - [(set ptr_rc:$ea_res, - (pre_truncsti16 GPRC:$rS, ptr_rc:$ptrreg, - iaddroff:$ptroff))]>, - RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">; -def STWU : DForm_1a<37, (outs ptr_rc:$ea_res), (ins GPRC:$rS, - symbolLo:$ptroff, ptr_rc:$ptrreg), - "stwu $rS, $ptroff($ptrreg)", LdStStoreUpd, - [(set ptr_rc:$ea_res, (pre_store GPRC:$rS, ptr_rc:$ptrreg, - iaddroff:$ptroff))]>, - RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">; -def STFSU : DForm_1a<37, (outs ptr_rc:$ea_res), (ins F4RC:$rS, - symbolLo:$ptroff, ptr_rc:$ptrreg), - "stfsu $rS, $ptroff($ptrreg)", LdStSTFDU, - [(set ptr_rc:$ea_res, (pre_store F4RC:$rS, ptr_rc:$ptrreg, - iaddroff:$ptroff))]>, - RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">; -def STFDU : DForm_1a<37, (outs ptr_rc:$ea_res), (ins F8RC:$rS, - symbolLo:$ptroff, ptr_rc:$ptrreg), - "stfdu $rS, $ptroff($ptrreg)", LdStSTFDU, - [(set ptr_rc:$ea_res, (pre_store F8RC:$rS, ptr_rc:$ptrreg, - iaddroff:$ptroff))]>, - RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">; +let PPC970_Unit = 2, mayStore = 1 in { +def STBU : DForm_1<39, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memri:$dst), + "stbu $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">; +def STHU : DForm_1<45, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memri:$dst), + "sthu $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">; +def STWU : DForm_1<37, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memri:$dst), + "stwu $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">; +def STFSU : DForm_1<53, (outs ptr_rc_nor0:$ea_res), (ins f4rc:$rS, memri:$dst), + "stfsu $rS, $dst", LdStSTFDU, []>, + RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">; +def STFDU : DForm_1<55, (outs ptr_rc_nor0:$ea_res), (ins f8rc:$rS, memri:$dst), + "stfdu $rS, $dst", LdStSTFDU, []>, + RegConstraint<"$dst.reg = $ea_res">, NoEncode<"$ea_res">; } +// Patterns to match the pre-inc stores. We can't put the patterns on +// the instruction definitions directly as ISel wants the address base +// and offset to be separate operands, not a single complex operand. +def : Pat<(pre_truncsti8 i32:$rS, iPTR:$ptrreg, iaddroff:$ptroff), + (STBU $rS, iaddroff:$ptroff, $ptrreg)>; +def : Pat<(pre_truncsti16 i32:$rS, iPTR:$ptrreg, iaddroff:$ptroff), + (STHU $rS, iaddroff:$ptroff, $ptrreg)>; +def : Pat<(pre_store i32:$rS, iPTR:$ptrreg, iaddroff:$ptroff), + (STWU $rS, iaddroff:$ptroff, $ptrreg)>; +def : Pat<(pre_store f32:$rS, iPTR:$ptrreg, iaddroff:$ptroff), + (STFSU $rS, iaddroff:$ptroff, $ptrreg)>; +def : Pat<(pre_store f64:$rS, iPTR:$ptrreg, iaddroff:$ptroff), + (STFDU $rS, iaddroff:$ptroff, $ptrreg)>; // Indexed (r+r) Stores. -// let PPC970_Unit = 2 in { -def STBX : XForm_8<31, 215, (outs), (ins GPRC:$rS, memrr:$dst), +def STBX : XForm_8<31, 215, (outs), (ins gprc:$rS, memrr:$dst), "stbx $rS, $dst", LdStStore, - [(truncstorei8 GPRC:$rS, xaddr:$dst)]>, + [(truncstorei8 i32:$rS, xaddr:$dst)]>, PPC970_DGroup_Cracked; -def STHX : XForm_8<31, 407, (outs), (ins GPRC:$rS, memrr:$dst), +def STHX : XForm_8<31, 407, (outs), (ins gprc:$rS, memrr:$dst), "sthx $rS, $dst", LdStStore, - [(truncstorei16 GPRC:$rS, xaddr:$dst)]>, + [(truncstorei16 i32:$rS, xaddr:$dst)]>, PPC970_DGroup_Cracked; -def STWX : XForm_8<31, 151, (outs), (ins GPRC:$rS, memrr:$dst), +def STWX : XForm_8<31, 151, (outs), (ins gprc:$rS, memrr:$dst), "stwx $rS, $dst", LdStStore, - [(store GPRC:$rS, xaddr:$dst)]>, + [(store i32:$rS, xaddr:$dst)]>, PPC970_DGroup_Cracked; -def STBUX : XForm_8<31, 247, (outs ptr_rc:$ea_res), - (ins GPRC:$rS, ptr_rc:$ptroff, ptr_rc:$ptrreg), - "stbux $rS, $ptroff, $ptrreg", LdStStoreUpd, - [(set ptr_rc:$ea_res, - (pre_truncsti8 GPRC:$rS, - ptr_rc:$ptrreg, xaddroff:$ptroff))]>, - RegConstraint<"$ptroff = $ea_res">, NoEncode<"$ea_res">, - PPC970_DGroup_Cracked; - -def STHUX : XForm_8<31, 439, (outs ptr_rc:$ea_res), - (ins GPRC:$rS, ptr_rc:$ptroff, ptr_rc:$ptrreg), - "sthux $rS, $ptroff, $ptrreg", LdStStoreUpd, - [(set ptr_rc:$ea_res, - (pre_truncsti16 GPRC:$rS, - ptr_rc:$ptrreg, xaddroff:$ptroff))]>, - RegConstraint<"$ptroff = $ea_res">, NoEncode<"$ea_res">, - PPC970_DGroup_Cracked; - -def STWUX : XForm_8<31, 183, (outs ptr_rc:$ea_res), - (ins GPRC:$rS, ptr_rc:$ptroff, ptr_rc:$ptrreg), - "stwux $rS, $ptroff, $ptrreg", LdStStoreUpd, - [(set ptr_rc:$ea_res, - (pre_store GPRC:$rS, ptr_rc:$ptrreg, xaddroff:$ptroff))]>, - RegConstraint<"$ptroff = $ea_res">, NoEncode<"$ea_res">, - PPC970_DGroup_Cracked; - -def STFSUX : XForm_8<31, 695, (outs ptr_rc:$ea_res), - (ins F4RC:$rS, ptr_rc:$ptroff, ptr_rc:$ptrreg), - "stfsux $rS, $ptroff, $ptrreg", LdStSTFDU, - [(set ptr_rc:$ea_res, - (pre_store F4RC:$rS, ptr_rc:$ptrreg, xaddroff:$ptroff))]>, - RegConstraint<"$ptroff = $ea_res">, NoEncode<"$ea_res">, - PPC970_DGroup_Cracked; - -def STFDUX : XForm_8<31, 759, (outs ptr_rc:$ea_res), - (ins F8RC:$rS, ptr_rc:$ptroff, ptr_rc:$ptrreg), - "stfdux $rS, $ptroff, $ptrreg", LdStSTFDU, - [(set ptr_rc:$ea_res, - (pre_store F8RC:$rS, ptr_rc:$ptrreg, xaddroff:$ptroff))]>, - RegConstraint<"$ptroff = $ea_res">, NoEncode<"$ea_res">, - PPC970_DGroup_Cracked; - -def STHBRX: XForm_8<31, 918, (outs), (ins GPRC:$rS, memrr:$dst), +def STHBRX: XForm_8<31, 918, (outs), (ins gprc:$rS, memrr:$dst), "sthbrx $rS, $dst", LdStStore, - [(PPCstbrx GPRC:$rS, xoaddr:$dst, i16)]>, + [(PPCstbrx i32:$rS, xoaddr:$dst, i16)]>, PPC970_DGroup_Cracked; -def STWBRX: XForm_8<31, 662, (outs), (ins GPRC:$rS, memrr:$dst), +def STWBRX: XForm_8<31, 662, (outs), (ins gprc:$rS, memrr:$dst), "stwbrx $rS, $dst", LdStStore, - [(PPCstbrx GPRC:$rS, xoaddr:$dst, i32)]>, + [(PPCstbrx i32:$rS, xoaddr:$dst, i32)]>, PPC970_DGroup_Cracked; -def STFIWX: XForm_28<31, 983, (outs), (ins F8RC:$frS, memrr:$dst), +def STFIWX: XForm_28<31, 983, (outs), (ins f8rc:$frS, memrr:$dst), "stfiwx $frS, $dst", LdStSTFD, - [(PPCstfiwx F8RC:$frS, xoaddr:$dst)]>; + [(PPCstfiwx f64:$frS, xoaddr:$dst)]>; -def STFSX : XForm_28<31, 663, (outs), (ins F4RC:$frS, memrr:$dst), +def STFSX : XForm_28<31, 663, (outs), (ins f4rc:$frS, memrr:$dst), "stfsx $frS, $dst", LdStSTFD, - [(store F4RC:$frS, xaddr:$dst)]>; -def STFDX : XForm_28<31, 727, (outs), (ins F8RC:$frS, memrr:$dst), + [(store f32:$frS, xaddr:$dst)]>; +def STFDX : XForm_28<31, 727, (outs), (ins f8rc:$frS, memrr:$dst), "stfdx $frS, $dst", LdStSTFD, - [(store F8RC:$frS, xaddr:$dst)]>; + [(store f64:$frS, xaddr:$dst)]>; +} + +// Indexed (r+r) Stores with Update (preinc). +let PPC970_Unit = 2, mayStore = 1 in { +def STBUX : XForm_8<31, 247, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memrr:$dst), + "stbux $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">, + PPC970_DGroup_Cracked; +def STHUX : XForm_8<31, 439, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memrr:$dst), + "sthux $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">, + PPC970_DGroup_Cracked; +def STWUX : XForm_8<31, 183, (outs ptr_rc_nor0:$ea_res), (ins gprc:$rS, memrr:$dst), + "stwux $rS, $dst", LdStStoreUpd, []>, + RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">, + PPC970_DGroup_Cracked; +def STFSUX: XForm_8<31, 695, (outs ptr_rc_nor0:$ea_res), (ins f4rc:$rS, memrr:$dst), + "stfsux $rS, $dst", LdStSTFDU, []>, + RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">, + PPC970_DGroup_Cracked; +def STFDUX: XForm_8<31, 759, (outs ptr_rc_nor0:$ea_res), (ins f8rc:$rS, memrr:$dst), + "stfdux $rS, $dst", LdStSTFDU, []>, + RegConstraint<"$dst.ptrreg = $ea_res">, NoEncode<"$ea_res">, + PPC970_DGroup_Cracked; } +// Patterns to match the pre-inc stores. We can't put the patterns on +// the instruction definitions directly as ISel wants the address base +// and offset to be separate operands, not a single complex operand. +def : Pat<(pre_truncsti8 i32:$rS, iPTR:$ptrreg, iPTR:$ptroff), + (STBUX $rS, $ptrreg, $ptroff)>; +def : Pat<(pre_truncsti16 i32:$rS, iPTR:$ptrreg, iPTR:$ptroff), + (STHUX $rS, $ptrreg, $ptroff)>; +def : Pat<(pre_store i32:$rS, iPTR:$ptrreg, iPTR:$ptroff), + (STWUX $rS, $ptrreg, $ptroff)>; +def : Pat<(pre_store f32:$rS, iPTR:$ptrreg, iPTR:$ptroff), + (STFSUX $rS, $ptrreg, $ptroff)>; +def : Pat<(pre_store f64:$rS, iPTR:$ptrreg, iPTR:$ptroff), + (STFDUX $rS, $ptrreg, $ptroff)>; + def SYNC : XForm_24_sync<31, 598, (outs), (ins), "sync", LdStSync, [(int_ppc_sync)]>; @@ -991,157 +1290,206 @@ def SYNC : XForm_24_sync<31, 598, (outs), (ins), // let PPC970_Unit = 1 in { // FXU Operations. -def ADDI : DForm_2<14, (outs GPRC:$rD), (ins GPRC:$rA, s16imm:$imm), - "addi $rD, $rA, $imm", IntSimple, - [(set GPRC:$rD, (add GPRC:$rA, immSExt16:$imm))]>; -def ADDIL : DForm_2<14, (outs GPRC:$rD), (ins GPRC:$rA, symbolLo:$imm), +def ADDI : DForm_2<14, (outs gprc:$rD), (ins gprc_nor0:$rA, symbolLo:$imm), "addi $rD, $rA, $imm", IntSimple, - [(set GPRC:$rD, (add GPRC:$rA, immSExt16:$imm))]>; -let Defs = [CARRY] in { -def ADDIC : DForm_2<12, (outs GPRC:$rD), (ins GPRC:$rA, s16imm:$imm), + [(set i32:$rD, (add i32:$rA, immSExt16:$imm))]>; +let BaseName = "addic" in { +let Defs = [CARRY] in +def ADDIC : DForm_2<12, (outs gprc:$rD), (ins gprc:$rA, s16imm:$imm), "addic $rD, $rA, $imm", IntGeneral, - [(set GPRC:$rD, (addc GPRC:$rA, immSExt16:$imm))]>, - PPC970_DGroup_Cracked; -def ADDICo : DForm_2<13, (outs GPRC:$rD), (ins GPRC:$rA, s16imm:$imm), + [(set i32:$rD, (addc i32:$rA, immSExt16:$imm))]>, + RecFormRel, PPC970_DGroup_Cracked; +let Defs = [CARRY, CR0] in +def ADDICo : DForm_2<13, (outs gprc:$rD), (ins gprc:$rA, s16imm:$imm), "addic. $rD, $rA, $imm", IntGeneral, - []>; + []>, isDOT, RecFormRel; } -def ADDIS : DForm_2<15, (outs GPRC:$rD), (ins GPRC:$rA, symbolHi:$imm), +def ADDIS : DForm_2<15, (outs gprc:$rD), (ins gprc_nor0:$rA, symbolHi:$imm), "addis $rD, $rA, $imm", IntSimple, - [(set GPRC:$rD, (add GPRC:$rA, imm16ShiftedSExt:$imm))]>; -def LA : DForm_2<14, (outs GPRC:$rD), (ins GPRC:$rA, symbolLo:$sym), + [(set i32:$rD, (add i32:$rA, imm16ShiftedSExt:$imm))]>; +let isCodeGenOnly = 1 in +def LA : DForm_2<14, (outs gprc:$rD), (ins gprc_nor0:$rA, symbolLo:$sym), "la $rD, $sym($rA)", IntGeneral, - [(set GPRC:$rD, (add GPRC:$rA, + [(set i32:$rD, (add i32:$rA, (PPClo tglobaladdr:$sym, 0)))]>; -def MULLI : DForm_2< 7, (outs GPRC:$rD), (ins GPRC:$rA, s16imm:$imm), +def MULLI : DForm_2< 7, (outs gprc:$rD), (ins gprc:$rA, s16imm:$imm), "mulli $rD, $rA, $imm", IntMulLI, - [(set GPRC:$rD, (mul GPRC:$rA, immSExt16:$imm))]>; -let Defs = [CARRY] in { -def SUBFIC : DForm_2< 8, (outs GPRC:$rD), (ins GPRC:$rA, s16imm:$imm), + [(set i32:$rD, (mul i32:$rA, immSExt16:$imm))]>; +let Defs = [CARRY] in +def SUBFIC : DForm_2< 8, (outs gprc:$rD), (ins gprc:$rA, s16imm:$imm), "subfic $rD, $rA, $imm", IntGeneral, - [(set GPRC:$rD, (subc immSExt16:$imm, GPRC:$rA))]>; -} + [(set i32:$rD, (subc immSExt16:$imm, i32:$rA))]>; let isReMaterializable = 1, isAsCheapAsAMove = 1, isMoveImm = 1 in { - def LI : DForm_2_r0<14, (outs GPRC:$rD), (ins symbolLo:$imm), + def LI : DForm_2_r0<14, (outs gprc:$rD), (ins symbolLo:$imm), "li $rD, $imm", IntSimple, - [(set GPRC:$rD, immSExt16:$imm)]>; - def LIS : DForm_2_r0<15, (outs GPRC:$rD), (ins symbolHi:$imm), + [(set i32:$rD, immSExt16:$imm)]>; + def LIS : DForm_2_r0<15, (outs gprc:$rD), (ins symbolHi:$imm), "lis $rD, $imm", IntSimple, - [(set GPRC:$rD, imm16ShiftedSExt:$imm)]>; + [(set i32:$rD, imm16ShiftedSExt:$imm)]>; } } let PPC970_Unit = 1 in { // FXU Operations. -def ANDIo : DForm_4<28, (outs GPRC:$dst), (ins GPRC:$src1, u16imm:$src2), +let Defs = [CR0] in { +def ANDIo : DForm_4<28, (outs gprc:$dst), (ins gprc:$src1, u16imm:$src2), "andi. $dst, $src1, $src2", IntGeneral, - [(set GPRC:$dst, (and GPRC:$src1, immZExt16:$src2))]>, + [(set i32:$dst, (and i32:$src1, immZExt16:$src2))]>, isDOT; -def ANDISo : DForm_4<29, (outs GPRC:$dst), (ins GPRC:$src1, u16imm:$src2), +def ANDISo : DForm_4<29, (outs gprc:$dst), (ins gprc:$src1, u16imm:$src2), "andis. $dst, $src1, $src2", IntGeneral, - [(set GPRC:$dst, (and GPRC:$src1,imm16ShiftedZExt:$src2))]>, + [(set i32:$dst, (and i32:$src1, imm16ShiftedZExt:$src2))]>, isDOT; -def ORI : DForm_4<24, (outs GPRC:$dst), (ins GPRC:$src1, u16imm:$src2), +} +def ORI : DForm_4<24, (outs gprc:$dst), (ins gprc:$src1, u16imm:$src2), "ori $dst, $src1, $src2", IntSimple, - [(set GPRC:$dst, (or GPRC:$src1, immZExt16:$src2))]>; -def ORIS : DForm_4<25, (outs GPRC:$dst), (ins GPRC:$src1, u16imm:$src2), + [(set i32:$dst, (or i32:$src1, immZExt16:$src2))]>; +def ORIS : DForm_4<25, (outs gprc:$dst), (ins gprc:$src1, u16imm:$src2), "oris $dst, $src1, $src2", IntSimple, - [(set GPRC:$dst, (or GPRC:$src1, imm16ShiftedZExt:$src2))]>; -def XORI : DForm_4<26, (outs GPRC:$dst), (ins GPRC:$src1, u16imm:$src2), + [(set i32:$dst, (or i32:$src1, imm16ShiftedZExt:$src2))]>; +def XORI : DForm_4<26, (outs gprc:$dst), (ins gprc:$src1, u16imm:$src2), "xori $dst, $src1, $src2", IntSimple, - [(set GPRC:$dst, (xor GPRC:$src1, immZExt16:$src2))]>; -def XORIS : DForm_4<27, (outs GPRC:$dst), (ins GPRC:$src1, u16imm:$src2), + [(set i32:$dst, (xor i32:$src1, immZExt16:$src2))]>; +def XORIS : DForm_4<27, (outs gprc:$dst), (ins gprc:$src1, u16imm:$src2), "xoris $dst, $src1, $src2", IntSimple, - [(set GPRC:$dst, (xor GPRC:$src1,imm16ShiftedZExt:$src2))]>; + [(set i32:$dst, (xor i32:$src1, imm16ShiftedZExt:$src2))]>; def NOP : DForm_4_zero<24, (outs), (ins), "nop", IntSimple, []>; -def CMPWI : DForm_5_ext<11, (outs CRRC:$crD), (ins GPRC:$rA, s16imm:$imm), - "cmpwi $crD, $rA, $imm", IntCompare>; -def CMPLWI : DForm_6_ext<10, (outs CRRC:$dst), (ins GPRC:$src1, u16imm:$src2), - "cmplwi $dst, $src1, $src2", IntCompare>; +let isCompare = 1, neverHasSideEffects = 1 in { + def CMPWI : DForm_5_ext<11, (outs crrc:$crD), (ins gprc:$rA, s16imm:$imm), + "cmpwi $crD, $rA, $imm", IntCompare>; + def CMPLWI : DForm_6_ext<10, (outs crrc:$dst), (ins gprc:$src1, u16imm:$src2), + "cmplwi $dst, $src1, $src2", IntCompare>; +} } +let PPC970_Unit = 1, neverHasSideEffects = 1 in { // FXU Operations. +defm NAND : XForm_6r<31, 476, (outs gprc:$rA), (ins gprc:$rS, gprc:$rB), + "nand", "$rA, $rS, $rB", IntSimple, + [(set i32:$rA, (not (and i32:$rS, i32:$rB)))]>; +defm AND : XForm_6r<31, 28, (outs gprc:$rA), (ins gprc:$rS, gprc:$rB), + "and", "$rA, $rS, $rB", IntSimple, + [(set i32:$rA, (and i32:$rS, i32:$rB))]>; +defm ANDC : XForm_6r<31, 60, (outs gprc:$rA), (ins gprc:$rS, gprc:$rB), + "andc", "$rA, $rS, $rB", IntSimple, + [(set i32:$rA, (and i32:$rS, (not i32:$rB)))]>; +defm OR : XForm_6r<31, 444, (outs gprc:$rA), (ins gprc:$rS, gprc:$rB), + "or", "$rA, $rS, $rB", IntSimple, + [(set i32:$rA, (or i32:$rS, i32:$rB))]>; +defm NOR : XForm_6r<31, 124, (outs gprc:$rA), (ins gprc:$rS, gprc:$rB), + "nor", "$rA, $rS, $rB", IntSimple, + [(set i32:$rA, (not (or i32:$rS, i32:$rB)))]>; +defm ORC : XForm_6r<31, 412, (outs gprc:$rA), (ins gprc:$rS, gprc:$rB), + "orc", "$rA, $rS, $rB", IntSimple, + [(set i32:$rA, (or i32:$rS, (not i32:$rB)))]>; +defm EQV : XForm_6r<31, 284, (outs gprc:$rA), (ins gprc:$rS, gprc:$rB), + "eqv", "$rA, $rS, $rB", IntSimple, + [(set i32:$rA, (not (xor i32:$rS, i32:$rB)))]>; +defm XOR : XForm_6r<31, 316, (outs gprc:$rA), (ins gprc:$rS, gprc:$rB), + "xor", "$rA, $rS, $rB", IntSimple, + [(set i32:$rA, (xor i32:$rS, i32:$rB))]>; +defm SLW : XForm_6r<31, 24, (outs gprc:$rA), (ins gprc:$rS, gprc:$rB), + "slw", "$rA, $rS, $rB", IntGeneral, + [(set i32:$rA, (PPCshl i32:$rS, i32:$rB))]>; +defm SRW : XForm_6r<31, 536, (outs gprc:$rA), (ins gprc:$rS, gprc:$rB), + "srw", "$rA, $rS, $rB", IntGeneral, + [(set i32:$rA, (PPCsrl i32:$rS, i32:$rB))]>; +defm SRAW : XForm_6rc<31, 792, (outs gprc:$rA), (ins gprc:$rS, gprc:$rB), + "sraw", "$rA, $rS, $rB", IntShift, + [(set i32:$rA, (PPCsra i32:$rS, i32:$rB))]>; +} let PPC970_Unit = 1 in { // FXU Operations. -def NAND : XForm_6<31, 476, (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB), - "nand $rA, $rS, $rB", IntSimple, - [(set GPRC:$rA, (not (and GPRC:$rS, GPRC:$rB)))]>; -def AND : XForm_6<31, 28, (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB), - "and $rA, $rS, $rB", IntSimple, - [(set GPRC:$rA, (and GPRC:$rS, GPRC:$rB))]>; -def ANDC : XForm_6<31, 60, (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB), - "andc $rA, $rS, $rB", IntSimple, - [(set GPRC:$rA, (and GPRC:$rS, (not GPRC:$rB)))]>; -def OR : XForm_6<31, 444, (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB), - "or $rA, $rS, $rB", IntSimple, - [(set GPRC:$rA, (or GPRC:$rS, GPRC:$rB))]>; -def NOR : XForm_6<31, 124, (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB), - "nor $rA, $rS, $rB", IntSimple, - [(set GPRC:$rA, (not (or GPRC:$rS, GPRC:$rB)))]>; -def ORC : XForm_6<31, 412, (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB), - "orc $rA, $rS, $rB", IntSimple, - [(set GPRC:$rA, (or GPRC:$rS, (not GPRC:$rB)))]>; -def EQV : XForm_6<31, 284, (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB), - "eqv $rA, $rS, $rB", IntSimple, - [(set GPRC:$rA, (not (xor GPRC:$rS, GPRC:$rB)))]>; -def XOR : XForm_6<31, 316, (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB), - "xor $rA, $rS, $rB", IntSimple, - [(set GPRC:$rA, (xor GPRC:$rS, GPRC:$rB))]>; -def SLW : XForm_6<31, 24, (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB), - "slw $rA, $rS, $rB", IntGeneral, - [(set GPRC:$rA, (PPCshl GPRC:$rS, GPRC:$rB))]>; -def SRW : XForm_6<31, 536, (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB), - "srw $rA, $rS, $rB", IntGeneral, - [(set GPRC:$rA, (PPCsrl GPRC:$rS, GPRC:$rB))]>; -let Defs = [CARRY] in { -def SRAW : XForm_6<31, 792, (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB), - "sraw $rA, $rS, $rB", IntShift, - [(set GPRC:$rA, (PPCsra GPRC:$rS, GPRC:$rB))]>; +let neverHasSideEffects = 1 in { +defm SRAWI : XForm_10rc<31, 824, (outs gprc:$rA), (ins gprc:$rS, u5imm:$SH), + "srawi", "$rA, $rS, $SH", IntShift, + [(set i32:$rA, (sra i32:$rS, (i32 imm:$SH)))]>; +defm CNTLZW : XForm_11r<31, 26, (outs gprc:$rA), (ins gprc:$rS), + "cntlzw", "$rA, $rS", IntGeneral, + [(set i32:$rA, (ctlz i32:$rS))]>; +defm EXTSB : XForm_11r<31, 954, (outs gprc:$rA), (ins gprc:$rS), + "extsb", "$rA, $rS", IntSimple, + [(set i32:$rA, (sext_inreg i32:$rS, i8))]>; +defm EXTSH : XForm_11r<31, 922, (outs gprc:$rA), (ins gprc:$rS), + "extsh", "$rA, $rS", IntSimple, + [(set i32:$rA, (sext_inreg i32:$rS, i16))]>; } +let isCompare = 1, neverHasSideEffects = 1 in { + def CMPW : XForm_16_ext<31, 0, (outs crrc:$crD), (ins gprc:$rA, gprc:$rB), + "cmpw $crD, $rA, $rB", IntCompare>; + def CMPLW : XForm_16_ext<31, 32, (outs crrc:$crD), (ins gprc:$rA, gprc:$rB), + "cmplw $crD, $rA, $rB", IntCompare>; } - -let PPC970_Unit = 1 in { // FXU Operations. -let Defs = [CARRY] in { -def SRAWI : XForm_10<31, 824, (outs GPRC:$rA), (ins GPRC:$rS, u5imm:$SH), - "srawi $rA, $rS, $SH", IntShift, - [(set GPRC:$rA, (sra GPRC:$rS, (i32 imm:$SH)))]>; -} -def CNTLZW : XForm_11<31, 26, (outs GPRC:$rA), (ins GPRC:$rS), - "cntlzw $rA, $rS", IntGeneral, - [(set GPRC:$rA, (ctlz GPRC:$rS))]>; -def EXTSB : XForm_11<31, 954, (outs GPRC:$rA), (ins GPRC:$rS), - "extsb $rA, $rS", IntSimple, - [(set GPRC:$rA, (sext_inreg GPRC:$rS, i8))]>; -def EXTSH : XForm_11<31, 922, (outs GPRC:$rA), (ins GPRC:$rS), - "extsh $rA, $rS", IntSimple, - [(set GPRC:$rA, (sext_inreg GPRC:$rS, i16))]>; - -def CMPW : XForm_16_ext<31, 0, (outs CRRC:$crD), (ins GPRC:$rA, GPRC:$rB), - "cmpw $crD, $rA, $rB", IntCompare>; -def CMPLW : XForm_16_ext<31, 32, (outs CRRC:$crD), (ins GPRC:$rA, GPRC:$rB), - "cmplw $crD, $rA, $rB", IntCompare>; } let PPC970_Unit = 3 in { // FPU Operations. //def FCMPO : XForm_17<63, 32, (outs CRRC:$crD), (ins FPRC:$fA, FPRC:$fB), // "fcmpo $crD, $fA, $fB", FPCompare>; -def FCMPUS : XForm_17<63, 0, (outs CRRC:$crD), (ins F4RC:$fA, F4RC:$fB), - "fcmpu $crD, $fA, $fB", FPCompare>; -def FCMPUD : XForm_17<63, 0, (outs CRRC:$crD), (ins F8RC:$fA, F8RC:$fB), - "fcmpu $crD, $fA, $fB", FPCompare>; +let isCompare = 1, neverHasSideEffects = 1 in { + def FCMPUS : XForm_17<63, 0, (outs crrc:$crD), (ins f4rc:$fA, f4rc:$fB), + "fcmpu $crD, $fA, $fB", FPCompare>; + def FCMPUD : XForm_17<63, 0, (outs crrc:$crD), (ins f8rc:$fA, f8rc:$fB), + "fcmpu $crD, $fA, $fB", FPCompare>; +} let Uses = [RM] in { - def FCTIWZ : XForm_26<63, 15, (outs F8RC:$frD), (ins F8RC:$frB), - "fctiwz $frD, $frB", FPGeneral, - [(set F8RC:$frD, (PPCfctiwz F8RC:$frB))]>; - def FRSP : XForm_26<63, 12, (outs F4RC:$frD), (ins F8RC:$frB), - "frsp $frD, $frB", FPGeneral, - [(set F4RC:$frD, (fround F8RC:$frB))]>; - def FSQRT : XForm_26<63, 22, (outs F8RC:$frD), (ins F8RC:$frB), - "fsqrt $frD, $frB", FPSqrt, - [(set F8RC:$frD, (fsqrt F8RC:$frB))]>; - def FSQRTS : XForm_26<59, 22, (outs F4RC:$frD), (ins F4RC:$frB), - "fsqrts $frD, $frB", FPSqrt, - [(set F4RC:$frD, (fsqrt F4RC:$frB))]>; + let neverHasSideEffects = 1 in { + defm FCTIWZ : XForm_26r<63, 15, (outs f8rc:$frD), (ins f8rc:$frB), + "fctiwz", "$frD, $frB", FPGeneral, + [(set f64:$frD, (PPCfctiwz f64:$frB))]>; + + defm FRSP : XForm_26r<63, 12, (outs f4rc:$frD), (ins f8rc:$frB), + "frsp", "$frD, $frB", FPGeneral, + [(set f32:$frD, (fround f64:$frB))]>; + + // The frin -> nearbyint mapping is valid only in fast-math mode. + let Interpretation64Bit = 1 in + defm FRIND : XForm_26r<63, 392, (outs f8rc:$frD), (ins f8rc:$frB), + "frin", "$frD, $frB", FPGeneral, + [(set f64:$frD, (fnearbyint f64:$frB))]>; + defm FRINS : XForm_26r<63, 392, (outs f4rc:$frD), (ins f4rc:$frB), + "frin", "$frD, $frB", FPGeneral, + [(set f32:$frD, (fnearbyint f32:$frB))]>; + } + + // These pseudos expand to rint but also set FE_INEXACT when the result does + // not equal the argument. + let usesCustomInserter = 1, Defs = [RM] in { // FIXME: Model FPSCR! + def FRINDrint : Pseudo<(outs f8rc:$frD), (ins f8rc:$frB), + "#FRINDrint", [(set f64:$frD, (frint f64:$frB))]>; + def FRINSrint : Pseudo<(outs f4rc:$frD), (ins f4rc:$frB), + "#FRINSrint", [(set f32:$frD, (frint f32:$frB))]>; + } + + let neverHasSideEffects = 1 in { + let Interpretation64Bit = 1 in + defm FRIPD : XForm_26r<63, 456, (outs f8rc:$frD), (ins f8rc:$frB), + "frip", "$frD, $frB", FPGeneral, + [(set f64:$frD, (fceil f64:$frB))]>; + defm FRIPS : XForm_26r<63, 456, (outs f4rc:$frD), (ins f4rc:$frB), + "frip", "$frD, $frB", FPGeneral, + [(set f32:$frD, (fceil f32:$frB))]>; + let Interpretation64Bit = 1 in + defm FRIZD : XForm_26r<63, 424, (outs f8rc:$frD), (ins f8rc:$frB), + "friz", "$frD, $frB", FPGeneral, + [(set f64:$frD, (ftrunc f64:$frB))]>; + defm FRIZS : XForm_26r<63, 424, (outs f4rc:$frD), (ins f4rc:$frB), + "friz", "$frD, $frB", FPGeneral, + [(set f32:$frD, (ftrunc f32:$frB))]>; + let Interpretation64Bit = 1 in + defm FRIMD : XForm_26r<63, 488, (outs f8rc:$frD), (ins f8rc:$frB), + "frim", "$frD, $frB", FPGeneral, + [(set f64:$frD, (ffloor f64:$frB))]>; + defm FRIMS : XForm_26r<63, 488, (outs f4rc:$frD), (ins f4rc:$frB), + "frim", "$frD, $frB", FPGeneral, + [(set f32:$frD, (ffloor f32:$frB))]>; + + defm FSQRT : XForm_26r<63, 22, (outs f8rc:$frD), (ins f8rc:$frB), + "fsqrt", "$frD, $frB", FPSqrt, + [(set f64:$frD, (fsqrt f64:$frB))]>; + defm FSQRTS : XForm_26r<59, 22, (outs f4rc:$frD), (ins f4rc:$frB), + "fsqrts", "$frD, $frB", FPSqrt, + [(set f32:$frD, (fsqrt f32:$frB))]>; + } } } @@ -1149,55 +1497,74 @@ let Uses = [RM] in { /// often coalesced away and we don't want the dispatch group builder to think /// that they will fill slots (which could cause the load of a LSU reject to /// sneak into a d-group with a store). -def FMR : XForm_26<63, 72, (outs F4RC:$frD), (ins F4RC:$frB), - "fmr $frD, $frB", FPGeneral, - []>, // (set F4RC:$frD, F4RC:$frB) - PPC970_Unit_Pseudo; +let neverHasSideEffects = 1 in +defm FMR : XForm_26r<63, 72, (outs f4rc:$frD), (ins f4rc:$frB), + "fmr", "$frD, $frB", FPGeneral, + []>, // (set f32:$frD, f32:$frB) + PPC970_Unit_Pseudo; -let PPC970_Unit = 3 in { // FPU Operations. +let PPC970_Unit = 3, neverHasSideEffects = 1 in { // FPU Operations. // These are artificially split into two different forms, for 4/8 byte FP. -def FABSS : XForm_26<63, 264, (outs F4RC:$frD), (ins F4RC:$frB), - "fabs $frD, $frB", FPGeneral, - [(set F4RC:$frD, (fabs F4RC:$frB))]>; -def FABSD : XForm_26<63, 264, (outs F8RC:$frD), (ins F8RC:$frB), - "fabs $frD, $frB", FPGeneral, - [(set F8RC:$frD, (fabs F8RC:$frB))]>; -def FNABSS : XForm_26<63, 136, (outs F4RC:$frD), (ins F4RC:$frB), - "fnabs $frD, $frB", FPGeneral, - [(set F4RC:$frD, (fneg (fabs F4RC:$frB)))]>; -def FNABSD : XForm_26<63, 136, (outs F8RC:$frD), (ins F8RC:$frB), - "fnabs $frD, $frB", FPGeneral, - [(set F8RC:$frD, (fneg (fabs F8RC:$frB)))]>; -def FNEGS : XForm_26<63, 40, (outs F4RC:$frD), (ins F4RC:$frB), - "fneg $frD, $frB", FPGeneral, - [(set F4RC:$frD, (fneg F4RC:$frB))]>; -def FNEGD : XForm_26<63, 40, (outs F8RC:$frD), (ins F8RC:$frB), - "fneg $frD, $frB", FPGeneral, - [(set F8RC:$frD, (fneg F8RC:$frB))]>; -} - +defm FABSS : XForm_26r<63, 264, (outs f4rc:$frD), (ins f4rc:$frB), + "fabs", "$frD, $frB", FPGeneral, + [(set f32:$frD, (fabs f32:$frB))]>; +let Interpretation64Bit = 1 in +defm FABSD : XForm_26r<63, 264, (outs f8rc:$frD), (ins f8rc:$frB), + "fabs", "$frD, $frB", FPGeneral, + [(set f64:$frD, (fabs f64:$frB))]>; +defm FNABSS : XForm_26r<63, 136, (outs f4rc:$frD), (ins f4rc:$frB), + "fnabs", "$frD, $frB", FPGeneral, + [(set f32:$frD, (fneg (fabs f32:$frB)))]>; +let Interpretation64Bit = 1 in +defm FNABSD : XForm_26r<63, 136, (outs f8rc:$frD), (ins f8rc:$frB), + "fnabs", "$frD, $frB", FPGeneral, + [(set f64:$frD, (fneg (fabs f64:$frB)))]>; +defm FNEGS : XForm_26r<63, 40, (outs f4rc:$frD), (ins f4rc:$frB), + "fneg", "$frD, $frB", FPGeneral, + [(set f32:$frD, (fneg f32:$frB))]>; +let Interpretation64Bit = 1 in +defm FNEGD : XForm_26r<63, 40, (outs f8rc:$frD), (ins f8rc:$frB), + "fneg", "$frD, $frB", FPGeneral, + [(set f64:$frD, (fneg f64:$frB))]>; + +// Reciprocal estimates. +defm FRE : XForm_26r<63, 24, (outs f8rc:$frD), (ins f8rc:$frB), + "fre", "$frD, $frB", FPGeneral, + [(set f64:$frD, (PPCfre f64:$frB))]>; +defm FRES : XForm_26r<59, 24, (outs f4rc:$frD), (ins f4rc:$frB), + "fres", "$frD, $frB", FPGeneral, + [(set f32:$frD, (PPCfre f32:$frB))]>; +defm FRSQRTE : XForm_26r<63, 26, (outs f8rc:$frD), (ins f8rc:$frB), + "frsqrte", "$frD, $frB", FPGeneral, + [(set f64:$frD, (PPCfrsqrte f64:$frB))]>; +defm FRSQRTES : XForm_26r<59, 26, (outs f4rc:$frD), (ins f4rc:$frB), + "frsqrtes", "$frD, $frB", FPGeneral, + [(set f32:$frD, (PPCfrsqrte f32:$frB))]>; +} // XL-Form instructions. condition register logical ops. // -def MCRF : XLForm_3<19, 0, (outs CRRC:$BF), (ins CRRC:$BFA), +let neverHasSideEffects = 1 in +def MCRF : XLForm_3<19, 0, (outs crrc:$BF), (ins crrc:$BFA), "mcrf $BF, $BFA", BrMCR>, PPC970_DGroup_First, PPC970_Unit_CRU; -def CREQV : XLForm_1<19, 289, (outs CRBITRC:$CRD), - (ins CRBITRC:$CRA, CRBITRC:$CRB), +def CREQV : XLForm_1<19, 289, (outs crbitrc:$CRD), + (ins crbitrc:$CRA, crbitrc:$CRB), "creqv $CRD, $CRA, $CRB", BrCR, []>; -def CROR : XLForm_1<19, 449, (outs CRBITRC:$CRD), - (ins CRBITRC:$CRA, CRBITRC:$CRB), +def CROR : XLForm_1<19, 449, (outs crbitrc:$CRD), + (ins crbitrc:$CRA, crbitrc:$CRB), "cror $CRD, $CRA, $CRB", BrCR, []>; -def CRSET : XLForm_1_ext<19, 289, (outs CRBITRC:$dst), (ins), +let isCodeGenOnly = 1 in { +def CRSET : XLForm_1_ext<19, 289, (outs crbitrc:$dst), (ins), "creqv $dst, $dst, $dst", BrCR, []>; -def CRUNSET: XLForm_1_ext<19, 193, (outs CRBITRC:$dst), (ins), +def CRUNSET: XLForm_1_ext<19, 193, (outs crbitrc:$dst), (ins), "crxor $dst, $dst, $dst", BrCR, []>; @@ -1210,27 +1577,28 @@ def CR6UNSET: XLForm_1_ext<19, 193, (outs), (ins), "crxor 6, 6, 6", BrCR, [(PPCcr6unset)]>; } +} // XFX-Form instructions. Instructions that deal with SPRs. // let Uses = [CTR] in { -def MFCTR : XFXForm_1_ext<31, 339, 9, (outs GPRC:$rT), (ins), +def MFCTR : XFXForm_1_ext<31, 339, 9, (outs gprc:$rT), (ins), "mfctr $rT", SprMFSPR>, PPC970_DGroup_First, PPC970_Unit_FXU; } -let Defs = [CTR], Pattern = [(PPCmtctr GPRC:$rS)] in { -def MTCTR : XFXForm_7_ext<31, 467, 9, (outs), (ins GPRC:$rS), +let Defs = [CTR], Pattern = [(PPCmtctr i32:$rS)] in { +def MTCTR : XFXForm_7_ext<31, 467, 9, (outs), (ins gprc:$rS), "mtctr $rS", SprMTSPR>, PPC970_DGroup_First, PPC970_Unit_FXU; } let Defs = [LR] in { -def MTLR : XFXForm_7_ext<31, 467, 8, (outs), (ins GPRC:$rS), +def MTLR : XFXForm_7_ext<31, 467, 8, (outs), (ins gprc:$rS), "mtlr $rS", SprMTSPR>, PPC970_DGroup_First, PPC970_Unit_FXU; } let Uses = [LR] in { -def MFLR : XFXForm_1_ext<31, 339, 8, (outs GPRC:$rT), (ins), +def MFLR : XFXForm_1_ext<31, 339, 8, (outs gprc:$rT), (ins), "mflr $rT", SprMFSPR>, PPC970_DGroup_First, PPC970_Unit_FXU; } @@ -1238,14 +1606,38 @@ def MFLR : XFXForm_1_ext<31, 339, 8, (outs GPRC:$rT), (ins), // Move to/from VRSAVE: despite being a SPR, the VRSAVE register is renamed like // a GPR on the PPC970. As such, copies in and out have the same performance // characteristics as an OR instruction. -def MTVRSAVE : XFXForm_7_ext<31, 467, 256, (outs), (ins GPRC:$rS), +def MTVRSAVE : XFXForm_7_ext<31, 467, 256, (outs), (ins gprc:$rS), "mtspr 256, $rS", IntGeneral>, PPC970_DGroup_Single, PPC970_Unit_FXU; -def MFVRSAVE : XFXForm_1_ext<31, 339, 256, (outs GPRC:$rT), (ins), +def MFVRSAVE : XFXForm_1_ext<31, 339, 256, (outs gprc:$rT), (ins), "mfspr $rT, 256", IntGeneral>, PPC970_DGroup_First, PPC970_Unit_FXU; -def MTCRF : XFXForm_5<31, 144, (outs crbitm:$FXM), (ins GPRC:$rS), +let isCodeGenOnly = 1 in { + def MTVRSAVEv : XFXForm_7_ext<31, 467, 256, + (outs VRSAVERC:$reg), (ins gprc:$rS), + "mtspr 256, $rS", IntGeneral>, + PPC970_DGroup_Single, PPC970_Unit_FXU; + def MFVRSAVEv : XFXForm_1_ext<31, 339, 256, (outs gprc:$rT), + (ins VRSAVERC:$reg), + "mfspr $rT, 256", IntGeneral>, + PPC970_DGroup_First, PPC970_Unit_FXU; +} + +// SPILL_VRSAVE - Indicate that we're dumping the VRSAVE register, +// so we'll need to scavenge a register for it. +let mayStore = 1 in +def SPILL_VRSAVE : Pseudo<(outs), (ins VRSAVERC:$vrsave, memri:$F), + "#SPILL_VRSAVE", []>; + +// RESTORE_VRSAVE - Indicate that we're restoring the VRSAVE register (previously +// spilled), so we'll need to scavenge a register for it. +let mayLoad = 1 in +def RESTORE_VRSAVE : Pseudo<(outs VRSAVERC:$vrsave), (ins memri:$F), + "#RESTORE_VRSAVE", []>; + +let neverHasSideEffects = 1 in { +def MTCRF : XFXForm_5<31, 144, (outs crbitm:$FXM), (ins gprc:$rS), "mtcrf $FXM, $rS", BrMCRX>, PPC970_MicroCode, PPC970_Unit_CRU; @@ -1259,215 +1651,205 @@ def MTCRF : XFXForm_5<31, 144, (outs crbitm:$FXM), (ins GPRC:$rS), // instruction to keep the register allocator from becoming confused. // // FIXME: Make this a real Pseudo instruction when the JIT switches to MC. -def MFCRpseud: XFXForm_3<31, 19, (outs GPRC:$rT), (ins crbitm:$FXM), +let isCodeGenOnly = 1 in +def MFCRpseud: XFXForm_3<31, 19, (outs gprc:$rT), (ins crbitm:$FXM), "#MFCRpseud", SprMFCR>, PPC970_MicroCode, PPC970_Unit_CRU; - -def MFCR : XFXForm_3<31, 19, (outs GPRC:$rT), (ins), - "mfcr $rT", SprMFCR>, - PPC970_MicroCode, PPC970_Unit_CRU; -def MFOCRF: XFXForm_5a<31, 19, (outs GPRC:$rT), (ins crbitm:$FXM), +def MFOCRF: XFXForm_5a<31, 19, (outs gprc:$rT), (ins crbitm:$FXM), "mfocrf $rT, $FXM", SprMFCR>, PPC970_DGroup_First, PPC970_Unit_CRU; +} // neverHasSideEffects = 1 -// Instructions to manipulate FPSCR. Only long double handling uses these. -// FPSCR is not modelled; we use the SDNode Flag to keep things in order. +let neverHasSideEffects = 1 in +def MFCR : XFXForm_3<31, 19, (outs gprc:$rT), (ins), + "mfcr $rT", SprMFCR>, + PPC970_MicroCode, PPC970_Unit_CRU; +// Pseudo instruction to perform FADD in round-to-zero mode. +let usesCustomInserter = 1, Uses = [RM] in { + def FADDrtz: Pseudo<(outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRB), "", + [(set f64:$FRT, (PPCfaddrtz f64:$FRA, f64:$FRB))]>; +} + +// The above pseudo gets expanded to make use of the following instructions +// to manipulate FPSCR. Note that FPSCR is not modeled at the DAG level. let Uses = [RM], Defs = [RM] in { def MTFSB0 : XForm_43<63, 70, (outs), (ins u5imm:$FM), - "mtfsb0 $FM", IntMTFSB0, - [(PPCmtfsb0 (i32 imm:$FM))]>, + "mtfsb0 $FM", IntMTFSB0, []>, PPC970_DGroup_Single, PPC970_Unit_FPU; def MTFSB1 : XForm_43<63, 38, (outs), (ins u5imm:$FM), - "mtfsb1 $FM", IntMTFSB0, - [(PPCmtfsb1 (i32 imm:$FM))]>, + "mtfsb1 $FM", IntMTFSB0, []>, PPC970_DGroup_Single, PPC970_Unit_FPU; - // MTFSF does not actually produce an FP result. We pretend it copies - // input reg B to the output. If we didn't do this it would look like the - // instruction had no outputs (because we aren't modelling the FPSCR) and - // it would be deleted. - def MTFSF : XFLForm<63, 711, (outs F8RC:$FRA), - (ins i32imm:$FM, F8RC:$rT, F8RC:$FRB), - "mtfsf $FM, $rT", "$FRB = $FRA", IntMTFSB0, - [(set F8RC:$FRA, (PPCmtfsf (i32 imm:$FM), - F8RC:$rT, F8RC:$FRB))]>, + def MTFSF : XFLForm<63, 711, (outs), (ins i32imm:$FM, f8rc:$rT), + "mtfsf $FM, $rT", IntMTFSB0, []>, PPC970_DGroup_Single, PPC970_Unit_FPU; } let Uses = [RM] in { - def MFFS : XForm_42<63, 583, (outs F8RC:$rT), (ins), + def MFFS : XForm_42<63, 583, (outs f8rc:$rT), (ins), "mffs $rT", IntMFFS, - [(set F8RC:$rT, (PPCmffs))]>, - PPC970_DGroup_Single, PPC970_Unit_FPU; - def FADDrtz: AForm_2<63, 21, - (outs F8RC:$FRT), (ins F8RC:$FRA, F8RC:$FRB), - "fadd $FRT, $FRA, $FRB", FPAddSub, - [(set F8RC:$FRT, (PPCfaddrtz F8RC:$FRA, F8RC:$FRB))]>, + [(set f64:$rT, (PPCmffs))]>, PPC970_DGroup_Single, PPC970_Unit_FPU; } -let PPC970_Unit = 1 in { // FXU Operations. - +let PPC970_Unit = 1, neverHasSideEffects = 1 in { // FXU Operations. // XO-Form instructions. Arithmetic instructions that can set overflow bit // -def ADD4 : XOForm_1<31, 266, 0, (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB), - "add $rT, $rA, $rB", IntSimple, - [(set GPRC:$rT, (add GPRC:$rA, GPRC:$rB))]>; -let Defs = [CARRY] in { -def ADDC : XOForm_1<31, 10, 0, (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB), - "addc $rT, $rA, $rB", IntGeneral, - [(set GPRC:$rT, (addc GPRC:$rA, GPRC:$rB))]>, - PPC970_DGroup_Cracked; -} -def DIVW : XOForm_1<31, 491, 0, (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB), - "divw $rT, $rA, $rB", IntDivW, - [(set GPRC:$rT, (sdiv GPRC:$rA, GPRC:$rB))]>, - PPC970_DGroup_First, PPC970_DGroup_Cracked; -def DIVWU : XOForm_1<31, 459, 0, (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB), - "divwu $rT, $rA, $rB", IntDivW, - [(set GPRC:$rT, (udiv GPRC:$rA, GPRC:$rB))]>, - PPC970_DGroup_First, PPC970_DGroup_Cracked; -def MULHW : XOForm_1<31, 75, 0, (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB), - "mulhw $rT, $rA, $rB", IntMulHW, - [(set GPRC:$rT, (mulhs GPRC:$rA, GPRC:$rB))]>; -def MULHWU : XOForm_1<31, 11, 0, (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB), - "mulhwu $rT, $rA, $rB", IntMulHWU, - [(set GPRC:$rT, (mulhu GPRC:$rA, GPRC:$rB))]>; -def MULLW : XOForm_1<31, 235, 0, (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB), - "mullw $rT, $rA, $rB", IntMulHW, - [(set GPRC:$rT, (mul GPRC:$rA, GPRC:$rB))]>; -def SUBF : XOForm_1<31, 40, 0, (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB), - "subf $rT, $rA, $rB", IntGeneral, - [(set GPRC:$rT, (sub GPRC:$rB, GPRC:$rA))]>; -let Defs = [CARRY] in { -def SUBFC : XOForm_1<31, 8, 0, (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB), - "subfc $rT, $rA, $rB", IntGeneral, - [(set GPRC:$rT, (subc GPRC:$rB, GPRC:$rA))]>, - PPC970_DGroup_Cracked; -} -def NEG : XOForm_3<31, 104, 0, (outs GPRC:$rT), (ins GPRC:$rA), - "neg $rT, $rA", IntSimple, - [(set GPRC:$rT, (ineg GPRC:$rA))]>; -let Uses = [CARRY], Defs = [CARRY] in { -def ADDE : XOForm_1<31, 138, 0, (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB), - "adde $rT, $rA, $rB", IntGeneral, - [(set GPRC:$rT, (adde GPRC:$rA, GPRC:$rB))]>; -def ADDME : XOForm_3<31, 234, 0, (outs GPRC:$rT), (ins GPRC:$rA), - "addme $rT, $rA", IntGeneral, - [(set GPRC:$rT, (adde GPRC:$rA, -1))]>; -def ADDZE : XOForm_3<31, 202, 0, (outs GPRC:$rT), (ins GPRC:$rA), - "addze $rT, $rA", IntGeneral, - [(set GPRC:$rT, (adde GPRC:$rA, 0))]>; -def SUBFE : XOForm_1<31, 136, 0, (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB), - "subfe $rT, $rA, $rB", IntGeneral, - [(set GPRC:$rT, (sube GPRC:$rB, GPRC:$rA))]>; -def SUBFME : XOForm_3<31, 232, 0, (outs GPRC:$rT), (ins GPRC:$rA), - "subfme $rT, $rA", IntGeneral, - [(set GPRC:$rT, (sube -1, GPRC:$rA))]>; -def SUBFZE : XOForm_3<31, 200, 0, (outs GPRC:$rT), (ins GPRC:$rA), - "subfze $rT, $rA", IntGeneral, - [(set GPRC:$rT, (sube 0, GPRC:$rA))]>; +defm ADD4 : XOForm_1r<31, 266, 0, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB), + "add", "$rT, $rA, $rB", IntSimple, + [(set i32:$rT, (add i32:$rA, i32:$rB))]>; +defm ADDC : XOForm_1rc<31, 10, 0, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB), + "addc", "$rT, $rA, $rB", IntGeneral, + [(set i32:$rT, (addc i32:$rA, i32:$rB))]>, + PPC970_DGroup_Cracked; +defm DIVW : XOForm_1r<31, 491, 0, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB), + "divw", "$rT, $rA, $rB", IntDivW, + [(set i32:$rT, (sdiv i32:$rA, i32:$rB))]>, + PPC970_DGroup_First, PPC970_DGroup_Cracked; +defm DIVWU : XOForm_1r<31, 459, 0, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB), + "divwu", "$rT, $rA, $rB", IntDivW, + [(set i32:$rT, (udiv i32:$rA, i32:$rB))]>, + PPC970_DGroup_First, PPC970_DGroup_Cracked; +defm MULHW : XOForm_1r<31, 75, 0, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB), + "mulhw", "$rT, $rA, $rB", IntMulHW, + [(set i32:$rT, (mulhs i32:$rA, i32:$rB))]>; +defm MULHWU : XOForm_1r<31, 11, 0, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB), + "mulhwu", "$rT, $rA, $rB", IntMulHWU, + [(set i32:$rT, (mulhu i32:$rA, i32:$rB))]>; +defm MULLW : XOForm_1r<31, 235, 0, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB), + "mullw", "$rT, $rA, $rB", IntMulHW, + [(set i32:$rT, (mul i32:$rA, i32:$rB))]>; +defm SUBF : XOForm_1r<31, 40, 0, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB), + "subf", "$rT, $rA, $rB", IntGeneral, + [(set i32:$rT, (sub i32:$rB, i32:$rA))]>; +defm SUBFC : XOForm_1rc<31, 8, 0, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB), + "subfc", "$rT, $rA, $rB", IntGeneral, + [(set i32:$rT, (subc i32:$rB, i32:$rA))]>, + PPC970_DGroup_Cracked; +defm NEG : XOForm_3r<31, 104, 0, (outs gprc:$rT), (ins gprc:$rA), + "neg", "$rT, $rA", IntSimple, + [(set i32:$rT, (ineg i32:$rA))]>; +let Uses = [CARRY] in { +defm ADDE : XOForm_1rc<31, 138, 0, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB), + "adde", "$rT, $rA, $rB", IntGeneral, + [(set i32:$rT, (adde i32:$rA, i32:$rB))]>; +defm ADDME : XOForm_3rc<31, 234, 0, (outs gprc:$rT), (ins gprc:$rA), + "addme", "$rT, $rA", IntGeneral, + [(set i32:$rT, (adde i32:$rA, -1))]>; +defm ADDZE : XOForm_3rc<31, 202, 0, (outs gprc:$rT), (ins gprc:$rA), + "addze", "$rT, $rA", IntGeneral, + [(set i32:$rT, (adde i32:$rA, 0))]>; +defm SUBFE : XOForm_1rc<31, 136, 0, (outs gprc:$rT), (ins gprc:$rA, gprc:$rB), + "subfe", "$rT, $rA, $rB", IntGeneral, + [(set i32:$rT, (sube i32:$rB, i32:$rA))]>; +defm SUBFME : XOForm_3rc<31, 232, 0, (outs gprc:$rT), (ins gprc:$rA), + "subfme", "$rT, $rA", IntGeneral, + [(set i32:$rT, (sube -1, i32:$rA))]>; +defm SUBFZE : XOForm_3rc<31, 200, 0, (outs gprc:$rT), (ins gprc:$rA), + "subfze", "$rT, $rA", IntGeneral, + [(set i32:$rT, (sube 0, i32:$rA))]>; } } // A-Form instructions. Most of the instructions executed in the FPU are of // this type. // -let PPC970_Unit = 3 in { // FPU Operations. +let PPC970_Unit = 3, neverHasSideEffects = 1 in { // FPU Operations. let Uses = [RM] in { - def FMADD : AForm_1<63, 29, - (outs F8RC:$FRT), (ins F8RC:$FRA, F8RC:$FRC, F8RC:$FRB), - "fmadd $FRT, $FRA, $FRC, $FRB", FPFused, - [(set F8RC:$FRT, - (fma F8RC:$FRA, F8RC:$FRC, F8RC:$FRB))]>; - def FMADDS : AForm_1<59, 29, - (outs F4RC:$FRT), (ins F4RC:$FRA, F4RC:$FRC, F4RC:$FRB), - "fmadds $FRT, $FRA, $FRC, $FRB", FPGeneral, - [(set F4RC:$FRT, - (fma F4RC:$FRA, F4RC:$FRC, F4RC:$FRB))]>; - def FMSUB : AForm_1<63, 28, - (outs F8RC:$FRT), (ins F8RC:$FRA, F8RC:$FRC, F8RC:$FRB), - "fmsub $FRT, $FRA, $FRC, $FRB", FPFused, - [(set F8RC:$FRT, - (fma F8RC:$FRA, F8RC:$FRC, (fneg F8RC:$FRB)))]>; - def FMSUBS : AForm_1<59, 28, - (outs F4RC:$FRT), (ins F4RC:$FRA, F4RC:$FRC, F4RC:$FRB), - "fmsubs $FRT, $FRA, $FRC, $FRB", FPGeneral, - [(set F4RC:$FRT, - (fma F4RC:$FRA, F4RC:$FRC, (fneg F4RC:$FRB)))]>; - def FNMADD : AForm_1<63, 31, - (outs F8RC:$FRT), (ins F8RC:$FRA, F8RC:$FRC, F8RC:$FRB), - "fnmadd $FRT, $FRA, $FRC, $FRB", FPFused, - [(set F8RC:$FRT, - (fneg (fma F8RC:$FRA, F8RC:$FRC, F8RC:$FRB)))]>; - def FNMADDS : AForm_1<59, 31, - (outs F4RC:$FRT), (ins F4RC:$FRA, F4RC:$FRC, F4RC:$FRB), - "fnmadds $FRT, $FRA, $FRC, $FRB", FPGeneral, - [(set F4RC:$FRT, - (fneg (fma F4RC:$FRA, F4RC:$FRC, F4RC:$FRB)))]>; - def FNMSUB : AForm_1<63, 30, - (outs F8RC:$FRT), (ins F8RC:$FRA, F8RC:$FRC, F8RC:$FRB), - "fnmsub $FRT, $FRA, $FRC, $FRB", FPFused, - [(set F8RC:$FRT, (fneg (fma F8RC:$FRA, F8RC:$FRC, - (fneg F8RC:$FRB))))]>; - def FNMSUBS : AForm_1<59, 30, - (outs F4RC:$FRT), (ins F4RC:$FRA, F4RC:$FRC, F4RC:$FRB), - "fnmsubs $FRT, $FRA, $FRC, $FRB", FPGeneral, - [(set F4RC:$FRT, (fneg (fma F4RC:$FRA, F4RC:$FRC, - (fneg F4RC:$FRB))))]>; + defm FMADD : AForm_1r<63, 29, + (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRC, f8rc:$FRB), + "fmadd", "$FRT, $FRA, $FRC, $FRB", FPFused, + [(set f64:$FRT, (fma f64:$FRA, f64:$FRC, f64:$FRB))]>; + defm FMADDS : AForm_1r<59, 29, + (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRC, f4rc:$FRB), + "fmadds", "$FRT, $FRA, $FRC, $FRB", FPGeneral, + [(set f32:$FRT, (fma f32:$FRA, f32:$FRC, f32:$FRB))]>; + defm FMSUB : AForm_1r<63, 28, + (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRC, f8rc:$FRB), + "fmsub", "$FRT, $FRA, $FRC, $FRB", FPFused, + [(set f64:$FRT, + (fma f64:$FRA, f64:$FRC, (fneg f64:$FRB)))]>; + defm FMSUBS : AForm_1r<59, 28, + (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRC, f4rc:$FRB), + "fmsubs", "$FRT, $FRA, $FRC, $FRB", FPGeneral, + [(set f32:$FRT, + (fma f32:$FRA, f32:$FRC, (fneg f32:$FRB)))]>; + defm FNMADD : AForm_1r<63, 31, + (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRC, f8rc:$FRB), + "fnmadd", "$FRT, $FRA, $FRC, $FRB", FPFused, + [(set f64:$FRT, + (fneg (fma f64:$FRA, f64:$FRC, f64:$FRB)))]>; + defm FNMADDS : AForm_1r<59, 31, + (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRC, f4rc:$FRB), + "fnmadds", "$FRT, $FRA, $FRC, $FRB", FPGeneral, + [(set f32:$FRT, + (fneg (fma f32:$FRA, f32:$FRC, f32:$FRB)))]>; + defm FNMSUB : AForm_1r<63, 30, + (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRC, f8rc:$FRB), + "fnmsub", "$FRT, $FRA, $FRC, $FRB", FPFused, + [(set f64:$FRT, (fneg (fma f64:$FRA, f64:$FRC, + (fneg f64:$FRB))))]>; + defm FNMSUBS : AForm_1r<59, 30, + (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRC, f4rc:$FRB), + "fnmsubs", "$FRT, $FRA, $FRC, $FRB", FPGeneral, + [(set f32:$FRT, (fneg (fma f32:$FRA, f32:$FRC, + (fneg f32:$FRB))))]>; } // FSEL is artificially split into 4 and 8-byte forms for the result. To avoid // having 4 of these, force the comparison to always be an 8-byte double (code // should use an FMRSD if the input comparison value really wants to be a float) // and 4/8 byte forms for the result and operand type.. -def FSELD : AForm_1<63, 23, - (outs F8RC:$FRT), (ins F8RC:$FRA, F8RC:$FRC, F8RC:$FRB), - "fsel $FRT, $FRA, $FRC, $FRB", FPGeneral, - [(set F8RC:$FRT, (PPCfsel F8RC:$FRA,F8RC:$FRC,F8RC:$FRB))]>; -def FSELS : AForm_1<63, 23, - (outs F4RC:$FRT), (ins F8RC:$FRA, F4RC:$FRC, F4RC:$FRB), - "fsel $FRT, $FRA, $FRC, $FRB", FPGeneral, - [(set F4RC:$FRT, (PPCfsel F8RC:$FRA,F4RC:$FRC,F4RC:$FRB))]>; +let Interpretation64Bit = 1 in +defm FSELD : AForm_1r<63, 23, + (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRC, f8rc:$FRB), + "fsel", "$FRT, $FRA, $FRC, $FRB", FPGeneral, + [(set f64:$FRT, (PPCfsel f64:$FRA, f64:$FRC, f64:$FRB))]>; +defm FSELS : AForm_1r<63, 23, + (outs f4rc:$FRT), (ins f8rc:$FRA, f4rc:$FRC, f4rc:$FRB), + "fsel", "$FRT, $FRA, $FRC, $FRB", FPGeneral, + [(set f32:$FRT, (PPCfsel f64:$FRA, f32:$FRC, f32:$FRB))]>; let Uses = [RM] in { - def FADD : AForm_2<63, 21, - (outs F8RC:$FRT), (ins F8RC:$FRA, F8RC:$FRB), - "fadd $FRT, $FRA, $FRB", FPAddSub, - [(set F8RC:$FRT, (fadd F8RC:$FRA, F8RC:$FRB))]>; - def FADDS : AForm_2<59, 21, - (outs F4RC:$FRT), (ins F4RC:$FRA, F4RC:$FRB), - "fadds $FRT, $FRA, $FRB", FPGeneral, - [(set F4RC:$FRT, (fadd F4RC:$FRA, F4RC:$FRB))]>; - def FDIV : AForm_2<63, 18, - (outs F8RC:$FRT), (ins F8RC:$FRA, F8RC:$FRB), - "fdiv $FRT, $FRA, $FRB", FPDivD, - [(set F8RC:$FRT, (fdiv F8RC:$FRA, F8RC:$FRB))]>; - def FDIVS : AForm_2<59, 18, - (outs F4RC:$FRT), (ins F4RC:$FRA, F4RC:$FRB), - "fdivs $FRT, $FRA, $FRB", FPDivS, - [(set F4RC:$FRT, (fdiv F4RC:$FRA, F4RC:$FRB))]>; - def FMUL : AForm_3<63, 25, - (outs F8RC:$FRT), (ins F8RC:$FRA, F8RC:$FRC), - "fmul $FRT, $FRA, $FRC", FPFused, - [(set F8RC:$FRT, (fmul F8RC:$FRA, F8RC:$FRC))]>; - def FMULS : AForm_3<59, 25, - (outs F4RC:$FRT), (ins F4RC:$FRA, F4RC:$FRC), - "fmuls $FRT, $FRA, $FRC", FPGeneral, - [(set F4RC:$FRT, (fmul F4RC:$FRA, F4RC:$FRC))]>; - def FSUB : AForm_2<63, 20, - (outs F8RC:$FRT), (ins F8RC:$FRA, F8RC:$FRB), - "fsub $FRT, $FRA, $FRB", FPAddSub, - [(set F8RC:$FRT, (fsub F8RC:$FRA, F8RC:$FRB))]>; - def FSUBS : AForm_2<59, 20, - (outs F4RC:$FRT), (ins F4RC:$FRA, F4RC:$FRB), - "fsubs $FRT, $FRA, $FRB", FPGeneral, - [(set F4RC:$FRT, (fsub F4RC:$FRA, F4RC:$FRB))]>; + defm FADD : AForm_2r<63, 21, + (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRB), + "fadd", "$FRT, $FRA, $FRB", FPAddSub, + [(set f64:$FRT, (fadd f64:$FRA, f64:$FRB))]>; + defm FADDS : AForm_2r<59, 21, + (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRB), + "fadds", "$FRT, $FRA, $FRB", FPGeneral, + [(set f32:$FRT, (fadd f32:$FRA, f32:$FRB))]>; + defm FDIV : AForm_2r<63, 18, + (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRB), + "fdiv", "$FRT, $FRA, $FRB", FPDivD, + [(set f64:$FRT, (fdiv f64:$FRA, f64:$FRB))]>; + defm FDIVS : AForm_2r<59, 18, + (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRB), + "fdivs", "$FRT, $FRA, $FRB", FPDivS, + [(set f32:$FRT, (fdiv f32:$FRA, f32:$FRB))]>; + defm FMUL : AForm_3r<63, 25, + (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRC), + "fmul", "$FRT, $FRA, $FRC", FPFused, + [(set f64:$FRT, (fmul f64:$FRA, f64:$FRC))]>; + defm FMULS : AForm_3r<59, 25, + (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRC), + "fmuls", "$FRT, $FRA, $FRC", FPGeneral, + [(set f32:$FRT, (fmul f32:$FRA, f32:$FRC))]>; + defm FSUB : AForm_2r<63, 20, + (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRB), + "fsub", "$FRT, $FRA, $FRB", FPAddSub, + [(set f64:$FRT, (fsub f64:$FRA, f64:$FRB))]>; + defm FSUBS : AForm_2r<59, 20, + (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRB), + "fsubs", "$FRT, $FRA, $FRB", FPGeneral, + [(set f32:$FRT, (fsub f32:$FRA, f32:$FRB))]>; } } +let neverHasSideEffects = 1 in { let PPC970_Unit = 1 in { // FXU Operations. + let isSelect = 1 in def ISEL : AForm_4<31, 15, - (outs GPRC:$rT), (ins GPRC:$rA, GPRC:$rB, pred:$cond), + (outs gprc:$rT), (ins gprc_nor0:$rA, gprc:$rB, crbitrc:$cond), "isel $rT, $rA, $rB, $cond", IntGeneral, []>; } @@ -1477,26 +1859,29 @@ let PPC970_Unit = 1 in { // FXU Operations. // let isCommutable = 1 in { // RLWIMI can be commuted if the rotate amount is zero. -def RLWIMI : MForm_2<20, - (outs GPRC:$rA), (ins GPRC:$rSi, GPRC:$rS, u5imm:$SH, u5imm:$MB, - u5imm:$ME), "rlwimi $rA, $rS, $SH, $MB, $ME", IntRotate, - []>, PPC970_DGroup_Cracked, RegConstraint<"$rSi = $rA">, - NoEncode<"$rSi">; +defm RLWIMI : MForm_2r<20, (outs gprc:$rA), + (ins gprc:$rSi, gprc:$rS, u5imm:$SH, u5imm:$MB, + u5imm:$ME), "rlwimi", "$rA, $rS, $SH, $MB, $ME", IntRotate, + []>, PPC970_DGroup_Cracked, RegConstraint<"$rSi = $rA">, + NoEncode<"$rSi">; } +let BaseName = "rlwinm" in { def RLWINM : MForm_2<21, - (outs GPRC:$rA), (ins GPRC:$rS, u5imm:$SH, u5imm:$MB, u5imm:$ME), + (outs gprc:$rA), (ins gprc:$rS, u5imm:$SH, u5imm:$MB, u5imm:$ME), "rlwinm $rA, $rS, $SH, $MB, $ME", IntGeneral, - []>; + []>, RecFormRel; +let Defs = [CR0] in def RLWINMo : MForm_2<21, - (outs GPRC:$rA), (ins GPRC:$rS, u5imm:$SH, u5imm:$MB, u5imm:$ME), - "rlwinm. $rA, $rS, $SH, $MB, $ME", IntGeneral, - []>, isDOT, PPC970_DGroup_Cracked; -def RLWNM : MForm_2<23, - (outs GPRC:$rA), (ins GPRC:$rS, GPRC:$rB, u5imm:$MB, u5imm:$ME), - "rlwnm $rA, $rS, $rB, $MB, $ME", IntGeneral, - []>; + (outs gprc:$rA), (ins gprc:$rS, u5imm:$SH, u5imm:$MB, u5imm:$ME), + "rlwinm. $rA, $rS, $SH, $MB, $ME", IntGeneral, + []>, isDOT, RecFormRel, PPC970_DGroup_Cracked; } - +defm RLWNM : MForm_2r<23, (outs gprc:$rA), + (ins gprc:$rS, gprc:$rB, u5imm:$MB, u5imm:$ME), + "rlwnm", "$rA, $rS, $rB, $MB, $ME", IntGeneral, + []>; +} +} // neverHasSideEffects = 1 //===----------------------------------------------------------------------===// // PowerPC Instruction Patterns @@ -1507,47 +1892,43 @@ def : Pat<(i32 imm:$imm), (ORI (LIS (HI16 imm:$imm)), (LO16 imm:$imm))>; // Implement the 'not' operation with the NOR instruction. -def NOT : Pat<(not GPRC:$in), - (NOR GPRC:$in, GPRC:$in)>; +def NOT : Pat<(not i32:$in), + (NOR $in, $in)>; // ADD an arbitrary immediate. -def : Pat<(add GPRC:$in, imm:$imm), - (ADDIS (ADDI GPRC:$in, (LO16 imm:$imm)), (HA16 imm:$imm))>; +def : Pat<(add i32:$in, imm:$imm), + (ADDIS (ADDI $in, (LO16 imm:$imm)), (HA16 imm:$imm))>; // OR an arbitrary immediate. -def : Pat<(or GPRC:$in, imm:$imm), - (ORIS (ORI GPRC:$in, (LO16 imm:$imm)), (HI16 imm:$imm))>; +def : Pat<(or i32:$in, imm:$imm), + (ORIS (ORI $in, (LO16 imm:$imm)), (HI16 imm:$imm))>; // XOR an arbitrary immediate. -def : Pat<(xor GPRC:$in, imm:$imm), - (XORIS (XORI GPRC:$in, (LO16 imm:$imm)), (HI16 imm:$imm))>; +def : Pat<(xor i32:$in, imm:$imm), + (XORIS (XORI $in, (LO16 imm:$imm)), (HI16 imm:$imm))>; // SUBFIC -def : Pat<(sub immSExt16:$imm, GPRC:$in), - (SUBFIC GPRC:$in, imm:$imm)>; +def : Pat<(sub immSExt16:$imm, i32:$in), + (SUBFIC $in, imm:$imm)>; // SHL/SRL -def : Pat<(shl GPRC:$in, (i32 imm:$imm)), - (RLWINM GPRC:$in, imm:$imm, 0, (SHL32 imm:$imm))>; -def : Pat<(srl GPRC:$in, (i32 imm:$imm)), - (RLWINM GPRC:$in, (SRL32 imm:$imm), imm:$imm, 31)>; +def : Pat<(shl i32:$in, (i32 imm:$imm)), + (RLWINM $in, imm:$imm, 0, (SHL32 imm:$imm))>; +def : Pat<(srl i32:$in, (i32 imm:$imm)), + (RLWINM $in, (SRL32 imm:$imm), imm:$imm, 31)>; // ROTL -def : Pat<(rotl GPRC:$in, GPRC:$sh), - (RLWNM GPRC:$in, GPRC:$sh, 0, 31)>; -def : Pat<(rotl GPRC:$in, (i32 imm:$imm)), - (RLWINM GPRC:$in, imm:$imm, 0, 31)>; +def : Pat<(rotl i32:$in, i32:$sh), + (RLWNM $in, $sh, 0, 31)>; +def : Pat<(rotl i32:$in, (i32 imm:$imm)), + (RLWINM $in, imm:$imm, 0, 31)>; // RLWNM -def : Pat<(and (rotl GPRC:$in, GPRC:$sh), maskimm32:$imm), - (RLWNM GPRC:$in, GPRC:$sh, (MB maskimm32:$imm), (ME maskimm32:$imm))>; +def : Pat<(and (rotl i32:$in, i32:$sh), maskimm32:$imm), + (RLWNM $in, $sh, (MB maskimm32:$imm), (ME maskimm32:$imm))>; // Calls -def : Pat<(PPCcall_Darwin (i32 tglobaladdr:$dst)), - (BL_Darwin tglobaladdr:$dst)>; -def : Pat<(PPCcall_Darwin (i32 texternalsym:$dst)), - (BL_Darwin texternalsym:$dst)>; -def : Pat<(PPCcall_SVR4 (i32 tglobaladdr:$dst)), - (BL_SVR4 tglobaladdr:$dst)>; -def : Pat<(PPCcall_SVR4 (i32 texternalsym:$dst)), - (BL_SVR4 texternalsym:$dst)>; +def : Pat<(PPCcall (i32 tglobaladdr:$dst)), + (BL tglobaladdr:$dst)>; +def : Pat<(PPCcall (i32 texternalsym:$dst)), + (BL texternalsym:$dst)>; def : Pat<(PPCtc_return (i32 tglobaladdr:$dst), imm:$imm), @@ -1570,28 +1951,28 @@ def : Pat<(PPChi tjumptable:$in, 0), (LIS tjumptable:$in)>; def : Pat<(PPClo tjumptable:$in, 0), (LI tjumptable:$in)>; def : Pat<(PPChi tblockaddress:$in, 0), (LIS tblockaddress:$in)>; def : Pat<(PPClo tblockaddress:$in, 0), (LI tblockaddress:$in)>; -def : Pat<(PPChi tglobaltlsaddr:$g, GPRC:$in), - (ADDIS GPRC:$in, tglobaltlsaddr:$g)>; -def : Pat<(PPClo tglobaltlsaddr:$g, GPRC:$in), - (ADDIL GPRC:$in, tglobaltlsaddr:$g)>; -def : Pat<(add GPRC:$in, (PPChi tglobaladdr:$g, 0)), - (ADDIS GPRC:$in, tglobaladdr:$g)>; -def : Pat<(add GPRC:$in, (PPChi tconstpool:$g, 0)), - (ADDIS GPRC:$in, tconstpool:$g)>; -def : Pat<(add GPRC:$in, (PPChi tjumptable:$g, 0)), - (ADDIS GPRC:$in, tjumptable:$g)>; -def : Pat<(add GPRC:$in, (PPChi tblockaddress:$g, 0)), - (ADDIS GPRC:$in, tblockaddress:$g)>; +def : Pat<(PPChi tglobaltlsaddr:$g, i32:$in), + (ADDIS $in, tglobaltlsaddr:$g)>; +def : Pat<(PPClo tglobaltlsaddr:$g, i32:$in), + (ADDI $in, tglobaltlsaddr:$g)>; +def : Pat<(add i32:$in, (PPChi tglobaladdr:$g, 0)), + (ADDIS $in, tglobaladdr:$g)>; +def : Pat<(add i32:$in, (PPChi tconstpool:$g, 0)), + (ADDIS $in, tconstpool:$g)>; +def : Pat<(add i32:$in, (PPChi tjumptable:$g, 0)), + (ADDIS $in, tjumptable:$g)>; +def : Pat<(add i32:$in, (PPChi tblockaddress:$g, 0)), + (ADDIS $in, tblockaddress:$g)>; // Standard shifts. These are represented separately from the real shifts above // so that we can distinguish between shifts that allow 5-bit and 6-bit shift // amounts. -def : Pat<(sra GPRC:$rS, GPRC:$rB), - (SRAW GPRC:$rS, GPRC:$rB)>; -def : Pat<(srl GPRC:$rS, GPRC:$rB), - (SRW GPRC:$rS, GPRC:$rB)>; -def : Pat<(shl GPRC:$rS, GPRC:$rB), - (SLW GPRC:$rS, GPRC:$rB)>; +def : Pat<(sra i32:$rS, i32:$rB), + (SRAW $rS, $rB)>; +def : Pat<(srl i32:$rS, i32:$rB), + (SRW $rS, $rB)>; +def : Pat<(shl i32:$rS, i32:$rB), + (SLW $rS, $rB)>; def : Pat<(zextloadi1 iaddr:$src), (LBZ iaddr:$src)>; @@ -1614,18 +1995,20 @@ def : Pat<(f64 (extloadf32 iaddr:$src)), def : Pat<(f64 (extloadf32 xaddr:$src)), (COPY_TO_REGCLASS (LFSX xaddr:$src), F8RC)>; -def : Pat<(f64 (fextend F4RC:$src)), - (COPY_TO_REGCLASS F4RC:$src, F8RC)>; - -// Memory barriers -def : Pat<(membarrier (i32 imm /*ll*/), - (i32 imm /*ls*/), - (i32 imm /*sl*/), - (i32 imm /*ss*/), - (i32 imm /*device*/)), - (SYNC)>; +def : Pat<(f64 (fextend f32:$src)), + (COPY_TO_REGCLASS $src, F8RC)>; def : Pat<(atomic_fence (imm), (imm)), (SYNC)>; +// Additional FNMSUB patterns: -a*c + b == -(a*c - b) +def : Pat<(fma (fneg f64:$A), f64:$C, f64:$B), + (FNMSUB $A, $C, $B)>; +def : Pat<(fma f64:$A, (fneg f64:$C), f64:$B), + (FNMSUB $A, $C, $B)>; +def : Pat<(fma (fneg f32:$A), f32:$C, f32:$B), + (FNMSUBS $A, $C, $B)>; +def : Pat<(fma f32:$A, (fneg f32:$C), f32:$B), + (FNMSUBS $A, $C, $B)>; + include "PPCInstrAltivec.td" include "PPCInstr64Bit.td" diff --git a/lib/Target/PowerPC/PPCMCInstLower.cpp b/lib/Target/PowerPC/PPCMCInstLower.cpp index 9b0df3e..f8cf3a5 100644 --- a/lib/Target/PowerPC/PPCMCInstLower.cpp +++ b/lib/Target/PowerPC/PPCMCInstLower.cpp @@ -14,6 +14,7 @@ #include "PPC.h" #include "llvm/ADT/SmallString.h" +#include "llvm/ADT/Twine.h" #include "llvm/CodeGen/AsmPrinter.h" #include "llvm/CodeGen/MachineFunction.h" #include "llvm/CodeGen/MachineModuleInfoImpls.h" @@ -51,7 +52,14 @@ static MCSymbol *GetSymbolFromOperand(const MachineOperand &MO, AsmPrinter &AP){ // before we return the symbol. if (MO.getTargetFlags() == PPCII::MO_DARWIN_STUB) { Name += "$stub"; - MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name.str()); + const char *PGP = AP.MAI->getPrivateGlobalPrefix(); + const char *Prefix = ""; + if (!Name.startswith(PGP)) { + // http://llvm.org/bugs/show_bug.cgi?id=15763 + // all stubs and lazy_ptrs should be local symbols, which need leading 'L' + Prefix = PGP; + } + MCSymbol *Sym = Ctx.GetOrCreateSymbol(Twine(Prefix) + Twine(Name)); MachineModuleInfoImpl::StubValueTy &StubSym = getMachOMMI(AP).getFnStubEntry(Sym); if (StubSym.getPointer()) diff --git a/lib/Target/PowerPC/PPCMachineFunctionInfo.h b/lib/Target/PowerPC/PPCMachineFunctionInfo.h index b1636a2..40d1f3a 100644 --- a/lib/Target/PowerPC/PPCMachineFunctionInfo.h +++ b/lib/Target/PowerPC/PPCMachineFunctionInfo.h @@ -47,6 +47,9 @@ class PPCFunctionInfo : public MachineFunctionInfo { /// SpillsCR - Indicates whether CR is spilled in the current function. bool SpillsCR; + /// Indicates whether VRSAVE is spilled in the current function. + bool SpillsVRSAVE; + /// LRStoreRequired - The bool indicates whether there is some explicit use of /// the LR/LR8 stack slot that is not obvious from scanning the code. This /// requires that the code generator produce a store of LR to the stack on @@ -81,6 +84,11 @@ class PPCFunctionInfo : public MachineFunctionInfo { /// CRSpillFrameIndex - FrameIndex for CR spill slot for 32-bit SVR4. int CRSpillFrameIndex; + /// If any of CR[2-4] need to be saved in the prologue and restored in the + /// epilogue then they are added to this array. This is used for the + /// 64-bit SVR4 ABI. + SmallVector<unsigned, 3> MustSaveCRs; + public: explicit PPCFunctionInfo(MachineFunction &MF) : FramePointerSaveIndex(0), @@ -88,6 +96,7 @@ public: HasSpills(false), HasNonRISpills(false), SpillsCR(false), + SpillsVRSAVE(false), LRStoreRequired(false), MinReservedArea(0), TailCallSPDelta(0), @@ -127,6 +136,9 @@ public: void setSpillsCR() { SpillsCR = true; } bool isCRSpilled() const { return SpillsCR; } + void setSpillsVRSAVE() { SpillsVRSAVE = true; } + bool isVRSAVESpilled() const { return SpillsVRSAVE; } + void setLRStoreRequired() { LRStoreRequired = true; } bool isLRStoreRequired() const { return LRStoreRequired; } @@ -147,6 +159,10 @@ public: int getCRSpillFrameIndex() const { return CRSpillFrameIndex; } void setCRSpillFrameIndex(int idx) { CRSpillFrameIndex = idx; } + + const SmallVector<unsigned, 3> & + getMustSaveCRs() const { return MustSaveCRs; } + void addMustSaveCR(unsigned Reg) { MustSaveCRs.push_back(Reg); } }; } // end of namespace llvm diff --git a/lib/Target/PowerPC/PPCRegisterInfo.cpp b/lib/Target/PowerPC/PPCRegisterInfo.cpp index e2c7221..2be6324 100644 --- a/lib/Target/PowerPC/PPCRegisterInfo.cpp +++ b/lib/Target/PowerPC/PPCRegisterInfo.cpp @@ -68,7 +68,7 @@ PPCRegisterInfo::PPCRegisterInfo(const PPCSubtarget &ST, ImmToIdxMap[PPC::LHZ8] = PPC::LHZX8; ImmToIdxMap[PPC::LWZ8] = PPC::LWZX8; ImmToIdxMap[PPC::STB8] = PPC::STBX8; ImmToIdxMap[PPC::STH8] = PPC::STHX8; ImmToIdxMap[PPC::STW8] = PPC::STWX8; ImmToIdxMap[PPC::STDU] = PPC::STDUX; - ImmToIdxMap[PPC::ADDI8] = PPC::ADD8; ImmToIdxMap[PPC::STD_32] = PPC::STDX_32; + ImmToIdxMap[PPC::ADDI8] = PPC::ADD8; } /// getPointerRegClass - Return the register class to use to hold pointers. @@ -76,6 +76,14 @@ PPCRegisterInfo::PPCRegisterInfo(const PPCSubtarget &ST, const TargetRegisterClass * PPCRegisterInfo::getPointerRegClass(const MachineFunction &MF, unsigned Kind) const { + // Note that PPCInstrInfo::FoldImmediate also directly uses this Kind value + // when it checks for ZERO folding. + if (Kind == 1) { + if (Subtarget.isPPC64()) + return &PPC::G8RC_NOX0RegClass; + return &PPC::GPRC_NOR0RegClass; + } + if (Subtarget.isPPC64()) return &PPC::G8RCRegClass; return &PPC::GPRCRegClass; @@ -99,12 +107,35 @@ PPCRegisterInfo::getCallPreservedMask(CallingConv::ID CC) const { return Subtarget.isPPC64() ? CSR_SVR464_RegMask : CSR_SVR432_RegMask; } +const uint32_t* +PPCRegisterInfo::getNoPreservedMask() const { + // The naming here is inverted: The CSR_NoRegs_Altivec has the + // Altivec registers masked so that they're not saved and restored around + // instructions with this preserved mask. + + if (!Subtarget.hasAltivec()) + return CSR_NoRegs_Altivec_RegMask; + + if (Subtarget.isDarwin()) + return CSR_NoRegs_Darwin_RegMask; + return CSR_NoRegs_RegMask; +} + BitVector PPCRegisterInfo::getReservedRegs(const MachineFunction &MF) const { BitVector Reserved(getNumRegs()); const PPCFrameLowering *PPCFI = static_cast<const PPCFrameLowering*>(MF.getTarget().getFrameLowering()); - Reserved.set(PPC::R0); + // The ZERO register is not really a register, but the representation of r0 + // when used in instructions that treat r0 as the constant 0. + Reserved.set(PPC::ZERO); + Reserved.set(PPC::ZERO8); + + // The FP register is also not really a register, but is the representation + // of the frame pointer register used by ISD::FRAMEADDR. + Reserved.set(PPC::FP); + Reserved.set(PPC::FP8); + Reserved.set(PPC::R1); Reserved.set(PPC::LR); Reserved.set(PPC::LR8); @@ -117,16 +148,14 @@ BitVector PPCRegisterInfo::getReservedRegs(const MachineFunction &MF) const { } // On PPC64, r13 is the thread pointer. Never allocate this register. - // Note that this is over conservative, as it also prevents allocation of R31 - // when the FP is not needed. if (Subtarget.isPPC64()) { Reserved.set(PPC::R13); - Reserved.set(PPC::R31); - Reserved.set(PPC::X0); Reserved.set(PPC::X1); Reserved.set(PPC::X13); - Reserved.set(PPC::X31); + + if (PPCFI->needsFP(MF)) + Reserved.set(PPC::X31); // The 64-bit SVR4 ABI reserves r2 for the TOC pointer. if (Subtarget.isSVR4ABI()) { @@ -149,6 +178,8 @@ PPCRegisterInfo::getRegPressureLimit(const TargetRegisterClass *RC, switch (RC->getID()) { default: return 0; + case PPC::G8RC_NOX0RegClassID: + case PPC::GPRC_NOR0RegClassID: case PPC::G8RCRegClassID: case PPC::GPRCRegClassID: { unsigned FP = TFI->hasFP(MF) ? 1 : 0; @@ -174,8 +205,7 @@ PPCRegisterInfo::getRegPressureLimit(const TargetRegisterClass *RC, /// stwxu R0, SP, Rnegsize ; add and update the SP with the negated size /// addi Rnew, SP, \#maxCalFrameSize ; get the top of the allocation /// -void PPCRegisterInfo::lowerDynamicAlloc(MachineBasicBlock::iterator II, - int SPAdj, RegScavenger *RS) const { +void PPCRegisterInfo::lowerDynamicAlloc(MachineBasicBlock::iterator II) const { // Get the instruction. MachineInstr &MI = *II; // Get the instruction's basic block. @@ -271,22 +301,19 @@ void PPCRegisterInfo::lowerDynamicAlloc(MachineBasicBlock::iterator II, /// stw rA, FI ; Store rA to the frame. /// void PPCRegisterInfo::lowerCRSpilling(MachineBasicBlock::iterator II, - unsigned FrameIndex, int SPAdj, - RegScavenger *RS) const { + unsigned FrameIndex) const { // Get the instruction. MachineInstr &MI = *II; // ; SPILL_CR <SrcReg>, <offset> // Get the instruction's basic block. MachineBasicBlock &MBB = *MI.getParent(); + MachineFunction &MF = *MBB.getParent(); DebugLoc dl = MI.getDebugLoc(); - // FIXME: Once LLVM supports creating virtual registers here, or the register - // scavenger can return multiple registers, stop using reserved registers - // here. - (void) SPAdj; - (void) RS; - bool LP64 = Subtarget.isPPC64(); - unsigned Reg = LP64 ? PPC::X0 : PPC::R0; + const TargetRegisterClass *G8RC = &PPC::G8RCRegClass; + const TargetRegisterClass *GPRC = &PPC::GPRCRegClass; + + unsigned Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC); unsigned SrcReg = MI.getOperand(0).getReg(); // We need to store the CR in the low 4-bits of the saved value. First, issue @@ -296,16 +323,20 @@ void PPCRegisterInfo::lowerCRSpilling(MachineBasicBlock::iterator II, // If the saved register wasn't CR0, shift the bits left so that they are in // CR0's slot. - if (SrcReg != PPC::CR0) + if (SrcReg != PPC::CR0) { + unsigned Reg1 = Reg; + Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC); + // rlwinm rA, rA, ShiftBits, 0, 31. BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::RLWINM8 : PPC::RLWINM), Reg) - .addReg(Reg, RegState::Kill) - .addImm(getPPCRegisterNumbering(SrcReg) * 4) + .addReg(Reg1, RegState::Kill) + .addImm(getEncodingValue(SrcReg) * 4) .addImm(0) .addImm(31); + } addFrameReference(BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::STW8 : PPC::STW)) - .addReg(Reg, getKillRegState(MI.getOperand(1).getImm())), + .addReg(Reg, RegState::Kill), FrameIndex); // Discard the pseudo instruction. @@ -313,22 +344,19 @@ void PPCRegisterInfo::lowerCRSpilling(MachineBasicBlock::iterator II, } void PPCRegisterInfo::lowerCRRestore(MachineBasicBlock::iterator II, - unsigned FrameIndex, int SPAdj, - RegScavenger *RS) const { + unsigned FrameIndex) const { // Get the instruction. MachineInstr &MI = *II; // ; <DestReg> = RESTORE_CR <offset> // Get the instruction's basic block. MachineBasicBlock &MBB = *MI.getParent(); + MachineFunction &MF = *MBB.getParent(); DebugLoc dl = MI.getDebugLoc(); - // FIXME: Once LLVM supports creating virtual registers here, or the register - // scavenger can return multiple registers, stop using reserved registers - // here. - (void) SPAdj; - (void) RS; - bool LP64 = Subtarget.isPPC64(); - unsigned Reg = LP64 ? PPC::X0 : PPC::R0; + const TargetRegisterClass *G8RC = &PPC::G8RCRegClass; + const TargetRegisterClass *GPRC = &PPC::GPRCRegClass; + + unsigned Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC); unsigned DestReg = MI.getOperand(0).getReg(); assert(MI.definesRegister(DestReg) && "RESTORE_CR does not define its destination"); @@ -339,15 +367,67 @@ void PPCRegisterInfo::lowerCRRestore(MachineBasicBlock::iterator II, // If the reloaded register isn't CR0, shift the bits right so that they are // in the right CR's slot. if (DestReg != PPC::CR0) { - unsigned ShiftBits = getPPCRegisterNumbering(DestReg)*4; + unsigned Reg1 = Reg; + Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC); + + unsigned ShiftBits = getEncodingValue(DestReg)*4; // rlwinm r11, r11, 32-ShiftBits, 0, 31. BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::RLWINM8 : PPC::RLWINM), Reg) - .addReg(Reg).addImm(32-ShiftBits).addImm(0) + .addReg(Reg1, RegState::Kill).addImm(32-ShiftBits).addImm(0) .addImm(31); } BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::MTCRF8 : PPC::MTCRF), DestReg) - .addReg(Reg); + .addReg(Reg, RegState::Kill); + + // Discard the pseudo instruction. + MBB.erase(II); +} + +void PPCRegisterInfo::lowerVRSAVESpilling(MachineBasicBlock::iterator II, + unsigned FrameIndex) const { + // Get the instruction. + MachineInstr &MI = *II; // ; SPILL_VRSAVE <SrcReg>, <offset> + // Get the instruction's basic block. + MachineBasicBlock &MBB = *MI.getParent(); + MachineFunction &MF = *MBB.getParent(); + DebugLoc dl = MI.getDebugLoc(); + + const TargetRegisterClass *GPRC = &PPC::GPRCRegClass; + unsigned Reg = MF.getRegInfo().createVirtualRegister(GPRC); + unsigned SrcReg = MI.getOperand(0).getReg(); + + BuildMI(MBB, II, dl, TII.get(PPC::MFVRSAVEv), Reg) + .addReg(SrcReg, getKillRegState(MI.getOperand(0).isKill())); + + addFrameReference(BuildMI(MBB, II, dl, TII.get(PPC::STW)) + .addReg(Reg, RegState::Kill), + FrameIndex); + + // Discard the pseudo instruction. + MBB.erase(II); +} + +void PPCRegisterInfo::lowerVRSAVERestore(MachineBasicBlock::iterator II, + unsigned FrameIndex) const { + // Get the instruction. + MachineInstr &MI = *II; // ; <DestReg> = RESTORE_VRSAVE <offset> + // Get the instruction's basic block. + MachineBasicBlock &MBB = *MI.getParent(); + MachineFunction &MF = *MBB.getParent(); + DebugLoc dl = MI.getDebugLoc(); + + const TargetRegisterClass *GPRC = &PPC::GPRCRegClass; + unsigned Reg = MF.getRegInfo().createVirtualRegister(GPRC); + unsigned DestReg = MI.getOperand(0).getReg(); + assert(MI.definesRegister(DestReg) && + "RESTORE_VRSAVE does not define its destination"); + + addFrameReference(BuildMI(MBB, II, dl, TII.get(PPC::LWZ), + Reg), FrameIndex); + + BuildMI(MBB, II, dl, TII.get(PPC::MTVRSAVEv), DestReg) + .addReg(Reg, RegState::Kill); // Discard the pseudo instruction. MBB.erase(II); @@ -374,6 +454,33 @@ PPCRegisterInfo::hasReservedSpillSlot(const MachineFunction &MF, return false; } +// Figure out if the offset in the instruction is shifted right two bits. This +// is true for instructions like "STD", which the machine implicitly adds two +// low zeros to. +static bool usesIXAddr(const MachineInstr &MI) { + unsigned OpC = MI.getOpcode(); + + switch (OpC) { + default: + return false; + case PPC::LWA: + case PPC::LD: + case PPC::STD: + return true; + } +} + +// Return the OffsetOperandNo given the FIOperandNum (and the instruction). +static unsigned getOffsetONFromFION(const MachineInstr &MI, + unsigned FIOperandNum) { + // Take into account whether it's an add or mem instruction + unsigned OffsetOperandNo = (FIOperandNum == 2) ? 1 : 2; + if (MI.isInlineAsm()) + OffsetOperandNo = FIOperandNum-1; + + return OffsetOperandNo; +} + void PPCRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, int SPAdj, unsigned FIOperandNum, @@ -391,10 +498,7 @@ PPCRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); DebugLoc dl = MI.getDebugLoc(); - // Take into account whether it's an add or mem instruction - unsigned OffsetOperandNo = (FIOperandNum == 2) ? 1 : 2; - if (MI.isInlineAsm()) - OffsetOperandNo = FIOperandNum-1; + unsigned OffsetOperandNo = getOffsetONFromFION(MI, FIOperandNum); // Get the frame index. int FrameIndex = MI.getOperand(FIOperandNum).getIndex(); @@ -409,16 +513,22 @@ PPCRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, // Special case for dynamic alloca. if (FPSI && FrameIndex == FPSI && (OpC == PPC::DYNALLOC || OpC == PPC::DYNALLOC8)) { - lowerDynamicAlloc(II, SPAdj, RS); + lowerDynamicAlloc(II); return; } - // Special case for pseudo-ops SPILL_CR and RESTORE_CR. + // Special case for pseudo-ops SPILL_CR and RESTORE_CR, etc. if (OpC == PPC::SPILL_CR) { - lowerCRSpilling(II, FrameIndex, SPAdj, RS); + lowerCRSpilling(II, FrameIndex); return; } else if (OpC == PPC::RESTORE_CR) { - lowerCRRestore(II, FrameIndex, SPAdj, RS); + lowerCRRestore(II, FrameIndex); + return; + } else if (OpC == PPC::SPILL_VRSAVE) { + lowerVRSAVESpilling(II, FrameIndex); + return; + } else if (OpC == PPC::RESTORE_VRSAVE) { + lowerVRSAVERestore(II, FrameIndex); return; } @@ -430,36 +540,12 @@ PPCRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, (is64Bit ? PPC::X1 : PPC::R1), false); - // Figure out if the offset in the instruction is shifted right two bits. This - // is true for instructions like "STD", which the machine implicitly adds two - // low zeros to. - bool isIXAddr = false; - switch (OpC) { - case PPC::LWA: - case PPC::LD: - case PPC::STD: - case PPC::STD_32: - isIXAddr = true; - break; - } + // Figure out if the offset in the instruction is shifted right two bits. + bool isIXAddr = usesIXAddr(MI); - bool noImmForm = false; - switch (OpC) { - case PPC::LVEBX: - case PPC::LVEHX: - case PPC::LVEWX: - case PPC::LVX: - case PPC::LVXL: - case PPC::LVSL: - case PPC::LVSR: - case PPC::STVEBX: - case PPC::STVEHX: - case PPC::STVEWX: - case PPC::STVX: - case PPC::STVXL: - noImmForm = true; - break; - } + // If the instruction is not present in ImmToIdxMap, then it has no immediate + // form (and must be r+r). + bool noImmForm = !MI.isInlineAsm() && !ImmToIdxMap.count(OpC); // Now add the frame object offset to the offset from r1. int Offset = MFI->getObjectOffset(FrameIndex); @@ -497,13 +583,15 @@ PPCRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, const TargetRegisterClass *G8RC = &PPC::G8RCRegClass; const TargetRegisterClass *GPRC = &PPC::GPRCRegClass; - unsigned SReg = MF.getRegInfo().createVirtualRegister(is64Bit ? G8RC : GPRC); + const TargetRegisterClass *RC = is64Bit ? G8RC : GPRC; + unsigned SRegHi = MF.getRegInfo().createVirtualRegister(RC), + SReg = MF.getRegInfo().createVirtualRegister(RC); // Insert a set of rA with the full offset value before the ld, st, or add - BuildMI(MBB, II, dl, TII.get(is64Bit ? PPC::LIS8 : PPC::LIS), SReg) + BuildMI(MBB, II, dl, TII.get(is64Bit ? PPC::LIS8 : PPC::LIS), SRegHi) .addImm(Offset >> 16); BuildMI(MBB, II, dl, TII.get(is64Bit ? PPC::ORI8 : PPC::ORI), SReg) - .addReg(SReg, RegState::Kill) + .addReg(SRegHi, RegState::Kill) .addImm(Offset); // Convert into indexed form of the instruction: @@ -545,3 +633,124 @@ unsigned PPCRegisterInfo::getEHExceptionRegister() const { unsigned PPCRegisterInfo::getEHHandlerRegister() const { return !Subtarget.isPPC64() ? PPC::R4 : PPC::X4; } + +/// Returns true if the instruction's frame index +/// reference would be better served by a base register other than FP +/// or SP. Used by LocalStackFrameAllocation to determine which frame index +/// references it should create new base registers for. +bool PPCRegisterInfo:: +needsFrameBaseReg(MachineInstr *MI, int64_t Offset) const { + assert(Offset < 0 && "Local offset must be negative"); + + unsigned FIOperandNum = 0; + while (!MI->getOperand(FIOperandNum).isFI()) { + ++FIOperandNum; + assert(FIOperandNum < MI->getNumOperands() && + "Instr doesn't have FrameIndex operand!"); + } + + unsigned OffsetOperandNo = getOffsetONFromFION(*MI, FIOperandNum); + + if (!usesIXAddr(*MI)) + Offset += MI->getOperand(OffsetOperandNo).getImm(); + else + Offset += MI->getOperand(OffsetOperandNo).getImm() << 2; + + // It's the load/store FI references that cause issues, as it can be difficult + // to materialize the offset if it won't fit in the literal field. Estimate + // based on the size of the local frame and some conservative assumptions + // about the rest of the stack frame (note, this is pre-regalloc, so + // we don't know everything for certain yet) whether this offset is likely + // to be out of range of the immediate. Return true if so. + + // We only generate virtual base registers for loads and stores that have + // an r+i form. Return false for everything else. + unsigned OpC = MI->getOpcode(); + if (!ImmToIdxMap.count(OpC)) + return false; + + // Don't generate a new virtual base register just to add zero to it. + if ((OpC == PPC::ADDI || OpC == PPC::ADDI8) && + MI->getOperand(2).getImm() == 0) + return false; + + MachineBasicBlock &MBB = *MI->getParent(); + MachineFunction &MF = *MBB.getParent(); + + const PPCFrameLowering *PPCFI = + static_cast<const PPCFrameLowering*>(MF.getTarget().getFrameLowering()); + unsigned StackEst = + PPCFI->determineFrameLayout(MF, false, true); + + // If we likely don't need a stack frame, then we probably don't need a + // virtual base register either. + if (!StackEst) + return false; + + // Estimate an offset from the stack pointer. + // The incoming offset is relating to the SP at the start of the function, + // but when we access the local it'll be relative to the SP after local + // allocation, so adjust our SP-relative offset by that allocation size. + Offset += StackEst; + + // The frame pointer will point to the end of the stack, so estimate the + // offset as the difference between the object offset and the FP location. + return !isFrameOffsetLegal(MI, Offset); +} + +/// Insert defining instruction(s) for BaseReg to +/// be a pointer to FrameIdx at the beginning of the basic block. +void PPCRegisterInfo:: +materializeFrameBaseRegister(MachineBasicBlock *MBB, + unsigned BaseReg, int FrameIdx, + int64_t Offset) const { + unsigned ADDriOpc = Subtarget.isPPC64() ? PPC::ADDI8 : PPC::ADDI; + + MachineBasicBlock::iterator Ins = MBB->begin(); + DebugLoc DL; // Defaults to "unknown" + if (Ins != MBB->end()) + DL = Ins->getDebugLoc(); + + const MCInstrDesc &MCID = TII.get(ADDriOpc); + MachineRegisterInfo &MRI = MBB->getParent()->getRegInfo(); + const MachineFunction &MF = *MBB->getParent(); + MRI.constrainRegClass(BaseReg, TII.getRegClass(MCID, 0, this, MF)); + + BuildMI(*MBB, Ins, DL, MCID, BaseReg) + .addFrameIndex(FrameIdx).addImm(Offset); +} + +void +PPCRegisterInfo::resolveFrameIndex(MachineBasicBlock::iterator I, + unsigned BaseReg, int64_t Offset) const { + MachineInstr &MI = *I; + + unsigned FIOperandNum = 0; + while (!MI.getOperand(FIOperandNum).isFI()) { + ++FIOperandNum; + assert(FIOperandNum < MI.getNumOperands() && + "Instr doesn't have FrameIndex operand!"); + } + + MI.getOperand(FIOperandNum).ChangeToRegister(BaseReg, false); + unsigned OffsetOperandNo = getOffsetONFromFION(MI, FIOperandNum); + + bool isIXAddr = usesIXAddr(MI); + if (!isIXAddr) + Offset += MI.getOperand(OffsetOperandNo).getImm(); + else + Offset += MI.getOperand(OffsetOperandNo).getImm() << 2; + + // Figure out if the offset in the instruction is shifted right two bits. + if (isIXAddr) + Offset >>= 2; // The actual encoded value has the low two bits zero. + + MI.getOperand(OffsetOperandNo).ChangeToImmediate(Offset); +} + +bool PPCRegisterInfo::isFrameOffsetLegal(const MachineInstr *MI, + int64_t Offset) const { + return MI->getOpcode() == PPC::DBG_VALUE || // DBG_VALUE is always Reg+Imm + (isInt<16>(Offset) && (!usesIXAddr(*MI) || (Offset & 3) == 0)); +} + diff --git a/lib/Target/PowerPC/PPCRegisterInfo.h b/lib/Target/PowerPC/PPCRegisterInfo.h index 5f89f63..7a48b4b 100644 --- a/lib/Target/PowerPC/PPCRegisterInfo.h +++ b/lib/Target/PowerPC/PPCRegisterInfo.h @@ -15,8 +15,8 @@ #ifndef POWERPC32_REGISTERINFO_H #define POWERPC32_REGISTERINFO_H +#include "llvm/ADT/DenseMap.h" #include "PPC.h" -#include <map> #define GET_REGINFO_HEADER #include "PPCGenRegisterInfo.inc" @@ -27,7 +27,7 @@ class TargetInstrInfo; class Type; class PPCRegisterInfo : public PPCGenRegisterInfo { - std::map<unsigned, unsigned> ImmToIdxMap; + DenseMap<unsigned, unsigned> ImmToIdxMap; const PPCSubtarget &Subtarget; const TargetInstrInfo &TII; public: @@ -44,6 +44,7 @@ public: /// Code Generation virtual methods... const uint16_t *getCalleeSavedRegs(const MachineFunction* MF = 0) const; const uint32_t *getCallPreservedMask(CallingConv::ID CC) const; + const uint32_t *getNoPreservedMask() const; BitVector getReservedRegs(const MachineFunction &MF) const; @@ -60,18 +61,35 @@ public: return true; } - void lowerDynamicAlloc(MachineBasicBlock::iterator II, - int SPAdj, RegScavenger *RS) const; - void lowerCRSpilling(MachineBasicBlock::iterator II, unsigned FrameIndex, - int SPAdj, RegScavenger *RS) const; - void lowerCRRestore(MachineBasicBlock::iterator II, unsigned FrameIndex, - int SPAdj, RegScavenger *RS) const; + virtual bool requiresVirtualBaseRegisters(const MachineFunction &MF) const { + return true; + } + + void lowerDynamicAlloc(MachineBasicBlock::iterator II) const; + void lowerCRSpilling(MachineBasicBlock::iterator II, + unsigned FrameIndex) const; + void lowerCRRestore(MachineBasicBlock::iterator II, + unsigned FrameIndex) const; + void lowerVRSAVESpilling(MachineBasicBlock::iterator II, + unsigned FrameIndex) const; + void lowerVRSAVERestore(MachineBasicBlock::iterator II, + unsigned FrameIndex) const; + bool hasReservedSpillSlot(const MachineFunction &MF, unsigned Reg, int &FrameIdx) const; void eliminateFrameIndex(MachineBasicBlock::iterator II, int SPAdj, unsigned FIOperandNum, RegScavenger *RS = NULL) const; + // Support for virtual base registers. + bool needsFrameBaseReg(MachineInstr *MI, int64_t Offset) const; + void materializeFrameBaseRegister(MachineBasicBlock *MBB, + unsigned BaseReg, int FrameIdx, + int64_t Offset) const; + void resolveFrameIndex(MachineBasicBlock::iterator I, + unsigned BaseReg, int64_t Offset) const; + bool isFrameOffsetLegal(const MachineInstr *MI, int64_t Offset) const; + // Debug information queries. unsigned getFrameRegister(const MachineFunction &MF) const; diff --git a/lib/Target/PowerPC/PPCRegisterInfo.td b/lib/Target/PowerPC/PPCRegisterInfo.td index 8ee9b1e..57a25f5 100644 --- a/lib/Target/PowerPC/PPCRegisterInfo.td +++ b/lib/Target/PowerPC/PPCRegisterInfo.td @@ -27,40 +27,40 @@ class PPCReg<string n> : Register<n> { // GPR - One of the 32 32-bit general-purpose registers class GPR<bits<5> num, string n> : PPCReg<n> { - field bits<5> Num = num; + let HWEncoding{4-0} = num; } // GP8 - One of the 32 64-bit general-purpose registers class GP8<GPR SubReg, string n> : PPCReg<n> { - field bits<5> Num = SubReg.Num; + let HWEncoding = SubReg.HWEncoding; let SubRegs = [SubReg]; let SubRegIndices = [sub_32]; } // SPR - One of the 32-bit special-purpose registers class SPR<bits<10> num, string n> : PPCReg<n> { - field bits<10> Num = num; + let HWEncoding{9-0} = num; } // FPR - One of the 32 64-bit floating-point registers class FPR<bits<5> num, string n> : PPCReg<n> { - field bits<5> Num = num; + let HWEncoding{4-0} = num; } // VR - One of the 32 128-bit vector registers class VR<bits<5> num, string n> : PPCReg<n> { - field bits<5> Num = num; + let HWEncoding{4-0} = num; } // CR - One of the 8 4-bit condition registers class CR<bits<3> num, string n, list<Register> subregs> : PPCReg<n> { - field bits<3> Num = num; + let HWEncoding{2-0} = num; let SubRegs = subregs; } // CRBIT - One of the 32 1-bit condition register fields class CRBIT<bits<5> num, string n> : PPCReg<n> { - field bits<5> Num = num; + let HWEncoding{4-0} = num; } // General-purpose registers @@ -86,6 +86,14 @@ foreach Index = 0-31 in { DwarfRegNum<[!add(Index, 77), !add(Index, 77)]>; } +// The reprsentation of r0 when treated as the constant 0. +def ZERO : GPR<0, "0">; +def ZERO8 : GP8<ZERO, "0">; + +// Representations of the frame pointer used by ISD::FRAMEADDR. +def FP : GPR<0 /* arbitrary */, "**FRAME POINTER**">; +def FP8 : GP8<FP, "**FRAME POINTER**">; + // Condition register bits def CR0LT : CRBIT< 0, "0">; def CR0GT : CRBIT< 1, "1">; @@ -164,11 +172,17 @@ def RM: SPR<512, "**ROUNDING MODE**">; // then nonvolatiles in reverse order since stmw/lmw save from rN to r31 def GPRC : RegisterClass<"PPC", [i32], 32, (add (sequence "R%u", 2, 12), (sequence "R%u", 30, 13), - R31, R0, R1, LR)>; + R31, R0, R1, FP)>; def G8RC : RegisterClass<"PPC", [i64], 64, (add (sequence "X%u", 2, 12), (sequence "X%u", 30, 14), - X31, X13, X0, X1, LR8)>; + X31, X13, X0, X1, FP8)>; + +// For some instructions r0 is special (representing the value 0 instead of +// the value in the r0 register), and we use these register subclasses to +// prevent r0 from being allocated for use by those instructions. +def GPRC_NOR0 : RegisterClass<"PPC", [i32], 32, (add (sub GPRC, R0), ZERO)>; +def G8RC_NOX0 : RegisterClass<"PPC", [i64], 64, (add (sub G8RC, X0), ZERO8)>; // Allocate volatiles first, then non-volatiles in reverse order. With the SVR4 // ABI the size of the Floating-point register save area is determined by the diff --git a/lib/Target/PowerPC/PPCScheduleA2.td b/lib/Target/PowerPC/PPCScheduleA2.td index ba63b5c..8d5838e 100644 --- a/lib/Target/PowerPC/PPCScheduleA2.td +++ b/lib/Target/PowerPC/PPCScheduleA2.td @@ -749,3 +749,18 @@ def PPCA2Itineraries : ProcessorItineraries< [15, 7], [FPR_Bypass, FPR_Bypass]> ]>; + +// ===---------------------------------------------------------------------===// +// A2 machine model for scheduling and other instruction cost heuristics. + +def PPCA2Model : SchedMachineModel { + let IssueWidth = 1; // 2 micro-ops are dispatched per cycle. + let MinLatency = -1; // OperandCycles are interpreted as MinLatency. + let LoadLatency = 6; // Optimistic load latency assuming bypass. + // This is overriden by OperandCycles if the + // Itineraries are queried instead. + let MispredictPenalty = 13; + + let Itineraries = PPCA2Itineraries; +} + diff --git a/lib/Target/PowerPC/PPCScheduleG5.td b/lib/Target/PowerPC/PPCScheduleG5.td index 7c02ea0..c64998d 100644 --- a/lib/Target/PowerPC/PPCScheduleG5.td +++ b/lib/Target/PowerPC/PPCScheduleG5.td @@ -92,3 +92,18 @@ def G5Itineraries : ProcessorItineraries< InstrItinData<VecVSL , [InstrStage<2, [VIU1]>]>, InstrItinData<VecVSR , [InstrStage<3, [VPU]>]> ]>; + +// ===---------------------------------------------------------------------===// +// e5500 machine model for scheduling and other instruction cost heuristics. + +def G5Model : SchedMachineModel { + let IssueWidth = 4; // 4 (non-branch) instructions are dispatched per cycle. + let MinLatency = 0; // Out-of-order dispatch. + let LoadLatency = 3; // Optimistic load latency assuming bypass. + // This is overriden by OperandCycles if the + // Itineraries are queried instead. + let MispredictPenalty = 16; + + let Itineraries = G5Itineraries; +} + diff --git a/lib/Target/PowerPC/PPCSubtarget.cpp b/lib/Target/PowerPC/PPCSubtarget.cpp index 18e4c07..a8f2b3f 100644 --- a/lib/Target/PowerPC/PPCSubtarget.cpp +++ b/lib/Target/PowerPC/PPCSubtarget.cpp @@ -38,8 +38,18 @@ PPCSubtarget::PPCSubtarget(const std::string &TT, const std::string &CPU, , HasAltivec(false) , HasQPX(false) , HasFSQRT(false) + , HasFRE(false) + , HasFRES(false) + , HasFRSQRTE(false) + , HasFRSQRTES(false) + , HasRecipPrec(false) , HasSTFIWX(false) + , HasLFIWAX(false) + , HasFPRND(false) + , HasFPCVT(false) , HasISEL(false) + , HasPOPCNTD(false) + , HasLDBRX(false) , IsBookE(false) , HasLazyResolverStubs(false) , IsJITCodeModel(false) diff --git a/lib/Target/PowerPC/PPCSubtarget.h b/lib/Target/PowerPC/PPCSubtarget.h index 15885bd..65b4d21 100644 --- a/lib/Target/PowerPC/PPCSubtarget.h +++ b/lib/Target/PowerPC/PPCSubtarget.h @@ -77,8 +77,15 @@ protected: bool HasAltivec; bool HasQPX; bool HasFSQRT; + bool HasFRE, HasFRES, HasFRSQRTE, HasFRSQRTES; + bool HasRecipPrec; bool HasSTFIWX; + bool HasLFIWAX; + bool HasFPRND; + bool HasFPCVT; bool HasISEL; + bool HasPOPCNTD; + bool HasLDBRX; bool IsBookE; bool HasLazyResolverStubs; bool IsJITCodeModel; @@ -154,11 +161,21 @@ public: // Specific obvious features. bool hasFSQRT() const { return HasFSQRT; } + bool hasFRE() const { return HasFRE; } + bool hasFRES() const { return HasFRES; } + bool hasFRSQRTE() const { return HasFRSQRTE; } + bool hasFRSQRTES() const { return HasFRSQRTES; } + bool hasRecipPrec() const { return HasRecipPrec; } bool hasSTFIWX() const { return HasSTFIWX; } + bool hasLFIWAX() const { return HasLFIWAX; } + bool hasFPRND() const { return HasFPRND; } + bool hasFPCVT() const { return HasFPCVT; } bool hasAltivec() const { return HasAltivec; } bool hasQPX() const { return HasQPX; } bool hasMFOCRF() const { return HasMFOCRF; } bool hasISEL() const { return HasISEL; } + bool hasPOPCNTD() const { return HasPOPCNTD; } + bool hasLDBRX() const { return HasLDBRX; } bool isBookE() const { return IsBookE; } const Triple &getTargetTriple() const { return TargetTriple; } diff --git a/lib/Target/PowerPC/PPCTargetMachine.cpp b/lib/Target/PowerPC/PPCTargetMachine.cpp index fe851c1..14dc794 100644 --- a/lib/Target/PowerPC/PPCTargetMachine.cpp +++ b/lib/Target/PowerPC/PPCTargetMachine.cpp @@ -86,8 +86,14 @@ public: return getTM<PPCTargetMachine>(); } + const PPCSubtarget &getPPCSubtarget() const { + return *getPPCTargetMachine().getSubtargetImpl(); + } + virtual bool addPreRegAlloc(); + virtual bool addILPOpts(); virtual bool addInstSelector(); + virtual bool addPreSched2(); virtual bool addPreEmitPass(); }; } // namespace @@ -103,13 +109,31 @@ bool PPCPassConfig::addPreRegAlloc() { return false; } +bool PPCPassConfig::addILPOpts() { + if (getPPCSubtarget().hasISEL()) { + addPass(&EarlyIfConverterID); + return true; + } + + return false; +} + bool PPCPassConfig::addInstSelector() { // Install an instruction selector. addPass(createPPCISelDag(getPPCTargetMachine())); return false; } +bool PPCPassConfig::addPreSched2() { + if (getOptLevel() != CodeGenOpt::None) + addPass(&IfConverterID); + + return true; +} + bool PPCPassConfig::addPreEmitPass() { + if (getOptLevel() != CodeGenOpt::None) + addPass(createPPCEarlyReturnPass()); // Must run branch selection immediately preceding the asm printer. addPass(createPPCBranchSelectionPass()); return false; diff --git a/lib/Target/PowerPC/PPCTargetTransformInfo.cpp b/lib/Target/PowerPC/PPCTargetTransformInfo.cpp index 5e9ad34..2504ba7 100644 --- a/lib/Target/PowerPC/PPCTargetTransformInfo.cpp +++ b/lib/Target/PowerPC/PPCTargetTransformInfo.cpp @@ -86,7 +86,9 @@ public: virtual unsigned getNumberOfRegisters(bool Vector) const; virtual unsigned getRegisterBitWidth(bool Vector) const; virtual unsigned getMaximumUnrollFactor() const; - virtual unsigned getArithmeticInstrCost(unsigned Opcode, Type *Ty) const; + virtual unsigned getArithmeticInstrCost(unsigned Opcode, Type *Ty, + OperandValueKind, + OperandValueKind) const; virtual unsigned getShuffleCost(ShuffleKind Kind, Type *Tp, int Index, Type *SubTp) const; virtual unsigned getCastInstrCost(unsigned Opcode, Type *Dst, @@ -122,9 +124,8 @@ llvm::createPPCTargetTransformInfoPass(const PPCTargetMachine *TM) { PPCTTI::PopcntSupportKind PPCTTI::getPopcntSupport(unsigned TyWidth) const { assert(isPowerOf2_32(TyWidth) && "Ty width must be power of 2"); - // FIXME: PPC currently does not have custom popcnt lowering even though - // there is hardware support. Once this is fixed, update this function - // to reflect the real capabilities of the hardware. + if (ST->hasPOPCNTD() && TyWidth <= 64) + return PSK_FastHardware; return PSK_Software; } @@ -167,11 +168,14 @@ unsigned PPCTTI::getMaximumUnrollFactor() const { return 2; } -unsigned PPCTTI::getArithmeticInstrCost(unsigned Opcode, Type *Ty) const { +unsigned PPCTTI::getArithmeticInstrCost(unsigned Opcode, Type *Ty, + OperandValueKind Op1Info, + OperandValueKind Op2Info) const { assert(TLI->InstructionOpcodeToISD(Opcode) && "Invalid opcode"); // Fallback to the default implementation. - return TargetTransformInfo::getArithmeticInstrCost(Opcode, Ty); + return TargetTransformInfo::getArithmeticInstrCost(Opcode, Ty, Op1Info, + Op2Info); } unsigned PPCTTI::getShuffleCost(ShuffleKind Kind, Type *Tp, int Index, diff --git a/lib/Target/PowerPC/README.txt b/lib/Target/PowerPC/README.txt index b6763aa..514f840 100644 --- a/lib/Target/PowerPC/README.txt +++ b/lib/Target/PowerPC/README.txt @@ -1,7 +1,6 @@ //===- README.txt - Notes for improving PowerPC-specific code gen ---------===// TODO: -* gpr0 allocation * lmw/stmw pass a la arm load store optimizer for prolog/epilog ===-------------------------------------------------------------------------=== @@ -127,25 +126,6 @@ produced this with bdnz, the loop would be a single dispatch group. ===-------------------------------------------------------------------------=== -Compile: - -void foo(int *P) { - if (P) *P = 0; -} - -into: - -_foo: - cmpwi cr0,r3,0 - beqlr cr0 - li r0,0 - stw r0,0(r3) - blr - -This is effectively a simple form of predication. - -===-------------------------------------------------------------------------=== - Lump the constant pool for each function into ONE pic object, and reference pieces of it as offsets from the start. For functions like this (contrived to have lots of constants obviously): @@ -204,12 +184,6 @@ http://gcc.gnu.org/ml/gcc-patches/2006-02/msg00133.html ===-------------------------------------------------------------------------=== -Implement Newton-Rhapson method for improving estimate instructions to the -correct accuracy, and implementing divide as multiply by reciprocal when it has -more than one use. Itanium would want this too. - -===-------------------------------------------------------------------------=== - Compile offsets from allocas: int *%test() { @@ -536,20 +510,6 @@ void func(unsigned int *ret, float dx, float dy, float dz, float dw) { ===-------------------------------------------------------------------------=== -Complete the signed i32 to FP conversion code using 64-bit registers -transformation, good for PI. See PPCISelLowering.cpp, this comment: - - // FIXME: disable this lowered code. This generates 64-bit register values, - // and we don't model the fact that the top part is clobbered by calls. We - // need to flag these together so that the value isn't live across a call. - //setOperationAction(ISD::SINT_TO_FP, MVT::i32, Custom); - -Also, if the registers are spilled to the stack, we have to ensure that all -64-bits of them are save/restored, otherwise we will miscompile the code. It -sounds like we need to get the 64-bit register classes going. - -===-------------------------------------------------------------------------=== - %struct.B = type { i8, [3 x i8] } define void @bar(%struct.B* %b) { |