X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=utils%2FTableGen%2FX86DisassemblerTables.cpp;h=ad36dc427a562a1c954c8677aea38cfb00f5a435;hb=c07a65e35ab86532fb638c3b1c0474a3b97d2d11;hp=0838ac412ee498b7bd0199a932b5686dc24fc256;hpb=c18f4efc5dd24adcc653806455fc7ae8508e9c66;p=oota-llvm.git diff --git a/utils/TableGen/X86DisassemblerTables.cpp b/utils/TableGen/X86DisassemblerTables.cpp index 0838ac412ee..ad36dc427a5 100644 --- a/utils/TableGen/X86DisassemblerTables.cpp +++ b/utils/TableGen/X86DisassemblerTables.cpp @@ -19,12 +19,54 @@ #include "llvm/ADT/STLExtras.h" #include "llvm/Support/ErrorHandling.h" #include "llvm/Support/Format.h" -#include "llvm/TableGen/TableGenBackend.h" #include using namespace llvm; using namespace X86Disassembler; +/// stringForContext - Returns a string containing the name of a particular +/// InstructionContext, usually for diagnostic purposes. +/// +/// @param insnContext - The instruction class to transform to a string. +/// @return - A statically-allocated string constant that contains the +/// name of the instruction class. +static inline const char* stringForContext(InstructionContext insnContext) { + switch (insnContext) { + default: + llvm_unreachable("Unhandled instruction class"); +#define ENUM_ENTRY(n, r, d) case n: return #n; break; +#define ENUM_ENTRY_K_B(n, r, d) ENUM_ENTRY(n, r, d) ENUM_ENTRY(n##_K_B, r, d)\ + ENUM_ENTRY(n##_KZ, r, d) ENUM_ENTRY(n##_K, r, d) ENUM_ENTRY(n##_B, r, d)\ + ENUM_ENTRY(n##_KZ_B, r, d) + INSTRUCTION_CONTEXTS +#undef ENUM_ENTRY +#undef ENUM_ENTRY_K_B + } +} + +/// stringForOperandType - Like stringForContext, but for OperandTypes. +static inline const char* stringForOperandType(OperandType type) { + switch (type) { + default: + llvm_unreachable("Unhandled type"); +#define ENUM_ENTRY(i, d) case i: return #i; + TYPES +#undef ENUM_ENTRY + } +} + +/// stringForOperandEncoding - like stringForContext, but for +/// OperandEncodings. +static inline const char* stringForOperandEncoding(OperandEncoding encoding) { + switch (encoding) { + default: + llvm_unreachable("Unhandled encoding"); +#define ENUM_ENTRY(i, d) case i: return #i; + ENCODINGS +#undef ENUM_ENTRY + } +} + /// inheritsFrom - Indicates whether all instructions in one class also belong /// to another class. /// @@ -33,13 +75,13 @@ using namespace X86Disassembler; /// @return - True if child is a subset of parent, false otherwise. static inline bool inheritsFrom(InstructionContext child, InstructionContext parent, - bool VEX_LIG = false) { + bool VEX_LIG = false, bool AdSize64 = false) { if (child == parent) return true; switch (parent) { case IC: - return(inheritsFrom(child, IC_64BIT) || + return(inheritsFrom(child, IC_64BIT, AdSize64) || inheritsFrom(child, IC_OPSIZE) || inheritsFrom(child, IC_ADSIZE) || inheritsFrom(child, IC_XD) || @@ -47,13 +89,19 @@ static inline bool inheritsFrom(InstructionContext child, case IC_64BIT: return(inheritsFrom(child, IC_64BIT_REXW) || inheritsFrom(child, IC_64BIT_OPSIZE) || - inheritsFrom(child, IC_64BIT_ADSIZE) || + (!AdSize64 && inheritsFrom(child, IC_64BIT_ADSIZE)) || inheritsFrom(child, IC_64BIT_XD) || inheritsFrom(child, IC_64BIT_XS)); case IC_OPSIZE: - return inheritsFrom(child, IC_64BIT_OPSIZE); + return inheritsFrom(child, IC_64BIT_OPSIZE) || + inheritsFrom(child, IC_OPSIZE_ADSIZE); case IC_ADSIZE: + return inheritsFrom(child, IC_OPSIZE_ADSIZE); + case IC_OPSIZE_ADSIZE: + return false; case IC_64BIT_ADSIZE: + return inheritsFrom(child, IC_64BIT_OPSIZE_ADSIZE); + case IC_64BIT_OPSIZE_ADSIZE: return false; case IC_XD: return inheritsFrom(child, IC_64BIT_XD); @@ -66,9 +114,12 @@ static inline bool inheritsFrom(InstructionContext child, case IC_64BIT_REXW: return(inheritsFrom(child, IC_64BIT_REXW_XS) || inheritsFrom(child, IC_64BIT_REXW_XD) || - inheritsFrom(child, IC_64BIT_REXW_OPSIZE)); + inheritsFrom(child, IC_64BIT_REXW_OPSIZE) || + (!AdSize64 && inheritsFrom(child, IC_64BIT_REXW_ADSIZE))); case IC_64BIT_OPSIZE: - return(inheritsFrom(child, IC_64BIT_REXW_OPSIZE)); + return inheritsFrom(child, IC_64BIT_REXW_OPSIZE) || + (!AdSize64 && inheritsFrom(child, IC_64BIT_OPSIZE_ADSIZE)) || + (!AdSize64 && inheritsFrom(child, IC_64BIT_REXW_ADSIZE)); case IC_64BIT_XD: return(inheritsFrom(child, IC_64BIT_REXW_XD)); case IC_64BIT_XS: @@ -79,33 +130,40 @@ static inline bool inheritsFrom(InstructionContext child, case IC_64BIT_REXW_XD: case IC_64BIT_REXW_XS: case IC_64BIT_REXW_OPSIZE: + case IC_64BIT_REXW_ADSIZE: return false; case IC_VEX: - return inheritsFrom(child, IC_VEX_L_W) || + return (VEX_LIG && inheritsFrom(child, IC_VEX_L_W)) || inheritsFrom(child, IC_VEX_W) || (VEX_LIG && inheritsFrom(child, IC_VEX_L)); case IC_VEX_XS: - return inheritsFrom(child, IC_VEX_L_W_XS) || + return (VEX_LIG && inheritsFrom(child, IC_VEX_L_W_XS)) || inheritsFrom(child, IC_VEX_W_XS) || (VEX_LIG && inheritsFrom(child, IC_VEX_L_XS)); case IC_VEX_XD: - return inheritsFrom(child, IC_VEX_L_W_XD) || + return (VEX_LIG && inheritsFrom(child, IC_VEX_L_W_XD)) || inheritsFrom(child, IC_VEX_W_XD) || (VEX_LIG && inheritsFrom(child, IC_VEX_L_XD)); case IC_VEX_OPSIZE: - return inheritsFrom(child, IC_VEX_L_W_OPSIZE) || + return (VEX_LIG && inheritsFrom(child, IC_VEX_L_W_OPSIZE)) || inheritsFrom(child, IC_VEX_W_OPSIZE) || (VEX_LIG && inheritsFrom(child, IC_VEX_L_OPSIZE)); case IC_VEX_W: + return VEX_LIG && inheritsFrom(child, IC_VEX_L_W); case IC_VEX_W_XS: + return VEX_LIG && inheritsFrom(child, IC_VEX_L_W_XS); case IC_VEX_W_XD: + return VEX_LIG && inheritsFrom(child, IC_VEX_L_W_XD); case IC_VEX_W_OPSIZE: - return false; + return VEX_LIG && inheritsFrom(child, IC_VEX_L_W_OPSIZE); case IC_VEX_L: + return inheritsFrom(child, IC_VEX_L_W); case IC_VEX_L_XS: + return inheritsFrom(child, IC_VEX_L_W_XS); case IC_VEX_L_XD: + return inheritsFrom(child, IC_VEX_L_W_XD); case IC_VEX_L_OPSIZE: - return false; + return inheritsFrom(child, IC_VEX_L_W_OPSIZE); case IC_VEX_L_W: case IC_VEX_L_W_XS: case IC_VEX_L_W_XD: @@ -122,13 +180,18 @@ static inline bool inheritsFrom(InstructionContext child, inheritsFrom(child, IC_EVEX_L_W_XD); case IC_EVEX_OPSIZE: return inheritsFrom(child, IC_EVEX_W_OPSIZE) || - inheritsFrom(child, IC_EVEX_W_OPSIZE); + inheritsFrom(child, IC_EVEX_L_W_OPSIZE); + case IC_EVEX_B: + return false; case IC_EVEX_W: case IC_EVEX_W_XS: case IC_EVEX_W_XD: case IC_EVEX_W_OPSIZE: return false; case IC_EVEX_L: + case IC_EVEX_L_K_B: + case IC_EVEX_L_KZ_B: + case IC_EVEX_L_B: case IC_EVEX_L_XS: case IC_EVEX_L_XD: case IC_EVEX_L_OPSIZE: @@ -152,43 +215,140 @@ static inline bool inheritsFrom(InstructionContext child, return inheritsFrom(child, IC_EVEX_W_K) || inheritsFrom(child, IC_EVEX_L_W_K); case IC_EVEX_XS_K: + case IC_EVEX_XS_K_B: + case IC_EVEX_XS_KZ_B: return inheritsFrom(child, IC_EVEX_W_XS_K) || inheritsFrom(child, IC_EVEX_L_W_XS_K); case IC_EVEX_XD_K: + case IC_EVEX_XD_K_B: + case IC_EVEX_XD_KZ_B: return inheritsFrom(child, IC_EVEX_W_XD_K) || inheritsFrom(child, IC_EVEX_L_W_XD_K); + case IC_EVEX_XS_B: + case IC_EVEX_XD_B: + case IC_EVEX_K_B: + case IC_EVEX_KZ: + return false; + case IC_EVEX_XS_KZ: + return inheritsFrom(child, IC_EVEX_W_XS_KZ) || + inheritsFrom(child, IC_EVEX_L_W_XS_KZ); + case IC_EVEX_XD_KZ: + return inheritsFrom(child, IC_EVEX_W_XD_KZ) || + inheritsFrom(child, IC_EVEX_L_W_XD_KZ); + case IC_EVEX_KZ_B: case IC_EVEX_OPSIZE_K: - return inheritsFrom(child, IC_EVEX_W_OPSIZE_K) || - inheritsFrom(child, IC_EVEX_W_OPSIZE_K); + case IC_EVEX_OPSIZE_B: + case IC_EVEX_OPSIZE_K_B: + case IC_EVEX_OPSIZE_KZ: + case IC_EVEX_OPSIZE_KZ_B: + return false; case IC_EVEX_W_K: + case IC_EVEX_W_B: + case IC_EVEX_W_K_B: + case IC_EVEX_W_KZ_B: case IC_EVEX_W_XS_K: case IC_EVEX_W_XD_K: case IC_EVEX_W_OPSIZE_K: + case IC_EVEX_W_OPSIZE_B: + case IC_EVEX_W_OPSIZE_K_B: return false; case IC_EVEX_L_K: case IC_EVEX_L_XS_K: case IC_EVEX_L_XD_K: + case IC_EVEX_L_XD_B: + case IC_EVEX_L_XD_K_B: case IC_EVEX_L_OPSIZE_K: + case IC_EVEX_L_OPSIZE_B: + case IC_EVEX_L_OPSIZE_K_B: + return false; + case IC_EVEX_W_KZ: + case IC_EVEX_W_XS_KZ: + case IC_EVEX_W_XD_KZ: + case IC_EVEX_W_XS_B: + case IC_EVEX_W_XD_B: + case IC_EVEX_W_XS_K_B: + case IC_EVEX_W_XD_K_B: + case IC_EVEX_W_XS_KZ_B: + case IC_EVEX_W_XD_KZ_B: + case IC_EVEX_W_OPSIZE_KZ: + case IC_EVEX_W_OPSIZE_KZ_B: + return false; + case IC_EVEX_L_KZ: + case IC_EVEX_L_XS_KZ: + case IC_EVEX_L_XS_B: + case IC_EVEX_L_XS_K_B: + case IC_EVEX_L_XS_KZ_B: + case IC_EVEX_L_XD_KZ: + case IC_EVEX_L_XD_KZ_B: + case IC_EVEX_L_OPSIZE_KZ: + case IC_EVEX_L_OPSIZE_KZ_B: return false; case IC_EVEX_L_W_K: + case IC_EVEX_L_W_B: + case IC_EVEX_L_W_K_B: case IC_EVEX_L_W_XS_K: - case IC_EVEX_L_W_XD_K: + case IC_EVEX_L_W_XS_B: + case IC_EVEX_L_W_XS_K_B: + case IC_EVEX_L_W_XS_KZ: + case IC_EVEX_L_W_XS_KZ_B: case IC_EVEX_L_W_OPSIZE_K: + case IC_EVEX_L_W_OPSIZE_B: + case IC_EVEX_L_W_OPSIZE_K_B: + case IC_EVEX_L_W_KZ: + case IC_EVEX_L_W_KZ_B: + case IC_EVEX_L_W_XD_K: + case IC_EVEX_L_W_XD_B: + case IC_EVEX_L_W_XD_K_B: + case IC_EVEX_L_W_XD_KZ: + case IC_EVEX_L_W_XD_KZ_B: + case IC_EVEX_L_W_OPSIZE_KZ: + case IC_EVEX_L_W_OPSIZE_KZ_B: return false; case IC_EVEX_L2_K: case IC_EVEX_L2_B: + case IC_EVEX_L2_K_B: + case IC_EVEX_L2_KZ_B: case IC_EVEX_L2_XS_K: + case IC_EVEX_L2_XS_K_B: + case IC_EVEX_L2_XS_B: + case IC_EVEX_L2_XD_B: case IC_EVEX_L2_XD_K: + case IC_EVEX_L2_XD_K_B: case IC_EVEX_L2_OPSIZE_K: case IC_EVEX_L2_OPSIZE_B: + case IC_EVEX_L2_OPSIZE_K_B: + case IC_EVEX_L2_KZ: + case IC_EVEX_L2_XS_KZ: + case IC_EVEX_L2_XS_KZ_B: + case IC_EVEX_L2_XD_KZ: + case IC_EVEX_L2_XD_KZ_B: + case IC_EVEX_L2_OPSIZE_KZ: + case IC_EVEX_L2_OPSIZE_KZ_B: return false; case IC_EVEX_L2_W_K: + case IC_EVEX_L2_W_B: + case IC_EVEX_L2_W_K_B: + case IC_EVEX_L2_W_KZ_B: case IC_EVEX_L2_W_XS_K: + case IC_EVEX_L2_W_XS_B: + case IC_EVEX_L2_W_XS_K_B: case IC_EVEX_L2_W_XD_K: + case IC_EVEX_L2_W_XD_B: case IC_EVEX_L2_W_OPSIZE_K: case IC_EVEX_L2_W_OPSIZE_B: + case IC_EVEX_L2_W_OPSIZE_K_B: + case IC_EVEX_L2_W_KZ: + case IC_EVEX_L2_W_XS_KZ: + case IC_EVEX_L2_W_XS_KZ_B: + case IC_EVEX_L2_W_XD_KZ: + case IC_EVEX_L2_W_XD_K_B: + case IC_EVEX_L2_W_XD_KZ_B: + case IC_EVEX_L2_W_OPSIZE_KZ: + case IC_EVEX_L2_W_OPSIZE_KZ_B: return false; default: + errs() << "Unknown instruction class: " << + stringForContext((InstructionContext)parent) << "\n"; llvm_unreachable("Unknown instruction class"); } } @@ -207,7 +367,8 @@ static inline bool outranks(InstructionContext upper, #define ENUM_ENTRY(n, r, d) r, #define ENUM_ENTRY_K_B(n, r, d) ENUM_ENTRY(n, r, d) \ - ENUM_ENTRY(n##_K_B, r, d) ENUM_ENTRY(n##_K, r, d) ENUM_ENTRY(n##_B, r, d) + ENUM_ENTRY(n##_K_B, r, d) ENUM_ENTRY(n##_KZ_B, r, d) \ + ENUM_ENTRY(n##_KZ, r, d) ENUM_ENTRY(n##_K, r, d) ENUM_ENTRY(n##_B, r, d) static int ranks[IC_max] = { INSTRUCTION_CONTEXTS }; @@ -217,77 +378,6 @@ static inline bool outranks(InstructionContext upper, return (ranks[upper] > ranks[lower]); } -/// stringForContext - Returns a string containing the name of a particular -/// InstructionContext, usually for diagnostic purposes. -/// -/// @param insnContext - The instruction class to transform to a string. -/// @return - A statically-allocated string constant that contains the -/// name of the instruction class. -static inline const char* stringForContext(InstructionContext insnContext) { - switch (insnContext) { - default: - llvm_unreachable("Unhandled instruction class"); -#define ENUM_ENTRY(n, r, d) case n: return #n; break; -#define ENUM_ENTRY_K_B(n, r, d) ENUM_ENTRY(n, r, d) ENUM_ENTRY(n##_K_B, r, d)\ - ENUM_ENTRY(n##_K, r, d) ENUM_ENTRY(n##_B, r, d) - INSTRUCTION_CONTEXTS -#undef ENUM_ENTRY -#undef ENUM_ENTRY_K_B - } -} - -/// stringForOperandType - Like stringForContext, but for OperandTypes. -static inline const char* stringForOperandType(OperandType type) { - switch (type) { - default: - llvm_unreachable("Unhandled type"); -#define ENUM_ENTRY(i, d) case i: return #i; - TYPES -#undef ENUM_ENTRY - } -} - -/// stringForOperandEncoding - like stringForContext, but for -/// OperandEncodings. -static inline const char* stringForOperandEncoding(OperandEncoding encoding) { - switch (encoding) { - default: - llvm_unreachable("Unhandled encoding"); -#define ENUM_ENTRY(i, d) case i: return #i; - ENCODINGS -#undef ENUM_ENTRY - } -} - -void DisassemblerTables::emitOneID(raw_ostream &o, unsigned &i, InstrUID id, - bool addComma) const { - if (id) - o.indent(i * 2) << format("0x%hx", id); - else - o.indent(i * 2) << 0; - - if (addComma) - o << ", "; - else - o << " "; - - o << "/* "; - o << InstructionSpecifiers[id].name; - o << "*/"; - - o << "\n"; -} - -/// emitEmptyTable - Emits the modRMEmptyTable, which is used as a ID table by -/// all ModR/M decisions for instructions that are invalid for all possible -/// ModR/M byte values. -/// -/// @param o - The output stream on which to emit the table. -/// @param i - The indentation level for that output stream. -static void emitEmptyTable(raw_ostream &o, unsigned &i) { - o.indent(i * 2) << "0x0, /* EmptyTable */\n"; -} - /// getDecisionType - Determines whether a ModRM decision with 255 entries can /// be compacted by eliminating redundant information. /// @@ -351,22 +441,6 @@ static const char* stringForDecisionType(ModRMDecisionType dt) { #undef ENUM_ENTRY } -/// stringForModifierType - Returns a statically-allocated string corresponding -/// to an opcode modifier type. -/// -/// @param mt - The modifier type. -/// @return - A pointer to the statically-allocated string (e.g., -/// "MODIFIER_NONE" for MODIFIER_NONE). -static const char* stringForModifierType(ModifierType mt) { -#define ENUM_ENTRY(n) case n: return #n; - switch(mt) { - default: - llvm_unreachable("Unknown modifier type"); - MODIFIER_TYPES - }; -#undef ENUM_ENTRY -} - DisassemblerTables::DisassemblerTables() { unsigned i; @@ -387,6 +461,7 @@ DisassemblerTables::~DisassemblerTables() { void DisassemblerTables::emitModRMDecision(raw_ostream &o1, raw_ostream &o2, unsigned &i1, unsigned &i2, + unsigned &ModRMTableNum, ModRMDecision &decision) const { static uint32_t sTableNumber = 0; static uint32_t sEntryNumber = 1; @@ -405,44 +480,56 @@ void DisassemblerTables::emitModRMDecision(raw_ostream &o1, raw_ostream &o2, return; } - o1 << "/* Table" << sTableNumber << " */\n"; - i1++; + std::vector ModRMDecision; switch (dt) { default: llvm_unreachable("Unknown decision type"); case MODRM_ONEENTRY: - emitOneID(o1, i1, decision.instructionIDs[0], true); + ModRMDecision.push_back(decision.instructionIDs[0]); break; case MODRM_SPLITRM: - emitOneID(o1, i1, decision.instructionIDs[0x00], true); // mod = 0b00 - emitOneID(o1, i1, decision.instructionIDs[0xc0], true); // mod = 0b11 + ModRMDecision.push_back(decision.instructionIDs[0x00]); + ModRMDecision.push_back(decision.instructionIDs[0xc0]); break; case MODRM_SPLITREG: for (unsigned index = 0; index < 64; index += 8) - emitOneID(o1, i1, decision.instructionIDs[index], true); + ModRMDecision.push_back(decision.instructionIDs[index]); for (unsigned index = 0xc0; index < 256; index += 8) - emitOneID(o1, i1, decision.instructionIDs[index], true); + ModRMDecision.push_back(decision.instructionIDs[index]); break; case MODRM_SPLITMISC: for (unsigned index = 0; index < 64; index += 8) - emitOneID(o1, i1, decision.instructionIDs[index], true); + ModRMDecision.push_back(decision.instructionIDs[index]); for (unsigned index = 0xc0; index < 256; ++index) - emitOneID(o1, i1, decision.instructionIDs[index], true); + ModRMDecision.push_back(decision.instructionIDs[index]); break; case MODRM_FULL: for (unsigned index = 0; index < 256; ++index) - emitOneID(o1, i1, decision.instructionIDs[index], true); + ModRMDecision.push_back(decision.instructionIDs[index]); break; } - i1--; + unsigned &EntryNumber = ModRMTable[ModRMDecision]; + if (EntryNumber == 0) { + EntryNumber = ModRMTableNum; + + ModRMTableNum += ModRMDecision.size(); + o1 << "/* Table" << EntryNumber << " */\n"; + i1++; + for (std::vector::const_iterator I = ModRMDecision.begin(), + E = ModRMDecision.end(); I != E; ++I) { + o1.indent(i1 * 2) << format("0x%hx", *I) << ", /* " + << InstructionSpecifiers[*I].name << " */\n"; + } + i1--; + } o2.indent(i2) << "{ /* struct ModRMDecision */" << "\n"; i2++; o2.indent(i2) << stringForDecisionType(dt) << "," << "\n"; - o2.indent(i2) << sEntryNumber << " /* Table" << sTableNumber << " */\n"; + o2.indent(i2) << EntryNumber << " /* Table" << EntryNumber << " */\n"; i2--; o2.indent(i2) << "}"; @@ -476,6 +563,7 @@ void DisassemblerTables::emitModRMDecision(raw_ostream &o1, raw_ostream &o2, void DisassemblerTables::emitOpcodeDecision(raw_ostream &o1, raw_ostream &o2, unsigned &i1, unsigned &i2, + unsigned &ModRMTableNum, OpcodeDecision &decision) const { o2.indent(i2) << "{ /* struct OpcodeDecision */" << "\n"; i2++; @@ -487,7 +575,8 @@ void DisassemblerTables::emitOpcodeDecision(raw_ostream &o1, raw_ostream &o2, o2 << "/* 0x" << format("%02hhx", index) << " */" << "\n"; - emitModRMDecision(o1, o2, i1, i2, decision.modRMDecisions[index]); + emitModRMDecision(o1, o2, i1, i2, ModRMTableNum, + decision.modRMDecisions[index]); if (index < 255) o2 << ","; @@ -503,6 +592,7 @@ void DisassemblerTables::emitOpcodeDecision(raw_ostream &o1, raw_ostream &o2, void DisassemblerTables::emitContextDecision(raw_ostream &o1, raw_ostream &o2, unsigned &i1, unsigned &i2, + unsigned &ModRMTableNum, ContextDecision &decision, const char* name) const { o2.indent(i2) << "static const struct ContextDecision " << name << " = {\n"; @@ -516,7 +606,8 @@ void DisassemblerTables::emitContextDecision(raw_ostream &o1, raw_ostream &o2, o2 << " */"; o2 << "\n"; - emitOpcodeDecision(o1, o2, i1, i2, decision.opcodeDecisions[index]); + emitOpcodeDecision(o1, o2, i1, i2, ModRMTableNum, + decision.opcodeDecisions[index]); if (index + 1 < IC_max) o2 << ", "; @@ -535,7 +626,8 @@ void DisassemblerTables::emitInstructionInfo(raw_ostream &o, o << "static const struct OperandSpecifier x86OperandSets[][" << X86_MAX_OPERANDS << "] = {\n"; - typedef std::vector > OperandListTy; + typedef SmallVector, + X86_MAX_OPERANDS> OperandListTy; std::map OperandSets; unsigned OperandSetNum = 0; @@ -544,12 +636,10 @@ void DisassemblerTables::emitInstructionInfo(raw_ostream &o, for (unsigned OperandIndex = 0; OperandIndex < X86_MAX_OPERANDS; ++OperandIndex) { - const char *Encoding = - stringForOperandEncoding((OperandEncoding)InstructionSpecifiers[Index] - .operands[OperandIndex].encoding); - const char *Type = - stringForOperandType((OperandType)InstructionSpecifiers[Index] - .operands[OperandIndex].type); + OperandEncoding Encoding = (OperandEncoding)InstructionSpecifiers[Index] + .operands[OperandIndex].encoding; + OperandType Type = (OperandType)InstructionSpecifiers[Index] + .operands[OperandIndex].type; OperandList.push_back(std::make_pair(Encoding, Type)); } unsigned &N = OperandSets[OperandList]; @@ -559,8 +649,9 @@ void DisassemblerTables::emitInstructionInfo(raw_ostream &o, o << " { /* " << (OperandSetNum - 1) << " */\n"; for (unsigned i = 0, e = OperandList.size(); i != e; ++i) { - o << " { " << OperandList[i].first << ", " - << OperandList[i].second << " },\n"; + const char *Encoding = stringForOperandEncoding(OperandList[i].first); + const char *Type = stringForOperandType(OperandList[i].second); + o << " { " << Encoding << ", " << Type << " },\n"; } o << " },\n"; } @@ -572,40 +663,24 @@ void DisassemblerTables::emitInstructionInfo(raw_ostream &o, i++; for (unsigned index = 0; index < NumInstructions; ++index) { - o.indent(i * 2) << "{ /* " << index << " */" << "\n"; + o.indent(i * 2) << "{ /* " << index << " */\n"; i++; - o.indent(i * 2) << stringForModifierType( - (ModifierType)InstructionSpecifiers[index].modifierType); - o << ",\n"; - - o.indent(i * 2) << "0x"; - o << format("%02hhx", (uint16_t)InstructionSpecifiers[index].modifierBase); - o << ",\n"; - OperandListTy OperandList; for (unsigned OperandIndex = 0; OperandIndex < X86_MAX_OPERANDS; ++OperandIndex) { - const char *Encoding = - stringForOperandEncoding((OperandEncoding)InstructionSpecifiers[index] - .operands[OperandIndex].encoding); - const char *Type = - stringForOperandType((OperandType)InstructionSpecifiers[index] - .operands[OperandIndex].type); + OperandEncoding Encoding = (OperandEncoding)InstructionSpecifiers[index] + .operands[OperandIndex].encoding; + OperandType Type = (OperandType)InstructionSpecifiers[index] + .operands[OperandIndex].type; OperandList.push_back(std::make_pair(Encoding, Type)); } o.indent(i * 2) << (OperandSets[OperandList] - 1) << ",\n"; - o.indent(i * 2) << "/* " << InstructionSpecifiers[index].name << " */"; - o << "\n"; + o.indent(i * 2) << "/* " << InstructionSpecifiers[index].name << " */\n"; i--; - o.indent(i * 2) << "}"; - - if (index + 1 < NumInstructions) - o << ","; - - o << "\n"; + o.indent(i * 2) << "},\n"; } i--; @@ -613,15 +688,43 @@ void DisassemblerTables::emitInstructionInfo(raw_ostream &o, } void DisassemblerTables::emitContextTable(raw_ostream &o, unsigned &i) const { + const unsigned int tableSize = 16384; o.indent(i * 2) << "static const uint8_t " CONTEXTS_STR - "[256] = {\n"; + "[" << tableSize << "] = {\n"; i++; - for (unsigned index = 0; index < 256; ++index) { + for (unsigned index = 0; index < tableSize; ++index) { o.indent(i * 2); - if ((index & ATTR_VEXL) && (index & ATTR_REXW) && (index & ATTR_OPSIZE)) + if (index & ATTR_EVEX) { + o << "IC_EVEX"; + if (index & ATTR_EVEXL2) + o << "_L2"; + else if (index & ATTR_EVEXL) + o << "_L"; + if (index & ATTR_REXW) + o << "_W"; + if (index & ATTR_OPSIZE) + o << "_OPSIZE"; + else if (index & ATTR_XD) + o << "_XD"; + else if (index & ATTR_XS) + o << "_XS"; + if (index & ATTR_EVEXKZ) + o << "_KZ"; + else if (index & ATTR_EVEXK) + o << "_K"; + if (index & ATTR_EVEXB) + o << "_B"; + } + else if ((index & ATTR_VEXL) && (index & ATTR_REXW) && (index & ATTR_OPSIZE)) o << "IC_VEX_L_W_OPSIZE"; + else if ((index & ATTR_VEXL) && (index & ATTR_REXW) && (index & ATTR_XD)) + o << "IC_VEX_L_W_XD"; + else if ((index & ATTR_VEXL) && (index & ATTR_REXW) && (index & ATTR_XS)) + o << "IC_VEX_L_W_XS"; + else if ((index & ATTR_VEXL) && (index & ATTR_REXW)) + o << "IC_VEX_L_W"; else if ((index & ATTR_VEXL) && (index & ATTR_OPSIZE)) o << "IC_VEX_L_OPSIZE"; else if ((index & ATTR_VEXL) && (index & ATTR_XD)) @@ -653,6 +756,9 @@ void DisassemblerTables::emitContextTable(raw_ostream &o, unsigned &i) const { else if ((index & ATTR_64BIT) && (index & ATTR_REXW) && (index & ATTR_OPSIZE)) o << "IC_64BIT_REXW_OPSIZE"; + else if ((index & ATTR_64BIT) && (index & ATTR_REXW) && + (index & ATTR_ADSIZE)) + o << "IC_64BIT_REXW_ADSIZE"; else if ((index & ATTR_64BIT) && (index & ATTR_XD) && (index & ATTR_OPSIZE)) o << "IC_64BIT_XD_OPSIZE"; else if ((index & ATTR_64BIT) && (index & ATTR_XS) && (index & ATTR_OPSIZE)) @@ -661,6 +767,9 @@ void DisassemblerTables::emitContextTable(raw_ostream &o, unsigned &i) const { o << "IC_64BIT_XS"; else if ((index & ATTR_64BIT) && (index & ATTR_XD)) o << "IC_64BIT_XD"; + else if ((index & ATTR_64BIT) && (index & ATTR_OPSIZE) && + (index & ATTR_ADSIZE)) + o << "IC_64BIT_OPSIZE_ADSIZE"; else if ((index & ATTR_64BIT) && (index & ATTR_OPSIZE)) o << "IC_64BIT_OPSIZE"; else if ((index & ATTR_64BIT) && (index & ATTR_ADSIZE)) @@ -677,6 +786,8 @@ void DisassemblerTables::emitContextTable(raw_ostream &o, unsigned &i) const { o << "IC_XS"; else if (index & ATTR_XD) o << "IC_XD"; + else if ((index & ATTR_OPSIZE) && (index & ATTR_ADSIZE)) + o << "IC_OPSIZE_ADSIZE"; else if (index & ATTR_OPSIZE) o << "IC_OPSIZE"; else if (index & ATTR_ADSIZE) @@ -684,7 +795,7 @@ void DisassemblerTables::emitContextTable(raw_ostream &o, unsigned &i) const { else o << "IC"; - if (index < 255) + if (index < tableSize - 1) o << ","; else o << " "; @@ -699,13 +810,15 @@ void DisassemblerTables::emitContextTable(raw_ostream &o, unsigned &i) const { } void DisassemblerTables::emitContextDecisions(raw_ostream &o1, raw_ostream &o2, - unsigned &i1, unsigned &i2) const { - emitContextDecision(o1, o2, i1, i2, *Tables[0], ONEBYTE_STR); - emitContextDecision(o1, o2, i1, i2, *Tables[1], TWOBYTE_STR); - emitContextDecision(o1, o2, i1, i2, *Tables[2], THREEBYTE38_STR); - emitContextDecision(o1, o2, i1, i2, *Tables[3], THREEBYTE3A_STR); - emitContextDecision(o1, o2, i1, i2, *Tables[4], THREEBYTEA6_STR); - emitContextDecision(o1, o2, i1, i2, *Tables[5], THREEBYTEA7_STR); + unsigned &i1, unsigned &i2, + unsigned &ModRMTableNum) const { + emitContextDecision(o1, o2, i1, i2, ModRMTableNum, *Tables[0], ONEBYTE_STR); + emitContextDecision(o1, o2, i1, i2, ModRMTableNum, *Tables[1], TWOBYTE_STR); + emitContextDecision(o1, o2, i1, i2, ModRMTableNum, *Tables[2], THREEBYTE38_STR); + emitContextDecision(o1, o2, i1, i2, ModRMTableNum, *Tables[3], THREEBYTE3A_STR); + emitContextDecision(o1, o2, i1, i2, ModRMTableNum, *Tables[4], XOP8_MAP_STR); + emitContextDecision(o1, o2, i1, i2, ModRMTableNum, *Tables[5], XOP9_MAP_STR); + emitContextDecision(o1, o2, i1, i2, ModRMTableNum, *Tables[6], XOPA_MAP_STR); } void DisassemblerTables::emit(raw_ostream &o) const { @@ -724,11 +837,17 @@ void DisassemblerTables::emit(raw_ostream &o) const { emitContextTable(o, i2); o << "\n"; + unsigned ModRMTableNum = 0; + o << "static const InstrUID modRMTable[] = {\n"; i1++; - emitEmptyTable(o1, i1); + std::vector EmptyTable(1, 0); + ModRMTable[EmptyTable] = ModRMTableNum; + ModRMTableNum += EmptyTable.size(); + o1 << "/* EmptyTable */\n"; + o1.indent(i1 * 2) << "0x0,\n"; i1--; - emitContextDecisions(o1, o2, i1, i2); + emitContextDecisions(o1, o2, i1, i2, ModRMTableNum); o << o1.str(); o << " 0x0\n"; @@ -754,9 +873,6 @@ void DisassemblerTables::setTableFields(ModRMDecision &decision, InstructionSpecifier &previousInfo = InstructionSpecifiers[decision.instructionIDs[index]]; - if(newInfo.filtered) - continue; // filtered instructions get lowest priority - if(previousInfo.name == "NOOP" && (newInfo.name == "XCHG16ar" || newInfo.name == "XCHG32ar" || newInfo.name == "XCHG32ar64" || @@ -766,8 +882,7 @@ void DisassemblerTables::setTableFields(ModRMDecision &decision, if (outranks(previousInfo.insnContext, newInfo.insnContext)) continue; - if (previousInfo.insnContext == newInfo.insnContext && - !previousInfo.filtered) { + if (previousInfo.insnContext == newInfo.insnContext) { errs() << "Error: Primary decode conflict: "; errs() << newInfo.name << " would overwrite " << previousInfo.name; errs() << "\n"; @@ -789,15 +904,19 @@ void DisassemblerTables::setTableFields(OpcodeType type, const ModRMFilter &filter, InstrUID uid, bool is32bit, - bool ignoresVEX_L) { + bool ignoresVEX_L, + unsigned addressSize) { ContextDecision &decision = *Tables[type]; for (unsigned index = 0; index < IC_max; ++index) { - if (is32bit && inheritsFrom((InstructionContext)index, IC_64BIT)) + if ((is32bit || addressSize == 16) && + inheritsFrom((InstructionContext)index, IC_64BIT)) continue; + bool adSize64 = addressSize == 64; if (inheritsFrom((InstructionContext)index, - InstructionSpecifiers[uid].insnContext, ignoresVEX_L)) + InstructionSpecifiers[uid].insnContext, ignoresVEX_L, + adSize64)) setTableFields(decision.opcodeDecisions[index].modRMDecisions[opcode], filter, uid,