#include "llvm/ADT/STLExtras.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/Format.h"
-#include "llvm/TableGen/TableGenBackend.h"
#include <map>
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.
///
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:
case IC_EVEX_XD_K:
return inheritsFrom(child, IC_EVEX_W_XD_K) ||
inheritsFrom(child, IC_EVEX_L_W_XD_K);
+ 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_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_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_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_XD_KZ:
+ 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_XS_K:
case IC_EVEX_L_W_XD_K:
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_XS_KZ:
+ case IC_EVEX_L_W_XD_KZ:
+ 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_B:
+ case IC_EVEX_L2_XD_B:
case IC_EVEX_L2_XD_K:
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_XD_KZ:
+ 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_XS_K:
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_XD_KZ:
+ 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");
}
}
#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
};
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
- }
-}
-
/// getDecisionType - Determines whether a ModRM decision with 255 entries can
/// be compacted by eliminating redundant information.
///
#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;
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) {
}
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
o << "IC";
- if (index < 255)
+ if (index < tableSize - 1)
o << ",";
else
o << " ";
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], THREEBYTEA6_STR);
- emitContextDecision(o1, o2, i1, i2, ModRMTableNum, *Tables[5], THREEBYTEA7_STR);
- emitContextDecision(o1, o2, i1, i2, ModRMTableNum, *Tables[6], XOP8_MAP_STR);
- emitContextDecision(o1, o2, i1, i2, ModRMTableNum, *Tables[7], XOP9_MAP_STR);
- emitContextDecision(o1, o2, i1, i2, ModRMTableNum, *Tables[8], XOPA_MAP_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 {
InstructionSpecifier &previousInfo =
InstructionSpecifiers[decision.instructionIDs[index]];
- if(newInfo.filtered)
- continue; // filtered instructions get lowest priority
+ // Instructions such as MOV8ao8 and MOV8ao8_16 differ only in the
+ // presence of the AdSize prefix. However, the disassembler doesn't
+ // care about that difference in the instruction definition; it
+ // handles 16-bit vs. 32-bit addressing for itself based purely
+ // on the 0x67 prefix and the CPU mode. So there's no need to
+ // disambiguate between them; just let them conflict/coexist.
+ if (previousInfo.name + "_16" == newInfo.name)
+ continue;
if(previousInfo.name == "NOOP" && (newInfo.name == "XCHG16ar" ||
newInfo.name == "XCHG32ar" ||
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";