X86RecognizableInstr.cpp

//===- X86RecognizableInstr.cpp - Disassembler instruction spec --*- C++ -*-===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file is part of the X86 Disassembler Emitter.
// It contains the implementation of a single recognizable instruction.
// Documentation for the disassembler emitter in general can be found in
//  X86DisasemblerEmitter.h.
//
//===----------------------------------------------------------------------===//

#include "X86DisassemblerShared.h"
#include "X86RecognizableInstr.h"
#include "X86ModRMFilters.h"

#include "llvm/Support/ErrorHandling.h"

#include <string>

using namespace llvm;

#define MRM_MAPPING     \
  MAP(C1, 33)           \
  MAP(C2, 34)           \
  MAP(C3, 35)           \
  MAP(C4, 36)           \
  MAP(C8, 37)           \
  MAP(C9, 38)           \
  MAP(E8, 39)           \
  MAP(F0, 40)           \
  MAP(F8, 41)           \
  MAP(F9, 42)           \
  MAP(D0, 45)           \
  MAP(D1, 46)

// A clone of X86 since we can't depend on something that is generated.
namespace X86Local {
  enum {
    Pseudo      = 0,
    RawFrm      = 1,
    AddRegFrm   = 2,
    MRMDestReg  = 3,
    MRMDestMem  = 4,
    MRMSrcReg   = 5,
    MRMSrcMem   = 6,
    MRM0r = 16, MRM1r = 17, MRM2r = 18, MRM3r = 19, 
    MRM4r = 20, MRM5r = 21, MRM6r = 22, MRM7r = 23,
    MRM0m = 24, MRM1m = 25, MRM2m = 26, MRM3m = 27,
    MRM4m = 28, MRM5m = 29, MRM6m = 30, MRM7m = 31,
    MRMInitReg  = 32,
#define MAP(from, to) MRM_##from = to,
    MRM_MAPPING
#undef MAP
    RawFrmImm8  = 43,
    RawFrmImm16 = 44,
    lastMRM
  };
  
  enum {
    TB  = 1,
    REP = 2,
    D8 = 3, D9 = 4, DA = 5, DB = 6,
    DC = 7, DD = 8, DE = 9, DF = 10,
    XD = 11,  XS = 12,
    T8 = 13,  P_TA = 14,
    A6 = 15,  A7 = 16, TF = 17
  };
}

// If rows are added to the opcode extension tables, then corresponding entries
// must be added here.  
//
// If the row corresponds to a single byte (i.e., 8f), then add an entry for
// that byte to ONE_BYTE_EXTENSION_TABLES.
//
// If the row corresponds to two bytes where the first is 0f, add an entry for 
// the second byte to TWO_BYTE_EXTENSION_TABLES.
//
// If the row corresponds to some other set of bytes, you will need to modify
// the code in RecognizableInstr::emitDecodePath() as well, and add new prefixes
// to the X86 TD files, except in two cases: if the first two bytes of such a 
// new combination are 0f 38 or 0f 3a, you just have to add maps called
// THREE_BYTE_38_EXTENSION_TABLES and THREE_BYTE_3A_EXTENSION_TABLES and add a
// switch(Opcode) just below the case X86Local::T8: or case X86Local::TA: line
// in RecognizableInstr::emitDecodePath().

#define ONE_BYTE_EXTENSION_TABLES \
  EXTENSION_TABLE(80)             \
  EXTENSION_TABLE(81)             \
  EXTENSION_TABLE(82)             \
  EXTENSION_TABLE(83)             \
  EXTENSION_TABLE(8f)             \
  EXTENSION_TABLE(c0)             \
  EXTENSION_TABLE(c1)             \
  EXTENSION_TABLE(c6)             \
  EXTENSION_TABLE(c7)             \
  EXTENSION_TABLE(d0)             \
  EXTENSION_TABLE(d1)             \
  EXTENSION_TABLE(d2)             \
  EXTENSION_TABLE(d3)             \
  EXTENSION_TABLE(f6)             \
  EXTENSION_TABLE(f7)             \
  EXTENSION_TABLE(fe)             \
  EXTENSION_TABLE(ff)
  
#define TWO_BYTE_EXTENSION_TABLES \
  EXTENSION_TABLE(00)             \
  EXTENSION_TABLE(01)             \
  EXTENSION_TABLE(18)             \
  EXTENSION_TABLE(71)             \
  EXTENSION_TABLE(72)             \
  EXTENSION_TABLE(73)             \
  EXTENSION_TABLE(ae)             \
  EXTENSION_TABLE(ba)             \
  EXTENSION_TABLE(c7)

using namespace X86Disassembler;

/// needsModRMForDecode - Indicates whether a particular instruction requires a
///   ModR/M byte for the instruction to be properly decoded.  For example, a 
///   MRMDestReg instruction needs the Mod field in the ModR/M byte to be set to
///   0b11.
///
/// @param form - The form of the instruction.
/// @return     - true if the form implies that a ModR/M byte is required, false
///               otherwise.
static bool needsModRMForDecode(uint8_t form) {
  if (form == X86Local::MRMDestReg    ||
     form == X86Local::MRMDestMem    ||
     form == X86Local::MRMSrcReg     ||
     form == X86Local::MRMSrcMem     ||
     (form >= X86Local::MRM0r && form <= X86Local::MRM7r) ||
     (form >= X86Local::MRM0m && form <= X86Local::MRM7m))
    return true;
  else
    return false;
}

/// isRegFormat - Indicates whether a particular form requires the Mod field of
///   the ModR/M byte to be 0b11.
///
/// @param form - The form of the instruction.
/// @return     - true if the form implies that Mod must be 0b11, false
///               otherwise.
static bool isRegFormat(uint8_t form) {
  if (form == X86Local::MRMDestReg ||
     form == X86Local::MRMSrcReg  ||
     (form >= X86Local::MRM0r && form <= X86Local::MRM7r))
    return true;
  else
    return false;
}

/// byteFromBitsInit - Extracts a value at most 8 bits in width from a BitsInit.
///   Useful for switch statements and the like.
///
/// @param init - A reference to the BitsInit to be decoded.
/// @return     - The field, with the first bit in the BitsInit as the lowest
///               order bit.
static uint8_t byteFromBitsInit(BitsInit &init) {
  int width = init.getNumBits();

  assert(width <= 8 && "Field is too large for uint8_t!");

  int     index;
  uint8_t mask = 0x01;

  uint8_t ret = 0;

  for (index = 0; index < width; index++) {
    if (static_cast<BitInit*>(init.getBit(index))->getValue())
      ret |= mask;

    mask <<= 1;
  }

  return ret;
}

/// byteFromRec - Extract a value at most 8 bits in with from a Record given the
///   name of the field.
///
/// @param rec  - The record from which to extract the value.
/// @param name - The name of the field in the record.
/// @return     - The field, as translated by byteFromBitsInit().
static uint8_t byteFromRec(const Record* rec, const std::string &name) {
  BitsInit* bits = rec->getValueAsBitsInit(name);
  return byteFromBitsInit(*bits);
}

RecognizableInstr::RecognizableInstr(DisassemblerTables &tables,
                                     const CodeGenInstruction &insn,
                                     InstrUID uid) {
  UID = uid;