about summary refs log tree commit diff stats
path: root/miasm/arch/x86
diff options
context:
space:
mode:
Diffstat (limited to 'miasm/arch/x86')
-rw-r--r--miasm/arch/x86/__init__.py1
-rw-r--r--miasm/arch/x86/arch.py4878
-rw-r--r--miasm/arch/x86/ctype.py137
-rw-r--r--miasm/arch/x86/disasm.py30
-rw-r--r--miasm/arch/x86/jit.py296
-rw-r--r--miasm/arch/x86/lifter_model_call.py80
-rw-r--r--miasm/arch/x86/regs.py454
-rw-r--r--miasm/arch/x86/sem.py6065
8 files changed, 0 insertions, 11941 deletions
diff --git a/miasm/arch/x86/__init__.py b/miasm/arch/x86/__init__.py
deleted file mode 100644
index bbad893b..00000000
--- a/miasm/arch/x86/__init__.py
+++ /dev/null
@@ -1 +0,0 @@
-__all__ = ["arch", "disasm", "regs", "sem"]
diff --git a/miasm/arch/x86/arch.py b/miasm/arch/x86/arch.py
deleted file mode 100644
index 5464a779..00000000
--- a/miasm/arch/x86/arch.py
+++ /dev/null
@@ -1,4878 +0,0 @@
-#-*- coding:utf-8 -*-
-
-from __future__ import print_function
-from builtins import range
-import re
-
-from future.utils import viewitems
-
-from miasm.core import utils
-from miasm.expression.expression import *
-from pyparsing import *
-from miasm.core.cpu import *
-from collections import defaultdict
-import miasm.arch.x86.regs as regs_module
-from miasm.arch.x86.regs import *
-from miasm.core.asm_ast import AstNode, AstInt, AstId, AstMem, AstOp
-from miasm.ir.ir import color_expr_html
-from miasm.core.utils import BRACKET_O, BRACKET_C
-
-
-log = logging.getLogger("x86_arch")
-console_handler = logging.StreamHandler()
-console_handler.setFormatter(logging.Formatter("[%(levelname)-8s]: %(message)s"))
-log.addHandler(console_handler)
-log.setLevel(logging.WARN)
-
-conditional_branch = ["JO", "JNO", "JB", "JAE",
-                      "JZ", "JNZ", "JBE", "JA",
-                      "JS", "JNS", "JPE", "JNP",
-                      #"L", "NL", "NG", "G"]
-                      "JL", "JGE", "JLE", "JG",
-                      "JCXZ", "JECXZ", "JRCXZ"]
-
-unconditional_branch = ['JMP', 'JMPF']
-
-f_isad = "AD"
-f_s08 = "S08"
-f_u08 = "U08"
-f_s16 = "S16"
-f_u16 = "U16"
-f_s32 = "S32"
-f_u32 = "U32"
-f_s64 = "S64"
-f_u64 = "U64"
-f_imm = 'IMM'
-
-f_imm2size = {f_s08: 8, f_s16: 16, f_s32: 32, f_s64: 64,
-              f_u08: 8, f_u16: 16, f_u32: 32, f_u64: 64}
-
-
-size2gpregs = {8: gpregs08, 16: gpregs16,
-               32: gpregs32, 64: gpregs64}
-
-
-replace_regs64 = {
-    AL: RAX[:8], CL: RCX[:8], DL: RDX[:8], BL: RBX[:8],
-    AH: RAX[8:16], CH: RCX[8:16], DH: RDX[8:16], BH: RBX[8:16],
-    SPL: RSP[0:8], BPL: RBP[0:8], SIL: RSI[0:8], DIL: RDI[0:8],
-    R8B: R8[0:8], R9B: R9[0:8], R10B: R10[0:8], R11B: R11[0:8],
-    R12B: R12[0:8], R13B: R13[0:8], R14B: R14[0:8], R15B: R15[0:8],
-
-    AX: RAX[:16], CX: RCX[:16], DX: RDX[:16], BX: RBX[:16],
-    SP: RSP[:16], BP: RBP[:16], SI: RSI[:16], DI: RDI[:16],
-    R8W:  R8[:16], R9W:  R9[:16], R10W: R10[:16], R11W: R11[:16],
-    R12W: R12[:16], R13W: R13[:16], R14W: R14[:16], R15W: R15[:16],
-
-    EAX: RAX[:32], ECX: RCX[:32], EDX: RDX[:32], EBX: RBX[:32],
-    ESP: RSP[:32], EBP: RBP[:32], ESI: RSI[:32], EDI: RDI[:32],
-    R8D: R8[:32], R9D: R9[:32], R10D: R10[:32], R11D: R11[:32],
-    R12D: R12[:32], R13D: R13[:32], R14D: R14[:32], R15D: R15[:32],
-
-    IP: RIP[:16], EIP: RIP[:32],
-
-    ExprId("ST", 64): float_st0,
-    ExprId("ST(0)", 64): float_st0,
-    ExprId("ST(1)", 64): float_st1,
-    ExprId("ST(2)", 64): float_st2,
-    ExprId("ST(3)", 64): float_st3,
-    ExprId("ST(4)", 64): float_st4,
-    ExprId("ST(5)", 64): float_st5,
-    ExprId("ST(6)", 64): float_st6,
-    ExprId("ST(7)", 64): float_st7,
-
-}
-
-replace_regs32 = {
-    AL: EAX[:8],   CL: ECX[:8],   DL: EDX[:8],   BL: EBX[:8],
-    AH: EAX[8:16], CH: ECX[8:16], DH: EDX[8:16], BH: EBX[8:16],
-
-    AX: EAX[:16], CX: ECX[:16], DX: EDX[:16], BX: EBX[:16],
-    SP: ESP[:16], BP: EBP[:16], SI: ESI[:16], DI: EDI[:16],
-
-    IP: EIP[:16],
-
-
-    ExprId("ST", 64): float_st0,
-    ExprId("ST(0)", 64): float_st0,
-    ExprId("ST(1)", 64): float_st1,
-    ExprId("ST(2)", 64): float_st2,
-    ExprId("ST(3)", 64): float_st3,
-    ExprId("ST(4)", 64): float_st4,
-    ExprId("ST(5)", 64): float_st5,
-    ExprId("ST(6)", 64): float_st6,
-    ExprId("ST(7)", 64): float_st7,
-
-}
-
-replace_regs16 = {
-    AL: AX[:8],   CL: CX[:8],   DL: DX[:8],   BL: BX[:8],
-    AH: AX[8:16], CH: CX[8:16], DH: DX[8:16], BH: BX[8:16],
-
-    AX: AX[:16],  CX: CX[:16],  DX: DX[:16],  BX: BX[:16],
-    SP: SP[:16],  BP: BP[:16],  SI: SI[:16],  DI: DI[:16],
-
-
-    ExprId("ST", 64): float_st0,
-    ExprId("ST(0)", 64): float_st0,
-    ExprId("ST(1)", 64): float_st1,
-    ExprId("ST(2)", 64): float_st2,
-    ExprId("ST(3)", 64): float_st3,
-    ExprId("ST(4)", 64): float_st4,
-    ExprId("ST(5)", 64): float_st5,
-    ExprId("ST(6)", 64): float_st6,
-    ExprId("ST(7)", 64): float_st7,
-
-}
-
-replace_regs = {16: replace_regs16,
-                32: replace_regs32,
-                64: replace_regs64}
-
-
-segm2enc = {CS: 1, SS: 2, DS: 3, ES: 4, FS: 5, GS: 6}
-enc2segm = dict((value, key) for key, value in viewitems(segm2enc))
-
-segm_info = reg_info_dct(enc2segm)
-
-
-
-enc2crx = {
-    0: cr0,
-    1: cr1,
-    2: cr2,
-    3: cr3,
-    4: cr4,
-    5: cr5,
-    6: cr6,
-    7: cr7,
-}
-
-crx_info = reg_info_dct(enc2crx)
-
-
-enc2drx = {
-    0: dr0,
-    1: dr1,
-    2: dr2,
-    3: dr3,
-    4: dr4,
-    5: dr5,
-    6: dr6,
-    7: dr7,
-}
-
-drx_info = reg_info_dct(enc2drx)
-
-
-
-# parser helper ###########
-PLUS = Suppress("+")
-MULT = Suppress("*")
-
-COLON = Suppress(":")
-
-
-LBRACK = Suppress("[")
-RBRACK = Suppress("]")
-
-
-gpreg = (
-    gpregs08.parser |
-    gpregs08_64.parser |
-    gpregs16.parser |
-    gpregs32.parser |
-    gpregs64.parser |
-    gpregs_xmm.parser |
-    gpregs_mm.parser |
-    gpregs_bnd.parser
-)
-
-
-def is_op_segm(expr):
-    """Returns True if is ExprOp and op == 'segm'"""
-    return expr.is_op('segm')
-
-def is_mem_segm(expr):
-    """Returns True if is ExprMem and ptr is_op_segm"""
-    return expr.is_mem() and is_op_segm(expr.ptr)
-
-
-def cb_deref_segmoff(tokens):
-    assert len(tokens) == 2
-    return AstOp('segm', tokens[0], tokens[1])
-
-
-def cb_deref_base_expr(tokens):
-    tokens = tokens[0]
-    assert isinstance(tokens, AstNode)
-    addr = tokens
-    return addr
-
-
-deref_mem_ad = (LBRACK + base_expr + RBRACK).setParseAction(cb_deref_base_expr)
-
-deref_ptr = (base_expr + COLON + base_expr).setParseAction(cb_deref_segmoff)
-
-
-PTR = Suppress('PTR')
-
-FAR = Suppress('FAR')
-
-
-BYTE = Literal('BYTE')
-WORD = Literal('WORD')
-DWORD = Literal('DWORD')
-QWORD = Literal('QWORD')
-TBYTE = Literal('TBYTE')
-XMMWORD = Literal('XMMWORD')
-
-MEMPREFIX2SIZE = {'BYTE': 8, 'WORD': 16, 'DWORD': 32,
-                  'QWORD': 64, 'TBYTE': 80, 'XMMWORD': 128}
-
-SIZE2MEMPREFIX = dict((value, key) for key, value in viewitems(MEMPREFIX2SIZE))
-
-def cb_deref_mem(tokens):
-    if len(tokens) == 2:
-        s, ptr = tokens
-        assert isinstance(ptr, AstNode)
-        return AstMem(ptr, MEMPREFIX2SIZE[s])
-    elif len(tokens) == 3:
-        s, segm, ptr = tokens
-        return AstMem(AstOp('segm', segm, ptr), MEMPREFIX2SIZE[s])
-    raise ValueError('len(tokens) > 3')
-
-mem_size = (BYTE | DWORD | QWORD | WORD | TBYTE | XMMWORD)
-deref_mem = (mem_size + PTR + Optional((base_expr + COLON))+ deref_mem_ad).setParseAction(cb_deref_mem)
-
-
-rmarg = (
-    gpregs08.parser |
-    gpregs08_64.parser |
-    gpregs16.parser |
-    gpregs32.parser |
-    gpregs64.parser |
-    gpregs_mm.parser |
-    gpregs_xmm.parser |
-    gpregs_bnd.parser
-)
-
-rmarg |= deref_mem
-
-
-mem_far = FAR + deref_mem
-
-
-cl_or_imm = r08_ecx.parser
-cl_or_imm |= base_expr
-
-
-
-class x86_arg(m_arg):
-    def asm_ast_to_expr(self, value, loc_db, size_hint=None, fixed_size=None):
-        if size_hint is None:
-            size_hint = self.parent.mode
-        if fixed_size is None:
-            fixed_size = set()
-        if isinstance(value, AstId):
-            if value.name in all_regs_ids_byname:
-                reg = all_regs_ids_byname[value.name]
-                fixed_size.add(reg.size)
-                return reg
-            if isinstance(value.name, ExprId):
-                fixed_size.add(value.name.size)
-                return value.name
-            if value.name in MEMPREFIX2SIZE:
-                return None
-            if value.name in ["FAR"]:
-                return None
-
-            loc_key = loc_db.get_or_create_name_location(value.name)
-            return ExprLoc(loc_key, size_hint)
-        if isinstance(value, AstOp):
-            # First pass to retrieve fixed_size
-            if value.op == "segm":
-                segm = self.asm_ast_to_expr(value.args[0], loc_db)
-                ptr = self.asm_ast_to_expr(value.args[1], loc_db, None, fixed_size)
-                return ExprOp('segm', segm, ptr)
-            args = [self.asm_ast_to_expr(arg, loc_db, None, fixed_size) for arg in value.args]
-            if len(fixed_size) == 0:
-                # No fixed size
-                pass
-            elif len(fixed_size) == 1:
-                # One fixed size, regen all
-                size = list(fixed_size)[0]
-                args = [self.asm_ast_to_expr(arg, loc_db, size, fixed_size) for arg in value.args]
-            else:
-                raise ValueError("Size conflict")
-            if None in args:
-                return None
-            return ExprOp(value.op, *args)
-        if isinstance(value, AstInt):
-            if 1 << size_hint < value.value:
-                size_hint *= 2
-            return ExprInt(value.value, size_hint)
-        if isinstance(value, AstMem):
-            fixed_size.add(value.size)
-            ptr = self.asm_ast_to_expr(value.ptr, loc_db, None, set())
-            if ptr is None:
-                return None
-            return ExprMem(ptr, value.size)
-        return None
-
-class r_al(reg_noarg, x86_arg):
-    reg_info = r08_eax
-    parser = reg_info.parser
-
-
-class r_ax(reg_noarg, x86_arg):
-    reg_info = r16_eax
-    parser = reg_info.parser
-
-
-class r_dx(reg_noarg, x86_arg):
-    reg_info = r16_edx
-    parser = reg_info.parser
-
-
-class r_eax(reg_noarg, x86_arg):
-    reg_info = r32_eax
-    parser = reg_info.parser
-
-
-class r_rax(reg_noarg, x86_arg):
-    reg_info = r64_eax
-    parser = reg_info.parser
-
-
-class r_cl(reg_noarg, x86_arg):
-    reg_info = r08_ecx
-    parser = reg_info.parser
-
-
-invmode = {16: 32, 32: 16}
-
-
-def opmode_prefix(mode):
-    size, opmode, admode = mode
-    if size in [16, 32]:
-        if opmode:
-            return invmode[size]
-        else:
-            return size
-    elif size == 64:
-        if opmode:
-            return 16
-        else:
-            return 32
-    raise NotImplementedError('not fully functional')
-
-
-def admode_prefix(mode):
-    size, opmode, admode = mode
-    if size in [16, 32]:
-        if admode:
-            return invmode[size]
-        else:
-            return size
-    elif size == 64:
-        return 64
-    raise NotImplementedError('not fully functional')
-
-
-def v_opmode_info(size, opmode, rex_w, stk):
-    if size in [16, 32]:
-        if opmode:
-            return invmode[size]
-        else:
-            return size
-    elif size == 64:
-        # Rex has the maximum priority
-        # Then opmode
-        # Then stacker
-        if rex_w == 1:
-            return 64
-        elif opmode == 1:
-            return 16
-        elif stk:
-            return 64
-        else:
-            return 32
-
-
-def v_opmode(p):
-    stk = hasattr(p, 'stk')
-    return v_opmode_info(p.mode, p.opmode, p.rex_w.value, stk)
-
-
-def v_admode_info(size, admode):
-    if size in [16, 32]:
-        if admode:
-            return invmode[size]
-        else:
-            return size
-    elif size == 64:
-        if admode == 1:
-            return 32
-        return 64
-
-
-def v_admode(p):
-    return v_admode_info(p.mode, p.admode)
-
-
-def offsize(p):
-    if p.opmode:
-        return 16
-    else:
-        return p.mode
-
-
-def get_prefix(s):
-    g = re.search(r'(\S+)(\s+)', s)
-    if not g:
-        return None, s
-    prefix, b = g.groups()
-    return prefix, s[len(prefix) + len(b):]
-
-
-repeat_mn = ["INS", "OUTS",
-             "MOVSB", "MOVSW", "MOVSD", "MOVSQ",
-             "SCASB", "SCASW", "SCASD", "SCASQ",
-             "LODSB", "LODSW", "LODSD", "LODSQ",
-             "STOSB", "STOSW", "STOSD", "STOSQ",
-             "CMPSB", "CMPSW", "CMPSD", "CMPSQ",
-             ]
-
-
-class group(object):
-
-    def __init__(self):
-        self.value = None
-
-
-class additional_info(object):
-
-    def __init__(self):
-        self.except_on_instr = False
-        self.g1 = group()
-        self.g2 = group()
-        self.vopmode = None
-        self.stk = False
-        self.v_opmode = None
-        self.v_admode = None
-        self.prefixed = b''
-
-
-class instruction_x86(instruction):
-    __slots__ = []
-
-    def __init__(self, *args, **kargs):
-        super(instruction_x86, self).__init__(*args, **kargs)
-
-    def v_opmode(self):
-        return self.additional_info.v_opmode
-
-    def v_admode(self):
-        return self.additional_info.v_admode
-
-    def dstflow(self):
-        if self.name in conditional_branch + unconditional_branch:
-            return True
-        if self.name.startswith('LOOP'):
-            return True
-        return self.name in ['CALL']
-
-    def dstflow2label(self, loc_db):
-        if self.additional_info.g1.value & 14 and self.name in repeat_mn:
-            return
-        expr = self.args[0]
-        if not expr.is_int():
-            return
-        addr = (int(expr) + int(self.offset)) & int(expr.mask)
-        loc_key = loc_db.get_or_create_offset_location(addr)
-        self.args[0] = ExprLoc(loc_key, expr.size)
-
-    def breakflow(self):
-        if self.name in conditional_branch + unconditional_branch:
-            return True
-        if self.name.startswith('LOOP'):
-            return True
-        if self.name.startswith('RET'):
-            return True
-        if self.name.startswith('INT'):
-            return True
-        if self.name.startswith('SYS'):
-            return True
-        return self.name in ['CALL', 'HLT', 'IRET', 'IRETD', 'IRETQ', 'ICEBP', 'UD2']
-
-    def splitflow(self):
-        if self.name in conditional_branch:
-            return True
-        if self.name in unconditional_branch:
-            return False
-        if self.name.startswith('LOOP'):
-            return True
-        if self.name.startswith('INT'):
-            return True
-        if self.name.startswith('SYS'):
-            return True
-        return self.name in ['CALL']
-
-    def setdstflow(self, a):
-        return
-
-    def is_subcall(self):
-        return self.name in ['CALL']
-
-    def getdstflow(self, loc_db):
-        if self.additional_info.g1.value & 14 and self.name in repeat_mn:
-            addr = int(self.offset)
-            loc_key = loc_db.get_or_create_offset_location(addr)
-            return [ExprLoc(loc_key, self.v_opmode())]
-        return [self.args[0]]
-
-    def get_symbol_size(self, symbol, loc_db):
-        return self.mode
-
-    def fixDstOffset(self):
-        expr = self.args[0]
-        if self.offset is None:
-            raise ValueError('symbol not resolved %s' % l)
-        if not isinstance(expr, ExprInt):
-            log.warning('dynamic dst %r', expr)
-            return
-        self.args[0] = ExprInt(int(expr) - self.offset, self.mode)
-
-    def get_info(self, c):
-        self.additional_info.g1.value = c.g1.value
-        self.additional_info.g2.value = c.g2.value
-        self.additional_info.stk = hasattr(c, 'stk')
-        self.additional_info.v_opmode = c.v_opmode()
-        self.additional_info.v_admode = c.v_admode()
-        self.additional_info.prefix = c.prefix
-        self.additional_info.prefixed = getattr(c, "prefixed", b"")
-
-    def __str__(self):
-        return self.to_string()
-
-    def to_string(self, loc_db=None):
-        o = super(instruction_x86, self).to_string(loc_db)
-        if self.additional_info.g1.value & 1:
-            o = "LOCK %s" % o
-        if self.additional_info.g1.value & 2:
-            if getattr(self.additional_info.prefixed, 'default', b"") != b"\xF2":
-                o = "REPNE %s" % o
-        if self.additional_info.g1.value & 8:
-            if getattr(self.additional_info.prefixed, 'default', b"") != b"\xF3":
-                o = "REP %s" % o
-        elif self.additional_info.g1.value & 4:
-            if getattr(self.additional_info.prefixed, 'default', b"") != b"\xF3":
-                o = "REPE %s" % o
-        return o
-
-    def to_html(self, loc_db=None):
-        o = super(instruction_x86, self).to_html(loc_db)
-        if self.additional_info.g1.value & 1:
-            text =  utils.set_html_text_color("LOCK", utils.COLOR_MNEMO)
-            o = "%s %s" % (text, o)
-        if self.additional_info.g1.value & 2:
-            if getattr(self.additional_info.prefixed, 'default', b"") != b"\xF2":
-                text =  utils.set_html_text_color("REPNE", utils.COLOR_MNEMO)
-                o = "%s %s" % (text, o)
-        if self.additional_info.g1.value & 8:
-            if getattr(self.additional_info.prefixed, 'default', b"") != b"\xF3":
-                text =  utils.set_html_text_color("REP", utils.COLOR_MNEMO)
-                o = "%s %s" % (text, o)
-        elif self.additional_info.g1.value & 4:
-            if getattr(self.additional_info.prefixed, 'default', b"") != b"\xF3":
-                text =  utils.set_html_text_color("REPE", utils.COLOR_MNEMO)
-                o = "%s %s" % (text, o)
-        return o
-
-
-    def get_args_expr(self):
-        args = []
-        for a in self.args:
-            a = a.replace_expr(replace_regs[self.mode])
-            args.append(a)
-        return args
-
-    @staticmethod
-    def arg2str(expr, index=None, loc_db=None):
-        if expr.is_id() or expr.is_int():
-            o = str(expr)
-        elif expr.is_loc():
-            if loc_db is not None:
-                o = loc_db.pretty_str(expr.loc_key)
-            else:
-                o = str(expr)
-        elif ((isinstance(expr, ExprOp) and expr.op == 'far' and
-               isinstance(expr.args[0], ExprMem)) or
-              isinstance(expr, ExprMem)):
-            if isinstance(expr, ExprOp):
-                prefix, expr = "FAR ", expr.args[0]
-            else:
-                prefix = ""
-            sz = SIZE2MEMPREFIX[expr.size]
-            segm = ""
-            if is_mem_segm(expr):
-                segm = "%s:" % expr.ptr.args[0]
-                expr = expr.ptr.args[1]
-            else:
-                expr = expr.ptr
-            if isinstance(expr, ExprOp):
-                s = str(expr).replace('(', '').replace(')', '')
-            else:
-                s = str(expr)
-            o = prefix + sz + ' PTR ' + str(segm) + '[%s]' % s
-        elif isinstance(expr, ExprOp) and expr.op == 'segm':
-            o = "%s:%s" % (expr.args[0], expr.args[1])
-        else:
-            raise ValueError('check this %r' % expr)
-        return "%s" % o
-
-
-    @staticmethod
-    def arg2html(expr, index=None, loc_db=None):
-        if expr.is_id() or expr.is_int() or expr.is_loc():
-            o = color_expr_html(expr, loc_db)
-        elif ((isinstance(expr, ExprOp) and expr.op == 'far' and
-               isinstance(expr.args[0], ExprMem)) or
-              isinstance(expr, ExprMem)):
-            if isinstance(expr, ExprOp):
-                prefix, expr = "FAR ", expr.args[0]
-            else:
-                prefix = ""
-            sz = SIZE2MEMPREFIX[expr.size]
-            sz =  '<font color="%s">%s</font>' % (utils.COLOR_MEM, sz)
-            segm = ""
-            if is_mem_segm(expr):
-                segm = "%s:" % expr.ptr.args[0]
-                expr = expr.ptr.args[1]
-            else:
-                expr = expr.ptr
-            if isinstance(expr, ExprOp):
-                s = color_expr_html(expr, loc_db)#.replace('(', '').replace(')', '')
-            else:
-                s = color_expr_html(expr, loc_db)
-            o = prefix + sz + ' PTR ' + str(segm) + BRACKET_O + str(s) + BRACKET_C
-        elif isinstance(expr, ExprOp) and expr.op == 'segm':
-            o = "%s:%s" % (
-                color_expr_html(expr.args[0], loc_db),
-                color_expr_html(expr.args[1], loc_db)
-            )
-        else:
-            raise ValueError('check this %r' % expr)
-        return "%s" % o
-
-
-
-class mn_x86(cls_mn):
-    name = "x86"
-    prefix_op_size = False
-    prefix_ad_size = False
-    regs = regs_module
-    all_mn = []
-    all_mn_mode = defaultdict(list)
-    all_mn_name = defaultdict(list)
-    all_mn_inst = defaultdict(list)
-    bintree = {}
-    num = 0
-    delayslot = 0
-    pc = {16: IP, 32: EIP, 64: RIP}
-    sp = {16: SP, 32: ESP, 64: RSP}
-    instruction = instruction_x86
-    max_instruction_len = 15
-
-    @classmethod
-    def getpc(cls, attrib):
-        return cls.pc[attrib]
-
-    @classmethod
-    def getsp(cls, attrib):
-        return cls.sp[attrib]
-
-    def v_opmode(self):
-        if hasattr(self, 'stk'):
-            stk = 1
-        else:
-            stk = 0
-        return v_opmode_info(self.mode, self.opmode, self.rex_w.value, stk)
-
-    def v_admode(self):
-        size, opmode, admode = self.mode, self.opmode, self.admode
-        if size in [16, 32]:
-            if admode:
-                return invmode[size]
-            else:
-                return size
-        elif size == 64:
-            if admode == 1:
-                return 32
-            return 64
-
-    def additional_info(self):
-        info = additional_info()
-        info.g1.value = self.g1.value
-        info.g2.value = self.g2.value
-        info.stk = hasattr(self, 'stk')
-        info.v_opmode = self.v_opmode()
-        info.prefixed = b""
-        if hasattr(self, 'prefixed'):
-            info.prefixed = self.prefixed.default
-        return info
-
-    @classmethod
-    def check_mnemo(cls, fields):
-        pass
-
-    @classmethod
-    def getmn(cls, name):
-        return name.upper()
-
-    @classmethod
-    def mod_fields(cls, fields):
-        prefix = [d_g1, d_g2, d_rex_p, d_rex_w, d_rex_r, d_rex_x, d_rex_b, d_vex, d_vex_l, d_vex_p, d_vex_v, d_vex_m]
-        return prefix + fields
-
-    @classmethod
-    def gen_modes(cls, subcls, name, bases, dct, fields):
-        dct['mode'] = None
-        return [(subcls, name, bases, dct, fields)]
-
-    @classmethod
-    def fromstring(cls, text, loc_db, mode):
-        pref = 0
-        prefix, new_s = get_prefix(text)
-        if prefix == "LOCK":
-            pref |= 1
-            text = new_s
-        elif prefix == "REPNE" or prefix == "REPNZ":
-            pref |= 2
-            text = new_s
-        elif prefix == "REPE" or prefix == "REPZ":
-            pref |= 4
-            text = new_s
-        elif prefix == "REP":
-            pref |= 8
-            text = new_s
-        c = super(mn_x86, cls).fromstring(text, loc_db, mode)
-        c.additional_info.g1.value = pref
-        return c
-
-    @classmethod
-    def pre_dis(cls, v, mode, offset):
-        offset_o = offset
-        pre_dis_info = {'opmode': 0,
-                        'admode': 0,
-                        'g1': 0,
-                        'g2': 0,
-                        'rex_p': 0,
-                        'rex_w': 0,
-                        'rex_r': 0,
-                        'rex_x': 0,
-                        'rex_b': 0,
-                        'vex_l': 0,
-                        'vex_p': 0,
-                        'vex_v': 0,
-                        'vex_m': 0,
-                        'vex' : 0,
-                        'prefix': b"",
-                        'prefixed': b"",
-                        }
-        while True:
-            c = v.getbytes(offset)
-            if c == b'\x66':
-                pre_dis_info['opmode'] = 1
-            elif c == b'\x67':
-                pre_dis_info['admode'] = 1
-            elif c == b'\xf0':
-                pre_dis_info['g1'] = 1
-            elif c == b'\xf2':
-                pre_dis_info['g1'] = 2
-            elif c == b'\xf3':
-                pre_dis_info['g1'] = 12
-
-            elif c == b'\x2e':
-                pre_dis_info['g2'] = 1
-            elif c == b'\x36':
-                pre_dis_info['g2'] = 2
-            elif c == b'\x3e':
-                pre_dis_info['g2'] = 3
-            elif c == b'\x26':
-                pre_dis_info['g2'] = 4
-            elif c == b'\x64':
-                pre_dis_info['g2'] = 5
-            elif c == b'\x65':
-                pre_dis_info['g2'] = 6
-
-            else:
-                break
-            pre_dis_info['prefix'] += c
-            offset += 1
-        vex3_prefix = b'\xc4'
-        vex2_prefix = b'\xc5'
-        rex_prefixes = b'@ABCDEFGHIJKLMNO'
-        if mode == 64 and c in rex_prefixes:
-            while c in rex_prefixes:
-                # multiple REX prefixes case - use last REX prefix
-                x = ord(c)
-                offset += 1
-                c = v.getbytes(offset)
-            pre_dis_info['rex_p'] = 1
-            pre_dis_info['rex_w'] = (x >> 3) & 1
-            pre_dis_info['rex_r'] = (x >> 2) & 1
-            pre_dis_info['rex_x'] = (x >> 1) & 1
-            pre_dis_info['rex_b'] = (x >> 0) & 1
-        elif mode == 64 and c == vex3_prefix:
-            offset += 1
-            c = ord(v.getbytes(offset))
-            pre_dis_info['vex'] = 1
-            pre_dis_info['rex_r'] = ((c >> 7) ^ 1) & 1
-            pre_dis_info['rex_x'] = ((c >> 6) ^ 1) & 1
-            pre_dis_info['rex_b'] = ((c >> 5) ^ 1) & 1
-            pre_dis_info['vex_m'] = (c & int('0b11111', 2))
-
-            offset += 1
-            c = ord(v.getbytes(offset))
-            pre_dis_info['rex_w'] = (c >> 7) & 1
-            pre_dis_info['vex_v'] = ((c >> 3) ^ 15) & 15
-            pre_dis_info['vex_l'] = (c >> 2) & 1
-            pre_dis_info['vex_p'] = c & int('0b11', 2)
-            offset += 1
-
-            if pre_dis_info['vex_p'] == 1:
-                pre_dis_info['opmode'] = 1
-            elif pre_dis_info['vex_p'] == 3:
-                pre_dis_info['g1'] = 2
-            elif pre_dis_info['vex_p'] == 2:
-                pre_dis_info['g1'] = 12
-
-        elif mode == 64 and c == vex2_prefix and v.getlen() > 2:
-            offset += 1
-            c = ord(v.getbytes(offset))
-            pre_dis_info['vex'] = 1
-            pre_dis_info['rex_r'] = ((c >> 7) ^ 1) & 1
-            pre_dis_info['rex_x'] = 0
-            pre_dis_info['rex_b'] = 0
-            pre_dis_info['rex_w'] = 0
-            pre_dis_info['vex_l'] = (c >> 2) & 1
-            pre_dis_info['vex_p'] = c & int('0b11', 2)
-            offset += 1
-
-            if pre_dis_info['vex_p'] == 1:
-                pre_dis_info['opmode'] = 1
-            elif pre_dis_info['vex_p'] == 3:
-                pre_dis_info['g1'] = 2
-            elif pre_dis_info['vex_p'] == 2:
-                pre_dis_info['g1'] = 12
-
-        elif pre_dis_info.get('g1', None) == 12 and c in [b'\xa6', b'\xa7', b'\xae', b'\xaf']:
-            pre_dis_info['g1'] = 4
-        return pre_dis_info, v, mode, offset, offset - offset_o
-
-    @classmethod
-    def get_cls_instance(cls, cc, mode, infos=None):
-        for opmode in [0, 1]:
-            for admode in [0, 1]:
-                c = cc()
-                c.init_class()
-
-                c.reset_class()
-                c.add_pre_dis_info()
-                c.dup_info(infos)
-                c.mode = mode
-                c.opmode = opmode
-                c.admode = admode
-
-                if not hasattr(c, 'stk') and hasattr(c, "fopmode") and c.fopmode.mode == 64:
-                    c.rex_w.value = 1
-                yield c
-
-    def post_dis(self):
-        if self.g2.value:
-            for a in self.args:
-                if not isinstance(a.expr, ExprMem):
-                    continue
-                m = a.expr
-                a.expr = ExprMem(
-                    ExprOp('segm', enc2segm[self.g2.value], m.ptr), m.size)
-        return self
-
-    def dup_info(self, infos):
-        if infos is not None:
-            self.g1.value = infos.g1.value
-            self.g2.value = infos.g2.value
-
-    def reset_class(self):
-        super(mn_x86, self).reset_class()
-        if hasattr(self, "opmode"):
-            del(self.opmode)
-        if hasattr(self, "admode"):
-            del(self.admode)
-
-    def add_pre_dis_info(self, pre_dis_info=None):
-        if pre_dis_info is None:
-            return True
-        if hasattr(self, "prefixed") and self.prefixed.default == b"\x66":
-            pre_dis_info['opmode'] = 0
-        self.opmode = pre_dis_info['opmode']
-        self.admode = pre_dis_info['admode']
-
-        if hasattr(self, 'no_xmm_pref') and\
-                pre_dis_info['prefix'] and\
-                pre_dis_info['prefix'][-1] in b'\x66\xf2\xf3':
-            return False
-        if (hasattr(self, "prefixed") and
-            not pre_dis_info['prefix'].endswith(self.prefixed.default)):
-            return False
-        if (self.rex_w.value is not None and
-            self.rex_w.value != pre_dis_info['rex_w']):
-            return False
-        else:
-            self.rex_w.value = pre_dis_info['rex_w']
-        self.rex_r.value = pre_dis_info['rex_r']
-        self.rex_b.value = pre_dis_info['rex_b']
-        self.rex_x.value = pre_dis_info['rex_x']
-        self.rex_p.value = pre_dis_info['rex_p']
-
-        self.vex.value = pre_dis_info['vex']
-        self.vex_l.value = pre_dis_info['vex_l']
-        self.vex_p.value = pre_dis_info['vex_p']
-        self.vex_v.value = pre_dis_info['vex_v']
-        self.vex_m.value = pre_dis_info['vex_m']
-
-        if hasattr(self, 'no_rex') and\
-           (self.rex_r.value or self.rex_b.value or
-            self.rex_x.value or self.rex_p.value):
-            return False
-
-        if self.vex.value == 0 and (hasattr(self, 'pref_0f')
-                                    or hasattr(self, 'pref_0f38')
-                                    or hasattr(self, 'pref_0f3a')):
-            return False
-
-        if hasattr(self, 'no_rep') and b'\xf3' in pre_dis_info['prefix']:
-            return False
-
-        if self.vex_m.value == 1 and not hasattr(self, 'pref_0f'):
-            return False
-        if self.vex_m.value == 2 and not hasattr(self, 'pref_0f38'):
-            return False
-        if self.vex_m.value == 3 and not hasattr(self, 'pref_0f3a'):
-            return False
-
-        self.g1.value = pre_dis_info['g1']
-        self.g2.value = pre_dis_info['g2']
-        self.prefix = pre_dis_info['prefix']
-        return True
-
-    def post_asm(self, v):
-        return v
-
-
-    def gen_prefix(self):
-        v = b""
-        rex = 0x40
-        if self.g1.value is None:
-            self.g1.value = 0
-        if self.g2.value is None:
-            self.g2.value = 0
-
-        if self.rex_w.value:
-            rex |= 0x8
-        if self.rex_r.value:
-            rex |= 0x4
-        if self.rex_x.value:
-            rex |= 0x2
-        if self.rex_b.value:
-            rex |= 0x1
-        if (rex != 0x40 and not self.vex.value) or self.rex_p.value == 1:
-            v = utils.int_to_byte(rex) + v
-            if hasattr(self, 'no_rex'):
-                return None
-
-        vex_byte1 = 0xc4
-        vex_byte2 = 0x00
-        vex_byte3 = 0x00
-
-        m_prefix = [field.fname for field in self.fields_order if 'pref_0f' in field.fname]
-        if m_prefix:
-            if m_prefix[0] == 'pref_0f':
-                vex_byte2 |= 0x01
-            elif m_prefix[0] == 'pref_0f38':
-                vex_byte2 |= 0x02
-            elif m_prefix[0] == 'pref_0f3a':
-                vex_byte2 |= 0x03
-
-        # TODO: L and p 
-        if m_prefix and m_prefix[0] == 'pref_0f' and not self.rex_w.value and not self.rex_b.value and ((hasattr(self, 'mod') and self.mod.value == 3) or not self.rex_x.value): # VEX2
-            print("test")
-            vex_version = 2
-            vex_byte1 = 0x00
-            vex_byte2 = 0xc5
-
-            if not hasattr(self, 'reg') or not self.rex_r.value:
-                vex_byte3 |= 0x80
-
-        else:
-            vex_version = 3
-            if not hasattr(self, 'reg') or not self.rex_r.value:
-                vex_byte2 |= 0x80
-            if (hasattr(self, 'mod') and self.mod.value == 3) or not self.rex_x.value:
-                vex_byte2 |= 0x40
-            if not self.rex_b.value:
-                vex_byte2 |= 0x20
-
-            if self.rex_w.value:
-                vex_byte3 |= 0x80
-
-        if self.vex.value == 1:
-            vex_byte3 |= ((15 - self.vex_v.value) << 3)
-            vex = (vex_byte1 << 16) | (vex_byte2 << 8) | vex_byte3
-            v = vex.to_bytes(vex_version, 'big') + v
-
-        if hasattr(self, 'prefixed'):
-            v = self.prefixed.default + v
-
-        if self.g1.value & 1:
-            v = b"\xf0" + v
-        if self.g1.value & 2:
-            if hasattr(self, 'no_xmm_pref'):
-                return None
-            v = b"\xf2" + v
-        if self.g1.value & 12:
-            if hasattr(self, 'no_xmm_pref'):
-                return None
-            v = b"\xf3" + v
-        if self.g2.value:
-            v = {
-                1: b'\x2e',
-                2: b'\x36',
-                3: b'\x3e',
-                4: b'\x26',
-                5: b'\x64',
-                6: b'\x65'
-            }[self.g2.value] + v
-        # mode prefix
-        if hasattr(self, "admode") and self.admode:
-            v = b"\x67" + v
-
-        if hasattr(self, "opmode") and self.opmode:
-            if hasattr(self, 'no_xmm_pref'):
-                return None
-            v = b"\x66" + v
-        return v
-
-    def encodefields(self, decoded):
-        v = super(mn_x86, self).encodefields(decoded)
-        prefix = self.gen_prefix()
-        if prefix is None:
-            return None
-        return prefix + v
-
-    def getnextflow(self, loc_db):
-        raise NotImplementedError('not fully functional')
-
-    def ir_pre_instruction(self):
-        return [ExprAssign(mRIP[self.mode],
-            ExprInt(self.offset + self.l, mRIP[self.mode].size))]
-
-    @classmethod
-    def filter_asm_candidates(cls, instr, candidates):
-
-        cand_same_mode = []
-        cand_diff_mode = []
-        out = []
-        for c, v in candidates:
-            if (hasattr(c, 'no_xmm_pref') and
-                (c.g1.value & 2 or c.g1.value & 4 or c.g1.value & 8 or c.opmode)):
-                continue
-            if hasattr(c, "fopmode") and v_opmode(c) != c.fopmode.mode:
-                continue
-            if hasattr(c, "fadmode") and v_admode(c) != c.fadmode.mode:
-                continue
-            # relative dstflow must not have opmode set
-            # (assign IP instead of EIP for instance)
-            if (instr.dstflow() and
-                instr.name not in ["JCXZ", "JECXZ", "JRCXZ"] and
-                len(instr.args) == 1 and
-                    isinstance(instr.args[0], ExprInt) and c.opmode):
-                continue
-
-            out.append((c, v))
-        candidates = out
-        for c, v in candidates:
-            if v_opmode(c) == instr.mode:
-                cand_same_mode += v
-        for c, v in candidates:
-            if v_opmode(c) != instr.mode:
-                cand_diff_mode += v
-        cand_same_mode.sort(key=len)
-        cand_diff_mode.sort(key=len)
-        return cand_same_mode + cand_diff_mode
-
-
-class bs_modname_size(bs_divert):
-    prio = 1
-
-    def divert(self, i, candidates):
-        out = []
-        for candidate in candidates:
-            cls, name, bases, dct, fields = candidate
-            fopmode = opmode_prefix(
-                (dct['mode'], dct['opmode'], dct['admode']))
-            mode = dct['mode']
-            size, opmode, admode = dct['mode'], dct['opmode'], dct['admode']
-            # no mode64 exinstance in name means no 64bit version of mnemo
-            if mode == 64:
-                if mode in self.args['name']:
-                    nfields = fields[:]
-                    f, i = getfieldindexby_name(nfields, 'rex_w')
-                    f = bs("1", l=0, cls=(bs_fbit,), fname="rex_w")
-                    osize = v_opmode_info(size, opmode, 1, 0)
-                    nfields[i] = f
-                    nfields = nfields[:-1]
-                    ndct = dict(dct)
-                    if osize in self.args['name']:
-                        ndct['name'] = self.args['name'][osize]
-                        out.append((cls, ndct['name'], bases, ndct, nfields))
-
-                    nfields = fields[:]
-                    nfields = nfields[:-1]
-                    f, i = getfieldindexby_name(nfields, 'rex_w')
-                    f = bs("0", l=0, cls=(bs_fbit,), fname="rex_w")
-                    osize = v_opmode_info(size, opmode, 0, 0)
-                    nfields[i] = f
-                    ndct = dict(dct)
-                    if osize in self.args['name']:
-                        ndct['name'] = self.args['name'][osize]
-                        out.append((cls, ndct['name'], bases, ndct, nfields))
-            else:
-                l = opmode_prefix((dct['mode'], dct['opmode'], dct['admode']))
-                osize = v_opmode_info(size, opmode, None, 0)
-                nfields = fields[:-1]
-                ndct = dict(dct)
-                if osize in self.args['name']:
-                    ndct['name'] = self.args['name'][osize]
-                    out.append((cls, ndct['name'], bases, ndct, nfields))
-        return out
-
-
-class bs_modname_jecx(bs_divert):
-    prio = 1
-
-    def divert(self, i, candidates):
-        out = []
-        for candidate in candidates:
-            cls, name, bases, dct, fields = candidate
-            fopmode = opmode_prefix(
-                (dct['mode'], dct['opmode'], dct['admode']))
-            mode = dct['mode']
-            size, opmode, admode = dct['mode'], dct['opmode'], dct['admode']
-
-            nfields = fields[:]
-            nfields = nfields[:-1]
-            args = dict(self.args)
-            ndct = dict(dct)
-            if mode == 64:
-                if admode:
-                    ndct['name'] = "JECXZ"
-                else:
-                    ndct['name'] = "JRCXZ"
-            elif mode == 32:
-                if admode:
-                    ndct['name'] = "JCXZ"
-                else:
-                    ndct['name'] = "JECXZ"
-            elif mode == 16:
-                if admode:
-                    ndct['name'] = "JECXZ"
-                else:
-                    ndct['name'] = "JCXZ"
-            else:
-                raise ValueError('unhandled mode')
-            out.append((cls, ndct['name'], bases, ndct, nfields))
-        return out
-
-
-class bs_modname_mode(bs_divert):
-    prio = 1
-
-    def divert(self, i, candidates):
-        out = []
-        for candidate in candidates:
-            cls, name, bases, dct, fields = candidate
-            fopmode = opmode_prefix(
-                (dct['mode'], dct['opmode'], dct['admode']))
-            size, opmode, admode = dct['mode'], dct['opmode'], dct['admode']
-
-            mode = dct['mode']
-            l = opmode_prefix((dct['mode'], dct['opmode'], dct['admode']))
-            osize = v_opmode_info(size, opmode, None, 0)
-            nfields = fields[:-1]
-            args = dict(self.args)
-            ndct = dict(dct)
-            if mode == 64 or osize == 32:
-                ndct['name'] = self.args['name'][mode]
-            else:
-                ndct['name'] = self.args['name'][16]
-            out.append((cls, ndct['name'], bases, ndct, nfields))
-        return out
-
-
-class x86_imm(imm_noarg):
-    parser = base_expr
-
-    def decodeval(self, v):
-        return swap_uint(self.l, v)
-
-    def encodeval(self, v):
-        return swap_uint(self.l, v)
-
-
-class x86_imm_fix_08(imm_noarg):
-    parser = base_expr
-    intsize = 8
-    intmask = (1 << intsize) - 1
-
-    def decodeval(self, v):
-        return self.ival
-
-    def encode(self):
-        v = self.expr2int(self.expr)
-        if v != self.ival:
-            return False
-        self.value = 0
-        return True
-
-
-class x86_08(x86_imm):
-    intsize = 8
-    intmask = (1 << intsize) - 1
-
-
-class x86_16(x86_imm):
-    intsize = 16
-    intmask = (1 << intsize) - 1
-
-
-class x86_32(x86_imm):
-    intsize = 32
-    intmask = (1 << intsize) - 1
-
-
-class x86_64(x86_imm):
-    intsize = 64
-    intmask = (1 << intsize) - 1
-
-
-class x86_08_ne(x86_imm):
-    intsize = 8
-    intmask = (1 << intsize) - 1
-
-    def encode(self):
-        return True
-
-    def decode(self, v):
-        v = swap_uint(self.l, v)
-        p = self.parent
-        admode = p.v_admode()
-        value = sign_ext(v, self.intsize, admode)
-        self.expr = ExprInt(value, admode)
-        return True
-
-
-class x86_16_ne(x86_08_ne):
-    intsize = 16
-    intmask = (1 << intsize) - 1
-
-
-class x86_32_ne(x86_08_ne):
-    intsize = 32
-    intmask = (1 << intsize) - 1
-
-
-class x86_64_ne(x86_08_ne):
-    intsize = 64
-    intmask = (1 << intsize) - 1
-
-
-class x86_s08to16(x86_imm):
-    in_size = 8
-    out_size = 16
-
-    def myexpr(self, x):
-        return ExprInt(x, 16)
-
-    def int2expr(self, v):
-        return self.myexpr(v)
-
-    def expr2int(self, e):
-        if not isinstance(e, ExprInt):
-            return None
-        v = int(e)
-        if v & ~((1 << self.l) - 1) != 0:
-            return None
-        return v
-
-    def decode(self, v):
-        v = v & self.lmask
-        v = self.decodeval(v)
-        if self.parent.v_opmode() == 64:
-            self.expr = ExprInt(sign_ext(v, self.in_size, 64), 64)
-        else:
-            if (1 << (self.l - 1)) & v:
-                v = sign_ext(v, self.l, self.out_size)
-            self.expr = self.myexpr(v)
-        return True
-
-    def encode(self):
-        if not isinstance(self.expr, ExprInt):
-            return False
-        v = int(self.expr)
-        opmode = self.parent.v_opmode()
-
-        out_size = self.out_size
-        if opmode != self.out_size:
-            if opmode == 32 and self.out_size == 64:
-                out_size = opmode
-                if v == sign_ext(
-                    int(v & ((1 << self.in_size) - 1)), self.in_size, out_size):
-                    pass
-                else:
-                    # test with rex_w
-                    self.parent.rex_w.value = 1
-                    opmode = self.parent.v_opmode()
-                    out_size = opmode
-                    if (v != sign_ext(
-                        int(v & ((1 << self.in_size) - 1)),
-                        self.in_size, out_size)):
-                        return False
-        if v != sign_ext(
-            int(v & ((1 << self.in_size) - 1)), self.in_size, out_size):
-            return False
-        v = self.encodeval(v)
-        self.value = (v & 0xffffffff) & self.lmask
-        return True
-
-    def decodeval(self, v):
-        return swap_uint(self.l, v)
-
-    def encodeval(self, v):
-        return swap_sint(self.l, v)
-
-
-class x86_s08to32(x86_s08to16):
-    in_size = 8
-    out_size = 32
-
-    def myexpr(self, x):
-        return ExprInt(x, 32)
-
-    def decode(self, v):
-        v = v & self.lmask
-        v = self.decodeval(v)
-        if self.parent.rex_w.value == 1:
-            v = ExprInt(sign_ext(v, self.in_size, 64), 64)
-        else:
-            v = ExprInt(sign_ext(v, self.in_size, 32), 32)
-
-        self.expr = v
-        return True
-
-
-class x86_s08to64(x86_s08to32):
-    in_size = 8
-    out_size = 64
-
-    def myexpr(self, x):
-        return ExprInt(x, 64)
-
-
-class x86_s32to64(x86_s08to32):
-    in_size = 32
-    out_size = 64
-
-    def myexpr(self, x):
-        return ExprInt(x, 64)
-
-
-class bs_eax(x86_arg):
-    reg_info = r_eax_all
-    rindex = 0
-    parser = reg_info.parser
-
-    def decode(self, v):
-        p = self.parent
-        expr = None
-        if hasattr(p, 'w8') and p.w8.value == 0:
-            expr = regs08_expr[self.rindex]
-        else:
-            expr = size2gpregs[p.v_opmode()].expr[self.rindex]
-        self.expr = expr
-        return True
-
-    def encode(self):
-        self.value = 0
-        p = self.parent
-        expr = self.expr
-        osize = p.v_opmode()
-        if hasattr(p, 'w8'):
-            if p.w8.value is None:
-                # XXX TODO: priority in w8 erase?
-                if expr.size == 8:
-                    p.w8.value = 0
-                else:
-                    p.w8.value = 1
-        if hasattr(p, 'w8') and p.w8.value == 0:
-            return expr == regs08_expr[self.rindex]
-        elif p.mode in [16, 32]:
-            return expr == size2gpregs[osize].expr[self.rindex]
-        elif p.mode == 64:
-            if expr == size2gpregs[64].expr[self.rindex]:
-                p.rex_w.value = 1
-                return True
-            elif expr == size2gpregs[osize].expr[self.rindex]:
-                return True
-            return False
-        return False
-
-class bs_seg(x86_arg):
-    reg_info = r_eax_all
-    rindex = 0
-    parser = reg_info.parser
-
-    def decode(self, v):
-        self.expr = self.reg_info.expr[0]
-        return True
-
-    def encode(self):
-        self.value = 0
-        return self.expr == self.reg_info.expr[0]
-
-
-class bs_edx(bs_eax):
-    reg_info = r_edx_all
-    rindex = 2
-    parser = reg_info.parser
-
-
-class bs_st(bs_eax):
-    reg_info = r_st_all
-    rindex = 0
-    parser = reg_info.parser
-
-
-class bs_cs(bs_seg):
-    reg_info = r_cs_all
-    rindex = 0
-    parser = reg_info.parser
-
-
-class bs_ds(bs_seg):
-    reg_info = r_ds_all
-    rindex = 0
-    parser = reg_info.parser
-
-
-class bs_es(bs_seg):
-    reg_info = r_es_all
-    rindex = 0
-    parser = reg_info.parser
-
-
-class bs_ss(bs_seg):
-    reg_info = r_ss_all
-    rindex = 0
-    parser = reg_info.parser
-
-
-class bs_fs(bs_seg):
-    reg_info = r_fs_all
-    rindex = 0
-    parser = reg_info.parser
-
-
-class bs_gs(bs_seg):
-    reg_info = r_gs_all
-    rindex = 0
-    parser = reg_info.parser
-
-
-class x86_reg_st(reg_noarg, x86_arg):
-    reg_info = r_st_all
-    parser = reg_info.parser
-
-
-class bs_sib_scale(bs_divert):
-    bsname = "sib_scale"
-
-    def divert(self, i, candidates):
-        out = []
-        done = False
-        for cls, name, bases, dct, fields in candidates:
-            if (not (admode_prefix(
-                (dct['mode'], dct['opmode'], dct['admode'])) != 16 and
-                'rm' in dct and dct['rm'] == 0b100 and
-                'mod' in dct and dct['mod'] != 0b11)):
-                ndct = dict(dct)
-                nfields = fields[:]
-                nfields[i] = None
-                ndct[self.args['fname']] = None
-                out.append((cls, ndct['name'], bases, ndct, nfields))
-                continue
-
-            nfields = fields[:]
-            args = dict(self.args)
-            ndct = dict(dct)
-            f = bs(**args)
-            nfields[i] = f
-            ndct[self.args['fname']] = None
-            out.append((cls, ndct['name'], bases, ndct, nfields))
-        return out
-
-
-class bs_sib_index(bs_sib_scale):
-    pass
-
-
-class bs_sib_base(bs_sib_scale):
-    pass
-
-
-class bs_disp(bs_divert):
-
-    def divert(self, i, candidates):
-        out = []
-        done = False
-        for cls, name, bases, dct, fields in candidates:
-            ndct = dict(dct)
-            nfields = fields[:]
-            if (admode_prefix(
-                (dct['mode'], dct['opmode'], dct['admode'])) == 16):
-                if 'mod' in dct and dct['mod'] == 0b00 and \
-                        'rm' in dct and dct['rm'] == 0b110:
-                    nfields[i] = bs(
-                        l=16, cls=(x86_16_ne,), fname=self.args['fname'])
-                    ndct[self.args['fname']] = True
-                    out.append((cls, ndct['name'], bases, ndct, nfields))
-                    continue
-                elif 'mod' in dct and dct['mod'] == 0b01:
-                    nfields[i] = bs(
-                        l=8, cls=(x86_08_ne,), fname=self.args['fname'])
-                    ndct[self.args['fname']] = True
-                    out.append((cls, ndct['name'], bases, ndct, nfields))
-                    continue
-                elif 'mod' in dct and dct['mod'] == 0b10:
-                    nfields[i] = bs(
-                        l=16, cls=(x86_16_ne,), fname=self.args['fname'])
-                    ndct[self.args['fname']] = True
-                    out.append((cls, ndct['name'], bases, ndct, nfields))
-                    continue
-            else:
-                if 'mod' in dct and dct['mod'] == 0b00 and \
-                        'rm' in dct and dct['rm'] == 0b101:
-                    nfields[i] = bs(
-                        l=32, cls=(x86_32_ne,), fname=self.args['fname'])
-                    ndct[self.args['fname']] = True
-                    out.append((cls, ndct['name'], bases, ndct, nfields))
-                    continue
-                elif 'mod' in dct and dct['mod'] == 0b01:
-                    nfields[i] = bs(
-                        l=8, cls=(x86_08_ne,), fname=self.args['fname'])
-                    ndct[self.args['fname']] = True
-                    out.append((cls, ndct['name'], bases, ndct, nfields))
-                    continue
-                elif 'mod' in dct and dct['mod'] == 0b10:
-                    nfields[i] = bs(
-                        l=32, cls=(x86_32_ne,), fname=self.args['fname'])
-                    ndct[self.args['fname']] = True
-                    out.append((cls, ndct['name'], bases, ndct, nfields))
-                    continue
-
-            nfields[i] = None
-            ndct[self.args['fname']] = None
-            out.append((cls, ndct['name'], bases, ndct, nfields))
-        return out
-
-
-def getmodrm(c):
-    return (c >> 6) & 3, (c >> 3) & 7, c & 7
-
-
-def setmodrm(mod, re, rm):
-    return ((mod & 3) << 6) | ((re & 7) << 3) | (rm & 7)
-
-
-def sib(c):
-    return modrm(c)
-
-db_afs_64 = []
-sib_64_s08_ebp = []
-
-
-def gen_modrm_form():
-    global db_afs_64, sib_64_s08_ebp
-    ebp = 5
-
-    sib_s08_ebp = [{f_isad: True} for i in range(0x100)]
-    sib_u32_ebp = [{f_isad: True} for i in range(0x100)]
-    sib_u32 = [{f_isad: True} for i in range(0x100)]
-
-    sib_u64 = []
-    for rex_x in range(2):
-        o = []
-        for rex_b in range(2):
-            x = [{f_isad: True} for i in range(0x100)]
-            o.append(x)
-        sib_u64.append(o)
-
-    sib_u64_ebp = []
-    for rex_x in range(2):
-        o = []
-        for rex_b in range(2):
-            x = [{f_isad: True} for i in range(0x100)]
-            o.append(x)
-        sib_u64_ebp.append(o)
-
-    sib_64_s08_ebp = []
-    for rex_x in range(2):
-        o = []
-        for rex_b in range(2):
-            x = [{f_isad: True} for i in range(0x100)]
-            o.append(x)
-        sib_64_s08_ebp.append(o)
-
-    for sib_rez in [sib_s08_ebp,
-                    sib_u32_ebp,
-                    sib_u32,
-                    sib_64_s08_ebp,
-                    sib_u64_ebp,
-                    sib_u64,
-                    ]:
-        for index in range(0x100):
-            ss, i, b = getmodrm(index)
-
-            if b == 0b101:
-                if sib_rez == sib_s08_ebp:
-                    sib_rez[index][f_imm] = f_s08
-                    sib_rez[index][ebp] = 1
-                elif sib_rez == sib_u32_ebp:
-                    sib_rez[index][f_imm] = f_u32
-                    sib_rez[index][ebp] = 1
-                elif sib_rez == sib_u32:
-                    sib_rez[index][f_imm] = f_u32
-                elif sib_rez == sib_u64_ebp:
-                    for rex_b in range(2):
-                        for rex_x in range(2):
-                            sib_rez[rex_x][rex_b][index][f_imm] = f_u32
-                            sib_rez[rex_x][rex_b][index][ebp + 8 * rex_b] = 1
-                elif sib_rez == sib_u64:
-                    for rex_b in range(2):
-                        for rex_x in range(2):
-                            sib_rez[rex_x][rex_b][index][f_imm] = f_u32
-                elif sib_rez == sib_64_s08_ebp:
-                    for rex_b in range(2):
-                        for rex_x in range(2):
-                            sib_rez[rex_x][rex_b][index][f_imm] = f_s08
-                            sib_rez[rex_x][rex_b][index][ebp + 8 * rex_b] = 1
-
-            else:
-                if sib_rez == sib_s08_ebp:
-                    sib_rez[index][b] = 1
-                    sib_rez[index][f_imm] = f_s08
-                elif sib_rez == sib_u32_ebp:
-                    sib_rez[index][b] = 1
-                    sib_rez[index][f_imm] = f_u32
-                elif sib_rez == sib_u32:
-                    sib_rez[index][b] = 1
-                elif sib_rez == sib_u64_ebp:
-                    for rex_b in range(2):
-                        for rex_x in range(2):
-                            sib_rez[rex_x][rex_b][index][b + 8 * rex_b] = 1
-                            sib_rez[rex_x][rex_b][index][f_imm] = f_u32
-                elif sib_rez == sib_u64:
-                    for rex_b in range(2):
-                        for rex_x in range(2):
-                            sib_rez[rex_x][rex_b][index][b + 8 * rex_b] = 1
-                elif sib_rez == sib_64_s08_ebp:
-                    for rex_b in range(2):
-                        for rex_x in range(2):
-                            sib_rez[rex_x][rex_b][index][f_imm] = f_s08
-                            sib_rez[rex_x][rex_b][index][b + 8 * rex_b] = 1
-
-            if i == 0b100 and sib_rez in [sib_s08_ebp, sib_u32_ebp, sib_u32]:
-                continue
-
-            if sib_rez in [sib_s08_ebp, sib_u32_ebp, sib_u32]:
-                tmp = i
-                if not tmp in sib_rez[index]:
-                    sib_rez[index][tmp] = 0  # 1 << ss
-                sib_rez[index][tmp] += 1 << ss
-            else:
-                for rex_b in range(2):
-                    for rex_x in range(2):
-                        tmp = i + 8 * rex_x
-                        if i == 0b100 and rex_x == 0:
-                            continue
-                        if not tmp in sib_rez[rex_x][rex_b][index]:
-                            sib_rez[rex_x][rex_b][index][tmp] = 0  # 1 << ss
-                        sib_rez[rex_x][rex_b][index][tmp] += 1 << ss
-
-    # 32bit
-    db_afs_32 = [None for i in range(0x100)]
-    for i in range(0x100):
-        index = i
-        mod, re, rm = getmodrm(i)
-
-        if mod == 0b00:
-            if rm == 0b100:
-                db_afs_32[index] = sib_u32
-            elif rm == 0b101:
-                db_afs_32[index] = {f_isad: True, f_imm: f_u32}
-            else:
-                db_afs_32[index] = {f_isad: True, rm: 1}
-        elif mod == 0b01:
-            if rm == 0b100:
-                db_afs_32[index] = sib_s08_ebp
-                continue
-            tmp = {f_isad: True, rm: 1, f_imm: f_s08}
-            db_afs_32[index] = tmp
-
-        elif mod == 0b10:
-            if rm == 0b100:
-                db_afs_32[index] = sib_u32_ebp
-            else:
-                db_afs_32[index] = {f_isad: True, rm: 1, f_imm: f_u32}
-        elif mod == 0b11:
-            db_afs_32[index] = {f_isad: False, rm: 1}
-
-    # 64bit
-    db_afs_64 = [None for i in range(0x400)]
-    for i in range(0x400):
-        index = i
-        rex_x = (index >> 9) & 1
-        rex_b = (index >> 8) & 1
-        mod, re, rm = getmodrm(i & 0xff)
-
-        if mod == 0b00:
-            if rm == 0b100:
-                db_afs_64[i] = sib_u64[rex_x][rex_b]
-            elif rm == 0b101:
-                db_afs_64[i] = {f_isad: True, f_imm: f_u32, 16: 1}
-            else:
-                db_afs_64[i] = {f_isad: True, rm + 8 * rex_b: 1}
-        elif mod == 0b01:
-            if rm == 0b100:
-                db_afs_64[i] = sib_64_s08_ebp[rex_x][rex_b]
-                continue
-            tmp = {f_isad: True, rm + 8 * rex_b: 1, f_imm: f_s08}
-            db_afs_64[i] = tmp
-
-        elif mod == 0b10:
-            if rm == 0b100:
-                db_afs_64[i] = sib_u64_ebp[rex_x][rex_b]
-            else:
-                db_afs_64[i] = {f_isad: True, rm + 8 * rex_b: 1, f_imm: f_u32}
-        elif mod == 0b11:
-            db_afs_64[i] = {f_isad: False, rm + 8 * rex_b: 1}
-
-    # 16bit
-    db_afs_16 = [None for i in range(0x100)]
-    _si = 6
-    _di = 7
-    _bx = 3
-    _bp = 5
-    for i in range(0x100):
-        index = i
-        mod, re, rm = getmodrm(i)
-
-        if mod == 0b00:
-            if rm == 0b100:
-                db_afs_16[index] = {f_isad: True, _si: 1}
-            elif rm == 0b101:
-                db_afs_16[index] = {f_isad: True, _di: 1}
-            elif rm == 0b110:
-                db_afs_16[index] = {
-                    f_isad: True, f_imm: f_u16}  # {f_isad:True,_bp:1}
-            elif rm == 0b111:
-                db_afs_16[index] = {f_isad: True, _bx: 1}
-            else:
-                db_afs_16[index] = {f_isad: True,
-                         [_si, _di][rm % 2]: 1,
-                    [_bx, _bp][(rm >> 1) % 2]: 1}
-        elif mod in [0b01, 0b10]:
-            if mod == 0b01:
-                my_imm = f_s08
-            else:
-                my_imm = f_u16
-
-            if rm == 0b100:
-                db_afs_16[index] = {f_isad: True, _si: 1, f_imm: my_imm}
-            elif rm == 0b101:
-                db_afs_16[index] = {f_isad: True, _di: 1, f_imm: my_imm}
-            elif rm == 0b110:
-                db_afs_16[index] = {f_isad: True, _bp: 1, f_imm: my_imm}
-            elif rm == 0b111:
-                db_afs_16[index] = {f_isad: True, _bx: 1, f_imm: my_imm}
-            else:
-                db_afs_16[index] = {f_isad: True,
-                         [_si, _di][rm % 2]: 1,
-                    [_bx, _bp][(rm >> 1) % 2]: 1,
-                    f_imm: my_imm}
-
-        elif mod == 0b11:
-            db_afs_16[index] = {f_isad: False, rm: 1}
-
-    byte2modrm = {}
-    byte2modrm[16] = db_afs_16
-    byte2modrm[32] = db_afs_32
-    byte2modrm[64] = db_afs_64
-
-    modrm2byte = {16: defaultdict(list),
-                  32: defaultdict(list),
-                  64: defaultdict(list),
-                  }
-    for size, db_afs in viewitems(byte2modrm):
-        for i, modrm in enumerate(db_afs):
-            if not isinstance(modrm, list):
-                # We only need sort for determinism
-                modrm = tuple(sorted(viewitems(modrm), key=str))
-                modrm2byte[size][modrm].append(i)
-                continue
-            for j, modrm_f in enumerate(modrm):
-                # We only need sort for determinism
-                modrm_f = tuple(sorted(viewitems(modrm_f), key=str))
-                modrm2byte[size][modrm_f].append((i, j))
-
-    return byte2modrm, modrm2byte
-
-byte2modrm, modrm2byte = gen_modrm_form()
-
-
-# ret is modr; ret is displacement
-def exprfindmod(e, o=None):
-    if o is None:
-        o = {}
-    if isinstance(e, ExprInt):
-        return e
-    if isinstance(e, ExprId):
-        i = size2gpregs[e.size].expr.index(e)
-        o[i] = 1
-        return None
-    elif isinstance(e, ExprOp):
-        out = None
-        if e.op == '+':
-            for a in e.args:
-                r = exprfindmod(a, o)
-                if out and r:
-                    raise ValueError('multiple displacement!')
-                out = r
-            return out
-        elif e.op == "*":
-            mul = int(e.args[1])
-            a = e.args[0]
-            i = size2gpregs[a.size].expr.index(a)
-            o[i] = mul
-        else:
-            raise ValueError('bad op')
-    return None
-
-def test_addr_size(ptr, size):
-    if isinstance(ptr, ExprInt):
-        return int(ptr) < (1 << size)
-    else:
-        return ptr.size == size
-
-SIZE2XMMREG = {64:gpregs_mm,
-               128:gpregs_xmm}
-SIZE2BNDREG = {64:gpregs_mm,
-               128:gpregs_bnd}
-
-def parse_mem(expr, parent, w8, sx=0, xmm=0, mm=0, bnd=0):
-    dct_expr = {}
-    opmode = parent.v_opmode()
-    if is_mem_segm(expr) and expr.ptr.args[0].is_int():
-        return None, None, False
-
-    if is_mem_segm(expr):
-        segm = expr.ptr.args[0]
-        ptr = expr.ptr.args[1]
-    else:
-        segm = None
-        ptr = expr.ptr
-
-    dct_expr[f_isad] = True
-    ad_size = ptr.size
-    admode = parent.v_admode()
-    if not test_addr_size(ptr, admode):
-        return None, None, False
-
-    if (w8 == 1 and expr.size != opmode and not sx and
-        not (hasattr(parent, 'sd') or hasattr(parent, 'wd'))):
-        return None, None, False
-
-    if hasattr(parent, 'wd'):
-        if expr.size == 16:
-            parent.wd.value = 1
-        elif expr.size == 32:
-            pass
-        else:
-            return None, None, False
-
-    if (not isinstance(ptr, ExprInt) and
-        parent.mode == 64 and
-        ptr.size == 32 and
-        parent.admode != 1):
-        return None, None, False
-    dct_expr = {f_isad: True}
-    disp = exprfindmod(ptr, dct_expr)
-    out = []
-    if disp is None:
-        # add 0 disp
-        disp = ExprInt(0, 32)
-    if disp is not None:
-        for signed, encoding, cast_size in [(True, f_s08, 8),
-                                           (True, f_s16, 16),
-                                           (True, f_s32, 32),
-                                           (False, f_u08, 8),
-                                           (False, f_u16, 16),
-                                           (False, f_u32, 32)]:
-            value = ExprInt(int(disp), cast_size)
-            if admode < value.size:
-                if signed:
-                    if int(disp) != sign_ext(int(value), admode, disp.size):
-                        continue
-                else:
-                    if int(disp) != int(value):
-                        continue
-            else:
-                if int(disp) != sign_ext(int(value), value.size, admode):
-                    continue
-            x1 = dict(dct_expr)
-            x1[f_imm] = (encoding, value)
-            out.append(x1)
-    else:
-        out = [dct_expr]
-    return out, segm, True
-
-def expr2modrm(expr, parent, w8, sx=0, xmm=0, mm=0, bnd=0):
-    dct_expr = {f_isad : False}
-
-    if mm or xmm or bnd:
-        if mm and expr.size != 64:
-            return None, None, False
-        elif xmm and expr.size != 128:
-            return None, None, False
-        elif bnd and expr.size != 128:
-            return None, None, False
-
-        if isinstance(expr, ExprId):
-            if bnd:
-                size2reg = SIZE2BNDREG
-            else:
-                size2reg = SIZE2XMMREG
-            selreg = size2reg[expr.size]
-            if not expr in selreg.expr:
-                return None, None, False
-            i = selreg.expr.index(expr)
-            dct_expr[i] = 1
-            return [dct_expr], None, True
-        else:
-            return parse_mem(expr, parent, w8, sx, xmm, mm)
-
-    elif expr.size == 64 and expr not in gpregs_mm.expr:
-        if hasattr(parent, 'sd'):
-            parent.sd.value = 1
-        elif hasattr(parent, 'wd'):
-            pass
-        elif hasattr(parent, 'stk'):
-            pass
-        else:
-            parent.rex_w.value = 1
-    opmode = parent.v_opmode()
-    if sx == 1:
-        opmode = 16
-    if sx == 2:
-        opmode = 32
-    if expr.size == 8 and w8 != 0:
-        return None, None, False
-
-    if w8 == 0 and expr.size != 8:
-        return None, None, False
-
-    if not isinstance(expr, ExprMem):
-        dct_expr[f_isad] = False
-        if xmm:
-            if expr in gpregs_xmm.expr:
-                i = gpregs_xmm.expr.index(expr)
-                dct_expr[i] = 1
-                return [dct_expr], None, True
-            else:
-                return None, None, False
-        if bnd:
-            if expr in gpregs_bnd.expr:
-                i = gpregs_bnd.expr.index(expr)
-                dct_expr[i] = 1
-                return [dct_expr], None, True
-            else:
-                return None, None, False
-        if mm:
-            if expr in gpregs_mm.expr:
-                i = gpregs_mm.expr.index(expr)
-                dct_expr[i] = 1
-                return [dct_expr], None, True
-            else:
-                return None, None, False
-        if w8 == 0:
-            if parent.mode == 64 and expr in gpregs08_64.expr:
-                r = gpregs08_64
-                parent.rex_p.value = 1
-            else:
-                parent.rex_p.value = 0
-                parent.rex_x.value = 0
-                r = size2gpregs[8]
-            if not expr in r.expr:
-                return None, None, False
-            i = r.expr.index(expr)
-            dct_expr[i] = 1
-            return [dct_expr], None, True
-        if opmode != expr.size:
-            return None, None, False
-        if not expr in size2gpregs[opmode].expr:
-            return None, None, False
-        i = size2gpregs[opmode].expr.index(expr)
-        if i > 7:
-            if parent.mode != 64:
-                return None, None, False
-        dct_expr[i] = 1
-        return [dct_expr], None, True
-    return parse_mem(expr, parent, w8, sx, xmm, mm, bnd)
-
-def modrm2expr(modrm, parent, w8, sx=0, xmm=0, mm=0, bnd=0):
-    o = []
-    if not modrm[f_isad]:
-        modrm_k = [key for key, value in viewitems(modrm) if value == 1]
-        if len(modrm_k) != 1:
-            raise ValueError('strange reg encoding %r' % modrm)
-        modrm_k = modrm_k[0]
-        if w8 == 0:
-            opmode = 8
-        elif sx == 1:
-            opmode = 16
-        elif sx == 2:
-            opmode = 32
-        else:
-            opmode = parent.v_opmode()
-        if xmm:
-            expr = gpregs_xmm.expr[modrm_k]
-        elif mm:
-            expr = gpregs_mm.expr[modrm_k]
-        elif bnd:
-            expr = gpregs_bnd.expr[modrm_k]
-        elif opmode == 8 and (parent.v_opmode() == 64 or parent.rex_p.value == 1):
-            expr = gpregs08_64.expr[modrm_k]
-        else:
-            expr = size2gpregs[opmode].expr[modrm_k]
-        return expr
-    admode = parent.v_admode()
-    opmode = parent.v_opmode()
-    for modrm_k, scale in viewitems(modrm):
-        if isinstance(modrm_k, int):
-            expr = size2gpregs[admode].expr[modrm_k]
-            if scale != 1:
-                expr = ExprInt(scale, admode) * expr
-            o.append(expr)
-    if f_imm in modrm:
-        if parent.disp.value is None:
-            return None
-        o.append(ExprInt(int(parent.disp.expr), admode))
-    if len(o) == 1:
-        expr = o[0]
-    else:
-        expr = ExprOp('+', *o)
-    if w8 == 0:
-        opmode = 8
-    elif sx == 1:
-        opmode = 16
-    elif sx == 2:
-        opmode = 32
-    if xmm:
-        opmode = 128
-    elif mm:
-        opmode = 64
-    elif bnd:
-        opmode = 128
-
-    expr = ExprMem(expr, size=opmode)
-    return expr
-
-
-class x86_rm_arg(x86_arg):
-    parser = rmarg
-
-    def fromstring(self, text, loc_db, parser_result=None):
-        start, stop = super(x86_rm_arg, self).fromstring(text, loc_db, parser_result)
-        p = self.parent
-        if start is None:
-            return None, None
-        return start, stop
-
-    def get_modrm(self):
-        p = self.parent
-        admode = p.v_admode()
-
-        if not admode in [16, 32, 64]:
-            raise ValueError('strange admode %r', admode)
-        v = setmodrm(p.mod.value, 0, p.rm.value)
-        v |= p.rex_b.value << 8
-        v |= p.rex_x.value << 9
-        if p.mode == 64:
-            # XXXx to check
-            admode = 64
-
-        xx = byte2modrm[admode][v]
-        if isinstance(xx, list):
-            if not p.sib_scale:
-                return False
-            v = setmodrm(p.sib_scale.value,
-                         p.sib_index.value,
-                         p.sib_base.value)
-            xx = xx[v]
-        return xx
-
-    def decode(self, v):
-        p = self.parent
-        xx = self.get_modrm()
-        self.expr = modrm2expr(xx, p, 1)
-        return self.expr is not None
-
-    def gen_cand(self, v_cand, admode):
-        if not admode in modrm2byte:
-            # XXX TODO: 64bit
-            return
-        if not v_cand:
-            return
-
-        p = self.parent
-        o_rex_x = p.rex_x.value
-        o_rex_b = p.rex_b.value
-        # add candidate without 0 imm
-        new_v_cand = []
-        moddd = False
-        for v in v_cand:
-            new_v_cand.append(v)
-            if f_imm in v and int(v[f_imm][1]) == 0:
-                v = dict(v)
-                del(v[f_imm])
-                new_v_cand.append(v)
-                moddd = True
-
-        v_cand = new_v_cand
-
-        out_c = []
-        for v in v_cand:
-            disp = None
-            # patch value in modrm
-            if f_imm in v:
-                size, disp = v[f_imm]
-                disp = int(disp)
-
-                v[f_imm] = size
-            vo = v
-            # We only need sort for determinism
-            v = tuple(sorted(viewitems(v), key=str))
-            admode = 64 if p.mode == 64 else admode
-            if not v in modrm2byte[admode]:
-                continue
-            xx = modrm2byte[admode][v]
-
-            # default case
-            for x in xx:
-                if type(x) == tuple:
-                    modrm, sib = x
-                else:
-                    modrm = x
-                    sib = None
-
-                # 16 bit cannot have sib
-                if sib is not None and admode == 16:
-                    continue
-                rex = modrm >> 8  # 0# XXX HACK REM temporary REX modrm>>8
-                if rex and admode != 64:
-                    continue
-
-                p.rex_x.value = (rex >> 1) & 1
-                p.rex_b.value = rex & 1
-
-                if o_rex_x is not None and p.rex_x.value != o_rex_x:
-                    continue
-                if o_rex_b is not None and p.rex_b.value != o_rex_b:
-                    continue
-
-                mod, re, rm = getmodrm(modrm)
-                # check re on parent
-                if hasattr(p, 'reg') and re != p.reg.value:
-                    continue
-
-                if sib is not None:
-                    s_scale, s_index, s_base = getmodrm(sib)
-                else:
-                    s_scale, s_index, s_base = None, None, None
-
-                p.mod.value = mod
-                p.rm.value = rm
-                p.sib_scale.value = s_scale
-                p.sib_index.value = s_index
-                p.sib_base.value = s_base
-                p.disp.value = disp
-                if disp is not None:
-                    p.disp.l = f_imm2size[vo[f_imm]]
-
-                yield True
-
-        return
-
-    def encode(self):
-        if isinstance(self.expr, ExprInt):
-            return
-        p = self.parent
-        admode = p.v_admode()
-        mode = self.expr.size
-        v_cand, segm, ok = expr2modrm(self.expr, p, 1)
-        if segm:
-            p.g2.value = segm2enc[segm]
-        for x in self.gen_cand(v_cand, admode):
-            yield x
-
-class x86_rm_mem(x86_rm_arg):
-    def fromstring(self, text, loc_db, parser_result=None):
-        self.expr = None
-        start, stop = super(x86_rm_mem, self).fromstring(text, loc_db, parser_result)
-        if not isinstance(self.expr, ExprMem):
-            return None, None
-        return start, stop
-
-
-class x86_rm_mem_far(x86_rm_arg):
-    parser = mem_far
-    def fromstring(self, text, loc_db, parser_result=None):
-        self.expr = None
-        start, stop = super(x86_rm_mem_far, self).fromstring(text, loc_db, parser_result)
-        if not isinstance(self.expr, ExprMem):
-            return None, None
-        self.expr = ExprOp('far', self.expr)
-        return start, stop
-
-    def decode(self, v):
-        ret = super(x86_rm_mem_far, self).decode(v)
-        if not ret:
-            return ret
-        if isinstance(self.expr, m2_expr.ExprMem):
-            self.expr = ExprOp('far', self.expr)
-        return True
-
-    def encode(self):
-        if not (isinstance(self.expr, m2_expr.ExprOp) and
-                self.expr.op == 'far'):
-            return
-
-        expr = self.expr.args[0]
-        if isinstance(expr, ExprInt):
-            return
-        p = self.parent
-        admode = p.v_admode()
-        mode = expr.size
-        v_cand, segm, ok = expr2modrm(expr, p, 1)
-        if segm:
-            p.g2.value = segm2enc[segm]
-        for x in self.gen_cand(v_cand, admode):
-            yield x
-
-class x86_rm_w8(x86_rm_arg):
-
-    def decode(self, v):
-        p = self.parent
-        xx = self.get_modrm()
-        self.expr = modrm2expr(xx, p, p.w8.value)
-        return self.expr is not None
-
-    def encode(self):
-        if isinstance(self.expr, ExprInt):
-            return
-        p = self.parent
-        if p.w8.value is None:
-            if self.expr.size == 8:
-                p.w8.value = 0
-            else:
-                p.w8.value = 1
-
-        v_cand, segm, ok = expr2modrm(self.expr, p, p.w8.value)
-        if segm:
-            p.g2.value = segm2enc[segm]
-        for x in self.gen_cand(v_cand, p.v_admode()):
-            yield x
-
-
-class x86_rm_sx(x86_rm_arg):
-
-    def decode(self, v):
-        p = self.parent
-        xx = self.get_modrm()
-        self.expr = modrm2expr(xx, p, p.w8.value, 1)
-        return self.expr is not None
-
-    def encode(self):
-        if isinstance(self.expr, ExprInt):
-            return
-        p = self.parent
-        if p.w8.value is None:
-            if self.expr.size == 8:
-                p.w8.value = 0
-            else:
-                p.w8.value = 1
-        v_cand, segm, ok = expr2modrm(self.expr, p, p.w8.value, 1)
-        if segm:
-            p.g2.value = segm2enc[segm]
-        for x in self.gen_cand(v_cand, p.v_admode()):
-            yield x
-
-
-class x86_rm_sxd(x86_rm_arg):
-
-    def decode(self, v):
-        p = self.parent
-        xx = self.get_modrm()
-        self.expr = modrm2expr(xx, p, 1, 2)
-        return self.expr is not None
-
-    def encode(self):
-        if isinstance(self.expr, ExprInt):
-            return
-        p = self.parent
-        v_cand, segm, ok = expr2modrm(self.expr, p, 1, 2)
-        if segm:
-            p.g2.value = segm2enc[segm]
-        for x in self.gen_cand(v_cand, p.v_admode()):
-            yield x
-
-
-class x86_rm_sd(x86_rm_arg):
-    out_size = 64
-    def get_s_value(self):
-        return self.parent.sd.value
-    def set_s_value(self, value):
-        self.parent.sd.value = value
-
-    def decode(self, v):
-        p = self.parent
-        xx = self.get_modrm()
-        expr = modrm2expr(xx, p, 1)
-        if not isinstance(expr, ExprMem):
-            return False
-        if self.get_s_value() == 0:
-            expr = ExprMem(expr.ptr, 32)
-        else:
-            expr = ExprMem(expr.ptr, self.out_size)
-        self.expr = expr
-        return self.expr is not None
-
-    def encode(self):
-        if isinstance(self.expr, ExprInt):
-            return
-        p = self.parent
-        if not self.expr.size in [32, 64]:
-            return
-        self.set_s_value(0)
-        v_cand, segm, ok = expr2modrm(self.expr, p, 1)
-        for x in self.gen_cand(v_cand, p.v_admode()):
-            yield x
-
-
-class x86_rm_wd(x86_rm_sd):
-    out_size = 16
-    def get_s_value(self):
-        return self.parent.wd.value
-    def set_s_value(self, value):
-        self.parent.wd.value = value
-
-    def encode(self):
-        if isinstance(self.expr, ExprInt):
-            return
-        p = self.parent
-        p.wd.value = 0
-        v_cand, segm, ok = expr2modrm(self.expr, p, 1)
-        for x in self.gen_cand(v_cand, p.v_admode()):
-            yield x
-
-
-class x86_rm_08(x86_rm_arg):
-    msize = 8
-
-    def decode(self, v):
-        p = self.parent
-        xx = self.get_modrm()
-        expr = modrm2expr(xx, p, 0)
-        if not isinstance(expr, ExprMem):
-            self.expr = expr
-            return True
-        self.expr = ExprMem(expr.ptr, self.msize)
-        return self.expr is not None
-
-    def encode(self):
-        if isinstance(self.expr, ExprInt):
-            return
-        p = self.parent
-        v_cand, segm, ok = expr2modrm(self.expr, p, 0, 0, 0, 0)
-        for x in self.gen_cand(v_cand, p.v_admode()):
-            yield x
-
-class x86_rm_reg_m08(x86_rm_arg):
-    msize = 8
-
-    def decode(self, v):
-        ret = x86_rm_arg.decode(self, v)
-        if not ret:
-            return ret
-        if not isinstance(self.expr, ExprMem):
-            return True
-        self.expr = ExprMem(self.expr.ptr, self.msize)
-        return self.expr is not None
-
-    def encode(self):
-        if isinstance(self.expr, ExprInt):
-            return
-        p = self.parent
-        if isinstance(self.expr, ExprMem):
-            expr = ExprMem(self.expr.ptr, 32)
-        else:
-            expr = self.expr
-        v_cand, segm, ok = expr2modrm(expr, p, 1, 0, 0, 0)
-        for x in self.gen_cand(v_cand, p.v_admode()):
-            yield x
-
-class x86_rm_reg_m16(x86_rm_reg_m08):
-    msize = 16
-
-class x86_rm_m64(x86_rm_arg):
-    msize = 64
-
-    def decode(self, v):
-        p = self.parent
-        xx = self.get_modrm()
-        expr = modrm2expr(xx, p, 1)
-        if not isinstance(expr, ExprMem):
-            return False
-        self.expr = ExprMem(expr.ptr, self.msize)
-        return self.expr is not None
-
-    def encode(self):
-        if isinstance(self.expr, ExprInt):
-            return
-        p = self.parent
-        v_cand, segm, ok = expr2modrm(self.expr, p, 0, 0, 0, 1)
-        for x in self.gen_cand(v_cand, p.v_admode()):
-            yield x
-
-
-class x86_rm_m80(x86_rm_m64):
-    msize = 80
-
-    def encode(self):
-        if isinstance(self.expr, ExprInt):
-            return
-        if not isinstance(self.expr, ExprMem) or self.expr.size != self.msize:
-            return
-        p = self.parent
-        mode = p.mode
-        if mode == 64:
-            mode = 32
-        self.expr = ExprMem(self.expr.ptr, mode)
-        v_cand, segm, ok = expr2modrm(self.expr, p, 1)
-        for x in self.gen_cand(v_cand, p.v_admode()):
-            yield x
-
-
-class x86_rm_m08(x86_rm_arg):
-    msize = 8
-
-    def decode(self, v):
-        p = self.parent
-        xx = self.get_modrm()
-        self.expr = modrm2expr(xx, p, 0)
-        return self.expr is not None
-
-    def encode(self):
-        if self.expr.size != 8:
-            return
-        p = self.parent
-        mode = p.mode
-        v_cand, segm, ok = expr2modrm(self.expr, p, 0)
-        for x in self.gen_cand(v_cand, p.v_admode()):
-            yield x
-
-
-class x86_rm_m16(x86_rm_m80):
-    msize = 16
-
-
-class x86_rm_mm(x86_rm_m80):
-    msize = 64
-    is_mm = True
-    is_xmm = False
-    is_bnd = False
-
-    def decode(self, v):
-        p = self.parent
-        xx = self.get_modrm()
-        expr = modrm2expr(xx, p, 0, 0, self.is_xmm, self.is_mm, self.is_bnd)
-        if isinstance(expr, ExprMem):
-            if self.msize is None:
-                return False
-            if expr.size != self.msize:
-                expr = ExprMem(expr.ptr, self.msize)
-        self.expr = expr
-        return True
-
-
-    def encode(self):
-        expr = self.expr
-        if isinstance(expr, ExprInt):
-            return
-        if isinstance(expr, ExprMem) and expr.size != self.msize:
-            return
-        p = self.parent
-        mode = p.mode
-        if mode == 64:
-            mode = 32
-        if isinstance(expr, ExprMem):
-            if self.is_xmm:
-                expr = ExprMem(expr.ptr, 128)
-            elif self.is_mm:
-                expr = ExprMem(expr.ptr, 64)
-
-        v_cand, segm, ok = expr2modrm(expr, p, 0, 0, self.is_xmm, self.is_mm,
-                                      self.is_bnd)
-        for x in self.gen_cand(v_cand, p.v_admode()):
-            yield x
-
-
-class x86_rm_mm_m64(x86_rm_mm):
-    msize = 64
-    is_mm = True
-    is_xmm = False
-
-class x86_rm_xmm(x86_rm_mm):
-    msize = 128
-    is_mm = False
-    is_xmm = True
-
-
-class x86_rm_xmm_m32(x86_rm_mm):
-    msize = 32
-    is_mm = False
-    is_xmm = True
-
-class x86_rm_xmm_m64(x86_rm_mm):
-    msize = 64
-    is_mm = False
-    is_xmm = True
-
-class x86_rm_xmm_m128(x86_rm_mm):
-    msize = 128
-    is_mm = False
-    is_xmm = True
-
-
-class x86_rm_xmm_reg(x86_rm_mm):
-    msize = None
-    is_mm = False
-    is_xmm = True
-
-class x86_rm_mm_reg(x86_rm_mm):
-    msize = None
-    is_mm = True
-    is_xmm = False
-
-
-class x86_rm_bnd(x86_rm_mm):
-    msize = 128
-    is_mm = False
-    is_xmm = False
-    is_bnd = True
-
-
-class x86_rm_bnd_reg(x86_rm_mm):
-    msize = None
-    is_mm = False
-    is_xmm = False
-    is_bnd = True
-
-
-class x86_rm_bnd_m64(x86_rm_mm):
-    msize = 64
-    is_mm = False
-    is_xmm = False
-    is_bnd = True
-
-
-class x86_rm_bnd_m128(x86_rm_mm):
-    msize = 128
-    is_mm = False
-    is_xmm = False
-    is_bnd = True
-
-
-class x86_rm_reg_noarg(object):
-    prio = default_prio + 1
-
-    parser = gpreg
-
-    def fromstring(self, text, loc_db, parser_result=None):
-        if not hasattr(self.parent, 'sx') and hasattr(self.parent, "w8"):
-            self.parent.w8.value = 1
-        if parser_result:
-            result, start, stop = parser_result[self.parser]
-            if result == [None]:
-                return None, None
-            self.expr = result
-            if self.expr.size == 8:
-                if hasattr(self.parent, 'sx') or not hasattr(self.parent, 'w8'):
-                    return None, None
-                self.parent.w8.value = 0
-            return start, stop
-        try:
-            result, start, stop = next(self.parser.scanString(text))
-        except StopIteration:
-            return None, None
-        expr = self.asm_ast_to_expr(result[0], loc_db)
-        if expr is None:
-            return None, None
-
-        self.expr = expr
-        if self.expr.size == 0:
-            if hasattr(self.parent, 'sx') or not hasattr(self.parent, 'w8'):
-                return None, None
-            self.parent.w8.value = 0
-
-        return start, stop
-
-    def getrexsize(self):
-        return self.parent.rex_r.value
-
-    def setrexsize(self, v):
-        self.parent.rex_r.value = v
-
-    def decode(self, v):
-        v = v & self.lmask
-        p = self.parent
-        opmode = p.v_opmode()
-        if not hasattr(p, 'sx') and (hasattr(p, 'w8') and p.w8.value == 0):
-            opmode = 8
-        r = size2gpregs[opmode]
-        if p.mode == 64 and self.getrexsize():
-            v |= 0x8
-        if p.v_opmode() == 64 or p.rex_p.value == 1:
-            if not hasattr(p, 'sx') and (hasattr(p, 'w8') and p.w8.value == 0):
-                r = gpregs08_64
-            elif p.rex_r.value == 1:
-                v |= 8
-        self.expr = r.expr[v]
-        return True
-
-    def encode(self):
-        if not isinstance(self.expr, ExprId):
-            return False
-        if self.expr in gpregs64.expr and not hasattr(self.parent, 'stk'):
-            self.parent.rex_w.value = 1
-        opmode = self.parent.v_opmode()
-        if not hasattr(self.parent, 'sx') and hasattr(self.parent, 'w8'):
-            self.parent.w8.value = 1
-        if self.expr.size == 8:
-            if hasattr(self.parent, 'sx') or not hasattr(self.parent, 'w8'):
-                return False
-            self.parent.w8.value = 0
-            opmode = 8
-        r = size2gpregs[opmode]
-        if self.expr in r.expr:
-            i = r.expr.index(self.expr)
-        elif (opmode == 8 and self.parent.mode == 64 and
-            self.expr in gpregs08_64.expr):
-            i = gpregs08_64.expr.index(self.expr)
-            self.parent.rex_p.value = 1
-        else:
-            log.debug("cannot encode reg %r", self.expr)
-            return False
-        if self.parent.v_opmode() == 64:
-            if i > 7:
-                self.setrexsize(1)
-                i -= 8
-        elif self.parent.mode == 64 and i > 7:
-            i -= 8
-            self.setrexsize(1)
-        self.value = i
-        return True
-
-
-class x86_rm_reg_mm(x86_rm_reg_noarg, x86_arg):
-    selreg = gpregs_mm
-    def decode(self, v):
-        if self.parent.mode == 64 and self.getrexsize():
-            v |= 0x8
-        self.expr = self.selreg.expr[v]
-        return True
-
-    def encode(self):
-        if not isinstance(self.expr, ExprId):
-            return False
-        if self.expr not in self.selreg.expr:
-            return False
-        i = self.selreg.expr.index(self.expr)
-        if self.parent.mode == 64 and i > 7:
-            i -= 8
-            self.setrexsize(1)
-        self.value = i
-        return True
-
-class x86_rm_reg_xmm(x86_rm_reg_mm):
-    selreg = gpregs_xmm
-
-class x86_rm_reg_bnd(x86_rm_reg_mm):
-    selreg = gpregs_bnd
-
-class x86_rm_reg(x86_rm_reg_noarg, x86_arg):
-    pass
-
-
-class x86_reg(x86_rm_reg):
-
-    def getrexsize(self):
-        return self.parent.rex_b.value
-
-    def setrexsize(self, v):
-        self.parent.rex_b.value = v
-
-class x86_vex_reg(x86_rm_reg):
-    # self.lmask = 15
-
-    def decode(self, v):
-        p = self.parent
-        
-        self.expr = size2gpregs[v_opmode(p)].expr[p.vex_v.value]
-        
-        return self.expr is not None
-
-    def encode(self):
-        opmode = self.parent.mode
-        size = self.expr.size
-
-        if opmode == 64 and size == 64:
-            self.parent.rex_w.value = 1
-        else:
-            self.parent.rex_w.value = 0
-
-        r = size2gpregs[size]
-        if self.expr in r.expr:
-            i = r.expr.index(self.expr)
-
-        self.parent.vex_v.value = i
-        self.parent.vex.value = 1
-        return True
-
-
-class x86_reg_modrm(x86_rm_reg):
-
-    def getrexsize(self):
-        return self.parent.rex_r.value
-
-    def setrexsize(self, v):
-        self.parent.rex_r.value = v
-
-
-
-class x86_reg_noarg(x86_rm_reg_noarg):
-
-    def getrexsize(self):
-        return self.parent.rex_b.value
-
-    def setrexsize(self, v):
-        self.parent.rex_b.value = v
-
-
-class x86_rm_segm(reg_noarg, x86_arg):
-    prio = default_prio + 1
-    reg_info = segmreg
-    parser = reg_info.parser
-
-
-class x86_rm_cr(reg_noarg, x86_arg):
-    prio = default_prio + 1
-    reg_info = crregs
-    parser = reg_info.parser
-
-
-class x86_rm_dr(reg_noarg, x86_arg):
-    prio = default_prio + 1
-    reg_info = drregs
-    parser = reg_info.parser
-
-
-class x86_rm_flt(reg_noarg, x86_arg):
-    prio = default_prio + 1
-    reg_info = fltregs
-    parser = reg_info.parser
-
-
-class bs_fbit(bsi):
-
-    def decode(self, v):
-        # value already decoded in pre_dis_info
-        return True
-
-
-class bs_cl1(bsi, x86_arg):
-    parser = cl_or_imm
-
-    def decode(self, v):
-        if v == 1:
-            self.expr = regs08_expr[1]
-        else:
-            self.expr = ExprInt(1, 8)
-        return True
-
-    def encode(self):
-        if self.expr == regs08_expr[1]:
-            self.value = 1
-        elif isinstance(self.expr, ExprInt) and int(self.expr) == 1:
-            self.value = 0
-        else:
-            return False
-        return True
-
-
-def sib_cond(cls, mode, v):
-    if admode_prefix((mode, v["opmode"], v["admode"])) == 16:
-        return None
-    if v['mod'] == 0b11:
-        return None
-    elif v['rm'] == 0b100:
-        return cls.ll
-    else:
-        return None
-    return v['rm'] == 0b100
-
-
-class bs_cond_scale(bs_cond):
-    # cond must return field len
-    ll = 2
-
-    @classmethod
-    def flen(cls, mode, v):
-        return sib_cond(cls, mode, v)
-
-    def encode(self):
-        if self.value is None:
-            self.value = 0
-            self.l = 0
-            return True
-        return super(bs_cond_scale, self).encode()
-
-    def decode(self, v):
-        self.value = v
-        return True
-
-
-class bs_cond_index(bs_cond_scale):
-    ll = 3
-
-    @classmethod
-    def flen(cls, mode, v):
-        return sib_cond(cls, mode, v)
-
-
-class bs_cond_disp(bs_cond):
-    # cond must return field len
-
-    @classmethod
-    def flen(cls, mode, v):
-        if admode_prefix((mode, v['opmode'], v['admode'])) == 16:
-            if v['mod'] == 0b00:
-                if v['rm'] == 0b110:
-                    return 16
-                else:
-                    return None
-            elif v['mod'] == 0b01:
-                return 8
-            elif v['mod'] == 0b10:
-                return 16
-            return None
-        # 32, 64
-        if 'sib_base' in v and v['sib_base'] == 0b101:
-            if v['mod'] == 0b00:
-                return 32
-            elif v['mod'] == 0b01:
-                return 8
-            elif v['mod'] == 0b10:
-                return 32
-            else:
-                return None
-
-        if v['mod'] == 0b00:
-            if v['rm'] == 0b101:
-                return 32
-            else:
-                return None
-        elif v['mod'] == 0b01:
-            return 8
-        elif v['mod'] == 0b10:
-            return 32
-        else:
-            return None
-
-    def encode(self):
-        if self.value is None:
-            self.value = 0
-            self.l = 0
-            return True
-        self.value = swap_uint(self.l, self.value)
-        return True
-
-    def decode(self, v):
-        admode = self.parent.v_admode()
-        v = swap_uint(self.l, v)
-        self.value = v
-        v = sign_ext(v, self.l, admode)
-        v = ExprInt(v, admode)
-        self.expr = v
-        return True
-
-
-class bs_cond_imm(bs_cond_scale, x86_arg):
-    parser = base_expr
-    max_size = 32
-
-    def fromstring(self, text, loc_db, parser_result=None):
-        if parser_result:
-            expr, start, stop = parser_result[self.parser]
-        else:
-            try:
-                expr, start, stop = next(self.parser.scanString(text))
-            except StopIteration:
-                expr = None
-        self.expr = expr
-
-        if len(self.parent.args) > 1:
-            l = self.parent.args[0].expr.size
-        else:
-            l = self.parent.v_opmode()
-        if isinstance(self.expr, ExprInt):
-            v = int(self.expr)
-            mask = ((1 << l) - 1)
-            self.expr = ExprInt(v & mask, l)
-
-        if self.expr is None:
-            log.debug('cannot fromstring int %r', text)
-            return None, None
-        return start, stop
-
-    @classmethod
-    def flen(cls, mode, v):
-        if 'w8' not in v or v['w8'] == 1:
-            if 'se' in v and v['se'] == 1:
-                return 8
-            else:
-                osize = v_opmode_info(mode, v['opmode'], v['rex_w'], 0)
-                osize = min(osize, cls.max_size)
-                return osize
-        return 8
-
-    def getmaxlen(self):
-        return 32
-
-    def encode(self):
-        if not isinstance(self.expr, ExprInt):
-            return
-        arg0_expr = self.parent.args[0].expr
-        self.parent.rex_w.value = 0
-        # special case for push
-        if len(self.parent.args) == 1:
-            v = int(self.expr)
-            l = self.parent.v_opmode()
-            l = min(l, self.max_size)
-
-            self.l = l
-            mask = ((1 << self.l) - 1)
-            if v != sign_ext(v & mask, self.l, l):
-                return
-            self.value = swap_uint(self.l, v & ((1 << self.l) - 1))
-            yield True
-            return
-
-        # assume 2 args; use first arg to guess op size
-        if arg0_expr.size == 64:
-            self.parent.rex_w.value = 1
-
-        l = self.parent.v_opmode()
-        v = int(self.expr)
-        if arg0_expr.size == 8:
-            if not hasattr(self.parent, 'w8'):
-                return
-            self.parent.w8.value = 0
-            l = 8
-            if hasattr(self.parent, 'se'):
-                self.parent.se.value = 0
-        elif hasattr(self.parent, 'se'):
-            if hasattr(self.parent, 'w8'):
-                self.parent.w8.value = 1
-            # try to generate signed extended version
-            if v == sign_ext(v & 0xFF, 8, arg0_expr.size):
-                self.parent.se.value = 1
-                self.l = 8
-                self.value = v & 0xFF
-                yield True
-            self.parent.se.value = 0
-        else:
-            if hasattr(self.parent, 'w8'):
-                self.parent.w8.value = 1
-        if l == 64:
-            self.l = self.getmaxlen()
-        else:
-            self.l = l
-
-        mask = ((1 << self.l) - 1)
-        if v != sign_ext(v & mask, self.l, l):
-            return
-        self.value = swap_uint(self.l, v & ((1 << self.l) - 1))
-        yield True
-
-    def decode(self, v):
-        opmode = self.parent.v_opmode()
-        v = swap_uint(self.l, v)
-        self.value = v
-        l_out = opmode
-        if hasattr(self.parent, 'w8') and self.parent.w8.value == 0:
-            l_out = 8
-        v = sign_ext(v, self.l, l_out)
-        self.expr = ExprInt(v, l_out)
-        return True
-
-
-class bs_cond_imm64(bs_cond_imm):
-    max_size = 64
-
-    def getmaxlen(self):
-        return 64
-
-    @classmethod
-    def flen(cls, mode, v):
-        if 'w8' not in v or v['w8'] == 1:
-            if 'se' in v and v['se'] == 1:
-                return 8
-            else:
-                osize = v_opmode_info(mode, v['opmode'], v['rex_w'], 0)
-                return osize
-        else:
-            return 8
-
-
-class bs_rel_off(bs_cond_imm):
-    parser = base_expr
-
-    def fromstring(self, text, loc_db, parser_result=None):
-        if parser_result:
-            expr, start, stop = parser_result[self.parser]
-        else:
-            try:
-                expr, start, stop = next(self.parser.scanString(text))
-            except StopIteration:
-                expr = None
-        self.expr = expr
-        l = self.parent.mode
-        if isinstance(self.expr, ExprInt):
-            v = int(self.expr)
-            mask = ((1 << l) - 1)
-            self.expr = ExprInt(v & mask, l)
-        return start, stop
-
-    @classmethod
-    def flen(cls, mode, v):
-        osize = v_opmode_info(mode, v['opmode'], v['rex_w'], 0)
-        if osize == 16:
-            return 16
-        else:
-            return 32
-
-    def encode(self):
-        if not isinstance(self.expr, ExprInt):
-            return
-        arg0_expr = self.parent.args[0].expr
-        if self.l == 0:
-            l = self.parent.v_opmode()
-            self.l = l
-        l = offsize(self.parent)
-        prefix = self.parent.gen_prefix()
-        parent_len = len(prefix) * 8 + self.parent.l + self.l
-        assert(parent_len % 8 == 0)
-
-        v = int(self.expr) - parent_len // 8
-        if prefix is None:
-            return
-        mask = ((1 << self.l) - 1)
-        if self.l > l:
-            return
-        if v != sign_ext(v & mask, self.l, l):
-            return
-        self.value = swap_uint(self.l, v & ((1 << self.l) - 1))
-        yield True
-
-    def decode(self, v):
-        v = swap_uint(self.l, v)
-        size = offsize(self.parent)
-        v = sign_ext(v, self.l, size)
-        v += self.parent.l
-        self.expr = ExprInt(v, size)
-        return True
-
-class bs_s08(bs_rel_off):
-    parser = base_expr
-
-    @classmethod
-    def flen(cls, mode, v):
-        return 8
-
-    def encode(self):
-        if not isinstance(self.expr, ExprInt):
-            return
-        arg0_expr = self.parent.args[0].expr
-        if self.l != 0:
-            l = self.l
-        else:
-            l = self.parent.v_opmode()
-            self.l = l
-        l = offsize(self.parent)
-        v = int(self.expr)
-        mask = ((1 << self.l) - 1)
-        if self.l > l:
-            return
-        if v != sign_ext(v & mask, self.l, l):
-            return
-        self.value = swap_uint(self.l, v & ((1 << self.l) - 1))
-        yield True
-
-    def decode(self, v):
-        v = swap_uint(self.l, v)
-        size = offsize(self.parent)
-        v = sign_ext(v, self.l, size)
-        self.expr = ExprInt(v, size)
-        return True
-
-
-class bs_rel_off08(bs_rel_off):
-
-    @classmethod
-    def flen(cls, mode, v):
-        return 8
-
-
-class bs_moff(bsi):
-
-    @classmethod
-    def flen(cls, mode, v):
-        osize = v_opmode_info(mode, v['opmode'], v['rex_w'], 0)
-        if osize == 16:
-            return 16
-        else:
-            return 32
-
-    def encode(self):
-        if not hasattr(self.parent, "mseg"):
-            return
-        m = self.parent.mseg.expr
-        if not (isinstance(m, ExprOp) and m.op == 'segm'):
-            return
-        if not isinstance(m.args[1], ExprInt):
-            return
-        l = self.parent.v_opmode()
-        if l == 16:
-            self.l = 16
-        else:
-            self.l = 32
-        v = int(m.args[1])
-        mask = ((1 << self.l) - 1)
-        if v != sign_ext(v & mask, self.l, l):
-            return
-        self.value = swap_uint(self.l, v & ((1 << self.l) - 1))
-        yield True
-
-    def decode(self, v):
-        opmode = self.parent.v_opmode()
-        if opmode == 64:
-            return False
-        v = swap_uint(self.l, v)
-        self.value = v
-        v = sign_ext(v, self.l, opmode)
-        self.expr = ExprInt(v, opmode)
-        return True
-
-
-class bs_movoff(x86_arg):
-    parser = deref_mem
-
-    def fromstring(self, text, loc_db, parser_result=None):
-        if parser_result:
-            e, start, stop = parser_result[self.parser]
-            if e is None:
-                return None, None
-            if not isinstance(e, ExprMem):
-                return None, None
-            self.expr = e
-            if self.expr is None:
-                return None, None
-            return start, stop
-        try:
-            v, start, stop = next(self.parser.scanString(text))
-        except StopIteration:
-            return None, None
-        if not isinstance(e, ExprMem):
-            return None, None
-        self.expr = v[0]
-        if self.expr is None:
-            log.debug('cannot fromstring int %r', text)
-            return None, None
-        return start, stop
-
-    @classmethod
-    def flen(cls, mode, v):
-        if mode == 64:
-            if v['admode']:
-                return 32
-            else:
-                return 64
-        asize = v_admode_info(mode, v['admode'])
-        return asize
-
-    def encode(self):
-        p = self.parent
-        if not isinstance(self.expr, ExprMem) or not isinstance(self.expr.ptr, ExprInt):
-            return
-        self.l = p.v_admode()
-        v = int(self.expr.ptr)
-        mask = ((1 << self.l) - 1)
-        if v != mask & v:
-            return
-        self.value = swap_uint(self.l, v & ((1 << self.l) - 1))
-        yield True
-
-    def decode(self, v):
-        if self.parent.mode == 64:
-            if self.parent.admode == 1:
-                l = 32
-            else:
-                l = 64
-        else:
-            l = self.parent.v_admode()
-        v = swap_uint(self.l, v)
-        self.value = v
-        v = sign_ext(v, self.l, l)
-        v = ExprInt(v, l)
-        size = self.parent.v_opmode()
-        if self.parent.w8.value == 0:
-            size = 8
-        self.expr = ExprMem(v, size)
-        return True
-
-
-class bs_msegoff(x86_arg):
-    parser = deref_ptr
-
-    def fromstring(self, text, loc_db, parser_result=None):
-        if parser_result:
-            e, start, stop = parser_result[self.parser]
-            if e is None:
-                return None, None
-            self.expr = e
-            if self.expr is None:
-                return None, None
-            return start, stop
-        try:
-            v, start, stop = next(self.parser.scanString(text))
-        except StopIteration:
-            return None, None
-        self.expr = v[0]
-        if self.expr is None:
-            log.debug('cannot fromstring int %r', text)
-            return None, None
-        return start, stop
-
-    def encode(self):
-        if not (isinstance(self.expr, ExprOp) and self.expr.op == 'segm'):
-            return
-        if not isinstance(self.expr.args[0], ExprInt):
-            return
-        if not isinstance(self.expr.args[1], ExprInt):
-            return
-        l = self.parent.v_opmode()
-        v = int(self.expr.args[0])
-        mask = ((1 << self.l) - 1)
-        if v != sign_ext(v & mask, self.l, l):
-            return
-        self.value = swap_uint(self.l, v & ((1 << self.l) - 1))
-        yield True
-
-    def decode(self, v):
-        opmode = self.parent.v_opmode()
-        v = swap_uint(self.l, v)
-        self.value = v
-        v = ExprInt(v, 16)
-        self.expr = ExprOp('segm', v, self.parent.off.expr)
-        return True
-
-
-d_rex_p = bs(l=0, cls=(bs_fbit,), fname="rex_p")
-d_rex_w = bs(l=0, cls=(bs_fbit,), fname="rex_w")
-d_rex_r = bs(l=0, cls=(bs_fbit,), fname="rex_r")
-d_rex_x = bs(l=0, cls=(bs_fbit,), fname="rex_x")
-d_rex_b = bs(l=0, cls=(bs_fbit,), fname="rex_b")
-
-d_vex = bs(l=0, cls=(bs_fbit,), fname="vex")
-d_vex_l = bs(l=0, cls=(bs_fbit,), fname="vex_l")
-d_vex_p = bs(l=0, cls=(bs_fbit,), fname="vex_p")
-d_vex_v = bs(l=0, cls=(bs_fbit,), fname="vex_v")
-d_vex_m = bs(l=0, cls=(bs_fbit,), fname="vex_m")
-
-pref_0f = bs(l=0, fname="pref_0f")
-pref_0f38 = bs(l=0, fname="pref_0f38")
-pref_0f3a = bs(l=0, fname="pref_0f3a")
-
-d_g1 = bs(l=0, cls=(bs_fbit,), fname="g1")
-d_g2 = bs(l=0, cls=(bs_fbit,), fname="g2")
-
-
-d_cl1 = bs(l=1, cls=(bs_cl1,), fname="cl1")
-
-
-w8 = bs(l=1, fname="w8")
-se = bs(l=1, fname="se")
-
-sx = bs(l=0, fname="sx")
-sxd = bs(l=0, fname="sx")
-
-
-xmmreg = bs(l=0, fname="xmmreg")
-mmreg = bs(l=0, fname="mmreg")
-
-pref_f2 = bs(l=0, fname="prefixed", default=b"\xf2")
-pref_f3 = bs(l=0, fname="prefixed", default=b"\xf3")
-pref_66 = bs(l=0, fname="prefixed", default=b"\x66")
-no_xmm_pref = bs(l=0, fname="no_xmm_pref")
-
-no_rex = bs(l=0, fname="no_rex")
-no_rep = bs(l=0, fname="no_rep")
-
-sib_scale = bs(l=2, cls=(bs_cond_scale,), fname = "sib_scale")
-sib_index = bs(l=3, cls=(bs_cond_index,), fname = "sib_index")
-sib_base = bs(l=3, cls=(bs_cond_index,), fname = "sib_base")
-
-disp = bs(l=0, cls=(bs_cond_disp,), fname = "disp")
-
-s08 = bs(l=8, cls=(bs_s08, ))
-
-u08 = bs(l=8, cls=(x86_08, x86_arg))
-u07 = bs(l=7, cls=(x86_08, x86_arg))
-u16 = bs(l=16, cls=(x86_16, x86_arg))
-u32 = bs(l=32, cls=(x86_32, x86_arg))
-s3264 = bs(l=32, cls=(x86_s32to64, x86_arg))
-
-u08_3 = bs(l=0, cls=(x86_imm_fix_08, x86_arg), ival = 3)
-
-d0 = bs("000", fname='reg')
-d1 = bs("001", fname='reg')
-d2 = bs("010", fname='reg')
-d3 = bs("011", fname='reg')
-d4 = bs("100", fname='reg')
-d5 = bs("101", fname='reg')
-d6 = bs("110", fname='reg')
-d7 = bs("111", fname='reg')
-
-sd = bs(l=1, fname="sd")
-wd = bs(l=1, fname="wd")
-
-stk = bs(l=0, fname="stk")
-
-
-class field_size(object):
-    prio = default_prio
-
-    def __init__(self, d=None):
-        if d is None:
-            d = {}
-        self.d = d
-
-    def get(self, opm, adm=None):
-        return self.d[opm]
-
-class bs_mem(object):
-    def encode(self):
-        return self.value != 0b11
-
-    def decode(self, v):
-        self.value = v
-        return v != 0b11
-
-class bs_reg(object):
-    def encode(self):
-        return self.value == 0b11
-
-    def decode(self, v):
-        self.value = v
-        return v == 0b11
-
-d_imm64 = bs(l=0, fname="imm64")
-
-d_eax = bs(l=0, cls=(bs_eax, ), fname='eax')
-d_edx = bs(l=0, cls=(bs_edx, ), fname='edx')
-d_st = bs(l=0, cls=(x86_reg_st, ), fname='st')
-d_imm = bs(l=0, cls=(bs_cond_imm,), fname="imm")
-d_imm64 = bs(l=0, cls=(bs_cond_imm64,), fname="imm")
-d_ax = bs(l=0, cls=(r_ax, ), fname='ax')
-d_dx = bs(l=0, cls=(r_dx, ), fname='dx')
-d_cl = bs(l=0, cls=(r_cl, ), fname='cl')
-
-d_cs = bs(l=0, cls=(bs_cs, ), fname='cs')
-d_ds = bs(l=0, cls=(bs_ds, ), fname='ds')
-d_es = bs(l=0, cls=(bs_es, ), fname='es')
-d_ss = bs(l=0, cls=(bs_ss, ), fname='ss')
-d_fs = bs(l=0, cls=(bs_fs, ), fname='fs')
-d_gs = bs(l=0, cls=(bs_gs, ), fname='gs')
-
-# Offset must be decoded in last position to have final instruction len
-rel_off = bs(l=0, cls=(bs_rel_off,), fname="off", order=-1)
-# Offset must be decoded in last position to have final instruction len
-rel_off08 = bs(l=8, cls=(bs_rel_off08,), fname="off", order=-1)
-moff = bs(l=0, cls=(bs_moff,), fname="off")
-msegoff = bs(l=16, cls=(bs_msegoff,), fname="mseg")
-movoff = bs(l=0, cls=(bs_movoff,), fname="off")
-mod = bs(l=2, fname="mod")
-mod_mem = bs(l=2, cls=(bs_mem,), fname="mod")
-mod_reg = bs(l=2, cls=(bs_reg,), fname="mod")
-
-rmreg = bs(l=3, cls=(x86_rm_reg, ), order =1, fname = "reg")
-reg = bs(l=3, cls=(x86_reg, ), order =1, fname = "reg")
-
-reg_modrm = bs(l=3, cls=(x86_reg_modrm, ), order =1, fname = "reg")
-
-vex_reg = bs(l=0, cls=(x86_vex_reg, ), order =1, fname = "vex_reg")
-
-regnoarg = bs(l=3, default_val="000", order=1, fname="reg")
-segm = bs(l=3, cls=(x86_rm_segm, ), order =1, fname = "reg")
-crreg = bs(l=3, cls=(x86_rm_cr, ), order =1, fname = "reg")
-drreg = bs(l=3, cls=(x86_rm_dr, ), order =1, fname = "reg")
-
-
-mm_reg = bs(l=3, cls=(x86_rm_reg_mm, ), order =1, fname = "reg")
-xmm_reg = bs(l=3, cls=(x86_rm_reg_xmm, ), order =1, fname = "reg")
-bnd_reg = bs(l=3, cls=(x86_rm_reg_bnd, ), order =1, fname = "reg")
-
-
-fltreg = bs(l=3, cls=(x86_rm_flt, ), order =1, fname = "reg")
-
-rm = bs(l=3, fname="rm")
-
-rm_arg = bs(l=0, cls=(x86_rm_arg,), fname='rmarg')
-rm_arg_w8 = bs(l=0, cls=(x86_rm_w8,), fname='rmarg')
-rm_arg_sx = bs(l=0, cls=(x86_rm_sx,), fname='rmarg')
-rm_arg_sxd = bs(l=0, cls=(x86_rm_sxd,), fname='rmarg')
-rm_arg_sd = bs(l=0, cls=(x86_rm_sd,), fname='rmarg')
-rm_arg_wd = bs(l=0, cls=(x86_rm_wd,), fname='rmarg')
-rm_arg_08 = bs(l=0, cls=(x86_rm_08,), fname='rmarg')
-rm_arg_reg_m08 = bs(l=0, cls=(x86_rm_reg_m08,), fname='rmarg')
-rm_arg_reg_m16 = bs(l=0, cls=(x86_rm_reg_m16,), fname='rmarg')
-rm_arg_m08 = bs(l=0, cls=(x86_rm_m08,), fname='rmarg')
-rm_arg_m64 = bs(l=0, cls=(x86_rm_m64,), fname='rmarg')
-rm_arg_m80 = bs(l=0, cls=(x86_rm_m80,), fname='rmarg')
-rm_arg_m16 = bs(l=0, cls=(x86_rm_m16,), fname='rmarg')
-
-rm_mem = bs(l=0, cls=(x86_rm_mem,), fname='rmarg')
-rm_mem_far = bs(l=0, cls=(x86_rm_mem_far,), fname='rmarg')
-
-rm_arg_mm = bs(l=0, cls=(x86_rm_mm,), fname='rmarg')
-rm_arg_mm_m64 = bs(l=0, cls=(x86_rm_mm_m64,), fname='rmarg')
-rm_arg_mm_reg = bs(l=0, cls=(x86_rm_mm_reg,), fname='rmarg')
-
-rm_arg_xmm = bs(l=0, cls=(x86_rm_xmm,), fname='rmarg')
-rm_arg_xmm_m32 = bs(l=0, cls=(x86_rm_xmm_m32,), fname='rmarg')
-rm_arg_xmm_m64 = bs(l=0, cls=(x86_rm_xmm_m64,), fname='rmarg')
-rm_arg_xmm_m128 = bs(l=0, cls=(x86_rm_xmm_m128,), fname='rmarg')
-rm_arg_xmm_reg = bs(l=0, cls=(x86_rm_xmm_reg,), fname='rmarg')
-
-rm_arg_bnd = bs(l=0, cls=(x86_rm_bnd,), fname='rmarg')
-rm_arg_bnd_m64 = bs(l=0, cls=(x86_rm_bnd_m64,), fname='rmarg')
-rm_arg_bnd_m128 = bs(l=0, cls=(x86_rm_bnd_m128,), fname='rmarg')
-rm_arg_bnd_reg = bs(l=0, cls=(x86_rm_bnd_reg,), fname='rmarg')
-
-
-swapargs = bs_swapargs(l=1, fname="swap", mn_mod=list(range(1 << 1)))
-
-
-class bs_op_mode(bsi):
-
-    def decode(self, v):
-        opmode = self.parent.v_opmode()
-        return opmode == self.mode
-
-
-class bs_ad_mode(bsi):
-
-    def decode(self, v):
-        admode = self.parent.v_admode()
-        return admode == self.mode
-
-
-class bs_op_mode_no64(bsi):
-
-    def encode(self):
-        if self.parent.mode == 64:
-            return False
-        return super(bs_op_mode_no64, self).encode()
-
-    def decode(self, v):
-        if self.parent.mode == 64:
-            return False
-        opmode = self.parent.v_opmode()
-        return opmode == self.mode
-
-
-class bs_op_mode64(bsi):
-    def encode(self):
-        if self.parent.mode != 64:
-            return False
-        return super(bs_op_mode64, self).encode()
-
-    def decode(self, v):
-        if self.parent.mode != 64:
-            return False
-        return True
-
-class bs_op_modeno64(bsi):
-    def encode(self):
-        if self.parent.mode == 64:
-            return False
-        return super(bs_op_modeno64, self).encode()
-
-    def decode(self, v):
-        if self.parent.mode == 64:
-            return False
-        return True
-
-
-
-bs_opmode16 = bs(l=0, cls=(bs_op_mode,), mode = 16, fname="fopmode")
-bs_opmode32 = bs(l=0, cls=(bs_op_mode,), mode = 32, fname="fopmode")
-bs_opmode64 = bs(l=0, cls=(bs_op_mode,), mode = 64, fname="fopmode")
-
-
-bs_admode16 = bs(l=0, cls=(bs_ad_mode,), mode = 16, fname="fadmode")
-bs_admode32 = bs(l=0, cls=(bs_ad_mode,), mode = 32, fname="fadmode")
-bs_admode64 = bs(l=0, cls=(bs_ad_mode,), mode = 64, fname="fadmode")
-
-bs_opmode16_no64 = bs(l=0, cls=(bs_op_mode_no64,), mode = 16, fname="fopmode")
-bs_opmode32_no64 = bs(l=0, cls=(bs_op_mode_no64,), mode = 32, fname="fopmode")
-
-bs_mode64 = bs(l=0, cls=(bs_op_mode64,))
-bs_modeno64 = bs(l=0, cls=(bs_op_modeno64,))
-
-
-cond_list = ["O", "NO", "B", "AE",
-             "Z", "NZ", "BE", "A",
-             "S", "NS", "PE", "NP",
-             #"L", "NL", "NG", "G"]
-             "L", "GE", "LE", "G"]
-cond = bs_mod_name(l=4, fname='cond', mn_mod=cond_list)
-
-
-def rmmod(r, rm_arg_x=rm_arg, modrm=mod):
-    return [modrm, r, rm, sib_scale, sib_index, sib_base, disp, rm_arg_x]
-
-#
-# mode | reg | rm #
-#
-
-#
-# scale | index | base #
-#
-
-#
-# Prefix | REX prefix | Opcode | mod/rm | sib | displacement | immediate #
-#
-
-
-def addop(name, fields, args=None, alias=False):
-    dct = {"fields": fields}
-    dct["alias"] = alias
-    if args is not None:
-        dct['args'] = args
-    type(name, (mn_x86,), dct)
-"""
-class ia32_aaa(mn_x86):
-    fields = [bs8(0x37)]
-"""
-addop("aaa", [bs8(0x37)])
-addop("aas", [bs8(0x3F)])
-addop("aad", [bs8(0xd5), u08])
-addop("aam", [bs8(0xd4), u08])
-
-addop("adc", [bs("0001010"), w8, d_eax, d_imm])
-addop("adc", [bs("100000"), se, w8] + rmmod(d2, rm_arg_w8) + [d_imm])
-addop("adc", [bs("000100"), swapargs, w8] +
-      rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-
-addop("add", [bs("0000010"), w8, d_eax, d_imm])
-addop("add", [bs("100000"), se, w8] + rmmod(d0, rm_arg_w8) + [d_imm])
-addop("add", [bs("000000"), swapargs, w8] +
-      rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-
-addop("and", [bs("0010010"), w8, d_eax, d_imm])
-addop("and", [bs("100000"), se, w8] + rmmod(d4, rm_arg_w8) + [d_imm])
-addop("and", [bs("001000"), swapargs, w8] +
-      rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-
-addop("bndmov", [bs8(0x0f), bs8(0x1a), pref_66, bs_modeno64] +
-      rmmod(bnd_reg, rm_arg_bnd_m64), [bnd_reg, rm_arg_bnd_m64])
-addop("bndmov", [bs8(0x0f), bs8(0x1a), pref_66, bs_mode64] +
-      rmmod(bnd_reg, rm_arg_bnd_m128), [bnd_reg, rm_arg_bnd_m128])
-addop("bndmov", [bs8(0x0f), bs8(0x1b), pref_66, bs_modeno64] +
-      rmmod(bnd_reg, rm_arg_bnd_m64), [rm_arg_bnd_m64, bnd_reg])
-addop("bndmov", [bs8(0x0f), bs8(0x1b), pref_66, bs_mode64] +
-      rmmod(bnd_reg, rm_arg_bnd_m128), [rm_arg_bnd_m128, bnd_reg])
-
-
-
-addop("bsf", [bs8(0x0f), bs8(0xbc), no_rep] + rmmod(rmreg))
-addop("bsr", [bs8(0x0f), bs8(0xbd), mod,
-    rmreg, rm, sib_scale, sib_index, sib_base, disp, rm_arg])
-
-addop("bswap", [bs8(0x0f), bs('11001'), reg])
-
-addop("bt", [bs8(0x0f), bs8(0xa3)] + rmmod(rmreg), [rm_arg, rmreg])
-addop("bt", [bs8(0x0f), bs8(0xba)] + rmmod(d4) + [u08])
-addop("btc", [bs8(0x0f), bs8(0xbb)] + rmmod(rmreg), [rm_arg, rmreg])
-addop("btc", [bs8(0x0f), bs8(0xba)] + rmmod(d7) + [u08])
-
-
-addop("btr", [bs8(0x0f), bs8(0xb3)] + rmmod(rmreg), [rm_arg, rmreg])
-addop("btr", [bs8(0x0f), bs8(0xba)] + rmmod(d6) + [u08])
-addop("bts", [bs8(0x0f), bs8(0xab)] + rmmod(rmreg), [rm_arg, rmreg])
-addop("bts", [bs8(0x0f), bs8(0xba)] + rmmod(d5) + [u08])
-
-addop("call", [bs8(0xe8), rel_off])
-addop("call", [bs8(0xff), stk] + rmmod(d2))
-addop("call", [bs8(0xff), stk] + rmmod(d3, rm_arg_x=rm_mem_far, modrm=mod_mem))
-addop("call", [bs8(0x9a), bs_modeno64, moff, msegoff])
-
-
-addop("cbw", [bs8(0x98), bs_opmode16])
-addop("cwde", [bs8(0x98), bs_opmode32])
-addop("cdqe", [bs8(0x98), bs_opmode64])
-
-addop("clc", [bs8(0xf8)])
-addop("cld", [bs8(0xfc)])
-addop("cli", [bs8(0xfa)])
-addop("clts", [bs8(0x0f), bs8(0x06)])
-addop("cmc", [bs8(0xf5)])
-
-addop("cmov", [bs8(0x0f), bs('0100'), cond] + rmmod(rmreg))
-
-addop("cmp", [bs("0011110"), w8, d_eax, d_imm])
-addop("cmp", [bs("100000"), se, w8] + rmmod(d7, rm_arg_w8) + [d_imm])
-addop("cmp", [bs("001110"), swapargs, w8] +
-      rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-
-
-addop("cmpsb", [bs8(0xa6)])
-addop("cmpsw", [bs8(0xa7), bs_opmode16])
-addop("cmpsd", [bs8(0xa7), bs_opmode32])
-addop("cmpsq", [bs8(0xa7), bs_opmode64])
-
-addop("cmpxchg", [bs8(0x0f), bs('1011000'), w8]
-      + rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-addop("cmpxchg8b", [bs8(0x0f), bs8(0xc7), bs_opmode16] + rmmod(d1, rm_arg_m64))
-addop("cmpxchg8b", [bs8(0x0f), bs8(0xc7), bs_opmode32] + rmmod(d1, rm_arg_m64))
-addop("cmpxchg16b", [bs8(0x0f), bs8(0xc7), bs_opmode64] + rmmod(d1, rm_arg_xmm_m128))
-
-# XXX TODO CMPXCHG8/16
-
-addop("comiss", [bs8(0x0f), bs8(0x2f), no_xmm_pref] +
-      rmmod(xmm_reg, rm_arg_xmm_m32), [xmm_reg, rm_arg_xmm_m32])
-addop("comisd", [bs8(0x0f), bs8(0x2f), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m64), [xmm_reg, rm_arg_xmm_m64])
-
-addop("cpuid", [bs8(0x0f), bs8(0xa2)])
-
-addop("cwd", [bs8(0x99), bs_opmode16])
-addop("cdq", [bs8(0x99), bs_opmode32])
-addop("cqo", [bs8(0x99), bs_opmode64])
-
-
-addop("daa", [bs8(0x27)])
-addop("das", [bs8(0x2f)])
-addop("dec", [bs('1111111'), w8] + rmmod(d1, rm_arg_w8))
-addop("dec", [bs('01001'), reg, bs_modeno64])
-addop("div", [bs('1111011'), w8] + rmmod(d6, rm_arg_w8))
-addop("enter", [bs8(0xc8), u16, u08])
-
-# float #####
-addop("fwait", [bs8(0x9b)])
-
-addop("f2xm1", [bs8(0xd9), bs8(0xf0)])
-addop("fabs", [bs8(0xd9), bs8(0xe1)])
-
-addop("fadd", [bs("11011"), sd, bs("00")] + rmmod(d0, rm_arg_sd))
-addop("fadd", [bs("11011"), swapargs, bs("00"),
-      bs("11000"), d_st, fltreg], [d_st, fltreg])
-addop("faddp", [bs8(0xde), bs("11000"), fltreg, d_st])
-addop("fiadd", [bs("11011"), wd, bs("10")] + rmmod(d0, rm_arg_wd))
-
-addop("fbld", [bs8(0xdf)] + rmmod(d4, rm_arg_m80))
-addop("fbldp", [bs8(0xdf)] + rmmod(d6, rm_arg_m80))
-addop("fchs", [bs8(0xd9), bs8(0xe0)])
-# addop("fclex", [bs8(0x9b), bs8(0xdb), bs8(0xe2)])
-addop("fnclex", [bs8(0xdb), bs8(0xe2)])
-
-addop("fcmovb", [bs8(0xda), bs("11000"), d_st, fltreg])
-addop("fcmove", [bs8(0xda), bs("11001"), d_st, fltreg])
-addop("fcmovbe", [bs8(0xda), bs("11010"), d_st, fltreg])
-addop("fcmovu", [bs8(0xda), bs("11011"), d_st, fltreg])
-addop("fcmovnb", [bs8(0xdb), bs("11000"), d_st, fltreg])
-addop("fcmovne", [bs8(0xdb), bs("11001"), d_st, fltreg])
-addop("fcmovnbe", [bs8(0xdb), bs("11010"), d_st, fltreg])
-addop("fcmovnu", [bs8(0xdb), bs("11011"), d_st, fltreg])
-
-addop("fcom", [bs("11011"), sd, bs("00")] + rmmod(d2, rm_arg_sd))
-addop("fcom", [bs("11011"), swapargs, bs("00"),
-      bs("11010"), d_st, fltreg], [d_st, fltreg])
-addop("fcomp", [bs("11011"), sd, bs("00")] + rmmod(d3, rm_arg_sd))
-addop("fcomp",
-      [bs("11011"), swapargs, bs("00"), bs("11011"),
-      d_st, fltreg], [d_st, fltreg])
-addop("fcompp", [bs8(0xde), bs8(0xd9)])
-
-addop("fcomi", [bs8(0xdb), bs("11110"), d_st, fltreg])
-addop("fcomip", [bs8(0xdf), bs("11110"), d_st, fltreg])
-addop("fucomi", [bs8(0xdb), bs("11101"), d_st, fltreg])
-addop("fucomip", [bs8(0xdf), bs("11101"), d_st, fltreg])
-
-addop("fcos", [bs8(0xd9), bs8(0xff)])
-addop("fdecstp", [bs8(0xd9), bs8(0xf6)])
-
-
-addop("fdiv", [bs("11011"), sd, bs("00")] + rmmod(d6, rm_arg_sd))
-addop("fdiv", [bs8(0xd8), bs("11110"), d_st, fltreg])
-addop("fdiv", [bs8(0xdc), bs("11111"), fltreg, d_st])
-addop("fdivp", [bs8(0xde), bs("11111"), fltreg, d_st])
-addop("fidiv", [bs("11011"), wd, bs("10")] + rmmod(d6, rm_arg_wd))
-
-addop("fdivr", [bs("11011"), sd, bs("00")] + rmmod(d7, rm_arg_sd))
-addop("fdivr", [bs8(0xd8), bs("11111"), d_st, fltreg])
-addop("fdivr", [bs8(0xdc), bs("11110"), fltreg, d_st])
-addop("fdivrp", [bs8(0xde), bs("11110"), fltreg, d_st])
-addop("fidivr", [bs("11011"), wd, bs("10")] + rmmod(d7, rm_arg_wd))
-
-addop("ffree", [bs8(0xdd), bs("11000"), fltreg])
-addop("ficom", [bs("11011"), wd, bs("10")] + rmmod(d2, rm_arg_wd))
-addop("ficomp", [bs("11011"), wd, bs("10")] + rmmod(d3, rm_arg_wd))
-addop("fild", [bs("11011"), wd, bs("11")] + rmmod(d0, rm_arg_wd))
-addop("fild", [bs8(0xdf)] + rmmod(d5, rm_arg_m64))
-
-addop("fincstp", [bs8(0xd9), bs8(0xf7)])
-
-addop("blsi", [pref_0f38, bs8(0xf3), vex_reg] + rmmod(bs("011"), rm_arg), [vex_reg, rm_arg])
-addop("andn", [pref_0f38, bs8(0xf2), vex_reg] + rmmod(rmreg, rm_arg), [rmreg, vex_reg, rm_arg])
-addop("bextr", [pref_0f38, bs8(0xf7), vex_reg] + rmmod(rmreg, rm_arg), [rmreg, rm_arg, vex_reg])
-addop("blsmsk", [pref_0f38, bs8(0xf3), vex_reg] + rmmod(bs("010"), rm_arg), [vex_reg, rm_arg])
-addop("blsr", [pref_0f38, bs8(0xf3), vex_reg] + rmmod(bs("001"), rm_arg), [vex_reg, rm_arg])
-addop("bzhi", [pref_0f38, bs8(0xf5), vex_reg] + rmmod(rmreg, rm_arg), [rmreg, rm_arg, vex_reg])
-addop("tzcnt", [bs8(0x0f), bs8(0xbc), pref_f3] + rmmod(rmreg, rm_arg), [rmreg, rm_arg])
-
-# addop("finit", [bs8(0x9b), bs8(0xdb), bs8(0xe3)])
-addop("fninit", [bs8(0xdb), bs8(0xe3)])
-
-addop("fist", [bs("11011"), wd, bs("11")] + rmmod(d2, rm_arg_wd))
-addop("fistp", [bs("11011"), wd, bs("11")] + rmmod(d3, rm_arg_wd))
-addop("fistp", [bs8(0xdf)] + rmmod(d7, rm_arg_m64))
-
-addop("fisttp", [bs("11011"), wd, bs("11")] + rmmod(d1, rm_arg_wd))
-addop("fisttp", [bs8(0xdd)] + rmmod(d1, rm_arg_m64))
-
-addop("fld", [bs("11011"), sd, bs("01")] + rmmod(d0, rm_arg_sd))
-addop("fld", [bs8(0xdb)] + rmmod(d5, rm_arg_m80))
-addop("fld", [bs8(0xd9), bs("11000"), fltreg])
-
-addop("fld1", [bs8(0xd9), bs8(0xe8)])
-addop("fldl2t", [bs8(0xd9), bs8(0xe9)])
-addop("fldl2e", [bs8(0xd9), bs8(0xea)])
-addop("fldpi", [bs8(0xd9), bs8(0xeb)])
-addop("fldlg2", [bs8(0xd9), bs8(0xec)])
-addop("fldln2", [bs8(0xd9), bs8(0xed)])
-addop("fldz", [bs8(0xd9), bs8(0xee)])
-
-addop("fldcw", [bs8(0xd9)] + rmmod(d5, rm_arg_m16))
-addop("fldenv", [bs8(0xd9)] + rmmod(d4, rm_arg_m80))  # XXX TODO: m14?
-
-addop("fmul", [bs("11011"), sd, bs("00")] + rmmod(d1, rm_arg_sd))
-addop("fmul", [bs("11011"), swapargs, bs("00"),
-      bs("11001"), d_st, fltreg], [d_st, fltreg])
-addop("fmulp", [bs8(0xde), bs("11001"), fltreg, d_st])
-addop("fimul", [bs("11011"), wd, bs("10")] + rmmod(d1, rm_arg_wd))
-
-addop("fnop", [bs8(0xd9), bs8(0xd0)])
-addop("fpatan", [bs8(0xd9), bs8(0xf3)])
-addop("fprem", [bs8(0xd9), bs8(0xf8)])
-addop("fprem1", [bs8(0xd9), bs8(0xf5)])
-addop("fptan", [bs8(0xd9), bs8(0xf2)])
-addop("frndint", [bs8(0xd9), bs8(0xfc)])
-addop("frstor", [bs8(0xdd)] + rmmod(d4, rm_arg_m80))  # XXX TODO: m94 ?
-# addop("fsave", [bs8(0x9b), bs8(0xdd)] + rmmod(d6, rm_arg_m80)) # XXX
-# TODO: m94 ?
-addop("fnsave", [bs8(0xdd)] + rmmod(d6, rm_arg_m80))  # XXX TODO: m94 ?
-
-addop("fscale", [bs8(0xd9), bs8(0xfd)])
-addop("fsin", [bs8(0xd9), bs8(0xfe)])
-addop("fsincos", [bs8(0xd9), bs8(0xfb)])
-addop("fsqrt", [bs8(0xd9), bs8(0xfa)])
-
-addop("fst", [bs("11011"), sd, bs("01")] + rmmod(d2, rm_arg_sd))
-addop("fst", [bs8(0xdd), bs("11010"), fltreg])
-addop("fstp", [bs("11011"), sd, bs("01")] + rmmod(d3, rm_arg_sd))
-addop("fstp", [bs8(0xdb)] + rmmod(d7, rm_arg_m80))
-addop("fstp", [bs8(0xdd), bs("11011"), fltreg])
-
-# addop("fstcw", [bs8(0x9b), bs8(0xd9)] + rmmod(d7, rm_arg_m16))
-addop("fnstcw", [bs8(0xd9)] + rmmod(d7, rm_arg_m16))
-# addop("fstenv", [bs8(0x9b), bs8(0xd9)] + rmmod(d6, rm_arg_m80)) # XXX
-# TODO: m14?
-addop("fnstenv", [bs8(0xd9)] + rmmod(d6, rm_arg_m80))  # XXX TODO: m14?
-# addop("fstsw", [bs8(0x9b), bs8(0xdd)] + rmmod(d7, rm_arg_m16))
-addop("fnstsw", [bs8(0xdd)] + rmmod(d7, rm_arg_m16))
-# addop("fstsw", [bs8(0x9b), bs8(0xdf), bs8(0xe0), d_ax])
-addop("fnstsw", [bs8(0xdf), bs8(0xe0), d_ax])
-
-addop("fsub", [bs("11011"), sd, bs("00")] + rmmod(d4, rm_arg_sd))
-addop("fsub", [bs8(0xd8), bs("11100"), d_st, fltreg])
-addop("fsub", [bs8(0xdc), bs("11101"), fltreg, d_st])
-addop("fsubp", [bs8(0xde), bs("11101"), fltreg, d_st])
-addop("fisub", [bs("11011"), wd, bs("10")] + rmmod(d4, rm_arg_wd))
-
-addop("fsubr", [bs("11011"), sd, bs("00")] + rmmod(d5, rm_arg_sd))
-addop("fsubr", [bs8(0xd8), bs("11101"), d_st, fltreg])
-addop("fsubr", [bs8(0xdc), bs("11100"), fltreg, d_st])
-addop("fsubrp", [bs8(0xde), bs("11100"), fltreg, d_st])
-addop("fisubr", [bs("11011"), wd, bs("10")] + rmmod(d5, rm_arg_wd))
-addop("ftst", [bs8(0xd9), bs8(0xe4)])
-
-
-addop("fucom", [bs8(0xdd), bs("11100"), fltreg])
-addop("fucomp", [bs8(0xdd), bs("11101"), fltreg])
-addop("fucompp", [bs8(0xda), bs8(0xe9)])
-
-addop("fxam", [bs8(0xd9), bs8(0xe5)])
-addop("fxch", [bs8(0xd9), bs("11001"), fltreg])
-addop("fxrstor", [bs8(0x0f), bs8(0xae)]
-      + rmmod(d1, rm_arg_m80))  # XXX TODO m512
-addop("fxsave", [bs8(0x0f), bs8(0xae)]
-      + rmmod(d0, rm_arg_m80))  # XXX TODO m512
-addop("stmxcsr", [bs8(0x0f), bs8(0xae)] + rmmod(d3))
-addop("ldmxcsr", [bs8(0x0f), bs8(0xae)] + rmmod(d2))
-
-addop("fxtract", [bs8(0xd9), bs8(0xf4)])
-addop("fyl2x", [bs8(0xd9), bs8(0xf1)])
-addop("fyl2xp1", [bs8(0xd9), bs8(0xf9)])
-
-addop("hlt", [bs8(0xf4)])
-addop("icebp", [bs8(0xf1)])
-
-addop("idiv", [bs('1111011'), w8] + rmmod(d7, rm_arg_w8))
-
-addop("imul", [bs('1111011'), w8] + rmmod(d5, rm_arg_w8))
-addop("imul", [bs8(0x0f), bs8(0xaf)] + rmmod(rmreg))
-
-addop("imul", [bs("011010"), se, bs('1')] + rmmod(rmreg) + [d_imm])
-
-addop("in", [bs("1110010"), w8, d_eax, u08])
-addop("in", [bs("1110110"), w8, d_eax, d_edx])
-
-addop("inc", [bs('1111111'), w8] + rmmod(d0, rm_arg_w8))
-addop("inc", [bs('01000'), reg, bs_modeno64])
-
-addop("insb", [bs8(0x6c)])
-addop("insw", [bs8(0x6d), bs_opmode16])
-addop("insd", [bs8(0x6d), bs_opmode32])
-addop("insd", [bs8(0x6d), bs_opmode64])
-
-addop("int", [bs8(0xcc), u08_3])
-addop("int", [bs8(0xcd), u08])
-addop("into", [bs8(0xce)])
-addop("invd", [bs8(0x0f), bs8(0x08)])
-addop("invlpg", [bs8(0x0f), bs8(0x01)] + rmmod(d7))
-
-addop("iret", [bs8(0xcf), bs_opmode16])
-addop("iretd", [bs8(0xcf), bs_opmode32])
-addop("iretq", [bs8(0xcf), bs_opmode64])
-
-addop("j", [bs('0111'), cond, rel_off08])
-
-addop("jcxz", [bs8(0xe3), rel_off08, bs_admode16])
-addop("jecxz", [bs8(0xe3), rel_off08, bs_admode32])
-addop("jrcxz", [bs8(0xe3), rel_off08, bs_admode64])
-
-addop("j", [bs8(0x0f), bs('1000'), cond, rel_off])
-addop("jmp", [bs8(0xeb), rel_off08])
-addop("jmp", [bs8(0xe9), rel_off])
-# TODO XXX replace stk force64?
-addop("jmp", [bs8(0xff), stk] + rmmod(d4))
-addop("jmp", [bs8(0xea), bs_modeno64, moff, msegoff])
-
-addop("jmp", [bs8(0xff)] + rmmod(d5, rm_arg_x=rm_mem_far, modrm=mod_mem))
-
-addop("lahf", [bs8(0x9f)])
-addop("lar", [bs8(0x0f), bs8(0x02)] + rmmod(rmreg))
-
-addop("lea", [bs8(0x8d)] + rmmod(rmreg, rm_arg_x=rm_mem, modrm=mod_mem))
-addop("les", [bs8(0xc4)] + rmmod(rmreg, rm_arg_x=rm_mem, modrm=mod_mem))
-addop("lds", [bs8(0xc5)] + rmmod(rmreg, rm_arg_x=rm_mem, modrm=mod_mem))
-addop("lss", [bs8(0x0f), bs8(0xb2)] + rmmod(rmreg, rm_arg_x=rm_mem, modrm=mod_mem))
-addop("lfs", [bs8(0x0f), bs8(0xb4)] + rmmod(rmreg, rm_arg_x=rm_mem, modrm=mod_mem))
-addop("lgs", [bs8(0x0f), bs8(0xb5)] + rmmod(rmreg, rm_arg_x=rm_mem, modrm=mod_mem))
-
-addop("lgdt", [bs8(0x0f), bs8(0x01)] + rmmod(d2, modrm=mod_mem))
-addop("lidt", [bs8(0x0f), bs8(0x01)] + rmmod(d3, modrm=mod_mem))
-
-addop("lfence", [bs8(0x0f), bs8(0xae), bs8(0xe8), no_xmm_pref])
-addop("mfence", [bs8(0x0f), bs8(0xae), bs8(0xf0)])
-addop("sfence", [bs8(0x0f), bs8(0xae), bs8(0xf8)])
-
-addop("leave", [bs8(0xc9), stk])
-
-addop("lodsb", [bs8(0xac)])
-addop("lodsw", [bs8(0xad), bs_opmode16])
-addop("lodsd", [bs8(0xad), bs_opmode32])
-addop("lodsq", [bs8(0xad), bs_opmode64])
-
-addop("loop", [bs8(0xe2), rel_off08])
-addop("loope", [bs8(0xe1), rel_off08])
-addop("loopne", [bs8(0xe0), rel_off08])
-addop("lsl", [bs8(0x0f), bs8(0x03)] + rmmod(rmreg))
-addop("monitor", [bs8(0x0f), bs8(0x01), bs8(0xc8)])
-
-addop("mov", [bs("100010"), swapargs, w8] +
-      rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-addop("mov", [bs("100011"), swapargs, bs('0')] + rmmod(segm), [rm_arg, segm])
-addop("mov", [bs("101000"), swapargs, w8, d_eax, movoff], [d_eax, movoff])
-addop("mov", [bs("1011"), w8, reg, d_imm64])
-addop("mov", [bs("1100011"), w8] + rmmod(d0, rm_arg_w8) + [d_imm])
-addop("mov", [bs8(0x0f), bs("001000"), swapargs, bs('0')]
-      + rmmod(crreg), [rm_arg, crreg])
-addop("mov", [bs8(0x0f), bs("001000"), swapargs, bs('1')]
-      + rmmod(drreg), [rm_arg, drreg])
-addop("movsb", [bs8(0xa4)])
-addop("movsw", [bs8(0xa5), bs_opmode16])
-addop("movsd", [bs8(0xa5), bs_opmode32])
-addop("movsq", [bs8(0xa5), bs_opmode64])
-
-addop("movsx", [bs8(0x0f), bs("1011111"), w8, sx] + rmmod(rmreg, rm_arg_sx))
-addop("movsxd", [bs8(0x63), sxd, bs_mode64] + rmmod(rmreg, rm_arg_sxd))
-
-addop("movups", [bs8(0x0f), bs("0001000"), swapargs, no_xmm_pref] +
-      rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-addop("movsd", [bs8(0x0f), bs("0001000"), swapargs, pref_f2]
-      + rmmod(xmm_reg, rm_arg_xmm_m64), [xmm_reg, rm_arg_xmm_m64])
-addop("movss", [bs8(0x0f), bs("0001000"), swapargs, pref_f3] +
-      rmmod(xmm_reg, rm_arg_xmm_m32), [xmm_reg, rm_arg_xmm_m32])
-addop("movupd", [bs8(0x0f), bs8(0x10), pref_66] + rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-addop("movupd", [bs8(0x0f), bs8(0x11), pref_66] + rmmod(xmm_reg, rm_arg_xmm), [rm_arg_xmm, xmm_reg])
-
-
-addop("movd", [bs8(0x0f), bs('011'), swapargs, bs('1110'), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg), [mm_reg, rm_arg])
-addop("movd", [bs8(0x0f), bs('011'), swapargs, bs('1110'), pref_66, bs_opmode32] +
-      rmmod(xmm_reg, rm_arg), [xmm_reg, rm_arg])
-addop("movq", [bs8(0x0f), bs('011'), swapargs, bs('1110'), pref_66, bs_opmode64] +
-      rmmod(xmm_reg, rm_arg), [xmm_reg, rm_arg])
-
-addop("movq", [bs8(0x0f), bs('011'), swapargs, bs('1111'), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64), [mm_reg, rm_arg_mm_m64])
-
-addop("movq", [bs8(0x0f), bs8(0x7e), pref_f3] +
-      rmmod(xmm_reg, rm_arg_xmm_m64), [xmm_reg, rm_arg_xmm_m64])
-addop("movq", [bs8(0x0f), bs8(0xd6), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m64), [rm_arg_xmm_m64, xmm_reg])
-
-addop("movmskps", [bs8(0x0f), bs8(0x50), no_xmm_pref] +
-      rmmod(reg_modrm, rm_arg_xmm_reg))
-addop("movmskpd", [bs8(0x0f), bs8(0x50), pref_66] +
-      rmmod(reg_modrm, rm_arg_xmm_reg))
-
-addop("movnti", [bs8(0x0f), bs8(0xc3)] + rmmod(rmreg), [rm_arg, rmreg])
-
-addop("addss", [bs8(0x0f), bs8(0x58), pref_f3] + rmmod(xmm_reg, rm_arg_xmm_m32))
-addop("addsd", [bs8(0x0f), bs8(0x58), pref_f2] + rmmod(xmm_reg, rm_arg_xmm_m64))
-
-addop("subss", [bs8(0x0f), bs8(0x5c), pref_f3] + rmmod(xmm_reg, rm_arg_xmm_m32))
-addop("subsd", [bs8(0x0f), bs8(0x5c), pref_f2] + rmmod(xmm_reg, rm_arg_xmm_m64))
-
-addop("mulss", [bs8(0x0f), bs8(0x59), pref_f3] + rmmod(xmm_reg, rm_arg_xmm_m32))
-addop("mulsd", [bs8(0x0f), bs8(0x59), pref_f2] + rmmod(xmm_reg, rm_arg_xmm_m64))
-
-addop("divss", [bs8(0x0f), bs8(0x5e), pref_f3] + rmmod(xmm_reg, rm_arg_xmm_m32))
-addop("divsd", [bs8(0x0f), bs8(0x5e), pref_f2] + rmmod(xmm_reg, rm_arg_xmm_m64))
-
-addop("roundss", [bs8(0x0f), bs8(0x3a), bs8(0x0a), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m32) + [u08])
-addop("roundsd", [bs8(0x0f), bs8(0x3a), bs8(0x0b), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m64) + [u08])
-
-addop("pminsw", [bs8(0x0f), bs8(0xea), no_xmm_pref] + rmmod(mm_reg, rm_arg_mm))
-addop("pminsw", [bs8(0x0f), bs8(0xea), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-addop("ucomiss", [bs8(0x0f), bs8(0x2e), no_xmm_pref] + rmmod(xmm_reg, rm_arg_xmm_m32))
-addop("ucomisd", [bs8(0x0f), bs8(0x2e), pref_66] + rmmod(xmm_reg, rm_arg_xmm_m64))
-
-
-addop("movzx", [bs8(0x0f), bs("1011011"), w8, sx] + rmmod(rmreg, rm_arg_sx))
-addop("mul", [bs('1111011'), w8] + rmmod(d4, rm_arg_w8))
-
-addop("neg", [bs('1111011'), w8] + rmmod(d3, rm_arg_w8))
-addop("nop", [bs8(0x0f), bs8(0x1f)] + rmmod(d0, rm_arg))  # XXX TODO m512
-addop("nop", [bs8(0x0f), bs8(0x1f)] + rmmod(d1, rm_arg))  # XXX TODO m512
-addop("nop", [bs8(0x0f), bs8(0x1f)] + rmmod(d2, rm_arg))  # XXX TODO m512
-addop("nop", [bs8(0x0f), bs8(0x1f)] + rmmod(d3, rm_arg))  # XXX TODO m512
-addop("nop", [bs8(0x0f), bs8(0x1f)] + rmmod(d4, rm_arg))  # XXX TODO m512
-addop("nop", [bs8(0x0f), bs8(0x1f)] + rmmod(d5, rm_arg))  # XXX TODO m512
-addop("nop", [bs8(0x0f), bs8(0x1f)] + rmmod(d6, rm_arg))  # XXX TODO m512
-addop("nop", [bs8(0x0f), bs8(0x1f)] + rmmod(d7, rm_arg))  # XXX TODO m512
-addop("not", [bs('1111011'), w8] + rmmod(d2, rm_arg_w8))
-addop("or", [bs("0000110"), w8, d_eax, d_imm])
-addop("or", [bs("100000"), se, w8] + rmmod(d1, rm_arg_w8) + [d_imm])
-addop("or", [bs("000010"), swapargs, w8] +
-      rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-addop("out", [bs("1110011"), w8, u08, d_eax])
-addop("out", [bs("1110111"), w8, d_edx, d_eax])
-
-addop("outsb", [bs8(0x6e)])
-addop("outsw", [bs8(0x6f), bs_opmode16])
-addop("outsd", [bs8(0x6f), bs_opmode32])
-addop("outsd", [bs8(0x6f), bs_opmode64])
-
-addop("setalc", [bs8(0xD6)])
-
-# addop("pause", [bs8(0xf3), bs8(0x90)])
-
-addop("popw", [bs8(0x8f), stk, bs_opmode16] + rmmod(d0))
-addop("popw", [bs("01011"), stk, reg, bs_opmode16])
-addop("popw", [bs8(0x1f), stk, d_ds, bs_opmode16])
-addop("popw", [bs8(0x07), stk, d_es, bs_opmode16])
-addop("popw", [bs8(0x17), stk, d_ss, bs_opmode16])
-addop("popw", [bs8(0x0f), stk, bs8(0xa1), d_fs, bs_opmode16])
-addop("popw", [bs8(0x0f), stk, bs8(0xa9), d_gs, bs_opmode16])
-
-addop("pop", [bs8(0x8f), stk, bs_opmode32] + rmmod(d0))
-addop("pop", [bs("01011"), stk, reg, bs_opmode32])
-addop("pop", [bs8(0x1f), stk, d_ds, bs_opmode32])
-addop("pop", [bs8(0x07), stk, d_es, bs_opmode32])
-addop("pop", [bs8(0x17), stk, d_ss, bs_opmode32])
-addop("pop", [bs8(0x0f), stk, bs8(0xa1), d_fs, bs_opmode32])
-addop("pop", [bs8(0x0f), stk, bs8(0xa9), d_gs, bs_opmode32])
-
-addop("pop", [bs8(0x8f), stk, bs_opmode64] + rmmod(d0))
-addop("pop", [bs("01011"), stk, reg, bs_opmode64])
-addop("pop", [bs8(0x1f), stk, d_ds, bs_opmode64])
-addop("pop", [bs8(0x07), stk, d_es, bs_opmode64])
-addop("pop", [bs8(0x17), stk, d_ss, bs_opmode64])
-addop("pop", [bs8(0x0f), stk, bs8(0xa1), d_fs, bs_opmode64])
-addop("pop", [bs8(0x0f), stk, bs8(0xa9), d_gs, bs_opmode64])
-
-
-addop("popa", [bs8(0x61), stk, bs_opmode16])
-addop("popad", [bs8(0x61), stk, bs_opmode32])
-
-addop("popfw", [bs8(0x9d), stk, bs_opmode16])
-addop("popfd", [bs8(0x9d), stk, bs_opmode32])
-addop("popfq", [bs8(0x9d), stk, bs_opmode64])
-
-addop("prefetch0", [bs8(0x0f), bs8(0x18)] + rmmod(d1, rm_arg_m08))
-addop("prefetch1", [bs8(0x0f), bs8(0x18)] + rmmod(d2, rm_arg_m08))
-addop("prefetch2", [bs8(0x0f), bs8(0x18)] + rmmod(d3, rm_arg_m08))
-addop("prefetchnta", [bs8(0x0f), bs8(0x18)] + rmmod(d0, rm_arg_m08))
-addop("prefetchw", [bs8(0x0f), bs8(0x0d)] + rmmod(d1, rm_arg_m08))
-
-addop("pushw", [bs8(0xff), stk, bs_opmode16] + rmmod(d6))
-addop("pushw", [bs("01010"), stk, reg, bs_opmode16])
-addop("pushw", [bs8(0x6a), s08, stk, bs_opmode16])
-addop("pushw", [bs8(0x68), d_imm, stk, bs_opmode16])
-addop("pushw", [bs8(0x0e), stk, d_cs, bs_opmode16])
-addop("pushw", [bs8(0x16), stk, d_ss, bs_opmode16])
-addop("pushw", [bs8(0x1e), stk, d_ds, bs_opmode16])
-addop("pushw", [bs8(0x06), stk, d_es, bs_opmode16])
-addop("pushw", [bs8(0x0f), stk, bs8(0xa0), d_fs, bs_opmode16])
-addop("pushw", [bs8(0x0f), stk, bs8(0xa8), d_gs, bs_opmode16])
-
-addop("push", [bs8(0xff), stk, bs_opmode32] + rmmod(d6))
-addop("push", [bs("01010"), stk, reg, bs_opmode32])
-addop("push", [bs8(0x6a), s08, stk, bs_opmode32])
-addop("push", [bs8(0x68), d_imm, stk, bs_opmode32])
-addop("push", [bs8(0x0e), stk, d_cs, bs_opmode32])
-addop("push", [bs8(0x16), stk, d_ss, bs_opmode32])
-addop("push", [bs8(0x1e), stk, d_ds, bs_opmode32])
-addop("push", [bs8(0x06), stk, d_es, bs_opmode32])
-addop("push", [bs8(0x0f), stk, bs8(0xa0), d_fs, bs_opmode32])
-addop("push", [bs8(0x0f), stk, bs8(0xa8), d_gs, bs_opmode32])
-
-addop("push", [bs8(0xff), stk, bs_opmode64] + rmmod(d6))
-addop("push", [bs("01010"), stk, reg, bs_opmode64])
-addop("push", [bs8(0x6a), s08, stk, bs_opmode64])
-addop("push", [bs8(0x68), d_imm, stk, bs_opmode64])
-addop("push", [bs8(0x0e), stk, d_cs, bs_opmode64])
-addop("push", [bs8(0x16), stk, d_ss, bs_opmode64])
-addop("push", [bs8(0x1e), stk, d_ds, bs_opmode64])
-addop("push", [bs8(0x06), stk, d_es, bs_opmode64])
-addop("push", [bs8(0x0f), stk, bs8(0xa0), d_fs, bs_opmode64])
-addop("push", [bs8(0x0f), stk, bs8(0xa8), d_gs, bs_opmode64])
-
-addop("pusha", [bs8(0x60), stk, bs_opmode16_no64])
-addop("pushad", [bs8(0x60), stk, bs_opmode32_no64])
-
-
-addop("pushfw", [bs8(0x9c), stk, bs_opmode16])
-addop("pushfd", [bs8(0x9c), stk, bs_opmode32])
-addop("pushfq", [bs8(0x9c), stk, bs_opmode64])
-
-addop("rcl", [bs('110100'), d_cl1, w8] +
-      rmmod(d2, rm_arg_w8), [rm_arg_w8, d_cl1])
-addop("rcl", [bs('1100000'), w8] + rmmod(d2, rm_arg_w8) + [u08])
-addop("rcr", [bs('110100'), d_cl1, w8] +
-      rmmod(d3, rm_arg_w8), [rm_arg_w8, d_cl1])
-addop("rcr", [bs('1100000'), w8] + rmmod(d3, rm_arg_w8) + [u08])
-addop("rol", [bs('110100'), d_cl1, w8]
-      + rmmod(d0, rm_arg_w8), [rm_arg_w8, d_cl1])
-addop("rol", [bs('1100000'), w8] + rmmod(d0, rm_arg_w8) + [u08])
-addop("ror", [bs('110100'), d_cl1, w8]
-      + rmmod(d1, rm_arg_w8), [rm_arg_w8, d_cl1])
-addop("ror", [bs('1100000'), w8] + rmmod(d1, rm_arg_w8) + [u08])
-
-addop("rdmsr", [bs8(0x0f), bs8(0x32)])
-addop("rdpmc", [bs8(0x0f), bs8(0x33)])
-addop("rdtsc", [bs8(0x0f), bs8(0x31)])
-addop("ret", [bs8(0xc3), stk])
-addop("ret", [bs8(0xc2), stk, u16])
-addop("retf", [bs8(0xcb), stk])
-addop("retf", [bs8(0xca), stk, u16])
-
-addop("rsm", [bs8(0x0f), bs8(0xaa)])
-addop("sahf", [bs8(0x9e)])
-
-# XXX tipo in doc: /4 instead of /6
-addop("sal", [bs('110100'), d_cl1, w8] +
-      rmmod(d6, rm_arg_w8), [rm_arg_w8, d_cl1])
-addop("sal", [bs('1100000'), w8] + rmmod(d6, rm_arg_w8) + [u08])
-addop("sar", [bs('110100'), d_cl1, w8] +
-      rmmod(d7, rm_arg_w8), [rm_arg_w8, d_cl1])
-addop("sar", [bs('1100000'), w8] + rmmod(d7, rm_arg_w8) + [u08])
-
-addop("scasb", [bs8(0xae)])
-addop("scasw", [bs8(0xaf), bs_opmode16])
-addop("scasd", [bs8(0xaf), bs_opmode32])
-addop("scasq", [bs8(0xaf), bs_opmode64])
-
-addop("shl", [bs('110100'), d_cl1, w8]
-      + rmmod(d4, rm_arg_w8), [rm_arg_w8, d_cl1])
-addop("shl", [bs('1100000'), w8] + rmmod(d4, rm_arg_w8) + [u08])
-addop("shr", [bs('110100'), d_cl1, w8]
-      + rmmod(d5, rm_arg_w8), [rm_arg_w8, d_cl1])
-addop("shr", [bs('1100000'), w8] + rmmod(d5, rm_arg_w8) + [u08])
-
-addop("sbb", [bs("0001110"), w8, d_eax, d_imm])
-addop("sbb", [bs("100000"), se, w8] + rmmod(d3, rm_arg_w8) + [d_imm])
-addop("sbb", [bs("000110"), swapargs, w8] +
-      rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-
-addop("set", [bs8(0x0f), bs('1001'), cond] + rmmod(regnoarg, rm_arg_08))
-addop("sgdt", [bs8(0x0f), bs8(0x01)] + rmmod(d0, modrm=mod_mem))
-addop("shld", [bs8(0x0f), bs8(0xa4)] +
-      rmmod(rmreg) + [u08], [rm_arg, rmreg, u08])
-addop("shld", [bs8(0x0f), bs8(0xa5)] +
-      rmmod(rmreg) + [d_cl], [rm_arg, rmreg, d_cl])
-addop("shrd", [bs8(0x0f), bs8(0xac)] +
-      rmmod(rmreg) + [u08], [rm_arg, rmreg, u08])
-addop("shrd", [bs8(0x0f), bs8(0xad)] +
-      rmmod(rmreg) + [d_cl], [rm_arg, rmreg, d_cl])
-addop("sidt", [bs8(0x0f), bs8(0x01)] + rmmod(d1, modrm=mod_mem))
-addop("sldt", [bs8(0x0f), bs8(0x00)] + rmmod(d0, rm_arg_x=rm_arg_reg_m16))
-addop("smsw", [bs8(0x0f), bs8(0x01)] + rmmod(d4))
-addop("stc", [bs8(0xf9)])
-addop("std", [bs8(0xfd)])
-addop("sti", [bs8(0xfb)])
-addop("stosb", [bs8(0xaa)])
-addop("stosw", [bs8(0xab), bs_opmode16])
-addop("stosd", [bs8(0xab), bs_opmode32])
-addop("stosq", [bs8(0xab), bs_opmode64])
-
-addop("str", [bs8(0x0f), bs8(0x00)] + rmmod(d1))
-
-addop("sub", [bs("0010110"), w8, d_eax, d_imm])
-addop("sub", [bs("100000"), se, w8] + rmmod(d5, rm_arg_w8) + [d_imm])
-addop("sub", [bs("001010"), swapargs, w8] +
-      rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-
-addop("syscall", [bs8(0x0f), bs8(0x05)])
-addop("sysenter", [bs8(0x0f), bs8(0x34)])
-addop("sysexit", [bs8(0x0f), bs8(0x35)])
-addop("sysret", [bs8(0x0f), bs8(0x07)])
-addop("test", [bs("1010100"), w8, d_eax, d_imm])
-addop("test", [bs("1111011"), w8] + rmmod(d0, rm_arg_w8) + [d_imm])
-addop("test", [bs("1000010"), w8] +
-      rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-addop("ud2", [bs8(0x0f), bs8(0x0b)])
-addop("verr", [bs8(0x0f), bs8(0x00)] + rmmod(d4))
-addop("verw", [bs8(0x0f), bs8(0x00)] + rmmod(d5))
-addop("wbinvd", [bs8(0x0f), bs8(0x09)])
-addop("wrmsr", [bs8(0x0f), bs8(0x30)])
-addop("xadd", [bs8(0x0f), bs("1100000"), w8]
-      + rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-
-addop("nop", [bs8(0x90), no_rex], alias=True)
-
-addop("xchg", [bs('10010'), d_eax, reg])
-addop("xchg", [bs('1000011'), w8] +
-      rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-addop("xlat", [bs8(0xd7)])
-
-
-addop("xor", [bs("0011010"), w8, d_eax, d_imm])
-addop("xor", [bs("100000"), se, w8] + rmmod(d6, rm_arg_w8) + [d_imm])
-addop("xor", [bs("001100"), swapargs, w8] +
-      rmmod(rmreg, rm_arg_w8), [rm_arg_w8, rmreg])
-
-
-addop("xgetbv", [bs8(0x0f), bs8(0x01), bs8(0xd0)])
-
-
-
-#### MMX/SSE/AVX operations
-#### Categories are the same than here: https://software.intel.com/sites/landingpage/IntrinsicsGuide/
-####
-
-### Arithmetic (integers)
-###
-
-## Move
-# SSE
-addop("movapd", [bs8(0x0f), bs("0010100"), swapargs]
-      + rmmod(xmm_reg, rm_arg_xmm) + [bs_opmode16], [xmm_reg, rm_arg_xmm])
-addop("movaps", [bs8(0x0f), bs("0010100"), swapargs]
-      + rmmod(xmm_reg, rm_arg_xmm_m128) + [bs_opmode32], [xmm_reg, rm_arg_xmm_m128])
-addop("movaps", [bs8(0x0f), bs("0010100"), swapargs]
-      + rmmod(xmm_reg, rm_arg_xmm_m128) + [bs_opmode64], [xmm_reg, rm_arg_xmm_m128])
-addop("movdqu", [bs8(0x0f), bs("011"), swapargs, bs("1111"), pref_f3]
-      + rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-addop("movdqa", [bs8(0x0f), bs("011"), swapargs, bs("1111"), pref_66]
-      + rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-
-addop("movhpd", [bs8(0x0f), bs("0001011"), swapargs, pref_66] +
-      rmmod(xmm_reg, rm_arg_m64), [xmm_reg, rm_arg_m64])
-addop("movhps", [bs8(0x0f), bs("0001011"), swapargs, no_xmm_pref] +
-      rmmod(xmm_reg, rm_arg_m64), [xmm_reg, rm_arg_m64])
-addop("movlpd", [bs8(0x0f), bs("0001001"), swapargs, pref_66] +
-      rmmod(xmm_reg, rm_arg_m64), [xmm_reg, rm_arg_m64])
-addop("movlps", [bs8(0x0f), bs("0001001"), swapargs, no_xmm_pref] +
-      rmmod(xmm_reg, rm_arg_m64), [xmm_reg, rm_arg_m64])
-
-addop("movhlps", [bs8(0x0f), bs8(0x12), no_xmm_pref] +
-      rmmod(xmm_reg, rm_arg_xmm_reg), [xmm_reg, rm_arg_xmm_reg])
-addop("movlhps", [bs8(0x0f), bs8(0x16), no_xmm_pref] +
-      rmmod(xmm_reg, rm_arg_xmm_reg), [xmm_reg, rm_arg_xmm_reg])
-
-addop("movdq2q", [bs8(0x0f), bs8(0xd6), pref_f2] +
-      rmmod(mm_reg, rm_arg_xmm_reg), [mm_reg, rm_arg_xmm_reg])
-addop("movq2dq", [bs8(0x0f), bs8(0xd6), pref_f3] +
-      rmmod(xmm_reg, rm_arg_mm))
-
-## Additions
-# SSE
-addop("paddb", [bs8(0x0f), bs8(0xfc), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-addop("paddw", [bs8(0x0f), bs8(0xfd), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-addop("paddd", [bs8(0x0f), bs8(0xfe), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-addop("paddq", [bs8(0x0f), bs8(0xd4), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-addop("paddb", [bs8(0x0f), bs8(0xfc), no_xmm_pref] + rmmod(mm_reg, rm_arg_mm))
-addop("paddw", [bs8(0x0f), bs8(0xfd), no_xmm_pref] + rmmod(mm_reg, rm_arg_mm))
-addop("paddd", [bs8(0x0f), bs8(0xfe), no_xmm_pref] + rmmod(mm_reg, rm_arg_mm))
-addop("paddq", [bs8(0x0f), bs8(0xd4), no_xmm_pref] + rmmod(mm_reg, rm_arg_mm))
-
-## Substractions
-# SSE
-addop("psubb", [bs8(0x0f), bs8(0xf8), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-addop("psubw", [bs8(0x0f), bs8(0xf9), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-addop("psubd", [bs8(0x0f), bs8(0xfa), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-addop("psubq", [bs8(0x0f), bs8(0xfb), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-addop("psubb", [bs8(0x0f), bs8(0xf8), no_xmm_pref] + rmmod(mm_reg, rm_arg_mm))
-addop("psubw", [bs8(0x0f), bs8(0xf9), no_xmm_pref] + rmmod(mm_reg, rm_arg_mm))
-addop("psubd", [bs8(0x0f), bs8(0xfa), no_xmm_pref] + rmmod(mm_reg, rm_arg_mm))
-addop("psubq", [bs8(0x0f), bs8(0xfb), no_xmm_pref] + rmmod(mm_reg, rm_arg_mm))
-
-### Arithmetic (floating-point)
-###
-
-## Additions
-# SSE
-addop("addps", [bs8(0x0f), bs8(0x58), no_xmm_pref] + rmmod(xmm_reg, rm_arg_xmm))
-addop("addpd", [bs8(0x0f), bs8(0x58), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-## Substractions
-# SSE
-addop("subps", [bs8(0x0f), bs8(0x5c), no_xmm_pref] + rmmod(xmm_reg, rm_arg_xmm))
-addop("subpd", [bs8(0x0f), bs8(0x5c), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-## Multiplications
-# SSE
-addop("mulps", [bs8(0x0f), bs8(0x59), no_xmm_pref] + rmmod(xmm_reg, rm_arg_xmm))
-addop("mulpd", [bs8(0x0f), bs8(0x59), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-## Divisions
-# SSE
-addop("divps", [bs8(0x0f), bs8(0x5e), no_xmm_pref] + rmmod(xmm_reg, rm_arg_xmm))
-addop("divpd", [bs8(0x0f), bs8(0x5e), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-### Logical (floating-point)
-###
-
-## XOR
-addop("xorps", [bs8(0x0f), bs8(0x57), no_xmm_pref] + rmmod(xmm_reg, rm_arg_xmm))
-addop("xorpd", [bs8(0x0f), bs8(0x57), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-## AND
-addop("andps", [bs8(0x0f), bs8(0x54), no_xmm_pref] + rmmod(xmm_reg, rm_arg_xmm))
-addop("andpd", [bs8(0x0f), bs8(0x54), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-addop("andnps", [bs8(0x0f), bs8(0x55), no_xmm_pref] + rmmod(xmm_reg, rm_arg_xmm))
-addop("andnpd", [bs8(0x0f), bs8(0x55), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-## OR
-addop("orps", [bs8(0x0f), bs8(0x56), no_xmm_pref] + rmmod(xmm_reg, rm_arg_xmm))
-addop("orpd", [bs8(0x0f), bs8(0x56), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-## AND
-# MMX
-addop("pand", [bs8(0x0f), bs8(0xdb), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm), [mm_reg, rm_arg_mm])
-# SSE
-addop("pand", [bs8(0x0f), bs8(0xdb), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-
-## ANDN
-# MMX
-addop("pandn", [bs8(0x0f), bs8(0xdf), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm), [mm_reg, rm_arg_mm])
-# SSE
-addop("pandn", [bs8(0x0f), bs8(0xdf), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-
-## OR
-# MMX
-addop("por", [bs8(0x0f), bs8(0xeb), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm), [mm_reg, rm_arg_mm])
-# SSE
-addop("por", [bs8(0x0f), bs8(0xeb), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-
-## XOR
-# MMX
-addop("pxor", [bs8(0x0f), bs8(0xef), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-# MMX
-addop("pxor", [bs8(0x0f), bs8(0xef), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-### Comparisons (floating-point)
-###
-addop("minps", [bs8(0x0f), bs8(0x5d), no_xmm_pref] + rmmod(xmm_reg,
-                                                           rm_arg_xmm_m128))
-addop("minss", [bs8(0x0f), bs8(0x5d), pref_f3] + rmmod(xmm_reg,
-                                                       rm_arg_xmm_m32))
-addop("minpd", [bs8(0x0f), bs8(0x5d), pref_66] + rmmod(xmm_reg,
-                                                       rm_arg_xmm_m128))
-addop("minsd", [bs8(0x0f), bs8(0x5d), pref_f2] + rmmod(xmm_reg,
-                                                       rm_arg_xmm_m64))
-addop("maxps", [bs8(0x0f), bs8(0x5f), no_xmm_pref] + rmmod(xmm_reg,
-                                                           rm_arg_xmm_m128))
-addop("maxpd", [bs8(0x0f), bs8(0x5f), pref_66] + rmmod(xmm_reg,
-                                                       rm_arg_xmm_m128))
-addop("maxsd", [bs8(0x0f), bs8(0x5f), pref_f2] + rmmod(xmm_reg, rm_arg_xmm_m64))
-addop("maxss", [bs8(0x0f), bs8(0x5f), pref_f3] + rmmod(xmm_reg, rm_arg_xmm_m32))
-
-for cond_name, value in [
-        ("eq", 0x00),
-        ("lt", 0x01),
-        ("le", 0x02),
-        ("unord", 0x03),
-        ("neq", 0x04),
-        ("nlt", 0x05),
-        ("nle", 0x06),
-        ("ord", 0x07),
-]:
-    addop("cmp%sps" % cond_name, [bs8(0x0f), bs8(0xc2), no_xmm_pref] +
-          rmmod(xmm_reg, rm_arg_xmm_m64) + [bs8(value)])
-    addop("cmp%spd" % cond_name, [bs8(0x0f), bs8(0xc2), pref_66] +
-          rmmod(xmm_reg, rm_arg_xmm_m64) + [bs8(value)])
-    addop("cmp%sss" % cond_name, [bs8(0x0f), bs8(0xc2), pref_f3] +
-          rmmod(xmm_reg, rm_arg_xmm_m32) + [bs8(value)])
-    addop("cmp%ssd" % cond_name, [bs8(0x0f), bs8(0xc2), pref_f2] +
-          rmmod(xmm_reg, rm_arg_xmm_m32) + [bs8(value)])
-
-
-
-addop("pshufb", [bs8(0x0f), bs8(0x38), bs8(0x00), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("pshufb", [bs8(0x0f), bs8(0x38), bs8(0x00), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("pshufd", [bs8(0x0f), bs8(0x70), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128) + [u08])
-addop("pshuflw", [bs8(0x0f), bs8(0x70), pref_f2] +
-      rmmod(xmm_reg, rm_arg_xmm_m128) + [u08])
-addop("pshufhw", [bs8(0x0f), bs8(0x70), pref_f3] +
-      rmmod(xmm_reg, rm_arg_xmm_m128) + [u08])
-
-
-### Convert
-### SS = single precision
-### SD = double precision
-###
-
-## SS -> SD
-##
-
-addop("cvtdq2pd", [bs8(0x0f), bs8(0xe6), pref_f3]
-      + rmmod(xmm_reg, rm_arg_xmm_m64))
-addop("cvtdq2ps", [bs8(0x0f), bs8(0x5b), no_xmm_pref]
-      + rmmod(xmm_reg, rm_arg_xmm))
-addop("cvtpd2dq", [bs8(0x0f), bs8(0xe6), pref_f2]
-      + rmmod(xmm_reg, rm_arg_xmm))
-addop("cvtpd2pi", [bs8(0x0f), bs8(0x2d), pref_66]
-      + rmmod(mm_reg, rm_arg_xmm))
-addop("cvtpd2ps", [bs8(0x0f), bs8(0x5a), pref_66]
-      + rmmod(xmm_reg, rm_arg_xmm))
-addop("cvtpi2pd", [bs8(0x0f), bs8(0x2a), pref_66]
-      + rmmod(xmm_reg, rm_arg_mm_m64))
-addop("cvtpi2ps", [bs8(0x0f), bs8(0x2a), no_xmm_pref]
-      + rmmod(xmm_reg, rm_arg_mm_m64))
-addop("cvtps2dq", [bs8(0x0f), bs8(0x5b), pref_66]
-      + rmmod(xmm_reg, rm_arg_xmm))
-addop("cvtps2pd", [bs8(0x0f), bs8(0x5a), no_xmm_pref]
-      + rmmod(xmm_reg, rm_arg_xmm_m64))
-addop("cvtps2pi", [bs8(0x0f), bs8(0x2d), no_xmm_pref]
-      + rmmod(mm_reg, rm_arg_xmm_m64))
-addop("cvtsd2si", [bs8(0x0f), bs8(0x2d), pref_f2]
-      + rmmod(reg, rm_arg_xmm_m64))
-addop("cvtsd2ss", [bs8(0x0f), bs8(0x5a), pref_f2]
-      + rmmod(xmm_reg, rm_arg_xmm_m64))
-addop("cvtsi2sd", [bs8(0x0f), bs8(0x2a), pref_f2]
-      + rmmod(xmm_reg, rm_arg))
-addop("cvtsi2ss", [bs8(0x0f), bs8(0x2a), xmmreg, pref_f3]
-      + rmmod(xmm_reg, rm_arg))
-addop("cvtss2sd", [bs8(0x0f), bs8(0x5a), pref_f3]
-      + rmmod(xmm_reg, rm_arg_xmm_m32))
-addop("cvtss2si", [bs8(0x0f), bs8(0x2d), pref_f3]
-      + rmmod(rmreg, rm_arg_xmm_m32))
-addop("cvttpd2pi",[bs8(0x0f), bs8(0x2c), pref_66]
-      + rmmod(mm_reg, rm_arg_xmm))
-addop("cvttpd2dq",[bs8(0x0f), bs8(0xe6), pref_66]
-      + rmmod(xmm_reg, rm_arg_xmm))
-addop("cvttps2dq",[bs8(0x0f), bs8(0x5b), pref_f3]
-      + rmmod(xmm_reg, rm_arg_xmm))
-addop("cvttps2pi",[bs8(0x0f), bs8(0x2c), no_xmm_pref]
-      + rmmod(mm_reg, rm_arg_xmm_m64))
-addop("cvttsd2si",[bs8(0x0f), bs8(0x2c), pref_f2]
-      + rmmod(reg, rm_arg_xmm_m64))
-addop("cvttss2si",[bs8(0x0f), bs8(0x2c), pref_f3]
-      + rmmod(reg, rm_arg_xmm_m32))
-
-addop("palignr", [bs8(0x0f), bs8(0x73), bs8(0x0f), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64) + [u08], [mm_reg, rm_arg_mm_m64, u08])
-addop("palignr", [bs8(0x0f), bs8(0x3a), bs8(0x0f), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128) + [u08], [xmm_reg, rm_arg_xmm_m128, u08])
-
-addop("psrlq", [bs8(0x0f), bs8(0x73), no_xmm_pref] +
-      rmmod(d2, rm_arg_mm) + [u08], [rm_arg_mm, u08])
-addop("psrlq", [bs8(0x0f), bs8(0x73), pref_66] +
-      rmmod(d2, rm_arg_xmm) + [u08], [rm_arg_xmm, u08])
-
-addop("psrlq", [bs8(0x0f), bs8(0xd3), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm), [mm_reg, rm_arg_mm])
-addop("psrlq", [bs8(0x0f), bs8(0xd3), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-
-
-addop("psrld", [bs8(0x0f), bs8(0x72), no_xmm_pref] +
-      rmmod(d2, rm_arg_mm) + [u08], [rm_arg_mm, u08])
-addop("psrld", [bs8(0x0f), bs8(0x72), pref_66] +
-      rmmod(d2, rm_arg_xmm) + [u08], [rm_arg_xmm, u08])
-
-addop("psrld", [bs8(0x0f), bs8(0xd2), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm), [mm_reg, rm_arg_mm])
-addop("psrld", [bs8(0x0f), bs8(0xd2), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-
-addop("psrldq", [bs8(0x0f), bs8(0x73), pref_66] +
-      rmmod(d3, rm_arg_xmm) + [u08], [rm_arg_xmm, u08])
-
-addop("psrlw", [bs8(0x0f), bs8(0x71), no_xmm_pref] +
-      rmmod(d2, rm_arg_mm) + [u08], [rm_arg_mm, u08])
-addop("psrlw", [bs8(0x0f), bs8(0x71), pref_66] +
-      rmmod(d2, rm_arg_xmm) + [u08], [rm_arg_xmm, u08])
-
-addop("psrlw", [bs8(0x0f), bs8(0xd1), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64), [mm_reg, rm_arg_mm_m64])
-addop("psrlw", [bs8(0x0f), bs8(0xd1), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128), [xmm_reg, rm_arg_xmm_m128])
-
-addop("psraw", [bs8(0x0f), bs8(0xe1), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64), [mm_reg, rm_arg_mm_m64])
-addop("psraw", [bs8(0x0f), bs8(0xe1), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128), [xmm_reg, rm_arg_xmm_m128])
-
-addop("psraw", [bs8(0x0f), bs8(0x71), no_xmm_pref] +
-      rmmod(d4, rm_arg_mm_m64) + [u08], [rm_arg_mm_m64, u08])
-addop("psraw", [bs8(0x0f), bs8(0x71), pref_66] +
-      rmmod(d4, rm_arg_xmm_m128) + [u08], [rm_arg_xmm_m128, u08])
-
-addop("psrad", [bs8(0x0f), bs8(0xe2), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64), [mm_reg, rm_arg_mm_m64])
-addop("psrad", [bs8(0x0f), bs8(0xe2), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128), [xmm_reg, rm_arg_xmm_m128])
-
-addop("psrad", [bs8(0x0f), bs8(0x72), no_xmm_pref] +
-      rmmod(d4, rm_arg_mm_m64) + [u08], [rm_arg_mm_m64, u08])
-addop("psrad", [bs8(0x0f), bs8(0x72), pref_66] +
-      rmmod(d4, rm_arg_xmm_m128) + [u08], [rm_arg_xmm_m128, u08])
-
-
-addop("psllq", [bs8(0x0f), bs8(0x73), no_xmm_pref] +
-      rmmod(d6, rm_arg_mm) + [u08], [rm_arg_mm, u08])
-addop("psllq", [bs8(0x0f), bs8(0x73), pref_66] +
-      rmmod(d6, rm_arg_xmm) + [u08], [rm_arg_xmm, u08])
-
-addop("psllq", [bs8(0x0f), bs8(0xf3), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm), [mm_reg, rm_arg_mm])
-addop("psllq", [bs8(0x0f), bs8(0xf3), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-
-
-addop("pslld", [bs8(0x0f), bs8(0x72), no_xmm_pref] +
-      rmmod(d6, rm_arg_mm) + [u08], [rm_arg_mm, u08])
-addop("pslld", [bs8(0x0f), bs8(0x72), pref_66] +
-      rmmod(d6, rm_arg_xmm) + [u08], [rm_arg_xmm, u08])
-
-addop("pslld", [bs8(0x0f), bs8(0xf2), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm), [mm_reg, rm_arg_mm])
-addop("pslld", [bs8(0x0f), bs8(0xf2), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-
-
-addop("psllw", [bs8(0x0f), bs8(0x71), no_xmm_pref] +
-      rmmod(d6, rm_arg_mm) + [u08], [rm_arg_mm, u08])
-addop("psllw", [bs8(0x0f), bs8(0x71), pref_66] +
-      rmmod(d6, rm_arg_xmm) + [u08], [rm_arg_xmm, u08])
-
-addop("psllw", [bs8(0x0f), bs8(0xf1), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm), [mm_reg, rm_arg_mm])
-addop("psllw", [bs8(0x0f), bs8(0xf1), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm), [xmm_reg, rm_arg_xmm])
-
-addop("pslldq", [bs8(0x0f), bs8(0x73), pref_66] +
-      rmmod(d7, rm_arg_xmm) + [u08], [rm_arg_xmm, u08])
-
-
-addop("pmaxub", [bs8(0x0f), bs8(0xde), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("pmaxub", [bs8(0x0f), bs8(0xde), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("pmaxuw", [bs8(0x0f), bs8(0x38), bs8(0x3e), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("pmaxud", [bs8(0x0f), bs8(0x38), bs8(0x3f), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("pmaxsw", [bs8(0x0f), bs8(0xee), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("pmaxsw", [bs8(0x0f), bs8(0xee), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-
-addop("pminub", [bs8(0x0f), bs8(0xda), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("pminub", [bs8(0x0f), bs8(0xda), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("pminuw", [bs8(0x0f), bs8(0x38), bs8(0x3a), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("pminud", [bs8(0x0f), bs8(0x38), bs8(0x3b), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-
-addop("pcmpeqb", [bs8(0x0f), bs8(0x74), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("pcmpeqb", [bs8(0x0f), bs8(0x74), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("pcmpeqw", [bs8(0x0f), bs8(0x75), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("pcmpeqw", [bs8(0x0f), bs8(0x75), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("pcmpeqd", [bs8(0x0f), bs8(0x76), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("pcmpeqd", [bs8(0x0f), bs8(0x76), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("pcmpgtb", [bs8(0x0f), bs8(0x64), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("pcmpgtb", [bs8(0x0f), bs8(0x64), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("pcmpgtw", [bs8(0x0f), bs8(0x65), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("pcmpgtw", [bs8(0x0f), bs8(0x65), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("pcmpgtd", [bs8(0x0f), bs8(0x66), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("pcmpgtd", [bs8(0x0f), bs8(0x66), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("pcmpeqq", [bs8(0x0f), bs8(0x38), bs8(0x29), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-addop("pcmpgtq", [bs8(0x0f), bs8(0x38), bs8(0x37), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-addop("punpckhbw", [bs8(0x0f), bs8(0x68), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("punpckhbw", [bs8(0x0f), bs8(0x68), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("punpckhwd", [bs8(0x0f), bs8(0x69), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("punpckhwd", [bs8(0x0f), bs8(0x69), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("punpckhdq", [bs8(0x0f), bs8(0x6a), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("punpckhdq", [bs8(0x0f), bs8(0x6a), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("punpckhqdq", [bs8(0x0f), bs8(0x6d), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-
-
-addop("punpcklbw", [bs8(0x0f), bs8(0x60), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("punpcklbw", [bs8(0x0f), bs8(0x60), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("punpcklwd", [bs8(0x0f), bs8(0x61), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("punpcklwd", [bs8(0x0f), bs8(0x61), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("punpckldq", [bs8(0x0f), bs8(0x62), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm))
-addop("punpckldq", [bs8(0x0f), bs8(0x62), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-addop("punpcklqdq", [bs8(0x0f), bs8(0x6c), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-
-addop("unpckhps", [bs8(0x0f), bs8(0x15), no_xmm_pref] +
-      rmmod(xmm_reg, rm_arg_xmm))
-addop("unpckhpd", [bs8(0x0f), bs8(0x15), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-
-addop("unpcklps", [bs8(0x0f), bs8(0x14), no_xmm_pref] +
-      rmmod(xmm_reg, rm_arg_xmm))
-addop("unpcklpd", [bs8(0x0f), bs8(0x14), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-
-
-
-addop("pinsrb", [bs8(0x0f), bs8(0x3a), bs8(0x20), pref_66] +
-      rmmod(xmm_reg, rm_arg_reg_m08) + [u08])
-addop("pinsrd", [bs8(0x0f), bs8(0x3a), bs8(0x22), pref_66, bs_opmode32] +
-      rmmod(xmm_reg, rm_arg) + [u08])
-addop("pinsrq", [bs8(0x0f), bs8(0x3a), bs8(0x22), pref_66] +
-      rmmod(xmm_reg, rm_arg_m64) + [bs_opmode64] + [u08])
-
-addop("pinsrw", [bs8(0x0f), bs8(0xc4), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_reg_m16) + [u08])
-addop("pinsrw", [bs8(0x0f), bs8(0xc4), pref_66] +
-      rmmod(xmm_reg, rm_arg_reg_m16) + [u08])
-
-
-addop("pextrb", [bs8(0x0f), bs8(0x3a), bs8(0x14), pref_66] +
-      rmmod(xmm_reg, rm_arg_reg_m08) + [u08], [rm_arg_reg_m08, xmm_reg, u08])
-addop("pextrd", [bs8(0x0f), bs8(0x3a), bs8(0x16), pref_66, bs_opmode32] +
-      rmmod(xmm_reg, rm_arg) + [u08], [rm_arg, xmm_reg, u08])
-addop("pextrq", [bs8(0x0f), bs8(0x3a), bs8(0x16), pref_66] +
-      rmmod(xmm_reg, rm_arg_m64) + [bs_opmode64] + [u08], [rm_arg_m64, xmm_reg, u08])
-
-
-addop("pextrw", [bs8(0x0f), bs8(0x3a), bs8(0x15), pref_66] +
-      rmmod(xmm_reg, rm_arg_reg_m16) + [u08], [rm_arg_reg_m16, xmm_reg, u08])
-addop("pextrw", [bs8(0x0f), bs8(0xc5), no_xmm_pref] +
-      rmmod(rmreg, rm_arg_mm) + [u08], [rmreg, rm_arg_mm, u08])
-addop("pextrw", [bs8(0x0f), bs8(0xc5), pref_66] +
-      rmmod(rmreg, rm_arg_xmm) + [u08], [rmreg, rm_arg_xmm, u08])
-
-
-addop("sqrtpd", [bs8(0x0f), bs8(0x51), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm))
-addop("sqrtps", [bs8(0x0f), bs8(0x51), no_xmm_pref] +
-      rmmod(xmm_reg, rm_arg_xmm))
-addop("sqrtsd", [bs8(0x0f), bs8(0x51), pref_f2] +
-      rmmod(xmm_reg, rm_arg_xmm_m64))
-addop("sqrtss", [bs8(0x0f), bs8(0x51), pref_f3] +
-      rmmod(xmm_reg, rm_arg_xmm_m32))
-
-addop("pmovmskb", [bs8(0x0f), bs8(0xd7), no_xmm_pref] +
-      rmmod(reg_modrm, rm_arg_mm_reg))
-addop("pmovmskb", [bs8(0x0f), bs8(0xd7), pref_66] +
-      rmmod(reg_modrm, rm_arg_xmm_reg))
-
-addop("shufps", [bs8(0x0f), bs8(0xc6), no_xmm_pref] +
-      rmmod(xmm_reg, rm_arg_xmm) + [u08])
-addop("shufpd", [bs8(0x0f), bs8(0xc6), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm) + [u08])
-
-addop("aesenc", [bs8(0x0f), bs8(0x38), bs8(0xdc), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-addop("aesdec", [bs8(0x0f), bs8(0x38), bs8(0xde), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-addop("aesenclast", [bs8(0x0f), bs8(0x38), bs8(0xdd), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-addop("aesdeclast", [bs8(0x0f), bs8(0x38), bs8(0xdf), pref_66] + rmmod(xmm_reg, rm_arg_xmm))
-
-addop("packsswb", [bs8(0x0f), bs8(0x63), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("packsswb", [bs8(0x0f), bs8(0x63), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("packssdw", [bs8(0x0f), bs8(0x6b), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("packssdw", [bs8(0x0f), bs8(0x6b), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-
-addop("packuswb", [bs8(0x0f), bs8(0x67), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("packuswb", [bs8(0x0f), bs8(0x67), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-
-addop("pmullw", [bs8(0x0f), bs8(0xd5), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("pmullw", [bs8(0x0f), bs8(0xd5), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("pmulhuw", [bs8(0x0f), bs8(0xe4), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("pmulhuw", [bs8(0x0f), bs8(0xe4), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("pmulhw", [bs8(0x0f), bs8(0xe5), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("pmulhw", [bs8(0x0f), bs8(0xe5), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("pmuludq", [bs8(0x0f), bs8(0xf4), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("pmuludq", [bs8(0x0f), bs8(0xf4), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-
-
-addop("psubusb", [bs8(0x0f), bs8(0xd8), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("psubusb", [bs8(0x0f), bs8(0xd8), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("psubusw", [bs8(0x0f), bs8(0xd9), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("psubusw", [bs8(0x0f), bs8(0xd9), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("psubsb", [bs8(0x0f), bs8(0xe8), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("psubsb", [bs8(0x0f), bs8(0xe8), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("psubsw", [bs8(0x0f), bs8(0xe9), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("psubsw", [bs8(0x0f), bs8(0xe9), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-
-
-addop("paddusb", [bs8(0x0f), bs8(0xdc), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("paddusb", [bs8(0x0f), bs8(0xdc), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("paddusw", [bs8(0x0f), bs8(0xdd), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("paddusw", [bs8(0x0f), bs8(0xdd), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("paddsb", [bs8(0x0f), bs8(0xec), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("paddsb", [bs8(0x0f), bs8(0xec), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("paddsw", [bs8(0x0f), bs8(0xed), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("paddsw", [bs8(0x0f), bs8(0xed), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-
-addop("pmaddwd", [bs8(0x0f), bs8(0xf5), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("pmaddwd", [bs8(0x0f), bs8(0xf5), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-
-addop("psadbw", [bs8(0x0f), bs8(0xf6), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("psadbw", [bs8(0x0f), bs8(0xf6), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-
-addop("pavgb", [bs8(0x0f), bs8(0xe0), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("pavgb", [bs8(0x0f), bs8(0xe0), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-addop("pavgw", [bs8(0x0f), bs8(0xe3), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_m64))
-addop("pavgw", [bs8(0x0f), bs8(0xe3), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_m128))
-
-addop("maskmovq", [bs8(0x0f), bs8(0xf7), no_xmm_pref] +
-      rmmod(mm_reg, rm_arg_mm_reg))
-addop("maskmovdqu", [bs8(0x0f), bs8(0xf7), pref_66] +
-      rmmod(xmm_reg, rm_arg_xmm_reg))
-
-addop("emms", [bs8(0x0f), bs8(0x77)])
-
-addop("incssp", [pref_f3, bs8(0x0f), bs8(0xae)] + rmmod(d5))
-addop("rdssp", [pref_f3, bs8(0x0f), bs8(0x1e)] + rmmod(d1, modrm=mod_reg))
-addop("saveprevssp", [pref_f3, bs8(0x0f), bs8(0x01), bs8(0xea)])
-addop("rstorssp", [pref_f3, bs8(0x0f), bs8(0x01)] + rmmod(d5, rm_arg_xmm, modrm=mod_mem))
-addop("wrss", [bs8(0x0f), bs8(0x38), bs8(0xf6)] + rmmod(rmreg, rm_arg), [rm_arg, rmreg])
-addop("wruss", [pref_66, bs8(0x0f), bs8(0x38), bs8(0xf5)] + rmmod(rmreg, rm_arg), [rm_arg, rmreg])
-addop("setssbsy", [pref_f3, bs8(0x0f), bs8(0x01), bs8(0xe8)])
-addop("clrssbsy", [pref_f3, bs8(0x0f), bs8(0xae)] + rmmod(d6, rm_arg_xmm))
-addop("endbr64", [pref_f3, bs8(0x0f), bs8(0x1e), bs8(0xfa)])
-addop("endbr32", [pref_f3, bs8(0x0f), bs8(0x1e), bs8(0xfb)])
-
-mn_x86.bintree = factor_one_bit(mn_x86.bintree)
-# mn_x86.bintree = factor_fields_all(mn_x86.bintree)
-"""
-mod reg r/m
- XX XXX XXX
-
-"""
-
-
-def print_size(e):
-    print(e, e.size)
-    return e
diff --git a/miasm/arch/x86/ctype.py b/miasm/arch/x86/ctype.py
deleted file mode 100644
index 2a61689a..00000000
--- a/miasm/arch/x86/ctype.py
+++ /dev/null
@@ -1,137 +0,0 @@
-from miasm.core.objc import CLeafTypes, ObjCDecl, PADDING_TYPE_NAME
-from miasm.core.ctypesmngr import CTypeId, CTypePtr
-
-
-class CTypeAMD64_unk(CLeafTypes):
-    """Define C types sizes/alignment for x86_64 architecture"""
-
-    obj_pad = ObjCDecl(PADDING_TYPE_NAME, 1, 1) # __padding__ is size 1/align 1
-
-    obj_char = ObjCDecl("char", 1, 1)
-    obj_short = ObjCDecl("short", 2, 2)
-    obj_int = ObjCDecl("int", 4, 4)
-    obj_long = ObjCDecl("long", 8, 8)
-
-    obj_uchar = ObjCDecl("uchar", 1, 1)
-    obj_ushort = ObjCDecl("ushort", 2, 2)
-    obj_uint = ObjCDecl("uint", 4, 4)
-    obj_ulong = ObjCDecl("ulong", 8, 8)
-
-    obj_void = ObjCDecl("void", 1, 1)
-
-    obj_enum = ObjCDecl("enum", 4, 4)
-
-    obj_float = ObjCDecl("float", 4, 4)
-    obj_double = ObjCDecl("double", 8, 8)
-    obj_ldouble = ObjCDecl("ldouble", 16, 16)
-
-    def __init__(self):
-        self.types = {
-            CTypeId(PADDING_TYPE_NAME): self.obj_pad,
-
-            CTypeId('char'): self.obj_char,
-            CTypeId('short'): self.obj_short,
-            CTypeId('int'): self.obj_int,
-            CTypeId('void'): self.obj_void,
-            CTypeId('long',): self.obj_long,
-            CTypeId('float'): self.obj_float,
-            CTypeId('double'): self.obj_double,
-
-            CTypeId('signed', 'char'): self.obj_char,
-            CTypeId('unsigned', 'char'): self.obj_uchar,
-
-            CTypeId('short', 'int'): self.obj_short,
-            CTypeId('signed', 'short'): self.obj_short,
-            CTypeId('signed', 'short', 'int'): self.obj_short,
-            CTypeId('unsigned', 'short'): self.obj_ushort,
-            CTypeId('unsigned', 'short', 'int'): self.obj_ushort,
-
-            CTypeId('unsigned', ): self.obj_uint,
-            CTypeId('unsigned', 'int'): self.obj_uint,
-            CTypeId('signed', 'int'): self.obj_int,
-
-            CTypeId('long', 'int'): self.obj_long,
-            CTypeId('long', 'long'): self.obj_long,
-            CTypeId('long', 'long', 'int'): self.obj_long,
-            CTypeId('signed', 'long', 'long'): self.obj_long,
-            CTypeId('unsigned', 'long', 'long'): self.obj_ulong,
-            CTypeId('signed', 'long', 'long', 'int'): self.obj_long,
-            CTypeId('unsigned', 'long', 'long', 'int'): self.obj_ulong,
-
-            CTypeId('signed', 'long'): self.obj_long,
-            CTypeId('unsigned', 'long'): self.obj_ulong,
-            CTypeId('signed', 'long', 'int'): self.obj_long,
-            CTypeId('unsigned', 'long', 'int'): self.obj_ulong,
-
-            CTypeId('long', 'double'): self.obj_ldouble,
-            CTypePtr(CTypeId('void')): self.obj_ulong,
-        }
-
-
-
-
-
-class CTypeX86_unk(CLeafTypes):
-    """Define C types sizes/alignment for x86_32 architecture"""
-
-    obj_pad = ObjCDecl(PADDING_TYPE_NAME, 1, 1) # __padding__ is size 1/align 1
-
-    obj_char = ObjCDecl("char", 1, 1)
-    obj_short = ObjCDecl("short", 2, 2)
-    obj_int = ObjCDecl("int", 4, 4)
-    obj_long = ObjCDecl("long", 4, 4)
-
-    obj_uchar = ObjCDecl("uchar", 1, 1)
-    obj_ushort = ObjCDecl("ushort", 2, 2)
-    obj_uint = ObjCDecl("uint", 4, 4)
-    obj_ulong = ObjCDecl("ulong", 4, 4)
-
-    obj_void = ObjCDecl("void", 1, 1)
-
-    obj_enum = ObjCDecl("enum", 4, 4)
-
-    obj_float = ObjCDecl("float", 4, 4)
-    obj_double = ObjCDecl("double", 8, 8)
-    obj_ldouble = ObjCDecl("ldouble", 16, 16)
-
-    def __init__(self):
-        self.types = {
-            CTypeId(PADDING_TYPE_NAME): self.obj_pad,
-
-            CTypeId('char'): self.obj_char,
-            CTypeId('short'): self.obj_short,
-            CTypeId('int'): self.obj_int,
-            CTypeId('void'): self.obj_void,
-            CTypeId('long',): self.obj_long,
-            CTypeId('float'): self.obj_float,
-            CTypeId('double'): self.obj_double,
-
-            CTypeId('signed', 'char'): self.obj_char,
-            CTypeId('unsigned', 'char'): self.obj_uchar,
-
-            CTypeId('short', 'int'): self.obj_short,
-            CTypeId('signed', 'short'): self.obj_short,
-            CTypeId('signed', 'short', 'int'): self.obj_short,
-            CTypeId('unsigned', 'short'): self.obj_ushort,
-            CTypeId('unsigned', 'short', 'int'): self.obj_ushort,
-
-            CTypeId('unsigned', ): self.obj_uint,
-            CTypeId('unsigned', 'int'): self.obj_uint,
-            CTypeId('signed', 'int'): self.obj_int,
-
-            CTypeId('long', 'int'): self.obj_long,
-            CTypeId('long', 'long'): self.obj_long,
-            CTypeId('long', 'long', 'int'): self.obj_long,
-            CTypeId('signed', 'long', 'long'): self.obj_long,
-            CTypeId('unsigned', 'long', 'long'): self.obj_ulong,
-            CTypeId('signed', 'long', 'long', 'int'): self.obj_long,
-            CTypeId('unsigned', 'long', 'long', 'int'): self.obj_ulong,
-
-            CTypeId('signed', 'long'): self.obj_long,
-            CTypeId('unsigned', 'long'): self.obj_ulong,
-            CTypeId('signed', 'long', 'int'): self.obj_long,
-            CTypeId('unsigned', 'long', 'int'): self.obj_ulong,
-
-            CTypeId('long', 'double'): self.obj_ldouble,
-            CTypePtr(CTypeId('void')): self.obj_uint,
-        }
diff --git a/miasm/arch/x86/disasm.py b/miasm/arch/x86/disasm.py
deleted file mode 100644
index 49b7158a..00000000
--- a/miasm/arch/x86/disasm.py
+++ /dev/null
@@ -1,30 +0,0 @@
-from miasm.core.asmblock import disasmEngine
-from miasm.arch.x86.arch import mn_x86
-
-
-cb_x86_funcs = []
-
-
-def cb_x86_disasm(mdis, cur_block, offset_to_dis):
-    for func in cb_x86_funcs:
-        func(mdis, cur_block, offset_to_dis)
-
-
-class dis_x86(disasmEngine):
-    attrib = None
-
-    def __init__(self, bs=None, **kwargs):
-        super(dis_x86, self).__init__(mn_x86, self.attrib, bs, **kwargs)
-        self.dis_block_callback = cb_x86_disasm
-
-
-class dis_x86_16(dis_x86):
-    attrib = 16
-
-
-class dis_x86_32(dis_x86):
-    attrib = 32
-
-
-class dis_x86_64(dis_x86):
-    attrib = 64
diff --git a/miasm/arch/x86/jit.py b/miasm/arch/x86/jit.py
deleted file mode 100644
index a90dec07..00000000
--- a/miasm/arch/x86/jit.py
+++ /dev/null
@@ -1,296 +0,0 @@
-from builtins import range
-import logging
-
-from miasm.jitter.jitload import Jitter, named_arguments
-from miasm.arch.x86.sem import Lifter_X86_16, Lifter_X86_32, Lifter_X86_64
-from miasm.jitter.codegen import CGen
-from miasm.ir.translators.C import TranslatorC
-
-log = logging.getLogger('jit_x86')
-hnd = logging.StreamHandler()
-hnd.setFormatter(logging.Formatter("[%(levelname)-8s]: %(message)s"))
-log.addHandler(hnd)
-log.setLevel(logging.CRITICAL)
-
-
-class x86_32_CGen(CGen):
-    def __init__(self, lifter):
-        self.lifter = lifter
-        self.PC = self.lifter.arch.regs.RIP
-        self.translator = TranslatorC(self.lifter.loc_db)
-        self.init_arch_C()
-
-    def gen_post_code(self, attrib, pc_value):
-        out = []
-        if attrib.log_regs:
-            # Update PC for dump_gpregs
-            out.append("%s = %s;" % (self.C_PC, pc_value))
-            out.append('dump_gpregs_32(jitcpu->cpu);')
-        return out
-
-class x86_64_CGen(x86_32_CGen):
-    def gen_post_code(self, attrib, pc_value):
-        out = []
-        if attrib.log_regs:
-            # Update PC for dump_gpregs
-            out.append("%s = %s;" % (self.C_PC, pc_value))
-            out.append('dump_gpregs_64(jitcpu->cpu);')
-        return out
-
-class jitter_x86_16(Jitter):
-
-    C_Gen = x86_32_CGen
-
-    def __init__(self, loc_db, *args, **kwargs):
-        Jitter.__init__(self, Lifter_X86_16(loc_db), *args, **kwargs)
-        self.vm.set_little_endian()
-        self.lifter.do_stk_segm = False
-        self.orig_irbloc_fix_regs_for_mode = self.lifter.irbloc_fix_regs_for_mode
-        self.lifter.irbloc_fix_regs_for_mode = self.lifterbloc_fix_regs_for_mode
-
-    def lifterbloc_fix_regs_for_mode(self, irblock, attrib=64):
-        return self.orig_irbloc_fix_regs_for_mode(irblock, 64)
-
-    def push_uint16_t(self, value):
-        self.cpu.SP -= self.lifter.sp.size // 8
-        self.vm.set_u16(self.cpu.SP, value)
-
-    def pop_uint16_t(self):
-        value = self.vm.get_u16(self.cpu.SP)
-        self.cpu.SP += self.lifter.sp.size // 8
-        return value
-
-    def get_stack_arg(self, index):
-        return self.vm.get_u16(self.cpu.SP + 4 * index)
-
-    def init_run(self, *args, **kwargs):
-        Jitter.init_run(self, *args, **kwargs)
-        self.cpu.IP = self.pc
-
-
-class jitter_x86_32(Jitter):
-
-    C_Gen = x86_32_CGen
-
-    def __init__(self, loc_db, *args, **kwargs):
-        Jitter.__init__(self, Lifter_X86_32(loc_db), *args, **kwargs)
-        self.vm.set_little_endian()
-        self.lifter.do_stk_segm = False
-
-        self.orig_irbloc_fix_regs_for_mode = self.lifter.irbloc_fix_regs_for_mode
-        self.lifter.irbloc_fix_regs_for_mode = self.lifterbloc_fix_regs_for_mode
-
-    def lifterbloc_fix_regs_for_mode(self, irblock, attrib=64):
-        return self.orig_irbloc_fix_regs_for_mode(irblock, 64)
-
-    def push_uint16_t(self, value):
-        self.cpu.ESP -= self.lifter.sp.size // 8
-        self.vm.set_u16(self.cpu.ESP, value)
-
-    def pop_uint16_t(self):
-        value = self.vm.get_u16(self.cpu.ESP)
-        self.cpu.ESP += self.lifter.sp.size // 8
-        return value
-
-    def push_uint32_t(self, value):
-        self.cpu.ESP -= self.lifter.sp.size // 8
-        self.vm.set_u32(self.cpu.ESP, value)
-
-    def pop_uint32_t(self):
-        value = self.vm.get_u32(self.cpu.ESP)
-        self.cpu.ESP += self.lifter.sp.size // 8
-        return value
-
-    def get_stack_arg(self, index):
-        return self.vm.get_u32(self.cpu.ESP + 4 * index)
-
-    def init_run(self, *args, **kwargs):
-        Jitter.init_run(self, *args, **kwargs)
-        self.cpu.EIP = self.pc
-
-    # calling conventions
-
-    # stdcall
-    @named_arguments
-    def func_args_stdcall(self, n_args):
-        ret_ad = self.pop_uint32_t()
-        args = [self.pop_uint32_t() for _ in range(n_args)]
-        return ret_ad, args
-
-    def func_ret_stdcall(self, ret_addr, ret_value1=None, ret_value2=None):
-        self.pc = self.cpu.EIP = ret_addr
-        if ret_value1 is not None:
-            self.cpu.EAX = ret_value1
-        if ret_value2 is not None:
-            self.cpu.EDX = ret_value2
-
-    def func_prepare_stdcall(self, ret_addr, *args):
-        for arg in reversed(args):
-            self.push_uint32_t(arg)
-        self.push_uint32_t(ret_addr)
-
-    get_arg_n_stdcall = get_stack_arg
-
-    # cdecl
-    @named_arguments
-    def func_args_cdecl(self, n_args):
-        ret_ad = self.pop_uint32_t()
-        args = [self.get_stack_arg(i) for i in range(n_args)]
-        return ret_ad, args
-
-    def func_ret_cdecl(self, ret_addr, ret_value1=None, ret_value2=None):
-        self.pc = self.cpu.EIP = ret_addr
-        if ret_value1 is not None:
-            self.cpu.EAX = ret_value1
-        if ret_value2 is not None:
-            self.cpu.EDX = ret_value2
-
-    get_arg_n_cdecl = get_stack_arg
-
-    # System V
-    func_args_systemv = func_args_cdecl
-    func_ret_systemv = func_ret_cdecl
-    func_prepare_systemv = func_prepare_stdcall
-    get_arg_n_systemv = get_stack_arg
-
-
-    # fastcall
-    @named_arguments
-    def func_args_fastcall(self, n_args):
-        args_regs = ['ECX', 'EDX']
-        ret_ad = self.pop_uint32_t()
-        args = []
-        for i in range(n_args):
-            args.append(self.get_arg_n_fastcall(i))
-        return ret_ad, args
-
-    def func_prepare_fastcall(self, ret_addr, *args):
-        args_regs = ['ECX', 'EDX']
-        for i in range(min(len(args), len(args_regs))):
-            setattr(self.cpu, args_regs[i], args[i])
-        remaining_args = args[len(args_regs):]
-        for arg in reversed(remaining_args):
-            self.push_uint32_t(arg)
-        self.push_uint32_t(ret_addr)
-
-    def get_arg_n_fastcall(self, index):
-        args_regs = ['ECX', 'EDX']
-        if index < len(args_regs):
-            return getattr(self.cpu, args_regs[index])
-        return self.get_stack_arg(index - len(args_regs))
-
-    def syscall_args_systemv(self, n_args):
-        # Documentation: http://man7.org/linux/man-pages/man2/syscall.2.html
-        # args: 
-        #   i386          ebx   ecx   edx   esi   edi   ebp   -
-        args = [self.cpu.EBX, self.cpu.ECX, self.cpu.EDX, self.cpu.ESI,
-                self.cpu.EDI, self.cpu.EBP][:n_args]
-        return args
-
-    def syscall_ret_systemv(self, value):
-        # Documentation: http://man7.org/linux/man-pages/man2/syscall.2.html
-        self.cpu.EAX = value
-
-
-class jitter_x86_64(Jitter):
-
-    C_Gen = x86_64_CGen
-    args_regs_systemv = ['RDI', 'RSI', 'RDX', 'RCX', 'R8', 'R9']
-    args_regs_stdcall = ['RCX', 'RDX', 'R8', 'R9']
-
-    def __init__(self, loc_db, *args, **kwargs):
-        Jitter.__init__(self, Lifter_X86_64(loc_db), *args, **kwargs)
-        self.vm.set_little_endian()
-        self.lifter.do_stk_segm = False
-
-        self.orig_irbloc_fix_regs_for_mode = self.lifter.irbloc_fix_regs_for_mode
-        self.lifter.irbloc_fix_regs_for_mode = self.lifterbloc_fix_regs_for_mode
-
-    def lifterbloc_fix_regs_for_mode(self, irblock, attrib=64):
-        return self.orig_irbloc_fix_regs_for_mode(irblock, 64)
-
-    def push_uint64_t(self, value):
-        self.cpu.RSP -= self.lifter.sp.size // 8
-        self.vm.set_u64(self.cpu.RSP, value)
-
-    def pop_uint64_t(self):
-        value = self.vm.get_u64(self.cpu.RSP)
-        self.cpu.RSP += self.lifter.sp.size // 8
-        return value
-
-    def get_stack_arg(self, index):
-        return self.vm.get_u64(self.cpu.RSP + 8 * index)
-
-    def init_run(self, *args, **kwargs):
-        Jitter.init_run(self, *args, **kwargs)
-        self.cpu.RIP = self.pc
-
-    # calling conventions
-
-    # stdcall
-    @named_arguments
-    def func_args_stdcall(self, n_args):
-        args_regs = self.args_regs_stdcall
-        ret_ad = self.pop_uint64_t()
-        args = []
-        for i in range(min(n_args, 4)):
-            args.append(self.cpu.get_gpreg()[args_regs[i]])
-        for i in range(max(0, n_args - 4)):
-            # Take into account the shadow registers on the stack 
-            # (Microsoft 64bit stdcall ABI)
-            # => Skip the first 4 stack parameters
-            args.append(self.get_stack_arg(4 + i))
-        return ret_ad, args
-
-    def func_prepare_stdcall(self, ret_addr, *args):
-        args_regs = self.args_regs_stdcall
-        for i in range(min(len(args), len(args_regs))):
-            setattr(self.cpu, args_regs[i], args[i])
-        remaining_args = args[len(args_regs):]
-        for arg in reversed(remaining_args):
-            self.push_uint64_t(arg)
-        self.push_uint64_t(ret_addr)
-
-    def func_ret_stdcall(self, ret_addr, ret_value=None):
-        self.pc = self.cpu.RIP = ret_addr
-        if ret_value is not None:
-            self.cpu.RAX = ret_value
-        return True
-
-    # cdecl
-    func_args_cdecl = func_args_stdcall
-    func_ret_cdecl = func_ret_stdcall
-    func_prepare_cdecl = func_prepare_stdcall
-
-    # System V
-
-    def get_arg_n_systemv(self, index):
-        args_regs = self.args_regs_systemv
-        if index < len(args_regs):
-            return getattr(self.cpu, args_regs[index])
-        return self.get_stack_arg(index - len(args_regs))
-
-    @named_arguments
-    def func_args_systemv(self, n_args):
-        ret_ad = self.pop_uint64_t()
-        args = [self.get_arg_n_systemv(index) for index in range(n_args)]
-        return ret_ad, args
-
-    func_ret_systemv = func_ret_cdecl
-
-    def func_prepare_systemv(self, ret_addr, *args):
-        args_regs = self.args_regs_systemv
-        self.push_uint64_t(ret_addr)
-        for i in range(min(len(args), len(args_regs))):
-            setattr(self.cpu, args_regs[i], args[i])
-        remaining_args = args[len(args_regs):]
-        for arg in reversed(remaining_args):
-            self.push_uint64_t(arg)
-
-    def syscall_args_systemv(self, n_args):
-        args = [self.cpu.RDI, self.cpu.RSI, self.cpu.RDX, self.cpu.R10,
-                self.cpu.R8, self.cpu.R9][:n_args]
-        return args
-
-    def syscall_ret_systemv(self, value):
-        self.cpu.RAX = value
diff --git a/miasm/arch/x86/lifter_model_call.py b/miasm/arch/x86/lifter_model_call.py
deleted file mode 100644
index e75f8c69..00000000
--- a/miasm/arch/x86/lifter_model_call.py
+++ /dev/null
@@ -1,80 +0,0 @@
-#-*- coding:utf-8 -*-
-
-from miasm.expression.expression import ExprAssign, ExprOp
-from miasm.ir.ir import AssignBlock
-from miasm.ir.analysis import LifterModelCall
-from miasm.arch.x86.sem import Lifter_X86_16, Lifter_X86_32, Lifter_X86_64
-
-
-class LifterModelCall_x86_16(Lifter_X86_16, LifterModelCall):
-
-    def __init__(self, loc_db):
-        Lifter_X86_16.__init__(self, loc_db)
-        self.ret_reg = self.arch.regs.AX
-
-    def get_out_regs(self, _):
-        return set([self.ret_reg, self.sp])
-
-class LifterModelCall_x86_32(Lifter_X86_32, LifterModelCall_x86_16):
-
-    def __init__(self, loc_db):
-        Lifter_X86_32.__init__(self, loc_db)
-        self.ret_reg = self.arch.regs.EAX
-
-    def sizeof_char(self):
-        return 8
-
-    def sizeof_short(self):
-        return 16
-
-    def sizeof_int(self):
-        return 32
-
-    def sizeof_long(self):
-        return 32
-
-    def sizeof_pointer(self):
-        return 32
-
-
-class LifterModelCall_x86_64(Lifter_X86_64, LifterModelCall_x86_16):
-
-    def __init__(self, loc_db):
-        Lifter_X86_64.__init__(self, loc_db)
-        self.ret_reg = self.arch.regs.RAX
-
-    def call_effects(self, ad, instr):
-        call_assignblk = AssignBlock(
-            [
-                ExprAssign(
-                    self.ret_reg,
-                    ExprOp(
-                        'call_func_ret',
-                        ad,
-                        self.sp,
-                        self.arch.regs.RCX,
-                        self.arch.regs.RDX,
-                        self.arch.regs.R8,
-                        self.arch.regs.R9,
-                    )
-                ),
-                ExprAssign(self.sp, ExprOp('call_func_stack', ad, self.sp)),
-            ],
-            instr
-        )
-        return [call_assignblk], []
-
-    def sizeof_char(self):
-        return 8
-
-    def sizeof_short(self):
-        return 16
-
-    def sizeof_int(self):
-        return 32
-
-    def sizeof_long(self):
-        return 64
-
-    def sizeof_pointer(self):
-        return 64
diff --git a/miasm/arch/x86/regs.py b/miasm/arch/x86/regs.py
deleted file mode 100644
index dc0b9264..00000000
--- a/miasm/arch/x86/regs.py
+++ /dev/null
@@ -1,454 +0,0 @@
-from builtins import range
-from miasm.expression.expression import ExprId
-from miasm.core.cpu import reg_info
-
-
-IP = ExprId('IP', 16)
-EIP = ExprId('EIP', 32)
-RIP = ExprId('RIP', 64)
-exception_flags = ExprId('exception_flags', 32)
-interrupt_num = ExprId('interrupt_num', 8)
-
-# GP
-
-
-regs08_str = ["AL", "CL", "DL", "BL", "AH", "CH", "DH", "BH"] + \
-    ["R%dB" % (i + 8) for i in range(8)]
-regs08_expr = [ExprId(x, 8) for x in regs08_str]
-
-regs08_64_str = ["AL", "CL", "DL", "BL", "SPL", "BPL", "SIL", "DIL"] + \
-    ["R%dB" % (i + 8) for i in range(8)]
-regs08_64_expr = [ExprId(x, 8) for x in regs08_64_str]
-
-
-regs16_str = ["AX", "CX", "DX", "BX", "SP", "BP", "SI", "DI"] + \
-    ["R%dW" % (i + 8) for i in range(8)]
-regs16_expr = [ExprId(x, 16) for x in regs16_str]
-
-regs32_str = ["EAX", "ECX", "EDX", "EBX", "ESP", "EBP", "ESI", "EDI"] + \
-    ["R%dD" % (i + 8) for i in range(8)] + ["EIP"]
-regs32_expr = [ExprId(x, 32) for x in regs32_str]
-
-regs64_str = ["RAX", "RCX", "RDX", "RBX", "RSP", "RBP", "RSI", "RDI",
-              "R8", "R9", "R10", "R11", "R12", "R13", "R14", "R15",
-              "RIP"]
-regs64_expr = [ExprId(x, 64) for x in regs64_str]
-
-
-regs_xmm_str = ["XMM%d" % i for i in range(16)]
-regs_xmm_expr = [ExprId(x, 128) for x in regs_xmm_str]
-
-regs_mm_str = ["MM%d" % i for i in range(16)]
-regs_mm_expr = [ExprId(x, 64) for x in regs_mm_str]
-
-regs_bnd_str = ["BND%d" % i for i in range(4)]
-regs_bnd_expr = [ExprId(x, 128) for x in regs_bnd_str]
-
-gpregs08 = reg_info(regs08_str, regs08_expr)
-gpregs08_64 = reg_info(regs08_64_str, regs08_64_expr)
-gpregs16 = reg_info(regs16_str, regs16_expr)
-gpregs32 = reg_info(regs32_str, regs32_expr)
-gpregs64 = reg_info(regs64_str, regs64_expr)
-
-gpregs_xmm = reg_info(regs_xmm_str, regs_xmm_expr)
-gpregs_mm = reg_info(regs_mm_str, regs_mm_expr)
-gpregs_bnd = reg_info(regs_bnd_str, regs_bnd_expr)
-
-r08_eax = reg_info([regs08_str[0]], [regs08_expr[0]])
-r16_eax = reg_info([regs16_str[0]], [regs16_expr[0]])
-r32_eax = reg_info([regs32_str[0]], [regs32_expr[0]])
-r64_eax = reg_info([regs64_str[0]], [regs64_expr[0]])
-
-r08_ecx = reg_info([regs08_str[1]], [regs08_expr[1]])
-
-r_eax_all = reg_info(
-    [regs08_str[0], regs16_str[0], regs32_str[0], regs64_str[0]],
-    [regs08_expr[0], regs16_expr[0], regs32_expr[0], regs64_expr[0]])
-r_edx_all = reg_info(
-    [regs08_str[2], regs16_str[2], regs32_str[2], regs64_str[2]],
-    [regs08_expr[2], regs16_expr[2], regs32_expr[2], regs64_expr[2]])
-
-r16_edx = reg_info([regs16_str[2]], [regs16_expr[2]])
-
-
-selectr_str = ["ES", "CS", "SS", "DS", "FS", "GS"]
-selectr_expr = [ExprId(x, 16) for x in selectr_str]
-segmreg = reg_info(selectr_str, selectr_expr)
-
-crregs32_str = ["CR%d" % i for i in range(8)]
-crregs32_expr = [ExprId(x, 32) for x in crregs32_str]
-crregs = reg_info(crregs32_str, crregs32_expr)
-
-
-drregs32_str = ["DR%d" % i for i in range(8)]
-drregs32_expr = [ExprId(x, 32) for x in drregs32_str]
-drregs = reg_info(drregs32_str, drregs32_expr)
-
-
-fltregs32_str = ["ST(%d)" % i for i in range(8)]
-fltregs32_expr = [ExprId(x, 64) for x in fltregs32_str]
-fltregs = reg_info(fltregs32_str, fltregs32_expr)
-
-r_st_all = reg_info(['ST'],
-                    [ExprId('ST', 64)])
-
-r_cs_all = reg_info(['CS'],
-                    [ExprId('CS', 16)])
-r_ds_all = reg_info(['DS'],
-                    [ExprId('DS', 16)])
-r_es_all = reg_info(['ES'],
-                    [ExprId('ES', 16)])
-r_ss_all = reg_info(['SS'],
-                    [ExprId('SS', 16)])
-r_fs_all = reg_info(['FS'],
-                    [ExprId('FS', 16)])
-r_gs_all = reg_info(['GS'],
-                    [ExprId('GS', 16)])
-
-
-AL = regs08_expr[0]
-CL = regs08_expr[1]
-DL = regs08_expr[2]
-BL = regs08_expr[3]
-AH = regs08_expr[4]
-CH = regs08_expr[5]
-DH = regs08_expr[6]
-BH = regs08_expr[7]
-R8B = regs08_expr[8]
-R9B = regs08_expr[9]
-R10B = regs08_expr[10]
-R11B = regs08_expr[11]
-R12B = regs08_expr[12]
-R13B = regs08_expr[13]
-R14B = regs08_expr[14]
-R15B = regs08_expr[15]
-
-SPL = regs08_64_expr[4]
-BPL = regs08_64_expr[5]
-SIL = regs08_64_expr[6]
-DIL = regs08_64_expr[7]
-
-
-AX = regs16_expr[0]
-CX = regs16_expr[1]
-DX = regs16_expr[2]
-BX = regs16_expr[3]
-SP = regs16_expr[4]
-BP = regs16_expr[5]
-SI = regs16_expr[6]
-DI = regs16_expr[7]
-R8W = regs16_expr[8]
-R9W = regs16_expr[9]
-R10W = regs16_expr[10]
-R11W = regs16_expr[11]
-R12W = regs16_expr[12]
-R13W = regs16_expr[13]
-R14W = regs16_expr[14]
-R15W = regs16_expr[15]
-
-
-EAX = regs32_expr[0]
-ECX = regs32_expr[1]
-EDX = regs32_expr[2]
-EBX = regs32_expr[3]
-ESP = regs32_expr[4]
-EBP = regs32_expr[5]
-ESI = regs32_expr[6]
-EDI = regs32_expr[7]
-R8D = regs32_expr[8]
-R9D = regs32_expr[9]
-R10D = regs32_expr[10]
-R11D = regs32_expr[11]
-R12D = regs32_expr[12]
-R13D = regs32_expr[13]
-R14D = regs32_expr[14]
-R15D = regs32_expr[15]
-
-
-RAX = regs64_expr[0]
-RCX = regs64_expr[1]
-RDX = regs64_expr[2]
-RBX = regs64_expr[3]
-RSP = regs64_expr[4]
-RBP = regs64_expr[5]
-RSI = regs64_expr[6]
-RDI = regs64_expr[7]
-R8 = regs64_expr[8]
-R9 = regs64_expr[9]
-R10 = regs64_expr[10]
-R11 = regs64_expr[11]
-R12 = regs64_expr[12]
-R13 = regs64_expr[13]
-R14 = regs64_expr[14]
-R15 = regs64_expr[15]
-
-
-reg_zf = 'zf'
-reg_nf = 'nf'
-reg_pf = 'pf'
-reg_of = 'of'
-reg_cf = 'cf'
-reg_tf = 'tf'
-reg_if = 'i_f'
-reg_df = 'df'
-reg_af = 'af'
-reg_iopl = 'iopl_f'
-reg_nt = 'nt'
-reg_rf = 'rf'
-reg_vm = 'vm'
-reg_ac = 'ac'
-reg_vif = 'vif'
-reg_vip = 'vip'
-reg_id = 'i_d'
-
-
-reg_es = "ES"
-reg_cs = "CS"
-reg_ss = "SS"
-reg_ds = "DS"
-reg_fs = "FS"
-reg_gs = "GS"
-
-reg_dr0 = 'DR0'
-reg_dr1 = 'DR1'
-reg_dr2 = 'DR2'
-reg_dr3 = 'DR3'
-reg_dr4 = 'DR4'
-reg_dr5 = 'DR5'
-reg_dr6 = 'DR6'
-reg_dr7 = 'DR7'
-
-reg_cr0 = 'CR0'
-reg_cr1 = 'CR1'
-reg_cr2 = 'CR2'
-reg_cr3 = 'CR3'
-reg_cr4 = 'CR4'
-reg_cr5 = 'CR5'
-reg_cr6 = 'CR6'
-reg_cr7 = 'CR7'
-
-reg_mm0 = 'MM0'
-reg_mm1 = 'MM1'
-reg_mm2 = 'MM2'
-reg_mm3 = 'MM3'
-reg_mm4 = 'MM4'
-reg_mm5 = 'MM5'
-reg_mm6 = 'MM6'
-reg_mm7 = 'MM7'
-
-reg_tsc = "tsc"
-
-reg_float_c0 = 'float_c0'
-reg_float_c1 = 'float_c1'
-reg_float_c2 = 'float_c2'
-reg_float_c3 = 'float_c3'
-reg_float_stack_ptr = "float_stack_ptr"
-reg_float_control = 'reg_float_control'
-reg_float_eip = 'reg_float_eip'
-reg_float_cs = 'reg_float_cs'
-reg_float_address = 'reg_float_address'
-reg_float_ds = 'reg_float_ds'
-
-
-dr0 = ExprId(reg_dr0, 32)
-dr1 = ExprId(reg_dr1, 32)
-dr2 = ExprId(reg_dr2, 32)
-dr3 = ExprId(reg_dr3, 32)
-dr4 = ExprId(reg_dr4, 32)
-dr5 = ExprId(reg_dr5, 32)
-dr6 = ExprId(reg_dr6, 32)
-dr7 = ExprId(reg_dr7, 32)
-
-cr0 = ExprId(reg_cr0, 32)
-cr1 = ExprId(reg_cr1, 32)
-cr2 = ExprId(reg_cr2, 32)
-cr3 = ExprId(reg_cr3, 32)
-cr4 = ExprId(reg_cr4, 32)
-cr5 = ExprId(reg_cr5, 32)
-cr6 = ExprId(reg_cr6, 32)
-cr7 = ExprId(reg_cr7, 32)
-
-mm0 = ExprId(reg_mm0, 64)
-mm1 = ExprId(reg_mm1, 64)
-mm2 = ExprId(reg_mm2, 64)
-mm3 = ExprId(reg_mm3, 64)
-mm4 = ExprId(reg_mm4, 64)
-mm5 = ExprId(reg_mm5, 64)
-mm6 = ExprId(reg_mm6, 64)
-mm7 = ExprId(reg_mm7, 64)
-
-XMM0 = regs_xmm_expr[0]
-XMM1 = regs_xmm_expr[1]
-XMM2 = regs_xmm_expr[2]
-XMM3 = regs_xmm_expr[3]
-XMM4 = regs_xmm_expr[4]
-XMM5 = regs_xmm_expr[5]
-XMM6 = regs_xmm_expr[6]
-XMM7 = regs_xmm_expr[7]
-XMM8 = regs_xmm_expr[8]
-XMM9 = regs_xmm_expr[9]
-XMM10 = regs_xmm_expr[10]
-XMM11 = regs_xmm_expr[11]
-XMM12 = regs_xmm_expr[12]
-XMM13 = regs_xmm_expr[13]
-XMM14 = regs_xmm_expr[14]
-XMM15 = regs_xmm_expr[15]
-
-# tmp1= ExprId(reg_tmp1)
-zf = ExprId(reg_zf, size=1)
-nf = ExprId(reg_nf, size=1)
-pf = ExprId(reg_pf, size=1)
-of = ExprId(reg_of, size=1)
-cf = ExprId(reg_cf, size=1)
-tf = ExprId(reg_tf, size=1)
-i_f = ExprId(reg_if, size=1)
-df = ExprId(reg_df, size=1)
-af = ExprId(reg_af, size=1)
-iopl = ExprId(reg_iopl, size=2)
-nt = ExprId(reg_nt, size=1)
-rf = ExprId(reg_rf, size=1)
-vm = ExprId(reg_vm, size=1)
-ac = ExprId(reg_ac, size=1)
-vif = ExprId(reg_vif, size=1)
-vip = ExprId(reg_vip, size=1)
-i_d = ExprId(reg_id, size=1)
-
-ES = ExprId(reg_es, size=16)
-CS = ExprId(reg_cs, size=16)
-SS = ExprId(reg_ss, size=16)
-DS = ExprId(reg_ds, size=16)
-FS = ExprId(reg_fs, size=16)
-GS = ExprId(reg_gs, size=16)
-
-tsc = ExprId(reg_tsc, size=64)
-
-float_c0 = ExprId(reg_float_c0, size=1)
-float_c1 = ExprId(reg_float_c1, size=1)
-float_c2 = ExprId(reg_float_c2, size=1)
-float_c3 = ExprId(reg_float_c3, size=1)
-float_stack_ptr = ExprId(reg_float_stack_ptr, size=3)
-float_control = ExprId(reg_float_control, 16)
-float_eip = ExprId(reg_float_eip, 32)
-float_cs = ExprId(reg_float_cs, size=16)
-float_address = ExprId(reg_float_address, 32)
-float_ds = ExprId(reg_float_ds, size=16)
-
-float_st0 = ExprId("float_st0", 64)
-float_st1 = ExprId("float_st1", 64)
-float_st2 = ExprId("float_st2", 64)
-float_st3 = ExprId("float_st3", 64)
-float_st4 = ExprId("float_st4", 64)
-float_st5 = ExprId("float_st5", 64)
-float_st6 = ExprId("float_st6", 64)
-float_st7 = ExprId("float_st7", 64)
-
-
-float_list = [float_st0, float_st1, float_st2, float_st3,
-              float_st4, float_st5, float_st6, float_st7]
-
-float_replace = {fltregs32_expr[i]: float_list[i] for i in range(8)}
-float_replace[r_st_all.expr[0]] = float_st0
-
-
-EAX_init = ExprId('EAX_init', 32)
-EBX_init = ExprId('EBX_init', 32)
-ECX_init = ExprId('ECX_init', 32)
-EDX_init = ExprId('EDX_init', 32)
-ESI_init = ExprId('ESI_init', 32)
-EDI_init = ExprId('EDI_init', 32)
-ESP_init = ExprId('ESP_init', 32)
-EBP_init = ExprId('EBP_init', 32)
-
-
-RAX_init = ExprId('RAX_init', 64)
-RBX_init = ExprId('RBX_init', 64)
-RCX_init = ExprId('RCX_init', 64)
-RDX_init = ExprId('RDX_init', 64)
-RSI_init = ExprId('RSI_init', 64)
-RDI_init = ExprId('RDI_init', 64)
-RSP_init = ExprId('RSP_init', 64)
-RBP_init = ExprId('RBP_init', 64)
-
-
-all_regs_ids = [
-    AL, CL, DL, BL, AH, CH, DH, BH,
-    R8B, R9B, R10B, R11B, R12B, R13B, R14B, R15B,
-    SPL, BPL, SIL, DIL,
-    AX, CX, DX, BX, SP, BP, SI, DI,
-    R8W, R9W, R10W, R11W, R12W, R13W, R14W, R15W,
-    IP,
-    EAX, ECX, EDX, EBX, ESP, EBP, ESI, EDI,
-    R8D, R9D, R10D, R11D, R12D, R13D, R14D, R15D,
-    EIP,
-
-    RAX, RBX, RCX, RDX, RSP, RBP, RIP, RSI, RDI,
-    R8, R9, R10, R11, R12, R13, R14, R15,
-    zf, nf, pf, of, cf, af, df,
-    tf, i_f, iopl, nt, rf, vm, ac, vif, vip, i_d,
-    float_control, float_eip, float_cs, float_address, float_ds,
-    tsc,
-    ES, CS, SS, DS, FS, GS,
-    float_st0, float_st1, float_st2, float_st3,
-    float_st4, float_st5, float_st6, float_st7,
-    float_c0, float_c1, float_c2, float_c3,
-    cr0, cr3,
-    dr0, dr1, dr2, dr3, dr4, dr5, dr6, dr7,
-    float_stack_ptr,
-    mm0, mm1, mm2, mm3, mm4, mm5, mm6, mm7,
-
-    XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7,
-    XMM8, XMM9, XMM10, XMM11, XMM12, XMM13, XMM14, XMM15,
-
-
-    exception_flags, interrupt_num,
-] + fltregs32_expr
-
-all_regs_ids_no_alias = [
-    RAX, RBX, RCX, RDX, RSP, RBP, RIP, RSI, RDI,
-    R8, R9, R10, R11, R12, R13, R14, R15,
-    zf, nf, pf, of, cf, af, df,
-    tf, i_f, iopl, nt, rf, vm, ac, vif, vip, i_d,
-    float_control, float_eip, float_cs, float_address, float_ds,
-    tsc,
-    ES, CS, SS, DS, FS, GS,
-    float_st0, float_st1, float_st2, float_st3,
-    float_st4, float_st5, float_st6, float_st7,
-    float_c0, float_c1, float_c2, float_c3,
-    cr0, cr3,
-    dr0, dr1, dr2, dr3, dr4, dr5, dr6, dr7,
-    float_stack_ptr,
-    mm0, mm1, mm2, mm3, mm4, mm5, mm6, mm7,
-    XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7,
-    XMM8, XMM9, XMM10, XMM11, XMM12, XMM13, XMM14, XMM15,
-
-
-    exception_flags, interrupt_num,
-] + fltregs32_expr
-
-attrib_to_regs = {
-    16: regs16_expr + all_regs_ids_no_alias[all_regs_ids_no_alias.index(zf):] + [IP],
-    32: regs32_expr + all_regs_ids_no_alias[all_regs_ids_no_alias.index(zf):] + [EIP],
-    64: all_regs_ids_no_alias,
-}
-
-all_regs_ids_byname = dict([(x.name, x) for x in all_regs_ids])
-
-all_regs_ids_init = [ExprId("%s_init" % x.name, x.size) for x in all_regs_ids]
-
-regs_init = {}
-for i, r in enumerate(all_regs_ids):
-    regs_init[r] = all_regs_ids_init[i]
-
-regs_flt_expr = [float_st0, float_st1, float_st2, float_st3,
-                 float_st4, float_st5, float_st6, float_st7,
-                 ]
-
-mRAX = {16: AX, 32: EAX, 64: RAX}
-mRBX = {16: BX, 32: EBX, 64: RBX}
-mRCX = {16: CX, 32: ECX, 64: RCX}
-mRDX = {16: DX, 32: EDX, 64: RDX}
-mRSI = {16: SI, 32: ESI, 64: RSI}
-mRDI = {16: DI, 32: EDI, 64: RDI}
-mRBP = {16: BP, 32: EBP, 64: RBP}
-mRSP = {16: SP, 32: ESP, 64: RSP}
-mRIP = {16: IP, 32: EIP, 64: RIP}
diff --git a/miasm/arch/x86/sem.py b/miasm/arch/x86/sem.py
deleted file mode 100644
index d19290b6..00000000
--- a/miasm/arch/x86/sem.py
+++ /dev/null
@@ -1,6065 +0,0 @@
-#
-# Copyright (C) 2011 EADS France, Fabrice Desclaux <fabrice.desclaux@eads.net>
-#
-# This program is free software; you can redistribute it and/or modify
-# it under the terms of the GNU General Public License as published by
-# the Free Software Foundation; either version 2 of the License, or
-# (at your option) any later version.
-#
-# This program is distributed in the hope that it will be useful,
-# but WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-# GNU General Public License for more details.
-#
-# You should have received a copy of the GNU General Public License along
-# with this program; if not, write to the Free Software Foundation, Inc.,
-# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-#
-
-from builtins import range
-
-from future.utils import viewitems
-
-import logging
-import miasm.expression.expression as m2_expr
-from miasm.expression.simplifications import expr_simp
-from miasm.arch.x86.regs import *
-from miasm.arch.x86.arch import mn_x86, repeat_mn, replace_regs, is_mem_segm
-from miasm.ir.ir import Lifter, IRBlock, AssignBlock
-from miasm.core.sembuilder import SemBuilder
-from miasm.jitter.csts import EXCEPT_DIV_BY_ZERO, EXCEPT_ILLEGAL_INSN, \
-    EXCEPT_PRIV_INSN, EXCEPT_SOFT_BP, EXCEPT_INT_XX, EXCEPT_INT_1, \
-    EXCEPT_SYSCALL
-import math
-import struct
-
-
-LOG_X86_SEM = logging.getLogger("x86_sem")
-CONSOLE_HANDLER = logging.StreamHandler()
-CONSOLE_HANDLER.setFormatter(logging.Formatter("[%(levelname)-8s]: %(message)s"))
-LOG_X86_SEM.addHandler(CONSOLE_HANDLER)
-LOG_X86_SEM.setLevel(logging.WARNING)
-
-
-# SemBuilder context
-ctx = {'mRAX': mRAX,
-       'mRBX': mRBX,
-       'mRCX': mRCX,
-       'mRDX': mRDX,
-       'zf': zf,
-       }
-sbuild = SemBuilder(ctx)
-
-
-
-"""
-http://www.emulators.com/docs/nx11_flags.htm
-
-CF(A+B) = (((A XOR B) XOR D) < 0) XOR (((A XOR D) AND NOT (A XOR B)) < 0)
-CF(A-B) = (((A XOR B) XOR D) < 0) XOR (((A XOR D) AND (A XOR B)) < 0)
-
-OF(A+B) = ((A XOR D) AND NOT (A XOR B)) < 0
-OF(A-B) = ((A XOR D) AND (A XOR B)) < 0
-"""
-
-
-# XXX TODO make default check against 0 or not 0 (same eq as in C)
-def update_flag_zf_eq(a, b):
-    return [m2_expr.ExprAssign(zf, m2_expr.ExprOp("FLAG_EQ_CMP", a, b))]
-
-
-def update_flag_zf(a):
-    return [
-        m2_expr.ExprAssign(
-            zf,
-            m2_expr.ExprCond(
-                a,
-                m2_expr.ExprInt(0, zf.size),
-                m2_expr.ExprInt(1, zf.size)
-            )
-        )
-    ]
-
-
-def update_flag_nf(arg):
-    return [
-        m2_expr.ExprAssign(
-            nf,
-            m2_expr.ExprOp("FLAG_SIGN_SUB", arg, m2_expr.ExprInt(0, arg.size))
-        )
-    ]
-
-
-def update_flag_pf(a):
-    return [m2_expr.ExprAssign(pf,
-                            m2_expr.ExprOp('parity',
-                                           a & m2_expr.ExprInt(0xFF, a.size)))]
-
-
-def update_flag_af(op1, op2, res):
-    return [m2_expr.ExprAssign(af, (op1 ^ op2 ^ res)[4:5])]
-
-
-def update_flag_znp(a):
-    e = []
-    e += update_flag_zf(a)
-    e += update_flag_nf(a)
-    e += update_flag_pf(a)
-    return e
-
-
-def update_flag_np(result):
-    e = []
-    e += update_flag_nf(result)
-    e += update_flag_pf(result)
-    return e
-
-
-def null_flag_co():
-    e = []
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprInt(0, of.size)))
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprInt(0, cf.size)))
-    return e
-
-
-def update_flag_arith(a):
-    e = []
-    e += update_flag_znp(a)
-    return e
-
-
-def update_flag_zfaddwc_eq(arg1, arg2, arg3):
-    return [m2_expr.ExprAssign(zf, m2_expr.ExprOp("FLAG_EQ_ADDWC", arg1, arg2, arg3))]
-
-def update_flag_zfsubwc_eq(arg1, arg2, arg3):
-    return [m2_expr.ExprAssign(zf, m2_expr.ExprOp("FLAG_EQ_SUBWC", arg1, arg2, arg3))]
-
-
-def update_flag_arith_add_znp(arg1, arg2):
-    """
-    Compute znp flags for (arg1 + arg2)
-    """
-    e = []
-    e += update_flag_zf_eq(arg1, -arg2)
-    e += [m2_expr.ExprAssign(nf, m2_expr.ExprOp("FLAG_SIGN_SUB", arg1, -arg2))]
-    e += update_flag_pf(arg1+arg2)
-    return e
-
-
-def update_flag_arith_addwc_znp(arg1, arg2, arg3):
-    """
-    Compute znp flags for (arg1 + arg2 + cf)
-    """
-    e = []
-    e += update_flag_zfaddwc_eq(arg1, arg2, arg3)
-    e += [m2_expr.ExprAssign(nf, m2_expr.ExprOp("FLAG_SIGN_ADDWC", arg1, arg2, arg3))]
-    e += update_flag_pf(arg1+arg2+arg3.zeroExtend(arg2.size))
-    return e
-
-
-
-
-def update_flag_arith_sub_znp(arg1, arg2):
-    """
-    Compute znp flags for (arg1 - arg2)
-    """
-    e = []
-    e += update_flag_zf_eq(arg1, arg2)
-    e += [m2_expr.ExprAssign(nf, m2_expr.ExprOp("FLAG_SIGN_SUB", arg1, arg2))]
-    e += update_flag_pf(arg1 - arg2)
-    return e
-
-
-def update_flag_arith_subwc_znp(arg1, arg2, arg3):
-    """
-    Compute znp flags for (arg1 - (arg2 + cf))
-    """
-    e = []
-    e += update_flag_zfsubwc_eq(arg1, arg2, arg3)
-    e += [m2_expr.ExprAssign(nf, m2_expr.ExprOp("FLAG_SIGN_SUBWC", arg1, arg2, arg3))]
-    e += update_flag_pf(arg1 - (arg2+arg3.zeroExtend(arg2.size)))
-    return e
-
-
-def check_ops_msb(a, b, c):
-    if not a or not b or not c or a != b or a != c:
-        raise ValueError('bad ops size %s %s %s' % (a, b, c))
-
-
-def arith_flag(a, b, c):
-    a_s, b_s, c_s = a.size, b.size, c.size
-    check_ops_msb(a_s, b_s, c_s)
-    a_s, b_s, c_s = a.msb(), b.msb(), c.msb()
-    return a_s, b_s, c_s
-
-# checked: ok for adc add because b & c before +cf
-
-
-def update_flag_add_cf(op1, op2, res):
-    "Compute cf in @res = @op1 + @op2"
-    #return [m2_expr.ExprAssign(cf, m2_expr.ExprOp("FLAG_SUB_CF", op1, -op2))]
-    return [m2_expr.ExprAssign(cf, m2_expr.ExprOp("FLAG_ADD_CF", op1, op2))]
-
-
-def update_flag_add_of(op1, op2, res):
-    "Compute of in @res = @op1 + @op2"
-    return [m2_expr.ExprAssign(of, m2_expr.ExprOp("FLAG_ADD_OF", op1, op2))]
-
-
-# checked: ok for sbb add because b & c before +cf
-def update_flag_sub_cf(op1, op2, res):
-    "Compote CF in @res = @op1 - @op2"
-    return [m2_expr.ExprAssign(cf, m2_expr.ExprOp("FLAG_SUB_CF", op1, op2))]
-
-
-def update_flag_sub_of(op1, op2, res):
-    "Compote OF in @res = @op1 - @op2"
-    return [m2_expr.ExprAssign(of, m2_expr.ExprOp("FLAG_SUB_OF", op1, op2))]
-
-
-def update_flag_addwc_cf(op1, op2, op3):
-    "Compute cf in @res = @op1 + @op2 + @op3"
-    return [m2_expr.ExprAssign(cf, m2_expr.ExprOp("FLAG_ADDWC_CF", op1, op2, op3))]
-
-
-def update_flag_addwc_of(op1, op2, op3):
-    "Compute of in @res = @op1 + @op2 + @op3"
-    return [m2_expr.ExprAssign(of, m2_expr.ExprOp("FLAG_ADDWC_OF", op1, op2, op3))]
-
-
-
-def update_flag_subwc_cf(op1, op2, op3):
-    "Compute cf in @res = @op1 + @op2 + @op3"
-    return [m2_expr.ExprAssign(cf, m2_expr.ExprOp("FLAG_SUBWC_CF", op1, op2, op3))]
-
-
-def update_flag_subwc_of(op1, op2, op3):
-    "Compute of in @res = @op1 + @op2 + @op3"
-    return [m2_expr.ExprAssign(of, m2_expr.ExprOp("FLAG_SUBWC_OF", op1, op2, op3))]
-
-
-
-
-def update_flag_arith_add_co(x, y, z):
-    e = []
-    e += update_flag_add_cf(x, y, z)
-    e += update_flag_add_of(x, y, z)
-    return e
-
-
-def update_flag_arith_sub_co(x, y, z):
-    e = []
-    e += update_flag_sub_cf(x, y, z)
-    e += update_flag_sub_of(x, y, z)
-    return e
-
-
-
-
-def update_flag_arith_addwc_co(arg1, arg2, arg3):
-    e = []
-    e += update_flag_addwc_cf(arg1, arg2, arg3)
-    e += update_flag_addwc_of(arg1, arg2, arg3)
-    return e
-
-
-def update_flag_arith_subwc_co(arg1, arg2, arg3):
-    e = []
-    e += update_flag_subwc_cf(arg1, arg2, arg3)
-    e += update_flag_subwc_of(arg1, arg2, arg3)
-    return e
-
-
-
-def set_float_cs_eip(instr):
-    e = []
-    # XXX TODO check float updt
-    e.append(m2_expr.ExprAssign(float_eip,
-                             m2_expr.ExprInt(instr.offset, float_eip.size)))
-    e.append(m2_expr.ExprAssign(float_cs, CS))
-    return e
-
-
-def mode2addrsize(mode):
-    """Returns the address size for a given @mode"""
-
-    mode2size = {16:32, 32:32, 64:64}
-    if mode not in mode2size:
-        raise RuntimeError("Unknown size %s", mode)
-    return mode2size[mode]
-
-
-def instr2addrsize(instr):
-    """Returns the address size for a given @instr"""
-
-    return mode2addrsize(instr.mode)
-
-
-def expraddr(mode, ptr):
-    """Returns memory address pointer with size according to current @mode"""
-    return ptr.zeroExtend(mode2addrsize(mode))
-
-
-def fix_mem_args_size(instr, *args):
-    out = []
-    for arg in args:
-        if not arg.is_mem():
-            out.append(arg)
-            continue
-        ptr = arg.ptr
-        size = arg.size
-        if ptr.is_op('segm'):
-            ptr = m2_expr.ExprOp(
-                'segm', ptr.args[0], expraddr(instr.mode, ptr.args[1]))
-        else:
-            ptr = expraddr(instr.mode, ptr)
-        out.append(m2_expr.ExprMem(ptr, size))
-    return out
-
-
-def mem2double(instr, arg):
-    """
-    Add float conversion if argument is an ExprMem
-    @arg: argument to transform
-    """
-    if isinstance(arg, m2_expr.ExprMem):
-        if arg.size > 64:
-            # TODO: move to 80 bits
-            arg = m2_expr.ExprMem(expraddr(instr.mode, arg.ptr), size=64)
-        return m2_expr.ExprOp('sint_to_fp', arg.signExtend(64))
-    else:
-        return arg
-
-
-def float_implicit_st0(arg1, arg2):
-    """
-    Generate full float operators if one argument is implicit (float_st0)
-    """
-    if arg2 is None:
-        arg2 = arg1
-        arg1 = float_st0
-    return arg1, arg2
-
-
-def gen_jcc(ir, instr, cond, dst, jmp_if):
-    """
-    Macro to generate jcc semantic
-    @ir: ir instance
-    @instr: instruction
-    @cond: condition of the jcc
-    @dst: the destination if jcc is taken
-    @jmp_if: jump if/notif cond
-    """
-
-    e = []
-    meip = mRIP[ir.IRDst.size]
-    loc_next = ir.get_next_loc_key(instr)
-    loc_next_expr = m2_expr.ExprLoc(loc_next, dst.size)
-
-    if jmp_if:
-        dstA, dstB = dst, loc_next_expr
-    else:
-        dstA, dstB = loc_next_expr, dst
-    mn_dst = m2_expr.ExprCond(cond,
-                              dstA.zeroExtend(ir.IRDst.size),
-                              dstB.zeroExtend(ir.IRDst.size))
-    e.append(m2_expr.ExprAssign(meip, mn_dst))
-    e.append(m2_expr.ExprAssign(ir.IRDst, mn_dst))
-    return e, []
-
-
-def gen_fcmov(ir, instr, cond, arg1, arg2, mov_if):
-    """Generate fcmov
-    @ir: ir instance
-    @instr: instruction instance
-    @cond: condition
-    @mov_if: invert condition if False"""
-
-    loc_do, loc_do_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_skip = ir.get_next_loc_key(instr)
-    loc_skip_expr = m2_expr.ExprLoc(loc_skip, ir.IRDst.size)
-    if mov_if:
-        dstA, dstB = loc_do_expr, loc_skip_expr
-    else:
-        dstA, dstB = loc_skip_expr, loc_do_expr
-    e = []
-    e_do, extra_irs = [m2_expr.ExprAssign(arg1, arg2)], []
-    e_do.append(m2_expr.ExprAssign(ir.IRDst, loc_skip_expr))
-    e.append(m2_expr.ExprAssign(ir.IRDst, m2_expr.ExprCond(cond, dstA, dstB)))
-    return e, [IRBlock(ir.loc_db, loc_do, [AssignBlock(e_do, instr)])]
-
-
-def gen_cmov(ir, instr, cond, dst, src, mov_if):
-    """Generate cmov
-    @ir: ir instance
-    @instr: instruction instance
-    @cond: condition
-    @mov_if: invert condition if False"""
-
-    loc_do, loc_do_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_skip = ir.get_next_loc_key(instr)
-    loc_skip_expr = m2_expr.ExprLoc(loc_skip, ir.IRDst.size)
-    if mov_if:
-        dstA, dstB = loc_do_expr, loc_skip_expr
-    else:
-        dstA, dstB = loc_skip_expr, loc_do_expr
-    e = []
-    if instr.mode == 64:
-        # Force destination set in order to zero high bit orders
-        # In 64 bit:
-        # cmovz eax, ebx
-        # if zf == 0 => high part of RAX is set to zero
-        e.append(m2_expr.ExprAssign(dst, dst))
-    e_do, extra_irs = mov(ir, instr, dst, src)
-    e_do.append(m2_expr.ExprAssign(ir.IRDst, loc_skip_expr))
-    e.append(m2_expr.ExprAssign(ir.IRDst, m2_expr.ExprCond(cond, dstA, dstB)))
-    return e, [IRBlock(ir.loc_db, loc_do, [AssignBlock(e_do, instr)])]
-
-
-def mov(_, instr, dst, src):
-    if dst in [ES, CS, SS, DS, FS, GS]:
-        src = src[:dst.size]
-    if src in [ES, CS, SS, DS, FS, GS]:
-        src = src.zeroExtend(dst.size)
-    e = [m2_expr.ExprAssign(dst, src)]
-    return e, []
-
-
-def movq(_, instr, dst, src):
-    src_final = (src.zeroExtend(dst.size)
-                 if dst.size >= src.size else
-                 src[:dst.size])
-    return [m2_expr.ExprAssign(dst, src_final)], []
-
-
-@sbuild.parse
-def xchg(arg1, arg2):
-    arg1 = arg2
-    arg2 = arg1
-
-
-
-def movzx(_, instr, dst, src):
-    e = [m2_expr.ExprAssign(dst, src.zeroExtend(dst.size))]
-    return e, []
-
-
-def movsx(_, instr, dst, src):
-    e = [m2_expr.ExprAssign(dst, src.signExtend(dst.size))]
-    return e, []
-
-
-def lea(_, instr, dst, src):
-    ptr = src.ptr
-    if is_mem_segm(src):
-        # Do not use segmentation here
-        ptr = ptr.args[1]
-
-    if ptr.size > dst.size:
-        ptr = ptr[:dst.size]
-    e = [m2_expr.ExprAssign(dst, ptr.zeroExtend(dst.size))]
-    return e, []
-
-
-def add(_, instr, dst, src):
-    e = []
-
-    result = dst + src
-
-    e += update_flag_arith_add_znp(dst, src)
-    e += update_flag_arith_add_co(dst, src, result)
-    e += update_flag_af(dst, src, result)
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-
-def xadd(_, instr, dst, src):
-    e = []
-
-    result = dst + src
-    e += update_flag_arith_add_znp(dst, src)
-    e += update_flag_arith_add_co(src, dst, result)
-    e += update_flag_af(dst, src, result)
-    if dst != src:
-        e.append(m2_expr.ExprAssign(src, dst))
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-
-def adc(_, instr, dst, src):
-    e = []
-
-    arg1 = dst
-    arg2 = src
-    result = arg1 + (arg2 + cf.zeroExtend(src.size))
-
-    e += update_flag_arith_addwc_znp(arg1, arg2, cf)
-    e += update_flag_arith_addwc_co(arg1, arg2, cf)
-    e += update_flag_af(arg1, arg2, result)
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-
-def sub(_, instr, dst, src):
-    e = []
-    arg1, arg2 = dst, src
-    result = dst - src
-
-    e += update_flag_arith_sub_znp(arg1, arg2)
-    e += update_flag_arith_sub_co(arg1, arg2, result)
-    e += update_flag_af(dst, src, result)
-
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-# a-(b+cf)
-
-
-def sbb(_, instr, dst, src):
-    e = []
-    arg1 = dst
-    arg2 = src
-    result = arg1 - (arg2 + cf.zeroExtend(src.size))
-
-    e += update_flag_arith_subwc_znp(arg1, arg2, cf)
-    e += update_flag_af(arg1, arg2, result)
-    e += update_flag_arith_subwc_co(arg1, arg2, cf)
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-
-def neg(_, instr, src):
-    e = []
-    dst = m2_expr.ExprInt(0, src.size)
-    arg1, arg2 = dst, src
-    result = arg1 - arg2
-
-    e += update_flag_arith_sub_znp(arg1, arg2)
-    e += update_flag_arith_sub_co(arg1, arg2, result)
-    e += update_flag_af(arg1, arg2, result)
-    e.append(m2_expr.ExprAssign(src, result))
-    return (e, [])
-
-
-def l_not(_, instr, dst):
-    e = []
-    result = (~dst)
-    e.append(m2_expr.ExprAssign(dst, result))
-    return (e, [])
-
-
-def l_cmp(_, instr, dst, src):
-    e = []
-    arg1, arg2 = dst, src
-    result = dst - src
-
-    e += update_flag_arith_sub_znp(arg1, arg2)
-    e += update_flag_arith_sub_co(arg1, arg2, result)
-    e += update_flag_af(dst, src, result)
-    return (e, [])
-
-
-def xor(_, instr, dst, src):
-    e = []
-    result = dst ^ src
-    e += [m2_expr.ExprAssign(zf, m2_expr.ExprOp('FLAG_EQ_CMP', dst, src))]
-    e += update_flag_np(result)
-    e += null_flag_co()
-    e.append(m2_expr.ExprAssign(dst, result))
-    return (e, [])
-
-
-def pxor(_, instr, dst, src):
-    e = []
-    result = dst ^ src
-    e.append(m2_expr.ExprAssign(dst, result))
-    return (e, [])
-
-
-def l_or(_, instr, dst, src):
-    e = []
-    result = dst | src
-    e += [m2_expr.ExprAssign(zf, m2_expr.ExprOp('FLAG_EQ', dst | src))]
-    e += update_flag_np(result)
-    e += null_flag_co()
-    e.append(m2_expr.ExprAssign(dst, result))
-    return (e, [])
-
-
-def l_and(_, instr, dst, src):
-    e = []
-    result = dst & src
-    e += [m2_expr.ExprAssign(zf, m2_expr.ExprOp('FLAG_EQ_AND', dst, src))]
-    e += update_flag_np(result)
-    e += null_flag_co()
-
-    e.append(m2_expr.ExprAssign(dst, result))
-    return (e, [])
-
-
-def l_test(_, instr, dst, src):
-    e = []
-    result = dst & src
-
-    e += [m2_expr.ExprAssign(zf, m2_expr.ExprOp('FLAG_EQ_CMP', result, m2_expr.ExprInt(0, result.size)))]
-    e += [m2_expr.ExprAssign(nf, m2_expr.ExprOp("FLAG_SIGN_SUB", result, m2_expr.ExprInt(0, result.size)))]
-    e += update_flag_pf(result)
-    e += null_flag_co()
-
-    return (e, [])
-
-
-def get_shift(dst, src):
-    if isinstance(src, m2_expr.ExprInt):
-        src = m2_expr.ExprInt(int(src), dst.size)
-    else:
-        src = src.zeroExtend(dst.size)
-    if dst.size == 64:
-        shift = src & m2_expr.ExprInt(63, src.size)
-    else:
-        shift = src & m2_expr.ExprInt(31, src.size)
-    shift = expr_simp(shift)
-    return shift
-
-
-def _rotate_tpl(ir, instr, dst, src, op, left=False):
-    '''Template to generate a rotater with operation @op
-    A temporary basic block is generated to handle 0-rotate
-    @op: operation to execute
-    @left (optional): indicates a left rotate if set, default is False
-    '''
-    # Compute results
-    shifter = get_shift(dst, src)
-    res = m2_expr.ExprOp(op, dst, shifter)
-
-    # CF is computed with 1-less round than `res`
-    new_cf = m2_expr.ExprOp(
-        op, dst, shifter - m2_expr.ExprInt(1, size=shifter.size))
-    new_cf = new_cf.msb() if left else new_cf[:1]
-
-    # OF is defined only for @b == 1
-    new_of = m2_expr.ExprCond(src - m2_expr.ExprInt(1, size=src.size),
-                              m2_expr.ExprInt(0, size=of.size),
-                              res.msb() ^ new_cf if left else (dst ^ res).msb())
-
-    # Build basic blocks
-    e_do = [m2_expr.ExprAssign(cf, new_cf),
-            m2_expr.ExprAssign(of, new_of),
-            m2_expr.ExprAssign(dst, res)
-            ]
-    e = []
-    if instr.mode == 64:
-        # Force destination set in order to zero high bit orders
-        # In 64 bit:
-        # rol eax, cl
-        # if cl == 0 => high part of RAX is set to zero
-        e.append(m2_expr.ExprAssign(dst, dst))
-    # Don't generate conditional shifter on constant
-    if isinstance(shifter, m2_expr.ExprInt):
-        if int(shifter) != 0:
-            return (e_do, [])
-        else:
-            return (e, [])
-    loc_do, loc_do_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_skip = ir.get_next_loc_key(instr)
-    loc_skip_expr = m2_expr.ExprLoc(loc_skip, ir.IRDst.size)
-    e_do.append(m2_expr.ExprAssign(ir.IRDst, loc_skip_expr))
-    e.append(m2_expr.ExprAssign(
-        ir.IRDst, m2_expr.ExprCond(shifter, loc_do_expr, loc_skip_expr)))
-    return (e, [IRBlock(ir.loc_db, loc_do, [AssignBlock(e_do, instr)])])
-
-
-def l_rol(ir, instr, dst, src):
-    return _rotate_tpl(ir, instr, dst, src, '<<<', left=True)
-
-
-def l_ror(ir, instr, dst, src):
-    return _rotate_tpl(ir, instr, dst, src, '>>>')
-
-
-def rotate_with_carry_tpl(ir, instr, op, dst, src):
-    # Compute results
-    shifter = get_shift(dst, src).zeroExtend(dst.size + 1)
-    result = m2_expr.ExprOp(op, m2_expr.ExprCompose(dst, cf), shifter)
-
-    new_cf = result[dst.size:dst.size +1]
-    new_dst = result[:dst.size]
-
-    result_trunc = result[:dst.size]
-    if op == '<<<':
-        of_value = result_trunc.msb() ^ new_cf
-    else:
-        of_value = (dst ^ result_trunc).msb()
-    # OF is defined only for @b == 1
-    new_of = m2_expr.ExprCond(src - m2_expr.ExprInt(1, size=src.size),
-                              m2_expr.ExprInt(0, size=of.size),
-                              of_value)
-
-
-    # Build basic blocks
-    e_do = [m2_expr.ExprAssign(cf, new_cf),
-            m2_expr.ExprAssign(of, new_of),
-            m2_expr.ExprAssign(dst, new_dst)
-            ]
-    e = [m2_expr.ExprAssign(dst, dst)]
-    # Don't generate conditional shifter on constant
-    if isinstance(shifter, m2_expr.ExprInt):
-        if int(shifter) != 0:
-            return (e_do, [])
-        else:
-            return (e, [])
-    loc_do, loc_do_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_skip = ir.get_next_loc_key(instr)
-    loc_skip_expr = m2_expr.ExprLoc(loc_skip, ir.IRDst.size)
-    e_do.append(m2_expr.ExprAssign(ir.IRDst, loc_skip_expr))
-    e.append(m2_expr.ExprAssign(
-        ir.IRDst, m2_expr.ExprCond(shifter, loc_do_expr, loc_skip_expr)))
-    return (e, [IRBlock(ir.loc_db, loc_do, [AssignBlock(e_do, instr)])])
-
-def rcl(ir, instr, dst, src):
-    return rotate_with_carry_tpl(ir, instr, '<<<', dst, src)
-
-def rcr(ir, instr, dst, src):
-    return rotate_with_carry_tpl(ir, instr, '>>>', dst, src)
-
-
-def _shift_tpl(op, ir, instr, a, b, c=None, op_inv=None, left=False,
-               custom_of=None):
-    """Template to generate a shifter with operation @op
-    A temporary basic block is generated to handle 0-shift
-    @op: operation to execute
-    @c (optional): if set, instruction has a bit provider
-    @op_inv (optional): opposite operation of @op. Must be provided if @c
-    @left (optional): indicates a left shift if set, default is False
-    @custom_of (optional): if set, override the computed value of OF
-    """
-    if c is not None:
-        shifter = get_shift(a, c)
-    else:
-        shifter = get_shift(a, b)
-
-    res = m2_expr.ExprOp(op, a, shifter)
-    cf_from_dst = m2_expr.ExprOp(op, a,
-                                 (shifter - m2_expr.ExprInt(1, a.size)))
-    cf_from_dst = cf_from_dst.msb() if left else cf_from_dst[:1]
-
-    new_cf = cf_from_dst
-    i1 = m2_expr.ExprInt(1, size=a.size)
-    if c is not None:
-        # There is a source for new bits
-        isize = m2_expr.ExprInt(a.size, size=a.size)
-        mask = m2_expr.ExprOp(op_inv, i1, (isize - shifter)) - i1
-
-        # An overflow can occurred, emulate the 'undefined behavior'
-        # Overflow behavior if (shift / size % 2)
-        base_cond_overflow = shifter if left else (
-            shifter - m2_expr.ExprInt(1, size=shifter.size))
-        cond_overflow = base_cond_overflow & m2_expr.ExprInt(a.size, shifter.size)
-        if left:
-            # Overflow occurs one round before right
-            mask = m2_expr.ExprCond(cond_overflow, mask, ~mask)
-        else:
-            mask = m2_expr.ExprCond(cond_overflow, ~mask, mask)
-
-        # Build res with dst and src
-        res = ((m2_expr.ExprOp(op, a, shifter) & mask) |
-               (m2_expr.ExprOp(op_inv, b, (isize - shifter)) & ~mask))
-
-        # Overflow case: cf come from src (bit number shifter % size)
-        cf_from_src = m2_expr.ExprOp(op, b,
-                                     (shifter.zeroExtend(b.size) &
-                                      m2_expr.ExprInt(a.size - 1, b.size)) - i1)
-        cf_from_src = cf_from_src.msb() if left else cf_from_src[:1]
-        new_cf = m2_expr.ExprCond(cond_overflow, cf_from_src, cf_from_dst)
-
-    # Overflow flag, only occurred when shifter is equal to 1
-    if custom_of is None:
-        value_of = a.msb() ^ a[-2:-1] if left else b[:1] ^ a.msb()
-    else:
-        value_of = custom_of
-
-    # Build basic blocks
-    e_do = [
-        m2_expr.ExprAssign(cf, new_cf),
-        m2_expr.ExprAssign(of, m2_expr.ExprCond(shifter - i1,
-                                             m2_expr.ExprInt(0, of.size),
-                                             value_of)),
-        m2_expr.ExprAssign(a, res),
-    ]
-    e_do += update_flag_znp(res)
-    e = []
-    if instr.mode == 64:
-        # Force destination set in order to zero high bit orders
-        # In 64 bit:
-        # shr eax, cl
-        # if cl == 0 => high part of RAX is set to zero
-        e.append(m2_expr.ExprAssign(a, a))
-    # Don't generate conditional shifter on constant
-    if isinstance(shifter, m2_expr.ExprInt):
-        if int(shifter) != 0:
-            return (e_do, [])
-        else:
-            return (e, [])
-    loc_do, loc_do_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_skip = ir.get_next_loc_key(instr)
-    loc_skip_expr = m2_expr.ExprLoc(loc_skip, ir.IRDst.size)
-    e_do.append(m2_expr.ExprAssign(ir.IRDst, loc_skip_expr))
-    e.append(m2_expr.ExprAssign(ir.IRDst, m2_expr.ExprCond(shifter, loc_do_expr,
-                                                        loc_skip_expr)))
-    return e, [IRBlock(ir.loc_db, loc_do, [AssignBlock(e_do, instr)])]
-
-
-def sar(ir, instr, dst, src):
-    # Fixup OF, always cleared if src != 0
-    i0 = m2_expr.ExprInt(0, size=of.size)
-    return _shift_tpl("a>>", ir, instr, dst, src, custom_of=i0)
-
-
-def shr(ir, instr, dst, src):
-    return _shift_tpl(">>", ir, instr, dst, src, custom_of=dst.msb())
-
-
-def shrd(ir, instr, dst, src1, src2):
-    return _shift_tpl(">>>", ir, instr, dst, src1, src2, "<<<")
-
-
-def shl(ir, instr, dst, src):
-    return _shift_tpl("<<", ir, instr, dst, src, left=True)
-
-
-def shld(ir, instr, dst, src1, src2):
-    return _shift_tpl("<<<", ir, instr, dst, src1, src2, ">>>", left=True)
-
-
-# XXX todo ###
-def cmc(_, instr):
-    e = [m2_expr.ExprAssign(cf, m2_expr.ExprCond(cf, m2_expr.ExprInt(0, cf.size),
-                                              m2_expr.ExprInt(1, cf.size)))]
-    return e, []
-
-
-def clc(_, instr):
-    e = [m2_expr.ExprAssign(cf, m2_expr.ExprInt(0, cf.size))]
-    return e, []
-
-
-def stc(_, instr):
-    e = [m2_expr.ExprAssign(cf, m2_expr.ExprInt(1, cf.size))]
-    return e, []
-
-
-def cld(_, instr):
-    e = [m2_expr.ExprAssign(df, m2_expr.ExprInt(0, df.size))]
-    return e, []
-
-
-def std(_, instr):
-    e = [m2_expr.ExprAssign(df, m2_expr.ExprInt(1, df.size))]
-    return e, []
-
-
-def cli(_, instr):
-    e = [m2_expr.ExprAssign(i_f, m2_expr.ExprInt(0, i_f.size))]
-    return e, []
-
-
-def sti(_, instr):
-    e = [m2_expr.ExprAssign(exception_flags, m2_expr.ExprInt(EXCEPT_PRIV_INSN, 32))]
-    return e, []
-
-
-def inc(_, instr, dst):
-    e = []
-    src = m2_expr.ExprInt(1, dst.size)
-    arg1, arg2 = dst, src
-    result = dst + src
-
-    e += update_flag_arith_add_znp(arg1, arg2)
-    e += update_flag_af(arg1, arg2, result)
-    e += update_flag_add_of(arg1, arg2, result)
-
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-
-def dec(_, instr, dst):
-    e = []
-    src = m2_expr.ExprInt(1, dst.size)
-    arg1, arg2 = dst, src
-    result = dst - src
-
-    e += update_flag_arith_sub_znp(arg1, arg2)
-    e += update_flag_af(arg1, arg2, result)
-    e += update_flag_sub_of(arg1, arg2, result)
-
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-
-def push_gen(ir, instr, src, size):
-    e = []
-    if not size in [16, 32, 64]:
-        raise ValueError('bad size stacker!')
-    if src.size < size:
-        src = src.zeroExtend(size)
-    off_size = src.size
-
-    sp = mRSP[instr.mode]
-    new_sp = sp - m2_expr.ExprInt(off_size // 8, sp.size)
-    e.append(m2_expr.ExprAssign(sp, new_sp))
-    if ir.do_stk_segm:
-        new_sp = ir.gen_segm_expr(SS, new_sp)
-    e.append(m2_expr.ExprAssign(ir.ExprMem(new_sp, off_size),
-                             src))
-    return e, []
-
-
-def push(ir, instr, src):
-    return push_gen(ir, instr, src, instr.mode)
-
-
-def pushw(ir, instr, src):
-    return push_gen(ir, instr, src, 16)
-
-
-def pop_gen(ir, instr, src, size):
-    e = []
-    if not size in [16, 32, 64]:
-        raise ValueError('bad size stacker!')
-
-    sp = mRSP[instr.mode]
-    new_sp = sp + m2_expr.ExprInt(src.size // 8, sp.size)
-    # Don't generate SP/ESP/RSP incrementation on POP SP/ESP/RSP
-    if not (src in mRSP.values()):
-        e.append(m2_expr.ExprAssign(sp, new_sp))
-    # XXX FIX XXX for pop [esp]
-    if isinstance(src, m2_expr.ExprMem):
-        src = expr_simp(src.replace_expr({sp: new_sp}))
-    result = sp
-    if ir.do_stk_segm:
-        result = ir.gen_segm_expr(SS, result)
-
-    e.append(m2_expr.ExprAssign(src, ir.ExprMem(result, src.size)))
-    return e, []
-
-
-def pop(ir, instr, src):
-    return pop_gen(ir, instr, src, instr.mode)
-
-
-def popw(ir, instr, src):
-    return pop_gen(ir, instr, src, 16)
-
-
-def sete(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_EQ", zf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setnz(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_EQ", ~zf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setl(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_S<", nf, of).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setg(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_S>", nf, of, zf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setge(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_S>=", nf, of).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def seta(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_U>", cf, zf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setae(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_U>=", cf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setb(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_U<", cf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setbe(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_U<=", cf, zf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setns(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_NEG", ~nf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def sets(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_NEG", nf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def seto(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            of.zeroExtend(dst.size)
-        )
-    )
-    return e, []
-
-
-def setp(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            pf.zeroExtend(dst.size)
-        )
-    )
-    return e, []
-
-
-def setnp(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprCond(
-                pf,
-                m2_expr.ExprInt(0, dst.size),
-                m2_expr.ExprInt(1, dst.size)
-            )
-        )
-    )
-    return e, []
-
-
-def setle(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_S<=", nf, of, zf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setna(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_U<=", cf, zf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setnbe(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_U>", cf, zf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setno(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprCond(
-                of,
-                m2_expr.ExprInt(0, dst.size),
-                m2_expr.ExprInt(1, dst.size)
-            )
-        )
-    )
-    return e, []
-
-
-def setnb(_, instr, dst):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst,
-            m2_expr.ExprOp("CC_U>=", cf).zeroExtend(dst.size),
-        )
-    )
-    return e, []
-
-
-def setalc(_, instr):
-    dst = mRAX[instr.mode][0:8]
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst, m2_expr.ExprCond(cf, m2_expr.ExprInt(0xff, dst.size),
-                                              m2_expr.ExprInt(0, dst.size))))
-    return e, []
-
-
-def bswap(_, instr, dst):
-    e = []
-    if dst.size == 16:
-        # BSWAP referencing a 16-bit register is undefined
-        # Seems to return 0 actually
-        result = m2_expr.ExprInt(0, 16)
-    elif dst.size == 32:
-        result = m2_expr.ExprCompose(
-            dst[24:32], dst[16:24], dst[8:16], dst[:8])
-    elif dst.size == 64:
-        result = m2_expr.ExprCompose(dst[56:64], dst[48:56], dst[40:48], dst[32:40],
-                                     dst[24:32], dst[16:24], dst[8:16], dst[:8])
-    else:
-        raise ValueError('the size DOES matter')
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-
-def cmps(ir, instr, size):
-    loc_df_0, loc_df_0_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_df_1, loc_df_1_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_next_expr = m2_expr.ExprLoc(ir.get_next_loc_key(instr), ir.IRDst.size)
-
-    src1 = mRSI[instr.mode][:instr.v_admode()]
-    src2 = mRDI[instr.mode][:instr.v_admode()]
-
-    if ir.do_str_segm:
-        if instr.additional_info.g2.value:
-            raise NotImplementedError("add segm support")
-        src1_sgm = ir.gen_segm_expr(DS, src1)
-        src2_sgm = ir.gen_segm_expr(ES, src2)
-    else:
-        src1_sgm = src1
-        src2_sgm = src2
-
-    offset = m2_expr.ExprInt(size // 8, src1.size)
-
-    e, _ = l_cmp(ir, instr,
-                 ir.ExprMem(src1_sgm, size),
-                 ir.ExprMem(src2_sgm, size))
-
-
-    e0 = []
-    e0.append(m2_expr.ExprAssign(src1, src1 + offset))
-    e0.append(m2_expr.ExprAssign(src2, src2 + offset))
-    e0.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    e0 = IRBlock(ir.loc_db, loc_df_0, [AssignBlock(e0, instr)])
-
-    e1 = []
-    e1.append(m2_expr.ExprAssign(src1, src1 - offset))
-    e1.append(m2_expr.ExprAssign(src2, src2 - offset))
-    e1.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    e1 = IRBlock(ir.loc_db, loc_df_1, [AssignBlock(e1, instr)])
-
-    e.append(m2_expr.ExprAssign(ir.IRDst,
-                             m2_expr.ExprCond(df, loc_df_1_expr, loc_df_0_expr)))
-    return e, [e0, e1]
-
-
-def scas(ir, instr, size):
-    loc_df_0, loc_df_0_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_df_1, loc_df_1_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_next_expr = m2_expr.ExprLoc(ir.get_next_loc_key(instr), ir.IRDst.size)
-
-    src = mRDI[instr.mode][:instr.v_admode()]
-
-    if ir.do_str_segm:
-        if instr.additional_info.g2.value:
-            raise NotImplementedError("add segm support")
-        src_sgm = ir.gen_segm_expr(ES, src)
-
-    else:
-        src_sgm = src
-
-    offset = m2_expr.ExprInt(size // 8, src.size)
-    e, extra = l_cmp(ir, instr,
-                     mRAX[instr.mode][:size],
-                     ir.ExprMem(src_sgm, size))
-
-    e0 = []
-    e0.append(m2_expr.ExprAssign(src, src + offset))
-
-    e0.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    e0 = IRBlock(ir.loc_db, loc_df_0, [AssignBlock(e0, instr)])
-
-    e1 = []
-    e1.append(m2_expr.ExprAssign(src, src - offset))
-    e1.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    e1 = IRBlock(ir.loc_db, loc_df_1, [AssignBlock(e1, instr)])
-
-    e.append(m2_expr.ExprAssign(ir.IRDst,
-                             m2_expr.ExprCond(df, loc_df_1_expr, loc_df_0_expr)))
-
-    return e, [e0, e1]
-
-
-def compose_eflag(s=32):
-    args = []
-
-    args = [cf, m2_expr.ExprInt(1, 1), pf, m2_expr.ExprInt(0, 1), af,
-            m2_expr.ExprInt(0, 1), zf, nf, tf, i_f, df, of, iopl]
-
-    if s == 32:
-        args += [nt, m2_expr.ExprInt(0, 1), rf, vm, ac, vif, vip, i_d]
-    elif s == 16:
-        args += [nt, m2_expr.ExprInt(0, 1)]
-    else:
-        raise ValueError('unk size')
-    if s == 32:
-        args.append(m2_expr.ExprInt(0, 10))
-    return m2_expr.ExprCompose(*args)
-
-
-def pushfd(ir, instr):
-    return push(ir, instr, compose_eflag())
-
-
-def pushfq(ir, instr):
-    return push(ir, instr, compose_eflag().zeroExtend(64))
-
-
-def pushfw(ir, instr):
-    return pushw(ir, instr, compose_eflag(16))
-
-
-def popfd(ir, instr):
-    tmp = ir.ExprMem(mRSP[instr.mode], 32)
-    e = []
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprSlice(tmp, 0, 1)))
-    e.append(m2_expr.ExprAssign(pf, m2_expr.ExprSlice(tmp, 2, 3)))
-    e.append(m2_expr.ExprAssign(af, m2_expr.ExprSlice(tmp, 4, 5)))
-    e.append(m2_expr.ExprAssign(zf, m2_expr.ExprSlice(tmp, 6, 7)))
-    e.append(m2_expr.ExprAssign(nf, m2_expr.ExprSlice(tmp, 7, 8)))
-    e.append(m2_expr.ExprAssign(tf, m2_expr.ExprSlice(tmp, 8, 9)))
-    e.append(m2_expr.ExprAssign(i_f, m2_expr.ExprSlice(tmp, 9, 10)))
-    e.append(m2_expr.ExprAssign(df, m2_expr.ExprSlice(tmp, 10, 11)))
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprSlice(tmp, 11, 12)))
-    e.append(m2_expr.ExprAssign(iopl, m2_expr.ExprSlice(tmp, 12, 14)))
-    e.append(m2_expr.ExprAssign(nt, m2_expr.ExprSlice(tmp, 14, 15)))
-    e.append(m2_expr.ExprAssign(rf, m2_expr.ExprSlice(tmp, 16, 17)))
-    e.append(m2_expr.ExprAssign(vm, m2_expr.ExprSlice(tmp, 17, 18)))
-    e.append(m2_expr.ExprAssign(ac, m2_expr.ExprSlice(tmp, 18, 19)))
-    e.append(m2_expr.ExprAssign(vif, m2_expr.ExprSlice(tmp, 19, 20)))
-    e.append(m2_expr.ExprAssign(vip, m2_expr.ExprSlice(tmp, 20, 21)))
-    e.append(m2_expr.ExprAssign(i_d, m2_expr.ExprSlice(tmp, 21, 22)))
-    e.append(m2_expr.ExprAssign(mRSP[instr.mode],
-                             mRSP[instr.mode] + m2_expr.ExprInt(instr.mode // 8, mRSP[instr.mode].size)))
-    e.append(m2_expr.ExprAssign(exception_flags,
-                             m2_expr.ExprCond(m2_expr.ExprSlice(tmp, 8, 9),
-                                              m2_expr.ExprInt(
-                                                  EXCEPT_SOFT_BP, 32),
-                                              exception_flags
-                                              )
-                             )
-             )
-    return e, []
-
-
-def _tpl_eflags(tmp):
-    """Extract eflags from @tmp
-    @tmp: Expr instance with a size >= 16
-    """
-    return [m2_expr.ExprAssign(dest, tmp[base:base + dest.size])
-            for base, dest in ((0, cf), (2, pf), (4, af), (6, zf), (7, nf),
-                               (8, tf), (9, i_f), (10, df), (11, of),
-                               (12, iopl), (14, nt))]
-
-
-def popfw(ir, instr):
-    tmp = ir.ExprMem(mRSP[instr.mode], 16)
-    e = _tpl_eflags(tmp)
-    e.append(
-        m2_expr.ExprAssign(mRSP[instr.mode], mRSP[instr.mode] + m2_expr.ExprInt(2, mRSP[instr.mode].size)))
-    return e, []
-
-pa_regs = [
-    mRAX, mRCX,
-    mRDX, mRBX,
-    mRSP, mRBP,
-    mRSI, mRDI
-]
-
-
-def pusha_gen(ir, instr, size):
-    e = []
-    cur_sp = mRSP[instr.mode]
-    for i, reg in enumerate(pa_regs):
-        stk_ptr = cur_sp + m2_expr.ExprInt(-(size // 8) * (i + 1), instr.mode)
-        e.append(m2_expr.ExprAssign(ir.ExprMem(stk_ptr, size), reg[size]))
-    e.append(m2_expr.ExprAssign(cur_sp, stk_ptr))
-    return e, []
-
-
-def pusha(ir, instr):
-    return pusha_gen(ir, instr, 16)
-
-
-def pushad(ir, instr):
-    return pusha_gen(ir, instr, 32)
-
-
-def popa_gen(ir, instr, size):
-    e = []
-    cur_sp = mRSP[instr.mode]
-    for i, reg in enumerate(reversed(pa_regs)):
-        if reg == mRSP:
-            continue
-        stk_ptr = cur_sp + m2_expr.ExprInt((size // 8) * i, instr.mode)
-        e.append(m2_expr.ExprAssign(reg[size], ir.ExprMem(stk_ptr, size)))
-
-    stk_ptr = cur_sp + m2_expr.ExprInt((size // 8) * (i + 1), instr.mode)
-    e.append(m2_expr.ExprAssign(cur_sp, stk_ptr))
-
-    return e, []
-
-
-def popa(ir, instr):
-    return popa_gen(ir, instr, 16)
-
-
-def popad(ir, instr):
-    return popa_gen(ir, instr, 32)
-
-
-def call(ir, instr, dst):
-    e = []
-    # opmode, admode = instr.opmode, instr.admode
-    s = dst.size
-    meip = mRIP[ir.IRDst.size]
-    opmode, admode = s, instr.v_admode()
-    myesp = mRSP[instr.mode][:opmode]
-    n = m2_expr.ExprLoc(ir.get_next_loc_key(instr), ir.IRDst.size)
-
-    if isinstance(dst, m2_expr.ExprOp):
-        if dst.op == "segm":
-            # Far call segm:addr
-            if instr.mode not in [16, 32]:
-                raise RuntimeError('not supported')
-            segm = dst.args[0]
-            base = dst.args[1]
-            m1 = segm.zeroExtend(CS.size)
-            m2 = base.zeroExtend(meip.size)
-        elif dst.op == "far":
-            # Far call far [eax]
-            addr = dst.args[0].ptr
-            m1 = ir.ExprMem(addr, CS.size)
-            m2 = ir.ExprMem(addr + m2_expr.ExprInt(2, addr.size), meip.size)
-        else:
-            raise RuntimeError("bad call operator")
-
-        e.append(m2_expr.ExprAssign(CS, m1))
-        e.append(m2_expr.ExprAssign(meip, m2))
-
-        e.append(m2_expr.ExprAssign(ir.IRDst, m2))
-
-        c = myesp + m2_expr.ExprInt(-s // 8, s)
-        e.append(m2_expr.ExprAssign(ir.ExprMem(c, size=s).zeroExtend(s),
-                                 CS.zeroExtend(s)))
-
-        c = myesp + m2_expr.ExprInt((-2 * s) // 8, s)
-        e.append(m2_expr.ExprAssign(ir.ExprMem(c, size=s).zeroExtend(s),
-                                 meip.zeroExtend(s)))
-
-        c = myesp + m2_expr.ExprInt((-2 * s) // 8, s)
-        e.append(m2_expr.ExprAssign(myesp, c))
-        return e, []
-
-    c = myesp + m2_expr.ExprInt(-s // 8, s)
-    e.append(m2_expr.ExprAssign(myesp, c))
-    if ir.do_stk_segm:
-        c = ir.gen_segm_expr(SS, c)
-
-    e.append(m2_expr.ExprAssign(ir.ExprMem(c, size=s), n))
-    e.append(m2_expr.ExprAssign(meip, dst.zeroExtend(ir.IRDst.size)))
-    e.append(m2_expr.ExprAssign(ir.IRDst, dst.zeroExtend(ir.IRDst.size)))
-    return e, []
-
-
-def ret(ir, instr, src=None):
-    e = []
-    meip = mRIP[ir.IRDst.size]
-    size, admode = instr.v_opmode(), instr.v_admode()
-    myesp = mRSP[instr.mode][:size]
-
-    if src is None:
-        value = (myesp + (m2_expr.ExprInt(size // 8, size)))
-    else:
-        src = m2_expr.ExprInt(int(src), size)
-        value = (myesp + (m2_expr.ExprInt(size // 8, size) + src))
-
-    e.append(m2_expr.ExprAssign(myesp, value))
-    result = myesp
-    if ir.do_stk_segm:
-        result = ir.gen_segm_expr(SS, result)
-
-    e.append(m2_expr.ExprAssign(meip, ir.ExprMem(
-        result, size=size).zeroExtend(size)))
-    e.append(m2_expr.ExprAssign(ir.IRDst,
-                             ir.ExprMem(result, size=size).zeroExtend(size)))
-    return e, []
-
-
-def retf(ir, instr, src=None):
-    e = []
-    meip = mRIP[ir.IRDst.size]
-    size, admode = instr.v_opmode(), instr.v_admode()
-    if src is None:
-        src = m2_expr.ExprInt(0, instr.mode)
-    myesp = mRSP[instr.mode][:size]
-
-    src = src.zeroExtend(size)
-
-    result = myesp
-    if ir.do_stk_segm:
-        result = ir.gen_segm_expr(SS, result)
-
-    e.append(m2_expr.ExprAssign(meip, ir.ExprMem(
-        result, size=size).zeroExtend(size)))
-    e.append(m2_expr.ExprAssign(ir.IRDst,
-                             ir.ExprMem(result, size=size).zeroExtend(size)))
-    # e.append(m2_expr.ExprAssign(meip, ir.ExprMem(c, size = s)))
-    result = myesp + m2_expr.ExprInt(size // 8, size)
-    if ir.do_stk_segm:
-        result = ir.gen_segm_expr(SS, result)
-
-    e.append(m2_expr.ExprAssign(CS, ir.ExprMem(result, size=16)))
-
-    value = myesp + (m2_expr.ExprInt((2 * size) // 8, size) + src)
-    e.append(m2_expr.ExprAssign(myesp, value))
-    return e, []
-
-
-def leave(ir, instr):
-    size = instr.mode
-    myesp = mRSP[size]
-    e = []
-    e.append(m2_expr.ExprAssign(mRBP[size], ir.ExprMem(mRBP[size], size=size)))
-    e.append(m2_expr.ExprAssign(myesp,
-                             m2_expr.ExprInt(size // 8, size) + mRBP[size]))
-    return e, []
-
-
-def enter(ir, instr, src1, src2):
-    size, admode = instr.v_opmode(), instr.v_admode()
-    myesp = mRSP[instr.mode][:size]
-    myebp = mRBP[instr.mode][:size]
-
-    src1 = src1.zeroExtend(size)
-
-    e = []
-    esp_tmp = myesp - m2_expr.ExprInt(size // 8, size)
-    e.append(m2_expr.ExprAssign(ir.ExprMem(esp_tmp, size=size),
-                             myebp))
-    e.append(m2_expr.ExprAssign(myebp, esp_tmp))
-    e.append(m2_expr.ExprAssign(myesp,
-                             myesp - (src1 + m2_expr.ExprInt(size // 8, size))))
-    return e, []
-
-
-def jmp(ir, instr, dst):
-    e = []
-    meip = mRIP[ir.IRDst.size]
-
-    if isinstance(dst, m2_expr.ExprOp):
-        if dst.op == "segm":
-            # Far jmp segm:addr
-            segm = dst.args[0]
-            base = dst.args[1]
-            m1 = segm.zeroExtend(CS.size)
-            m2 = base.zeroExtend(meip.size)
-        elif dst.op == "far":
-            # Far jmp far [eax]
-            addr = dst.args[0].ptr
-            m1 = ir.ExprMem(addr, CS.size)
-            m2 = ir.ExprMem(addr + m2_expr.ExprInt(2, addr.size), meip.size)
-        else:
-            raise RuntimeError("bad jmp operator")
-
-        e.append(m2_expr.ExprAssign(CS, m1))
-        e.append(m2_expr.ExprAssign(meip, m2))
-        e.append(m2_expr.ExprAssign(ir.IRDst, m2))
-
-    else:
-        # Classic jmp
-        e.append(m2_expr.ExprAssign(meip, dst))
-        e.append(m2_expr.ExprAssign(ir.IRDst, dst))
-
-        if isinstance(dst, m2_expr.ExprMem):
-            dst = meip
-    return e, []
-
-
-def jz(ir, instr, dst):
-    #return gen_jcc(ir, instr, zf, dst, True)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_EQ", zf), dst, True)
-
-
-def jcxz(ir, instr, dst):
-    return gen_jcc(ir, instr, mRCX[instr.mode][:16], dst, False)
-
-
-def jecxz(ir, instr, dst):
-    return gen_jcc(ir, instr, mRCX[instr.mode][:32], dst, False)
-
-
-def jrcxz(ir, instr, dst):
-    return gen_jcc(ir, instr, mRCX[instr.mode], dst, False)
-
-
-def jnz(ir, instr, dst):
-    #return gen_jcc(ir, instr, zf, dst, False)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_EQ", zf), dst, False)
-
-
-
-def jp(ir, instr, dst):
-    return gen_jcc(ir, instr, pf, dst, True)
-
-
-def jnp(ir, instr, dst):
-    return gen_jcc(ir, instr, pf, dst, False)
-
-
-def ja(ir, instr, dst):
-    #return gen_jcc(ir, instr, cf | zf, dst, False)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_U>", cf, zf), dst, True)
-
-
-def jae(ir, instr, dst):
-    #return gen_jcc(ir, instr, cf, dst, False)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_U>=", cf), dst, True)
-
-
-def jb(ir, instr, dst):
-    #return gen_jcc(ir, instr, cf, dst, True)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_U<", cf), dst, True)
-
-
-def jbe(ir, instr, dst):
-    #return gen_jcc(ir, instr, cf | zf, dst, True)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_U<=", cf, zf), dst, True)
-
-
-def jge(ir, instr, dst):
-    #return gen_jcc(ir, instr, nf - of, dst, False)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_S>=", nf, of), dst, True)
-
-
-def jg(ir, instr, dst):
-    #return gen_jcc(ir, instr, zf | (nf - of), dst, False)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_S>", nf, of, zf), dst, True)
-
-
-def jl(ir, instr, dst):
-    #return gen_jcc(ir, instr, nf - of, dst, True)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_S<", nf, of), dst, True)
-
-
-def jle(ir, instr, dst):
-    #return gen_jcc(ir, instr, zf | (nf - of), dst, True)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_S<=", nf, of, zf), dst, True)
-
-
-
-def js(ir, instr, dst):
-    #return gen_jcc(ir, instr, nf, dst, True)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_NEG", nf), dst, True)
-
-
-
-def jns(ir, instr, dst):
-    #return gen_jcc(ir, instr, nf, dst, False)
-    return gen_jcc(ir, instr, m2_expr.ExprOp("CC_NEG", nf), dst, False)
-
-
-def jo(ir, instr, dst):
-    return gen_jcc(ir, instr, of, dst, True)
-
-
-def jno(ir, instr, dst):
-    return gen_jcc(ir, instr, of, dst, False)
-
-
-def loop(ir, instr, dst):
-    e = []
-    meip = mRIP[ir.IRDst.size]
-    admode = instr.v_admode()
-    myecx = mRCX[instr.mode][:admode]
-
-    n = m2_expr.ExprLoc(ir.get_next_loc_key(instr), ir.IRDst.size)
-    c = myecx - m2_expr.ExprInt(1, myecx.size)
-    dst_o = m2_expr.ExprCond(c,
-                             dst.zeroExtend(ir.IRDst.size),
-                             n.zeroExtend(ir.IRDst.size))
-    e.append(m2_expr.ExprAssign(myecx, c))
-    e.append(m2_expr.ExprAssign(meip, dst_o))
-    e.append(m2_expr.ExprAssign(ir.IRDst, dst_o))
-    return e, []
-
-
-def loopne(ir, instr, dst):
-    e = []
-    meip = mRIP[ir.IRDst.size]
-    admode = instr.v_admode()
-    myecx = mRCX[instr.mode][:admode]
-
-    n = m2_expr.ExprLoc(ir.get_next_loc_key(instr), ir.IRDst.size)
-
-    c = m2_expr.ExprCond(myecx - m2_expr.ExprInt(1, size=myecx.size),
-                         m2_expr.ExprInt(1, 1),
-                         m2_expr.ExprInt(0, 1))
-    c &= zf ^ m2_expr.ExprInt(1, 1)
-
-    e.append(m2_expr.ExprAssign(myecx, myecx - m2_expr.ExprInt(1, myecx.size)))
-    dst_o = m2_expr.ExprCond(c,
-                             dst.zeroExtend(ir.IRDst.size),
-                             n.zeroExtend(ir.IRDst.size))
-    e.append(m2_expr.ExprAssign(meip, dst_o))
-    e.append(m2_expr.ExprAssign(ir.IRDst, dst_o))
-    return e, []
-
-
-def loope(ir, instr, dst):
-    e = []
-    meip = mRIP[ir.IRDst.size]
-    admode = instr.v_admode()
-    myecx = mRCX[instr.mode][:admode]
-
-    n = m2_expr.ExprLoc(ir.get_next_loc_key(instr), ir.IRDst.size)
-    c = m2_expr.ExprCond(myecx - m2_expr.ExprInt(1, size=myecx.size),
-                         m2_expr.ExprInt(1, 1),
-                         m2_expr.ExprInt(0, 1))
-    c &= zf
-    e.append(m2_expr.ExprAssign(myecx, myecx - m2_expr.ExprInt(1, myecx.size)))
-    dst_o = m2_expr.ExprCond(c,
-                             dst.zeroExtend(ir.IRDst.size),
-                             n.zeroExtend(ir.IRDst.size))
-    e.append(m2_expr.ExprAssign(meip, dst_o))
-    e.append(m2_expr.ExprAssign(ir.IRDst, dst_o))
-    return e, []
-
-# XXX size to do; eflag
-
-
-def div(ir, instr, src1):
-    e = []
-    size = src1.size
-    if size == 8:
-        src2 = mRAX[instr.mode][:16]
-    elif size in [16, 32, 64]:
-        s1, s2 = mRDX[size], mRAX[size]
-        src2 = m2_expr.ExprCompose(s2, s1)
-    else:
-        raise ValueError('div arg not impl', src1)
-
-    c_d = m2_expr.ExprOp('udiv', src2, src1.zeroExtend(src2.size))
-    c_r = m2_expr.ExprOp('umod', src2, src1.zeroExtend(src2.size))
-
-    # if 8 bit div, only ax is assigned
-    if size == 8:
-        e.append(m2_expr.ExprAssign(src2, m2_expr.ExprCompose(c_d[:8], c_r[:8])))
-    else:
-        e.append(m2_expr.ExprAssign(s1, c_r[:size]))
-        e.append(m2_expr.ExprAssign(s2, c_d[:size]))
-
-    loc_div, loc_div_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_except, loc_except_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_next = ir.get_next_loc_key(instr)
-    loc_next_expr = m2_expr.ExprLoc(loc_next, ir.IRDst.size)
-
-    do_div = []
-    do_div += e
-    do_div.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    blk_div = IRBlock(ir.loc_db, loc_div, [AssignBlock(do_div, instr)])
-
-    do_except = []
-    do_except.append(m2_expr.ExprAssign(exception_flags, m2_expr.ExprInt(
-        EXCEPT_DIV_BY_ZERO, exception_flags.size)))
-    do_except.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    blk_except = IRBlock(ir.loc_db, loc_except, [AssignBlock(do_except, instr)])
-
-    e = []
-    e.append(m2_expr.ExprAssign(ir.IRDst,
-                             m2_expr.ExprCond(src1, loc_div_expr, loc_except_expr)))
-
-    return e, [blk_div, blk_except]
-
-
-# XXX size to do; eflag
-
-def idiv(ir, instr, src1):
-    e = []
-    size = src1.size
-
-    if size == 8:
-        src2 = mRAX[instr.mode][:16]
-    elif size in [16, 32, 64]:
-        s1, s2 = mRDX[size], mRAX[size]
-        src2 = m2_expr.ExprCompose(s2, s1)
-    else:
-        raise ValueError('div arg not impl', src1)
-
-    c_d = m2_expr.ExprOp('sdiv', src2, src1.signExtend(src2.size))
-    c_r = m2_expr.ExprOp('smod', src2, src1.signExtend(src2.size))
-
-    # if 8 bit div, only ax is assigned
-    if size == 8:
-        e.append(m2_expr.ExprAssign(src2, m2_expr.ExprCompose(c_d[:8], c_r[:8])))
-    else:
-        e.append(m2_expr.ExprAssign(s1, c_r[:size]))
-        e.append(m2_expr.ExprAssign(s2, c_d[:size]))
-
-    loc_div, loc_div_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_except, loc_except_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_next = ir.get_next_loc_key(instr)
-    loc_next_expr = m2_expr.ExprLoc(loc_next, ir.IRDst.size)
-
-    do_div = []
-    do_div += e
-    do_div.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    blk_div = IRBlock(ir.loc_db, loc_div, [AssignBlock(do_div, instr)])
-
-    do_except = []
-    do_except.append(m2_expr.ExprAssign(exception_flags, m2_expr.ExprInt(
-        EXCEPT_DIV_BY_ZERO, exception_flags.size)))
-    do_except.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    blk_except = IRBlock(ir.loc_db, loc_except, [AssignBlock(do_except, instr)])
-
-    e = []
-    e.append(m2_expr.ExprAssign(ir.IRDst,
-                             m2_expr.ExprCond(src1, loc_div_expr, loc_except_expr)))
-
-    return e, [blk_div, blk_except]
-
-
-# XXX size to do; eflag
-
-
-def mul(_, instr, src1):
-    e = []
-    size = src1.size
-    if src1.size in [16, 32, 64]:
-        result = m2_expr.ExprOp('*',
-                                mRAX[size].zeroExtend(size * 2),
-                                src1.zeroExtend(size * 2))
-        e.append(m2_expr.ExprAssign(mRAX[size], result[:size]))
-        e.append(m2_expr.ExprAssign(mRDX[size], result[size:size * 2]))
-
-    elif src1.size == 8:
-        result = m2_expr.ExprOp('*',
-                                mRAX[instr.mode][:8].zeroExtend(16),
-                                src1.zeroExtend(16))
-        e.append(m2_expr.ExprAssign(mRAX[instr.mode][:16], result))
-    else:
-        raise ValueError('unknow size')
-
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprCond(result[size:size * 2],
-                                                  m2_expr.ExprInt(1, 1),
-                                                  m2_expr.ExprInt(0, 1))))
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprCond(result[size:size * 2],
-                                                  m2_expr.ExprInt(1, 1),
-                                                  m2_expr.ExprInt(0, 1))))
-
-    return e, []
-
-
-def imul(_, instr, src1, src2=None, src3=None):
-    e = []
-    size = src1.size
-    if src2 is None:
-        if size in [16, 32, 64]:
-            result = m2_expr.ExprOp('*',
-                                    mRAX[size].signExtend(size * 2),
-                                    src1.signExtend(size * 2))
-            e.append(m2_expr.ExprAssign(mRAX[size], result[:size]))
-            e.append(m2_expr.ExprAssign(mRDX[size], result[size:size * 2]))
-        elif size == 8:
-            dst = mRAX[instr.mode][:16]
-            result = m2_expr.ExprOp('*',
-                                    mRAX[instr.mode][:8].signExtend(16),
-                                    src1.signExtend(16))
-
-            e.append(m2_expr.ExprAssign(dst, result))
-        value = m2_expr.ExprCond(result - result[:size].signExtend(size * 2),
-                                 m2_expr.ExprInt(1, 1),
-                                 m2_expr.ExprInt(0, 1))
-        e.append(m2_expr.ExprAssign(cf, value))
-        value = m2_expr.ExprCond(result - result[:size].signExtend(size * 2),
-                                 m2_expr.ExprInt(1, 1),
-                                 m2_expr.ExprInt(0, 1))
-        e.append(m2_expr.ExprAssign(of, value))
-
-    else:
-        if src3 is None:
-            src3 = src2
-            src2 = src1
-        result = m2_expr.ExprOp('*',
-                                src2.signExtend(size * 2),
-                                src3.signExtend(size * 2))
-        e.append(m2_expr.ExprAssign(src1, result[:size]))
-
-        value = m2_expr.ExprCond(result - result[:size].signExtend(size * 2),
-                                 m2_expr.ExprInt(1, 1),
-                                 m2_expr.ExprInt(0, 1))
-        e.append(m2_expr.ExprAssign(cf, value))
-        value = m2_expr.ExprCond(result - result[:size].signExtend(size * 2),
-                                 m2_expr.ExprInt(1, 1),
-                                 m2_expr.ExprInt(0, 1))
-        e.append(m2_expr.ExprAssign(of, value))
-    return e, []
-
-
-def cbw(_, instr):
-    # Only in 16 bit
-    e = []
-    tempAL = mRAX[instr.v_opmode()][:8]
-    tempAX = mRAX[instr.v_opmode()][:16]
-    e.append(m2_expr.ExprAssign(tempAX, tempAL.signExtend(16)))
-    return e, []
-
-
-def cwde(_, instr):
-    # Only in 32/64 bit
-    e = []
-    tempAX = mRAX[instr.v_opmode()][:16]
-    tempEAX = mRAX[instr.v_opmode()][:32]
-    e.append(m2_expr.ExprAssign(tempEAX, tempAX.signExtend(32)))
-    return e, []
-
-
-def cdqe(_, instr):
-    # Only in 64 bit
-    e = []
-    tempEAX = mRAX[instr.mode][:32]
-    tempRAX = mRAX[instr.mode][:64]
-    e.append(m2_expr.ExprAssign(tempRAX, tempEAX.signExtend(64)))
-    return e, []
-
-
-def cwd(_, instr):
-    # Only in 16 bit
-    e = []
-    tempAX = mRAX[instr.mode][:16]
-    tempDX = mRDX[instr.mode][:16]
-    result = tempAX.signExtend(32)
-    e.append(m2_expr.ExprAssign(tempAX, result[:16]))
-    e.append(m2_expr.ExprAssign(tempDX, result[16:32]))
-    return e, []
-
-
-def cdq(_, instr):
-    # Only in 32/64 bit
-    e = []
-    tempEAX = mRAX[instr.v_opmode()]
-    tempEDX = mRDX[instr.v_opmode()]
-    result = tempEAX.signExtend(64)
-    e.append(m2_expr.ExprAssign(tempEDX, result[32:64]))
-    return e, []
-
-
-def cqo(_, instr):
-    # Only in 64 bit
-    e = []
-    tempRAX = mRAX[instr.mode][:64]
-    tempRDX = mRDX[instr.mode][:64]
-    result = tempRAX.signExtend(128)
-    e.append(m2_expr.ExprAssign(tempRAX, result[:64]))
-    e.append(m2_expr.ExprAssign(tempRDX, result[64:128]))
-    return e, []
-
-
-def stos(ir, instr, size):
-    loc_df_0, loc_df_0_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_df_1, loc_df_1_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_next_expr = m2_expr.ExprLoc(ir.get_next_loc_key(instr), ir.IRDst.size)
-
-    addr_o = mRDI[instr.mode][:instr.v_admode()]
-    addr = addr_o
-    addr_p = addr + m2_expr.ExprInt(size // 8, addr.size)
-    addr_m = addr - m2_expr.ExprInt(size // 8, addr.size)
-    if ir.do_str_segm:
-        mss = ES
-        if instr.additional_info.g2.value:
-            raise NotImplementedError("add segm support")
-        addr = ir.gen_segm_expr(mss, addr)
-
-
-    b = mRAX[instr.mode][:size]
-
-    e0 = []
-    e0.append(m2_expr.ExprAssign(addr_o, addr_p))
-    e0.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    e0 = IRBlock(ir.loc_db, loc_df_0, [AssignBlock(e0, instr)])
-
-    e1 = []
-    e1.append(m2_expr.ExprAssign(addr_o, addr_m))
-    e1.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    e1 = IRBlock(ir.loc_db, loc_df_1, [AssignBlock(e1, instr)])
-
-    e = []
-    e.append(m2_expr.ExprAssign(ir.ExprMem(addr, size), b))
-    e.append(m2_expr.ExprAssign(ir.IRDst,
-                             m2_expr.ExprCond(df, loc_df_1_expr, loc_df_0_expr)))
-    return e, [e0, e1]
-
-
-def lods(ir, instr, size):
-    loc_df_0, loc_df_0_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_df_1, loc_df_1_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_next_expr = m2_expr.ExprLoc(ir.get_next_loc_key(instr), ir.IRDst.size)
-    e = []
-
-    addr_o = mRSI[instr.mode][:instr.v_admode()]
-    addr = addr_o
-    addr_p = addr + m2_expr.ExprInt(size // 8, addr.size)
-    addr_m = addr - m2_expr.ExprInt(size // 8, addr.size)
-    if ir.do_str_segm:
-        mss = DS
-        if instr.additional_info.g2.value:
-            raise NotImplementedError("add segm support")
-        addr = ir.gen_segm_expr(mss, addr)
-
-
-    b = mRAX[instr.mode][:size]
-
-    e0 = []
-    e0.append(m2_expr.ExprAssign(addr_o, addr_p))
-    e0.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    e0 = IRBlock(ir.loc_db, loc_df_0, [AssignBlock(e0, instr)])
-
-    e1 = []
-    e1.append(m2_expr.ExprAssign(addr_o, addr_m))
-    e1.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    e1 = IRBlock(ir.loc_db, loc_df_1, [AssignBlock(e1, instr)])
-
-    e = []
-    if instr.mode == 64 and b.size == 32:
-        e.append(m2_expr.ExprAssign(mRAX[instr.mode],
-                                 ir.ExprMem(addr, size).zeroExtend(64)))
-    else:
-        e.append(m2_expr.ExprAssign(b, ir.ExprMem(addr, size)))
-
-    e.append(m2_expr.ExprAssign(ir.IRDst,
-                             m2_expr.ExprCond(df, loc_df_1_expr, loc_df_0_expr)))
-    return e, [e0, e1]
-
-
-def movs(ir, instr, size):
-    loc_df_0, loc_df_0_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_df_1, loc_df_1_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_next_expr = m2_expr.ExprLoc(ir.get_next_loc_key(instr), ir.IRDst.size)
-
-    dst = mRDI[instr.mode][:instr.v_admode()]
-    src = mRSI[instr.mode][:instr.v_admode()]
-
-    e = []
-    if ir.do_str_segm:
-        if instr.additional_info.g2.value:
-            raise NotImplementedError("add segm support")
-        src_sgm = ir.gen_segm_expr(DS, src)
-        dst_sgm = ir.gen_segm_expr(ES, dst)
-
-    else:
-        src_sgm = src
-        dst_sgm = dst
-
-    offset = m2_expr.ExprInt(size // 8, src.size)
-
-    e.append(m2_expr.ExprAssign(ir.ExprMem(dst_sgm, size),
-                             ir.ExprMem(src_sgm, size)))
-
-    e0 = []
-    e0.append(m2_expr.ExprAssign(src, src + offset))
-    e0.append(m2_expr.ExprAssign(dst, dst + offset))
-    e0.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    e0 = IRBlock(ir.loc_db, loc_df_0, [AssignBlock(e0, instr)])
-
-    e1 = []
-    e1.append(m2_expr.ExprAssign(src, src - offset))
-    e1.append(m2_expr.ExprAssign(dst, dst - offset))
-    e1.append(m2_expr.ExprAssign(ir.IRDst, loc_next_expr))
-    e1 = IRBlock(ir.loc_db, loc_df_1, [AssignBlock(e1, instr)])
-
-    e.append(m2_expr.ExprAssign(ir.IRDst,
-                             m2_expr.ExprCond(df, loc_df_1_expr, loc_df_0_expr)))
-    return e, [e0, e1]
-
-
-def movsd(_, instr, dst, src):
-    # 64 bits access
-    if dst.is_id() and src.is_id():
-        src = src[:64]
-        dst = dst[:64]
-    elif dst.is_mem() and src.is_id():
-        dst = m2_expr.ExprMem(dst.ptr, 64)
-        src = src[:64]
-    else:
-        src = m2_expr.ExprMem(src.ptr, 64)
-        # Erase dst high bits
-        src = src.zeroExtend(dst.size)
-    return [m2_expr.ExprAssign(dst, src)], []
-
-
-def movsd_dispatch(ir, instr, dst=None, src=None):
-    if dst is None and src is None:
-        return movs(ir, instr, 32)
-    else:
-        return movsd(ir, instr, dst, src)
-
-
-def float_prev(flt, popcount=1):
-    if not flt in float_list:
-        return None
-    i = float_list.index(flt)
-    if i < popcount:
-        # Drop value (ex: FSTP ST(0))
-        return None
-    flt = float_list[i - popcount]
-    return flt
-
-
-def float_pop(avoid_flt=None, popcount=1):
-    """
-    Generate floatpop semantic (@popcount times), avoiding the avoid_flt@ float
-    @avoid_flt: float avoided in the generated semantic
-    @popcount: pop count
-    """
-    avoid_flt = float_prev(avoid_flt, popcount)
-    e = []
-    for i in range(8 - popcount):
-        if avoid_flt != float_list[i]:
-            e.append(m2_expr.ExprAssign(float_list[i],
-                                     float_list[i + popcount]))
-    fill_value = m2_expr.ExprOp("sint_to_fp", m2_expr.ExprInt(0, 64))
-    for i in range(8 - popcount, 8):
-        e.append(m2_expr.ExprAssign(float_list[i],
-                                 fill_value))
-    e.append(
-        m2_expr.ExprAssign(float_stack_ptr,
-                        float_stack_ptr - m2_expr.ExprInt(popcount, 3)))
-    return e
-
-# XXX TODO
-
-
-def fcom(_, instr, dst=None, src=None):
-
-    if dst is None and src is None:
-        dst, src = float_st0, float_st1
-    elif src is None:
-        src = mem2double(instr, dst)
-        dst = float_st0
-
-    e = []
-
-    e.append(m2_expr.ExprAssign(float_c0, m2_expr.ExprOp('fcom_c0', dst, src)))
-    e.append(m2_expr.ExprAssign(float_c1, m2_expr.ExprOp('fcom_c1', dst, src)))
-    e.append(m2_expr.ExprAssign(float_c2, m2_expr.ExprOp('fcom_c2', dst, src)))
-    e.append(m2_expr.ExprAssign(float_c3, m2_expr.ExprOp('fcom_c3', dst, src)))
-
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def ftst(_, instr):
-    dst = float_st0
-
-    e = []
-    src = m2_expr.ExprOp('sint_to_fp', m2_expr.ExprInt(0, 64))
-    e.append(m2_expr.ExprAssign(float_c0, m2_expr.ExprOp('fcom_c0', dst, src)))
-    e.append(m2_expr.ExprAssign(float_c1, m2_expr.ExprOp('fcom_c1', dst, src)))
-    e.append(m2_expr.ExprAssign(float_c2, m2_expr.ExprOp('fcom_c2', dst, src)))
-    e.append(m2_expr.ExprAssign(float_c3, m2_expr.ExprOp('fcom_c3', dst, src)))
-
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fxam(ir, instr):
-    """
-    NaN:
-        C3, C2, C0 = 001;
-    Normal:
-        C3, C2, C0 = 010;
-    Infinity:
-        C3, C2, C0 = 011;
-    Zero:
-        C3, C2, C0 = 100;
-    Empty:
-        C3, C2, C0 = 101;
-    Denormal:
-        C3, C2, C0 = 110;
-
-    C1 = sign bit of ST; (* 0 for positive, 1 for negative *)
-    """
-    dst = float_st0
-
-    # Empty not handled
-    locs = {}
-    for name in ["NaN", "Normal", "Infinity", "Zero", "Denormal"]:
-        locs[name] = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_next = ir.get_next_loc_key(instr)
-    loc_next_expr = m2_expr.ExprLoc(loc_next, ir.IRDst.size)
-
-    # if Denormal:
-    #     if zero:
-    #         do_zero
-    #     else:
-    #         do_denormal
-    # else:
-    #     if Nan:
-    #         do_nan
-    #     else:
-    #         if infinity:
-    #             do_infinity
-    #         else:
-    #             do_normal
-
-    irdst = m2_expr.ExprCond(
-        m2_expr.expr_is_IEEE754_denormal(dst),
-        m2_expr.ExprCond(m2_expr.expr_is_IEEE754_zero(dst),
-                 locs["Zero"][1],
-                 locs["Denormal"][1],
-        ),
-        m2_expr.ExprCond(m2_expr.expr_is_NaN(dst),
-                 locs["NaN"][1],
-                 m2_expr.ExprCond(m2_expr.expr_is_infinite(dst),
-                          locs["Infinity"][1],
-                          locs["Normal"][1],
-                 )
-        )
-    )
-    base = [m2_expr.ExprAssign(ir.IRDst, irdst),
-         m2_expr.ExprAssign(float_c1, dst.msb())
-    ]
-    base += set_float_cs_eip(instr)
-
-    out = [
-        IRBlock(ir.loc_db, locs["Zero"][0], [AssignBlock({
-            float_c0: m2_expr.ExprInt(0, float_c0.size),
-            float_c2: m2_expr.ExprInt(0, float_c2.size),
-            float_c3: m2_expr.ExprInt(1, float_c3.size),
-            ir.IRDst: loc_next_expr,
-        }, instr)]),
-        IRBlock(ir.loc_db, locs["Denormal"][0], [AssignBlock({
-            float_c0: m2_expr.ExprInt(0, float_c0.size),
-            float_c2: m2_expr.ExprInt(1, float_c2.size),
-            float_c3: m2_expr.ExprInt(1, float_c3.size),
-            ir.IRDst: loc_next_expr,
-        }, instr)]),
-        IRBlock(ir.loc_db, locs["NaN"][0], [AssignBlock({
-            float_c0: m2_expr.ExprInt(1, float_c0.size),
-            float_c2: m2_expr.ExprInt(0, float_c2.size),
-            float_c3: m2_expr.ExprInt(0, float_c3.size),
-            ir.IRDst: loc_next_expr,
-        }, instr)]),
-        IRBlock(ir.loc_db, locs["Infinity"][0], [AssignBlock({
-            float_c0: m2_expr.ExprInt(1, float_c0.size),
-            float_c2: m2_expr.ExprInt(1, float_c2.size),
-            float_c3: m2_expr.ExprInt(0, float_c3.size),
-            ir.IRDst: loc_next_expr,
-        }, instr)]),
-        IRBlock(ir.loc_db, locs["Normal"][0], [AssignBlock({
-            float_c0: m2_expr.ExprInt(0, float_c0.size),
-            float_c2: m2_expr.ExprInt(1, float_c2.size),
-            float_c3: m2_expr.ExprInt(0, float_c3.size),
-            ir.IRDst: loc_next_expr,
-        }, instr)]),
-    ]
-    return base, out
-
-
-def ficom(_, instr, dst, src=None):
-
-    dst, src = float_implicit_st0(dst, src)
-
-    e = []
-
-    e.append(m2_expr.ExprAssign(float_c0,
-                             m2_expr.ExprOp('fcom_c0', dst,
-                                            src.zeroExtend(dst.size))))
-    e.append(m2_expr.ExprAssign(float_c1,
-                             m2_expr.ExprOp('fcom_c1', dst,
-                                            src.zeroExtend(dst.size))))
-    e.append(m2_expr.ExprAssign(float_c2,
-                             m2_expr.ExprOp('fcom_c2', dst,
-                                            src.zeroExtend(dst.size))))
-    e.append(m2_expr.ExprAssign(float_c3,
-                             m2_expr.ExprOp('fcom_c3', dst,
-                                            src.zeroExtend(dst.size))))
-
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fcomi(_, instr, dst=None, src=None):
-    # TODO unordered float
-    if dst is None and src is None:
-        dst, src = float_st0, float_st1
-    elif src is None:
-        src = dst
-        dst = float_st0
-
-    e = []
-
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprOp('fcom_c0', dst, src)))
-    e.append(m2_expr.ExprAssign(pf, m2_expr.ExprOp('fcom_c2', dst, src)))
-    e.append(m2_expr.ExprAssign(zf, m2_expr.ExprOp('fcom_c3', dst, src)))
-
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprInt(0, 1)))
-    e.append(m2_expr.ExprAssign(nf, m2_expr.ExprInt(0, 1)))
-    e.append(m2_expr.ExprAssign(af, m2_expr.ExprInt(0, 1)))
-
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fcomip(ir, instr, dst=None, src=None):
-    e, extra = fcomi(ir, instr, dst, src)
-    e += float_pop()
-    e += set_float_cs_eip(instr)
-    return e, extra
-
-
-def fucomi(ir, instr, dst=None, src=None):
-    # TODO unordered float
-    return fcomi(ir, instr, dst, src)
-
-
-def fucomip(ir, instr, dst=None, src=None):
-    # TODO unordered float
-    return fcomip(ir, instr, dst, src)
-
-
-def fcomp(ir, instr, dst=None, src=None):
-    e, extra = fcom(ir, instr, dst, src)
-    e += float_pop()
-    e += set_float_cs_eip(instr)
-    return e, extra
-
-
-def fcompp(ir, instr, dst=None, src=None):
-    e, extra = fcom(ir, instr, dst, src)
-    e += float_pop(popcount=2)
-    e += set_float_cs_eip(instr)
-    return e, extra
-
-
-def ficomp(ir, instr, dst, src=None):
-    e, extra = ficom(ir, instr, dst, src)
-    e += float_pop()
-    e += set_float_cs_eip(instr)
-    return e, extra
-
-
-def fucom(ir, instr, dst=None, src=None):
-    # TODO unordered float
-    return fcom(ir, instr, dst, src)
-
-
-def fucomp(ir, instr, dst=None, src=None):
-    # TODO unordered float
-    return fcomp(ir, instr, dst, src)
-
-
-def fucompp(ir, instr, dst=None, src=None):
-    # TODO unordered float
-    return fcompp(ir, instr, dst, src)
-
-
-def comiss(_, instr, dst, src):
-    # TODO unordered float
-
-    e = []
-
-    dst = m2_expr.ExprOp('sint_to_fp', dst[:32])
-    src = m2_expr.ExprOp('sint_to_fp', src[:32])
-
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprOp('fcom_c0', dst, src)))
-    e.append(m2_expr.ExprAssign(pf, m2_expr.ExprOp('fcom_c2', dst, src)))
-    e.append(m2_expr.ExprAssign(zf, m2_expr.ExprOp('fcom_c3', dst, src)))
-
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprInt(0, 1)))
-    e.append(m2_expr.ExprAssign(nf, m2_expr.ExprInt(0, 1)))
-    e.append(m2_expr.ExprAssign(af, m2_expr.ExprInt(0, 1)))
-
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def comisd(_, instr, dst, src):
-    # TODO unordered float
-
-    e = []
-
-    dst = m2_expr.ExprOp('sint_to_fp', dst[:64])
-    src = m2_expr.ExprOp('sint_to_fp', src[:64])
-
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprOp('fcom_c0', dst, src)))
-    e.append(m2_expr.ExprAssign(pf, m2_expr.ExprOp('fcom_c2', dst, src)))
-    e.append(m2_expr.ExprAssign(zf, m2_expr.ExprOp('fcom_c3', dst, src)))
-
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprInt(0, 1)))
-    e.append(m2_expr.ExprAssign(nf, m2_expr.ExprInt(0, 1)))
-    e.append(m2_expr.ExprAssign(af, m2_expr.ExprInt(0, 1)))
-
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fld(_, instr, src):
-
-    if src.size == 32:
-        src = m2_expr.ExprOp("fpconvert_fp64", src)
-    if isinstance(src, m2_expr.ExprMem) and src.size > 64:
-        raise NotImplementedError('convert from 80bits')
-
-    e = []
-    e.append(m2_expr.ExprAssign(float_st7, float_st6))
-    e.append(m2_expr.ExprAssign(float_st6, float_st5))
-    e.append(m2_expr.ExprAssign(float_st5, float_st4))
-    e.append(m2_expr.ExprAssign(float_st4, float_st3))
-    e.append(m2_expr.ExprAssign(float_st3, float_st2))
-    e.append(m2_expr.ExprAssign(float_st2, float_st1))
-    e.append(m2_expr.ExprAssign(float_st1, float_st0))
-    e.append(m2_expr.ExprAssign(float_st0, src))
-    e.append(
-        m2_expr.ExprAssign(float_stack_ptr,
-                        float_stack_ptr + m2_expr.ExprInt(1, 3)))
-
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fst(_, instr, dst):
-    e = []
-
-    if isinstance(dst, m2_expr.ExprMem) and dst.size > 64:
-        raise NotImplementedError('convert to 80bits')
-    src = float_st0
-
-    if dst.size == 32:
-        src = m2_expr.ExprOp("fpconvert_fp32", src)
-    e.append(m2_expr.ExprAssign(dst, src))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fstp(ir, instr, dst):
-    e = []
-
-    if isinstance(dst, m2_expr.ExprMem) and dst.size > 64:
-        raise NotImplementedError('convert to 80bits')
-
-    if isinstance(dst, m2_expr.ExprMem):
-        src = float_st0
-        if dst.size == 32:
-            src = m2_expr.ExprOp("fpconvert_fp32", src)
-        e.append(m2_expr.ExprAssign(dst, src))
-    else:
-        src = float_st0
-        if float_list.index(dst) > 1:
-            # a = st0 -> st0 is dropped
-            # a = st1 -> st0 = st0, useless
-            e.append(m2_expr.ExprAssign(float_prev(dst), src))
-
-    e += set_float_cs_eip(instr)
-    e += float_pop(dst)
-    return e, []
-
-
-def fist(_, instr, dst):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fp_to_sint%d' % dst.size,
-                                                 float_st0)))
-
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fistp(ir, instr, dst):
-    e, extra = fist(ir, instr, dst)
-    e += float_pop(dst)
-    return e, extra
-
-
-def fisttp(_, instr, dst):
-    e = []
-    e.append(m2_expr.ExprAssign(
-        dst,
-        m2_expr.ExprOp('fp_to_sint%d' % dst.size,
-                       m2_expr.ExprOp('fpround_towardszero', float_st0)
-        )))
-
-    e += set_float_cs_eip(instr)
-    e += float_pop(dst)
-    return e, []
-
-
-def fild(ir, instr, src):
-    # XXXXX
-    src = m2_expr.ExprOp('sint_to_fp', src.signExtend(64))
-    e = []
-    e += set_float_cs_eip(instr)
-    e_fld, extra = fld(ir, instr, src)
-    e += e_fld
-    return e, extra
-
-
-def fldz(ir, instr):
-    return fld(ir, instr, m2_expr.ExprOp('sint_to_fp', m2_expr.ExprInt(0, 64)))
-
-
-def fld1(ir, instr):
-    return fld(ir, instr, m2_expr.ExprOp('sint_to_fp', m2_expr.ExprInt(1, 64)))
-
-
-def fldl2t(ir, instr):
-    value_f = math.log(10) / math.log(2)
-    value = struct.unpack('Q', struct.pack('d', value_f))[0]
-    return fld(ir, instr, m2_expr.ExprOp(
-        'sint_to_fp',
-        m2_expr.ExprInt(value, 64)
-    ))
-
-
-def fldpi(ir, instr):
-    value_f = math.pi
-    value = struct.unpack('Q', struct.pack('d', value_f))[0]
-    return fld(ir, instr, m2_expr.ExprOp(
-        'sint_to_fp',
-        m2_expr.ExprInt(value, 64)
-    ))
-
-
-def fldln2(ir, instr):
-    value_f = math.log(2)
-    value = struct.unpack('Q', struct.pack('d', value_f))[0]
-    return fld(ir, instr, m2_expr.ExprOp('mem_64_to_double',
-                                         m2_expr.ExprInt(value, 64)))
-
-
-def fldl2e(ir, instr):
-    x = struct.pack('d', 1 / math.log(2))
-    x = struct.unpack('Q', x)[0]
-    return fld(ir, instr, m2_expr.ExprOp('mem_64_to_double',
-                                         m2_expr.ExprInt(x, 64)))
-
-
-def fldlg2(ir, instr):
-    x = struct.pack('d', math.log10(2))
-    x = struct.unpack('Q', x)[0]
-    return fld(ir, instr, m2_expr.ExprOp('mem_64_to_double',
-                                         m2_expr.ExprInt(x, 64)))
-
-
-def fadd(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fadd', dst, src)))
-
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fiadd(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fiadd', dst, src)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fisub(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fisub', dst, src)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fisubr(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fisub', src, dst)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fpatan(_, instr):
-    e = []
-    a = float_st1
-    e.append(m2_expr.ExprAssign(float_prev(a),
-                             m2_expr.ExprOp('fpatan', float_st0, float_st1)))
-    e += set_float_cs_eip(instr)
-    e += float_pop(a)
-    return e, []
-
-
-def fprem(_, instr):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('fprem', float_st0, float_st1)))
-    # Remaining bits (ex: used in argument reduction in tan)
-    quotient = m2_expr.ExprOp('fp_to_sint32', m2_expr.ExprOp('fpround_towardszero', m2_expr.ExprOp('fdiv', float_st0, float_st1)))
-    e += [m2_expr.ExprAssign(float_c0, quotient[2:3]),
-          m2_expr.ExprAssign(float_c3, quotient[1:2]),
-          m2_expr.ExprAssign(float_c1, quotient[0:1]),
-          # Consider the reduction is always completed
-          m2_expr.ExprAssign(float_c2, m2_expr.ExprInt(0, 1)),
-          ]
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fprem1(_, instr):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('fprem1', float_st0, float_st1)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def faddp(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(float_prev(dst), m2_expr.ExprOp('fadd', dst, src)))
-    e += set_float_cs_eip(instr)
-    e += float_pop(dst)
-    return e, []
-
-
-def fninit(_, instr):
-    e = []
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fyl2x(_, instr):
-    e = []
-    a = float_st1
-    e.append(
-        m2_expr.ExprAssign(float_prev(a), m2_expr.ExprOp('fyl2x', float_st0, float_st1)))
-    e += set_float_cs_eip(instr)
-    e += float_pop(a)
-    return e, []
-
-
-def fnstenv(ir, instr, dst):
-    e = []
-    # XXX TODO tag word, ...
-    status_word = m2_expr.ExprCompose(m2_expr.ExprInt(0, 8),
-                                      float_c0, float_c1, float_c2,
-                                      float_stack_ptr, float_c3,
-                                      m2_expr.ExprInt(0, 1))
-
-    s = instr.mode
-    # The behaviour in 64bit is identical to 32 bit
-    # This will truncate addresses
-    size = min(32, s)
-    ad = ir.ExprMem(dst.ptr, size=16)
-    e.append(m2_expr.ExprAssign(ad, float_control))
-    ad = ir.ExprMem(
-        dst.ptr + m2_expr.ExprInt(
-            (size // 8) * 1,
-            dst.ptr.size
-        ),
-        size=16
-    )
-    e.append(m2_expr.ExprAssign(ad, status_word))
-    ad = ir.ExprMem(
-        dst.ptr + m2_expr.ExprInt(
-            (size // 8) * 3,
-            dst.ptr.size
-        ),
-        size=size
-    )
-    e.append(m2_expr.ExprAssign(ad, float_eip[:size]))
-    ad = ir.ExprMem(
-        dst.ptr + m2_expr.ExprInt(
-            (size // 8) * 4,
-            dst.ptr.size
-        ),
-        size=16
-    )
-    e.append(m2_expr.ExprAssign(ad, float_cs))
-    ad = ir.ExprMem(
-        dst.ptr + m2_expr.ExprInt(
-            (size // 8) * 5,
-            dst.ptr.size
-        ),
-        size=size
-    )
-    e.append(m2_expr.ExprAssign(ad, float_address[:size]))
-    ad = ir.ExprMem(
-        dst.ptr + m2_expr.ExprInt(
-            (size // 8) * 6,
-            dst.ptr.size
-        ),
-        size=16
-    )
-    e.append(m2_expr.ExprAssign(ad, float_ds))
-    return e, []
-
-
-def fldenv(ir, instr, src):
-    e = []
-    # Inspired from fnstenv (same TODOs / issues)
-
-    s = instr.mode
-    # The behaviour in 64bit is identical to 32 bit
-    # This will truncate addresses
-    size = min(32, s)
-
-    # Float control
-    ad = ir.ExprMem(src.ptr, size=16)
-    e.append(m2_expr.ExprAssign(float_control, ad))
-
-    # Status word
-    ad = ir.ExprMem(
-        src.ptr + m2_expr.ExprInt(
-            size // (8 * 1),
-            size=src.ptr.size
-        ),
-        size=16
-    )
-    e += [
-        m2_expr.ExprAssign(x, y) for x, y in ((float_c0, ad[8:9]),
-                                              (float_c1, ad[9:10]),
-                                              (float_c2, ad[10:11]),
-                                              (float_stack_ptr, ad[11:14]),
-                                              (float_c3, ad[14:15]))
-    ]
-
-    # EIP, CS, Address, DS
-    for offset, target in (
-            (3, float_eip[:size]),
-            (4, float_cs),
-            (5, float_address[:size]),
-            (6, float_ds)
-    ):
-        ad = ir.ExprMem(
-            src.ptr + m2_expr.ExprInt(
-                size // ( 8 * offset),
-                size=src.ptr.size
-            ),
-            size=target.size
-        )
-        e.append(m2_expr.ExprAssign(target, ad))
-
-    return e, []
-
-
-def fsub(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fsub', dst, src)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fsubp(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(float_prev(dst), m2_expr.ExprOp('fsub', dst, src)))
-    e += set_float_cs_eip(instr)
-    e += float_pop(dst)
-    return e, []
-
-
-def fsubr(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fsub', src, dst)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fsubrp(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(float_prev(dst), m2_expr.ExprOp('fsub', src, dst)))
-    e += set_float_cs_eip(instr)
-    e += float_pop(dst)
-    return e, []
-
-
-def fmul(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fmul', dst, src)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fimul(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fimul', dst, src)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fdiv(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fdiv', dst, src)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fdivr(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fdiv', src, dst)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fdivrp(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(float_prev(dst), m2_expr.ExprOp('fdiv', src, dst)))
-    e += set_float_cs_eip(instr)
-    e += float_pop(dst)
-    return e, []
-
-
-def fidiv(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fidiv', dst, src)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fidivr(_, instr, dst, src=None):
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('fidiv', src, dst)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fdivp(_, instr, dst, src=None):
-    # Invalid emulation
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(float_prev(dst), m2_expr.ExprOp('fdiv', dst, src)))
-    e += set_float_cs_eip(instr)
-    e += float_pop(dst)
-    return e, []
-
-
-def fmulp(_, instr, dst, src=None):
-    # Invalid emulation
-    dst, src = float_implicit_st0(dst, src)
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(float_prev(dst), m2_expr.ExprOp('fmul', dst, src)))
-    e += set_float_cs_eip(instr)
-    e += float_pop(dst)
-    return e, []
-
-
-def ftan(_, instr, src):
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('ftan', src)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fxch(_, instr, src):
-    e = []
-    src = mem2double(instr, src)
-    e.append(m2_expr.ExprAssign(float_st0, src))
-    e.append(m2_expr.ExprAssign(src, float_st0))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fptan(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(float_st7, float_st6))
-    e.append(m2_expr.ExprAssign(float_st6, float_st5))
-    e.append(m2_expr.ExprAssign(float_st5, float_st4))
-    e.append(m2_expr.ExprAssign(float_st4, float_st3))
-    e.append(m2_expr.ExprAssign(float_st3, float_st2))
-    e.append(m2_expr.ExprAssign(float_st2, float_st1))
-    e.append(m2_expr.ExprAssign(float_st1, m2_expr.ExprOp('ftan', float_st0)))
-    e.append(
-        m2_expr.ExprAssign(
-            float_st0,
-            m2_expr.ExprOp(
-                'sint_to_fp',
-                m2_expr.ExprInt(1, 64)
-            )
-        )
-    )
-    e.append(
-        m2_expr.ExprAssign(float_stack_ptr,
-                        float_stack_ptr + m2_expr.ExprInt(1, 3)))
-    return e, []
-
-
-def frndint(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('frndint', float_st0)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fsin(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('fsin', float_st0)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fcos(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('fcos', float_st0)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fsincos(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(float_st7, float_st6))
-    e.append(m2_expr.ExprAssign(float_st6, float_st5))
-    e.append(m2_expr.ExprAssign(float_st5, float_st4))
-    e.append(m2_expr.ExprAssign(float_st4, float_st3))
-    e.append(m2_expr.ExprAssign(float_st3, float_st2))
-    e.append(m2_expr.ExprAssign(float_st2, float_st1))
-    e.append(m2_expr.ExprAssign(float_st1, m2_expr.ExprOp('fsin', float_st0)))
-    e.append(m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('fcos', float_st0)))
-    e.append(
-        m2_expr.ExprAssign(float_stack_ptr,
-                        float_stack_ptr + m2_expr.ExprInt(1, 3)))
-    return e, []
-
-
-def fscale(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('fscale', float_st0,
-                                                       float_st1)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def f2xm1(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('f2xm1', float_st0)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fchs(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('fchs', float_st0)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fsqrt(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('fsqrt', float_st0)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fabs(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(float_st0, m2_expr.ExprOp('fabs', float_st0)))
-    e += set_float_cs_eip(instr)
-    return e, []
-
-
-def fnstsw(_, instr, dst):
-    args = [
-        # Exceptions -> 0
-        m2_expr.ExprInt(0, 8),
-        float_c0,
-        float_c1,
-        float_c2,
-        float_stack_ptr,
-        float_c3,
-        # B: FPU is not busy -> 0
-        m2_expr.ExprInt(0, 1)]
-    e = [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*args))]
-    return e, []
-
-
-def fnstcw(_, instr, dst):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, float_control))
-    return e, []
-
-
-def fldcw(_, instr, src):
-    e = []
-    e.append(m2_expr.ExprAssign(float_control, src))
-    return e, []
-
-
-def fwait(_, instr):
-    return [], []
-
-
-def fcmovb(ir, instr, arg1, arg2):
-    return gen_fcmov(ir, instr, cf, arg1, arg2, True)
-
-
-def fcmove(ir, instr, arg1, arg2):
-    return gen_fcmov(ir, instr, zf, arg1, arg2, True)
-
-
-def fcmovbe(ir, instr, arg1, arg2):
-    return gen_fcmov(ir, instr, cf | zf, arg1, arg2, True)
-
-
-def fcmovu(ir, instr, arg1, arg2):
-    return gen_fcmov(ir, instr, pf, arg1, arg2, True)
-
-
-def fcmovnb(ir, instr, arg1, arg2):
-    return gen_fcmov(ir, instr, cf, arg1, arg2, False)
-
-
-def fcmovne(ir, instr, arg1, arg2):
-    return gen_fcmov(ir, instr, zf, arg1, arg2, False)
-
-
-def fcmovnbe(ir, instr, arg1, arg2):
-    return gen_fcmov(ir, instr, cf | zf, arg1, arg2, False)
-
-
-def fcmovnu(ir, instr, arg1, arg2):
-    return gen_fcmov(ir, instr, pf, arg1, arg2, False)
-
-
-def nop(_, instr, a=None):
-    return [], []
-
-
-def prefetch0(_, instr, src=None):
-    # see 4-198 on this documentation
-    # https://www-ssl.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf
-    return [], []
-
-
-def prefetch1(_, instr, src=None):
-    # see 4-198 on this documentation
-    # https://www-ssl.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf
-    return [], []
-
-
-def prefetch2(_, instr, src=None):
-    # see 4-198 on this documentation
-    # https://www-ssl.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf
-    return [], []
-
-
-def prefetchw(_, instr, src=None):
-    # see 4-201 on this documentation
-    # https://www-ssl.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf
-    return [], []
-
-def prefetchnta(_, instr, src=None):
-    # see 4-201 on this documentation
-    # https://www-ssl.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf
-    return [], []
-
-
-def lfence(_, instr, src=None):
-    # see 3-485 on this documentation
-    # https://www-ssl.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf
-    return [], []
-
-
-def mfence(_, instr, src=None):
-    # see 3-516 on this documentation
-    # https://www-ssl.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf
-    return [], []
-
-
-def sfence(_, instr, src=None):
-    # see 3-356 on this documentation
-    # https://www-ssl.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf
-    return [], []
-
-
-def ud2(_, instr, src=None):
-    e = [m2_expr.ExprAssign(exception_flags, m2_expr.ExprInt(
-        EXCEPT_ILLEGAL_INSN, exception_flags.size))]
-    return e, []
-
-
-def hlt(_, instr):
-    e = []
-    except_int = EXCEPT_PRIV_INSN
-    e.append(m2_expr.ExprAssign(exception_flags, m2_expr.ExprInt(except_int, 32)))
-    return e, []
-
-
-def rdtsc(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(tsc, tsc + m2_expr.ExprInt(1, 64)))
-    e.append(m2_expr.ExprAssign(mRAX[32], tsc[:32]))
-    e.append(m2_expr.ExprAssign(mRDX[32], tsc[32:]))
-    return e, []
-
-
-def daa(_, instr):
-    e = []
-    r_al = mRAX[instr.mode][:8]
-
-    cond1 = m2_expr.expr_is_unsigned_greater(r_al[:4], m2_expr.ExprInt(0x9, 4)) | af
-    e.append(m2_expr.ExprAssign(af, cond1))
-
-    cond2 = m2_expr.expr_is_unsigned_greater(m2_expr.ExprInt(6, 8), r_al)
-    cond3 = m2_expr.expr_is_unsigned_greater(r_al, m2_expr.ExprInt(0x99, 8)) | cf
-
-    cf_c1 = m2_expr.ExprCond(cond1,
-                             cf | (cond2),
-                             m2_expr.ExprInt(0, 1))
-    new_cf = m2_expr.ExprCond(cond3,
-                              m2_expr.ExprInt(1, 1),
-                              m2_expr.ExprInt(0, 1))
-    e.append(m2_expr.ExprAssign(cf, new_cf))
-
-    al_c1 = m2_expr.ExprCond(cond1,
-                             r_al + m2_expr.ExprInt(6, 8),
-                             r_al)
-
-    new_al = m2_expr.ExprCond(cond3,
-                              al_c1 + m2_expr.ExprInt(0x60, 8),
-                              al_c1)
-    e.append(m2_expr.ExprAssign(r_al, new_al))
-    e += update_flag_znp(new_al)
-    return e, []
-
-
-def das(_, instr):
-    e = []
-    r_al = mRAX[instr.mode][:8]
-
-    cond1 = m2_expr.expr_is_unsigned_greater(r_al[:4], m2_expr.ExprInt(0x9, 4)) | af
-    e.append(m2_expr.ExprAssign(af, cond1))
-
-    cond2 = m2_expr.expr_is_unsigned_greater(m2_expr.ExprInt(6, 8), r_al)
-    cond3 = m2_expr.expr_is_unsigned_greater(r_al, m2_expr.ExprInt(0x99, 8)) | cf
-
-    cf_c1 = m2_expr.ExprCond(cond1,
-                             cf | (cond2),
-                             m2_expr.ExprInt(0, 1))
-    new_cf = m2_expr.ExprCond(cond3,
-                              m2_expr.ExprInt(1, 1),
-                              cf_c1)
-    e.append(m2_expr.ExprAssign(cf, new_cf))
-
-    al_c1 = m2_expr.ExprCond(cond1,
-                             r_al - m2_expr.ExprInt(6, 8),
-                             r_al)
-
-    new_al = m2_expr.ExprCond(cond3,
-                              al_c1 - m2_expr.ExprInt(0x60, 8),
-                              al_c1)
-    e.append(m2_expr.ExprAssign(r_al, new_al))
-    e += update_flag_znp(new_al)
-    return e, []
-
-
-def aam(ir, instr, src):
-    e = []
-    assert src.is_int()
-
-    value = int(src)
-    if value:
-        tempAL = mRAX[instr.mode][0:8]
-        newEAX = m2_expr.ExprCompose(
-            m2_expr.ExprOp("umod", tempAL, src),
-            m2_expr.ExprOp("udiv", tempAL, src),
-            mRAX[instr.mode][16:]
-        )
-        e += [m2_expr.ExprAssign(mRAX[instr.mode], newEAX)]
-        e += update_flag_arith(newEAX)
-        e.append(m2_expr.ExprAssign(af, m2_expr.ExprInt(0, 1)))
-    else:
-        e.append(
-            m2_expr.ExprAssign(
-                exception_flags,
-                m2_expr.ExprInt(EXCEPT_DIV_BY_ZERO, exception_flags.size)
-            )
-        )
-    return e, []
-
-
-def aad(_, instr, src):
-    e = []
-    tempAL = mRAX[instr.mode][0:8]
-    tempAH = mRAX[instr.mode][8:16]
-    newEAX = m2_expr.ExprCompose((tempAL + (tempAH * src)) & m2_expr.ExprInt(0xFF, 8),
-                                 m2_expr.ExprInt(0, 8),
-                                 mRAX[instr.mode][16:])
-    e += [m2_expr.ExprAssign(mRAX[instr.mode], newEAX)]
-    e += update_flag_arith(newEAX)
-    e.append(m2_expr.ExprAssign(af, m2_expr.ExprInt(0, 1)))
-    return e, []
-
-
-def _tpl_aaa(_, instr, op):
-    """Templating for aaa, aas with operation @op
-    @op: operation to apply
-    """
-    e = []
-    r_al = mRAX[instr.mode][:8]
-    r_ah = mRAX[instr.mode][8:16]
-    r_ax = mRAX[instr.mode][:16]
-    i0 = m2_expr.ExprInt(0, 1)
-    i1 = m2_expr.ExprInt(1, 1)
-    # cond: if (al & 0xf) > 9 OR af == 1
-    cond = (r_al & m2_expr.ExprInt(0xf, 8)) - m2_expr.ExprInt(9, 8)
-    cond = ~cond.msb() & m2_expr.ExprCond(cond, i1, i0)
-    cond |= af & i1
-
-    to_add = m2_expr.ExprInt(0x106, size=r_ax.size)
-    if op == "-":
-        # Avoid ExprOp("-", A, B), should be ExprOp("+", A, ExprOp("-", B))
-        first_part = r_ax - to_add
-    else:
-        first_part = m2_expr.ExprOp(op, r_ax, to_add)
-    new_ax = first_part & m2_expr.ExprInt(0xff0f,
-                                          size=r_ax.size)
-    # set AL
-    e.append(m2_expr.ExprAssign(r_ax, m2_expr.ExprCond(cond, new_ax, r_ax)))
-    e.append(m2_expr.ExprAssign(af, cond))
-    e.append(m2_expr.ExprAssign(cf, cond))
-    return e, []
-
-
-def aaa(ir, instr):
-    return _tpl_aaa(ir, instr, "+")
-
-
-def aas(ir, instr):
-    return _tpl_aaa(ir, instr, "-")
-
-
-def bsr_bsf(ir, instr, dst, src, op_func):
-    """
-    IF SRC == 0
-        ZF = 1
-        DEST is left unchanged
-    ELSE
-        ZF = 0
-        DEST = @op_func(SRC)
-    """
-    loc_src_null, loc_src_null_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_src_not_null, loc_src_not_null_expr = ir.gen_loc_key_and_expr(ir.IRDst.size)
-    loc_next = ir.get_next_loc_key(instr)
-    loc_next_expr = m2_expr.ExprLoc(loc_next, ir.IRDst.size)
-
-    aff_dst = m2_expr.ExprAssign(ir.IRDst, loc_next_expr)
-    e = [m2_expr.ExprAssign(ir.IRDst, m2_expr.ExprCond(src,
-                                                    loc_src_not_null_expr,
-                                                    loc_src_null_expr))]
-    e_src_null = []
-    e_src_null.append(m2_expr.ExprAssign(zf, m2_expr.ExprInt(1, zf.size)))
-    # XXX destination is undefined
-    e_src_null.append(aff_dst)
-
-    e_src_not_null = []
-    e_src_not_null.append(m2_expr.ExprAssign(zf, m2_expr.ExprInt(0, zf.size)))
-    e_src_not_null.append(m2_expr.ExprAssign(dst, op_func(src)))
-    e_src_not_null.append(aff_dst)
-
-    return e, [IRBlock(ir.loc_db, loc_src_null, [AssignBlock(e_src_null, instr)]),
-               IRBlock(ir.loc_db, loc_src_not_null, [AssignBlock(e_src_not_null, instr)])]
-
-
-def bsf(ir, instr, dst, src):
-    return bsr_bsf(ir, instr, dst, src,
-                   lambda src: m2_expr.ExprOp("cnttrailzeros", src))
-
-
-def bsr(ir, instr, dst, src):
-    return bsr_bsf(
-        ir, instr, dst, src,
-        lambda src: m2_expr.ExprInt(src.size - 1, src.size) - m2_expr.ExprOp("cntleadzeros", src)
-    )
-
-
-def arpl(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(exception_flags, m2_expr.ExprInt(1 << 7, 32)))
-    return e, []
-
-
-def ins(_, instr, size):
-    e = []
-    e.append(m2_expr.ExprAssign(exception_flags, m2_expr.ExprInt(1 << 7, 32)))
-    return e, []
-
-
-def sidt(ir, instr, dst):
-    e = []
-    if not isinstance(dst, m2_expr.ExprMem) or dst.size != 32:
-        raise ValueError('not exprmem 32bit instance!!')
-    ptr = dst.ptr
-    LOG_X86_SEM.warning("DEFAULT SIDT ADDRESS %s!!", dst)
-    e.append(m2_expr.ExprAssign(ir.ExprMem(ptr, 32),
-                             m2_expr.ExprInt(0xe40007ff, 32)))
-    e.append(
-        m2_expr.ExprAssign(ir.ExprMem(ptr + m2_expr.ExprInt(4, ptr.size), 16),
-                        m2_expr.ExprInt(0x8245, 16)))
-    return e, []
-
-
-def sldt(_, instr, dst):
-    LOG_X86_SEM.warning("DEFAULT SLDT ADDRESS %s!!", dst)
-    e = [m2_expr.ExprAssign(dst, m2_expr.ExprInt(0, dst.size))]
-    return e, []
-
-
-def cmovz(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, zf, dst, src, True)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_EQ", zf), dst, src, True)
-
-
-def cmovnz(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, zf, dst, src, False)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_EQ", zf), dst, src, False)
-
-
-def cmovpe(ir, instr, dst, src):
-    return gen_cmov(ir, instr, pf, dst, src, True)
-
-
-def cmovnp(ir, instr, dst, src):
-    return gen_cmov(ir, instr, pf, dst, src, False)
-
-
-def cmovge(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, nf ^ of, dst, src, False)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_S>=", nf, of), dst, src, True)
-
-
-def cmovg(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, zf | (nf ^ of), dst, src, False)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_S>", nf, of, zf), dst, src, True)
-
-
-def cmovl(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, nf ^ of, dst, src, True)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_S<", nf, of), dst, src, True)
-
-
-def cmovle(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, zf | (nf ^ of), dst, src, True)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_S<=", nf, of, zf), dst, src, True)
-
-
-def cmova(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, cf | zf, dst, src, False)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_U>", cf, zf), dst, src, True)
-
-
-def cmovae(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, cf, dst, src, False)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_U>=", cf), dst, src, True)
-
-
-def cmovbe(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, cf | zf, dst, src, True)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_U<=", cf, zf), dst, src, True)
-
-
-def cmovb(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, cf, dst, src, True)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_U<", cf), dst, src, True)
-
-
-def cmovo(ir, instr, dst, src):
-    return gen_cmov(ir, instr, of, dst, src, True)
-
-
-def cmovno(ir, instr, dst, src):
-    return gen_cmov(ir, instr, of, dst, src, False)
-
-
-def cmovs(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, nf, dst, src, True)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_NEG", nf), dst, src, True)
-
-
-def cmovns(ir, instr, dst, src):
-    #return gen_cmov(ir, instr, nf, dst, src, False)
-    return gen_cmov(ir, instr, m2_expr.ExprOp("CC_NEG", nf), dst, src, False)
-
-
-def icebp(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(exception_flags,
-                             m2_expr.ExprInt(EXCEPT_SOFT_BP, 32)))
-    return e, []
-# XXX
-
-
-def l_int(_, instr, src):
-    e = []
-    # XXX
-    assert src.is_int()
-    value = int(src)
-    if value == 1:
-        except_int = EXCEPT_INT_1
-    elif value == 3:
-        except_int = EXCEPT_SOFT_BP
-    else:
-        except_int = EXCEPT_INT_XX
-    e.append(m2_expr.ExprAssign(exception_flags,
-                             m2_expr.ExprInt(except_int, 32)))
-    e.append(m2_expr.ExprAssign(interrupt_num, src))
-    return e, []
-
-
-def l_sysenter(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(exception_flags,
-                             m2_expr.ExprInt(EXCEPT_PRIV_INSN, 32)))
-    return e, []
-
-
-def l_syscall(_, instr):
-    e = []
-    e.append(m2_expr.ExprAssign(exception_flags,
-                             m2_expr.ExprInt(EXCEPT_SYSCALL, 32)))
-    return e, []
-
-# XXX
-
-
-def l_out(_, instr, src1, src2):
-    e = []
-    e.append(m2_expr.ExprAssign(exception_flags,
-                             m2_expr.ExprInt(EXCEPT_PRIV_INSN, 32)))
-    return e, []
-
-# XXX
-
-
-def l_outs(_, instr, size):
-    e = []
-    e.append(m2_expr.ExprAssign(exception_flags,
-                             m2_expr.ExprInt(EXCEPT_PRIV_INSN, 32)))
-    return e, []
-
-# XXX actually, xlat performs al = (ds:[e]bx + ZeroExtend(al))
-
-
-def xlat(ir, instr):
-    e = []
-    ptr = mRAX[instr.mode][0:8].zeroExtend(mRBX[instr.mode].size)
-    src = ir.ExprMem(mRBX[instr.mode] + ptr, 8)
-    e.append(m2_expr.ExprAssign(mRAX[instr.mode][0:8], src))
-    return e, []
-
-
-def cpuid(_, instr):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(mRAX[instr.mode],
-                        m2_expr.ExprOp('x86_cpuid', mRAX[instr.mode], m2_expr.ExprInt(0, instr.mode))))
-    e.append(
-        m2_expr.ExprAssign(mRBX[instr.mode],
-                        m2_expr.ExprOp('x86_cpuid', mRAX[instr.mode], m2_expr.ExprInt(1, instr.mode))))
-    e.append(
-        m2_expr.ExprAssign(mRCX[instr.mode],
-                        m2_expr.ExprOp('x86_cpuid', mRAX[instr.mode], m2_expr.ExprInt(2, instr.mode))))
-    e.append(
-        m2_expr.ExprAssign(mRDX[instr.mode],
-                        m2_expr.ExprOp('x86_cpuid', mRAX[instr.mode], m2_expr.ExprInt(3, instr.mode))))
-    return e, []
-
-
-def bittest_get(ir, instr, src, index):
-    index = index.zeroExtend(src.size)
-    if isinstance(src, m2_expr.ExprMem):
-        b_mask = {16: 4, 32: 5, 64: 6}
-        b_decal = {16: 1, 32: 3, 64: 7}
-        ptr = src.ptr
-        segm = is_mem_segm(src)
-        if segm:
-            ptr = ptr.args[1]
-
-        off_bit = index.zeroExtend(
-            src.size) & m2_expr.ExprInt((1 << b_mask[src.size]) - 1,
-                                        src.size)
-        off_byte = ((index.zeroExtend(ptr.size) >> m2_expr.ExprInt(3, ptr.size)) &
-                    m2_expr.ExprInt(((1 << src.size) - 1) ^ b_decal[src.size], ptr.size))
-
-        addr = ptr + off_byte
-        if segm:
-            addr = ir.gen_segm_expr(src.ptr.args[0], addr)
-
-        d = ir.ExprMem(addr, src.size)
-    else:
-        off_bit = m2_expr.ExprOp(
-            '&', index, m2_expr.ExprInt(src.size - 1, src.size))
-        d = src
-    return d, off_bit
-
-
-def bt(ir, instr, src, index):
-    e = []
-    index = index.zeroExtend(src.size)
-    d, off_bit = bittest_get(ir, instr, src, index)
-    d = d >> off_bit
-    e.append(m2_expr.ExprAssign(cf, d[:1]))
-    return e, []
-
-
-def btc(ir, instr, src, index):
-    e = []
-    d, off_bit = bittest_get(ir, instr, src, index)
-    e.append(m2_expr.ExprAssign(cf, (d >> off_bit)[:1]))
-
-    m = m2_expr.ExprInt(1, src.size) << off_bit
-    e.append(m2_expr.ExprAssign(d, d ^ m))
-
-    return e, []
-
-
-def bts(ir, instr, src, index):
-    e = []
-    d, off_bit = bittest_get(ir, instr, src, index)
-    e.append(m2_expr.ExprAssign(cf, (d >> off_bit)[:1]))
-    m = m2_expr.ExprInt(1, src.size) << off_bit
-    e.append(m2_expr.ExprAssign(d, d | m))
-
-    return e, []
-
-
-def btr(ir, instr, src, index):
-    e = []
-    d, off_bit = bittest_get(ir, instr, src, index)
-    e.append(m2_expr.ExprAssign(cf, (d >> off_bit)[:1]))
-    m = ~(m2_expr.ExprInt(1, src.size) << off_bit)
-    e.append(m2_expr.ExprAssign(d, d & m))
-
-    return e, []
-
-
-def into(_, instr):
-    return [], []
-
-
-def l_in(_, instr, src1, src2):
-    e = []
-    e.append(m2_expr.ExprAssign(exception_flags,
-                             m2_expr.ExprInt(EXCEPT_PRIV_INSN, 32)))
-    return e, []
-
-
-@sbuild.parse
-def cmpxchg(arg1, arg2):
-    accumulator = mRAX[instr.v_opmode()][:arg1.size]
-    if (accumulator - arg1):
-        zf = i1(0)
-        accumulator = arg1
-    else:
-        zf = i1(1)
-        arg1 = arg2
-
-
-@sbuild.parse
-def cmpxchg8b(arg1):
-    accumulator = {mRAX[32], mRDX[32]}
-    if accumulator - arg1:
-        zf = i1(0)
-        mRAX[32] = arg1[:32]
-        mRDX[32] = arg1[32:]
-    else:
-        zf = i1(1)
-        arg1 = {mRBX[32], mRCX[32]}
-
-
-@sbuild.parse
-def cmpxchg16b(arg1):
-    accumulator = {mRAX[64], mRDX[64]}
-    if accumulator - arg1:
-        zf = i1(0)
-        mRAX[64] = arg1[:64]
-        mRDX[64] = arg1[64:]
-    else:
-        zf = i1(1)
-        arg1 = {mRBX[64], mRCX[64]}
-
-
-def lds(ir, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, ir.ExprMem(src.ptr, size=dst.size)))
-    DS_value = ir.ExprMem(src.ptr + m2_expr.ExprInt(dst.size // 8, src.ptr.size),
-                          size=16)
-    e.append(m2_expr.ExprAssign(DS, DS_value))
-    return e, []
-
-
-def les(ir, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, ir.ExprMem(src.ptr, size=dst.size)))
-    ES_value = ir.ExprMem(src.ptr + m2_expr.ExprInt(dst.size // 8, src.ptr.size),
-                          size=16)
-    e.append(m2_expr.ExprAssign(ES, ES_value))
-    return e, []
-
-
-def lss(ir, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, ir.ExprMem(src.ptr, size=dst.size)))
-    SS_value = ir.ExprMem(src.ptr + m2_expr.ExprInt(dst.size // 8, src.ptr.size),
-                          size=16)
-    e.append(m2_expr.ExprAssign(SS, SS_value))
-    return e, []
-
-
-def lfs(ir, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, ir.ExprMem(src.ptr, size=dst.size)))
-    FS_value = ir.ExprMem(src.ptr + m2_expr.ExprInt(dst.size // 8, src.ptr.size),
-                          size=16)
-    e.append(m2_expr.ExprAssign(FS, FS_value))
-    return e, []
-
-
-def lgs(ir, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, ir.ExprMem(src.ptr, size=dst.size)))
-    GS_value = ir.ExprMem(src.ptr + m2_expr.ExprInt(dst.size // 8, src.ptr.size),
-                          size=16)
-    e.append(m2_expr.ExprAssign(GS, GS_value))
-    return e, []
-
-
-def lahf(_, instr):
-    e = []
-    args = [cf, m2_expr.ExprInt(1, 1), pf, m2_expr.ExprInt(0, 1), af,
-            m2_expr.ExprInt(0, 1), zf, nf]
-    e.append(
-        m2_expr.ExprAssign(mRAX[instr.mode][8:16], m2_expr.ExprCompose(*args)))
-    return e, []
-
-
-def sahf(_, instr):
-    tmp = mRAX[instr.mode][8:16]
-    e = []
-    e.append(m2_expr.ExprAssign(cf, tmp[0:1]))
-    e.append(m2_expr.ExprAssign(pf, tmp[2:3]))
-    e.append(m2_expr.ExprAssign(af, tmp[4:5]))
-    e.append(m2_expr.ExprAssign(zf, tmp[6:7]))
-    e.append(m2_expr.ExprAssign(nf, tmp[7:8]))
-    return e, []
-
-
-def lar(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('access_segment', src)))
-    e.append(m2_expr.ExprAssign(zf, m2_expr.ExprOp('access_segment_ok', src)))
-    return e, []
-
-
-def lsl(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('load_segment_limit', src)))
-    e.append(m2_expr.ExprAssign(zf, m2_expr.ExprOp('load_segment_limit_ok', src)))
-    return e, []
-
-
-def fclex(_, instr):
-    # XXX TODO
-    return [], []
-
-
-def fnclex(_, instr):
-    # XXX TODO
-    return [], []
-
-
-def l_str(_, instr, dst):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('load_tr_segment_selector',
-                                                 m2_expr.ExprInt(0, 32))))
-    return e, []
-
-
-def movd(_, instr, dst, src):
-    e = []
-    if dst in regs_mm_expr:
-        e.append(m2_expr.ExprAssign(
-            dst, m2_expr.ExprCompose(src, m2_expr.ExprInt(0, 32))))
-    elif dst in regs_xmm_expr:
-        e.append(m2_expr.ExprAssign(
-            dst, m2_expr.ExprCompose(src, m2_expr.ExprInt(0, 96))))
-    else:
-        e.append(m2_expr.ExprAssign(dst, src[:32]))
-    return e, []
-
-
-def movdqu(_, instr, dst, src):
-    # XXX TODO alignment check
-    return [m2_expr.ExprAssign(dst, src)], []
-
-
-def movapd(_, instr, dst, src):
-    # XXX TODO alignment check
-    return [m2_expr.ExprAssign(dst, src)], []
-
-
-def andps(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('&', dst, src)))
-    return e, []
-
-
-def andnps(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('&', dst ^ dst.mask, src)))
-    return e, []
-
-
-def orps(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('|', dst, src)))
-    return e, []
-
-
-def xorps(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprOp('^', dst, src)))
-    return e, []
-
-
-def rdmsr(ir, instr):
-    e = [m2_expr.ExprAssign(exception_flags,m2_expr.ExprInt(EXCEPT_PRIV_INSN, 32))]
-    return e, []
-
-
-def wrmsr(ir, instr):
-    e = [m2_expr.ExprAssign(exception_flags,m2_expr.ExprInt(EXCEPT_PRIV_INSN, 32))]
-    return e, []
-
-# MMX/SSE/AVX operations
-#
-
-def vec_op_clip(op, size, callback=None):
-    """
-    Generate simd operations
-    @op: the operator
-    @size: size of an element
-    """
-    def vec_op_clip_instr(ir, instr, dst, src):
-        if op == '-':
-            result = dst[:size] - src[:size]
-        else:
-            result = m2_expr.ExprOp(op, dst[:size], src[:size])
-        if callback is not None:
-            result = callback(result)
-        return [m2_expr.ExprAssign(dst[:size], result)], []
-    return vec_op_clip_instr
-
-# Generic vertical operation
-
-
-def vec_vertical_sem(op, elt_size, reg_size, dst, src, apply_on_output):
-    assert reg_size % elt_size == 0
-    n = reg_size // elt_size
-    if op == '-':
-        ops = [
-            apply_on_output((dst[i * elt_size:(i + 1) * elt_size]
-                             - src[i * elt_size:(i + 1) * elt_size]))
-            for i in range(0, n)
-        ]
-    else:
-        ops = [
-            apply_on_output(m2_expr.ExprOp(op, dst[i * elt_size:(i + 1) * elt_size],
-                                           src[i * elt_size:(i + 1) * elt_size]))
-            for i in range(0, n)
-        ]
-
-    return m2_expr.ExprCompose(*ops)
-
-
-def __vec_vertical_instr_gen(op, elt_size, sem, apply_on_output):
-    def vec_instr(ir, instr, dst, src):
-        e = []
-        if isinstance(src, m2_expr.ExprMem):
-            src = ir.ExprMem(src.ptr, dst.size)
-        reg_size = dst.size
-        e.append(m2_expr.ExprAssign(dst, sem(op, elt_size, reg_size, dst, src,
-                                          apply_on_output)))
-        return e, []
-    return vec_instr
-
-
-def vec_vertical_instr(op, elt_size, apply_on_output=lambda x: x):
-    return __vec_vertical_instr_gen(op, elt_size, vec_vertical_sem,
-                                    apply_on_output)
-
-
-def _keep_mul_high(expr, signed=False):
-    assert expr.is_op("*") and len(expr.args) == 2
-
-    if signed:
-        arg1 = expr.args[0].signExtend(expr.size * 2)
-        arg2 = expr.args[1].signExtend(expr.size * 2)
-    else:
-        arg1 = expr.args[0].zeroExtend(expr.size * 2)
-        arg2 = expr.args[1].zeroExtend(expr.size * 2)
-    return m2_expr.ExprOp("*", arg1, arg2)[expr.size:]
-
-# Op, signed => associated comparison
-_min_max_func = {
-    ("min", False): m2_expr.expr_is_unsigned_lower,
-    ("min", True): m2_expr.expr_is_signed_lower,
-    ("max", False): m2_expr.expr_is_unsigned_greater,
-    ("max", True): m2_expr.expr_is_signed_greater,
-}
-def _min_max(expr, signed):
-    assert (expr.is_op("min") or expr.is_op("max")) and len(expr.args) == 2
-    return m2_expr.ExprCond(
-        _min_max_func[(expr.op, signed)](expr.args[1], expr.args[0]),
-        expr.args[1],
-        expr.args[0],
-    )
-
-def _float_min_max(expr):
-    assert (expr.is_op("fmin") or expr.is_op("fmax")) and len(expr.args) == 2
-    src1 = expr.args[0]
-    src2 = expr.args[1]
-    if expr.is_op("fmin"):
-        comp = m2_expr.expr_is_float_lower(src1, src2)
-    elif expr.is_op("fmax"):
-        comp = m2_expr.expr_is_float_lower(src2, src1)
-
-    # x86 documentation (for MIN):
-    # IF ((SRC1 = 0.0) and (SRC2 = 0.0)) THEN DEST <-SRC2;
-    # ELSE IF (SRC1 = SNaN) THEN DEST <-SRC2; FI;
-    # ELSE IF (SRC2 = SNaN) THEN DEST <-SRC2; FI;
-    # ELSE IF (SRC1 < SRC2) THEN DEST <-SRC1;
-    # ELSE DEST<-SRC2;
-    #
-    # But this includes the NaN output of "SRC1 < SRC2"
-    # Associated text is more detailed, and this is the version impl here
-    return m2_expr.ExprCond(
-        m2_expr.expr_is_sNaN(src2), src2,
-        m2_expr.ExprCond(
-            m2_expr.expr_is_NaN(src2) | m2_expr.expr_is_NaN(src1), src2,
-            m2_expr.ExprCond(comp, src1, src2)
-        )
-    )
-
-
-# Integer arithmetic
-#
-
-# Additions
-#
-
-# SSE
-paddb = vec_vertical_instr('+', 8)
-paddw = vec_vertical_instr('+', 16)
-paddd = vec_vertical_instr('+', 32)
-paddq = vec_vertical_instr('+', 64)
-
-# Substractions
-#
-
-# SSE
-psubb = vec_vertical_instr('-', 8)
-psubw = vec_vertical_instr('-', 16)
-psubd = vec_vertical_instr('-', 32)
-psubq = vec_vertical_instr('-', 64)
-
-# Multiplications
-#
-
-# SSE
-pmullb = vec_vertical_instr('*', 8)
-pmullw = vec_vertical_instr('*', 16)
-pmulld = vec_vertical_instr('*', 32)
-pmullq = vec_vertical_instr('*', 64)
-pmulhub = vec_vertical_instr('*', 8, _keep_mul_high)
-pmulhuw = vec_vertical_instr('*', 16, _keep_mul_high)
-pmulhud = vec_vertical_instr('*', 32, _keep_mul_high)
-pmulhuq = vec_vertical_instr('*', 64, _keep_mul_high)
-pmulhb = vec_vertical_instr('*', 8, lambda x: _keep_mul_high(x, signed=True))
-pmulhw = vec_vertical_instr('*', 16, lambda x: _keep_mul_high(x, signed=True))
-pmulhd = vec_vertical_instr('*', 32, lambda x: _keep_mul_high(x, signed=True))
-pmulhq = vec_vertical_instr('*', 64, lambda x: _keep_mul_high(x, signed=True))
-
-def pmuludq(ir, instr, dst, src):
-    e = []
-    if dst.size == 64:
-        e.append(m2_expr.ExprAssign(
-            dst,
-            src[:32].zeroExtend(64) * dst[:32].zeroExtend(64)
-        ))
-    elif dst.size == 128:
-        e.append(m2_expr.ExprAssign(
-            dst[:64],
-            src[:32].zeroExtend(64) * dst[:32].zeroExtend(64)
-        ))
-        e.append(m2_expr.ExprAssign(
-            dst[64:],
-            src[64:96].zeroExtend(64) * dst[64:96].zeroExtend(64)
-        ))
-    else:
-        raise RuntimeError("Unsupported size %d" % dst.size)
-    return e, []
-
-# Mix
-#
-
-# SSE
-def pmaddwd(ir, instr, dst, src):
-    sizedst = 32
-    sizesrc = 16
-    out = []
-    for start in range(0, dst.size, sizedst):
-        base = start
-        mul1 = src[base: base + sizesrc].signExtend(sizedst) * dst[base: base + sizesrc].signExtend(sizedst)
-        base += sizesrc
-        mul2 = src[base: base + sizesrc].signExtend(sizedst) * dst[base: base + sizesrc].signExtend(sizedst)
-        out.append(mul1 + mul2)
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out))], []
-
-
-def _absolute(expr):
-    """Return abs(@expr)"""
-    signed = expr.msb()
-    value_unsigned = (expr ^ expr.mask) + m2_expr.ExprInt(1, expr.size)
-    return m2_expr.ExprCond(signed, value_unsigned, expr)
-
-
-def psadbw(ir, instr, dst, src):
-    sizedst = 16
-    sizesrc = 8
-    out_dst = []
-    for start in range(0, dst.size, 64):
-        out = []
-        for src_start in range(0, 64, sizesrc):
-            beg = start + src_start
-            end = beg + sizesrc
-            # Not clear in the doc equations, but in the text, src and dst are:
-            # "8 unsigned byte integers"
-            out.append(_absolute(dst[beg: end].zeroExtend(sizedst) - src[beg: end].zeroExtend(sizedst)))
-        out_dst.append(m2_expr.ExprOp("+", *out))
-        out_dst.append(m2_expr.ExprInt(0, 64 - sizedst))
-
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out_dst))], []
-
-def _average(expr):
-    assert expr.is_op("avg") and len(expr.args) == 2
-
-    arg1 = expr.args[0].zeroExtend(expr.size * 2)
-    arg2 = expr.args[1].zeroExtend(expr.size * 2)
-    one = m2_expr.ExprInt(1, arg1.size)
-    # avg(unsigned) = (a + b + 1) >> 1, addition being at least on one more bit
-    return ((arg1 + arg2 + one) >> one)[:expr.size]
-
-pavgb = vec_vertical_instr('avg', 8, _average)
-pavgw = vec_vertical_instr('avg', 16, _average)
-
-# Comparisons
-#
-
-# SSE
-pminsw = vec_vertical_instr('min', 16, lambda x: _min_max(x, signed=True))
-pminub = vec_vertical_instr('min', 8, lambda x: _min_max(x, signed=False))
-pminuw = vec_vertical_instr('min', 16, lambda x: _min_max(x, signed=False))
-pminud = vec_vertical_instr('min', 32, lambda x: _min_max(x, signed=False))
-pmaxub = vec_vertical_instr('max', 8, lambda x: _min_max(x, signed=False))
-pmaxuw = vec_vertical_instr('max', 16, lambda x: _min_max(x, signed=False))
-pmaxud = vec_vertical_instr('max', 32, lambda x: _min_max(x, signed=False))
-pmaxsw = vec_vertical_instr('max', 16, lambda x: _min_max(x, signed=True))
-
-# Floating-point arithmetic
-#
-
-# SSE
-addss = vec_op_clip('fadd', 32)
-addsd = vec_op_clip('fadd', 64)
-addps = vec_vertical_instr('fadd', 32)
-addpd = vec_vertical_instr('fadd', 64)
-subss = vec_op_clip('fsub', 32)
-subsd = vec_op_clip('fsub', 64)
-subps = vec_vertical_instr('fsub', 32)
-subpd = vec_vertical_instr('fsub', 64)
-mulss = vec_op_clip('fmul', 32)
-mulsd = vec_op_clip('fmul', 64)
-mulps = vec_vertical_instr('fmul', 32)
-mulpd = vec_vertical_instr('fmul', 64)
-divss = vec_op_clip('fdiv', 32)
-divsd = vec_op_clip('fdiv', 64)
-divps = vec_vertical_instr('fdiv', 32)
-divpd = vec_vertical_instr('fdiv', 64)
-
-# Comparisons (floating-point)
-
-minps = vec_vertical_instr('fmin', 32, _float_min_max)
-minpd = vec_vertical_instr('fmin', 64, _float_min_max)
-minss = vec_op_clip('fmin', 32, _float_min_max)
-minsd = vec_op_clip('fmin', 64, _float_min_max)
-maxps = vec_vertical_instr('fmax', 32, _float_min_max)
-maxpd = vec_vertical_instr('fmax', 64, _float_min_max)
-maxss = vec_op_clip('fmax', 32, _float_min_max)
-maxsd = vec_op_clip('fmax', 64, _float_min_max)
-
-def _float_compare_to_mask(expr):
-    if expr.op == 'unord':
-        to_ext = m2_expr.expr_is_NaN(expr.args[0]) | m2_expr.expr_is_NaN(expr.args[1])
-    elif expr.op == 'ord':
-        to_ext = ~m2_expr.expr_is_NaN(expr.args[0]) & ~m2_expr.expr_is_NaN(expr.args[1])
-    else:
-        if expr.op == '==fu':
-            to_ext = m2_expr.expr_is_float_equal(expr.args[0], expr.args[1])
-            on_NaN = m2_expr.ExprInt(0, 1)
-        elif expr.op == '<fu':
-            to_ext = m2_expr.expr_is_float_lower(expr.args[0], expr.args[1])
-            on_NaN = m2_expr.ExprInt(0, 1)
-        elif expr.op == '<=fu':
-            to_ext = (m2_expr.expr_is_float_equal(expr.args[0], expr.args[1]) |
-                      m2_expr.expr_is_float_lower(expr.args[0], expr.args[1]))
-            on_NaN = m2_expr.ExprInt(0, 1)
-        elif expr.op == '!=fu':
-            to_ext = ~m2_expr.expr_is_float_equal(expr.args[0], expr.args[1])
-            on_NaN = m2_expr.ExprInt(1, 1)
-        elif expr.op == '!<fu':
-            to_ext = ~m2_expr.expr_is_float_lower(expr.args[0], expr.args[1])
-            on_NaN = m2_expr.ExprInt(1, 1)
-        elif expr.op == '!<=fu':
-            to_ext = ~(m2_expr.expr_is_float_equal(expr.args[0], expr.args[1]) |
-                      m2_expr.expr_is_float_lower(expr.args[0], expr.args[1]))
-            on_NaN = m2_expr.ExprInt(1, 1)
-
-        to_ext = m2_expr.ExprCond(
-            m2_expr.expr_is_NaN(expr.args[0]) | m2_expr.expr_is_NaN(expr.args[1]),
-            on_NaN,
-            to_ext
-        )
-    return to_ext.signExtend(expr.size)
-
-cmpeqps = vec_vertical_instr('==fu', 32, lambda x: _float_compare_to_mask(x))
-cmpeqpd = vec_vertical_instr('==fu', 64, lambda x: _float_compare_to_mask(x))
-cmpeqss = vec_op_clip('==fu', 32, lambda x: _float_compare_to_mask(x))
-cmpeqsd = vec_op_clip('==fu', 64, lambda x: _float_compare_to_mask(x))
-cmpltps = vec_vertical_instr('<fu', 32, lambda x: _float_compare_to_mask(x))
-cmpltpd = vec_vertical_instr('<fu', 64, lambda x: _float_compare_to_mask(x))
-cmpltss = vec_op_clip('<fu', 32, lambda x: _float_compare_to_mask(x))
-cmpltsd = vec_op_clip('<fu', 64, lambda x: _float_compare_to_mask(x))
-cmpleps = vec_vertical_instr('<=fu', 32, lambda x: _float_compare_to_mask(x))
-cmplepd = vec_vertical_instr('<=fu', 64, lambda x: _float_compare_to_mask(x))
-cmpless = vec_op_clip('<=fu', 32, lambda x: _float_compare_to_mask(x))
-cmplesd = vec_op_clip('<=fu', 64, lambda x: _float_compare_to_mask(x))
-cmpunordps = vec_vertical_instr('unord', 32, lambda x: _float_compare_to_mask(x))
-cmpunordpd = vec_vertical_instr('unord', 64, lambda x: _float_compare_to_mask(x))
-cmpunordss = vec_op_clip('unord', 32, lambda x: _float_compare_to_mask(x))
-cmpunordsd = vec_op_clip('unord', 64, lambda x: _float_compare_to_mask(x))
-cmpneqps = vec_vertical_instr('!=fu', 32, lambda x: _float_compare_to_mask(x))
-cmpneqpd = vec_vertical_instr('!=fu', 64, lambda x: _float_compare_to_mask(x))
-cmpneqss = vec_op_clip('!=fu', 32, lambda x: _float_compare_to_mask(x))
-cmpneqsd = vec_op_clip('!=fu', 64, lambda x: _float_compare_to_mask(x))
-cmpnltps = vec_vertical_instr('!<fu', 32, lambda x: _float_compare_to_mask(x))
-cmpnltpd = vec_vertical_instr('!<fu', 64, lambda x: _float_compare_to_mask(x))
-cmpnltss = vec_op_clip('!<fu', 32, lambda x: _float_compare_to_mask(x))
-cmpnltsd = vec_op_clip('!<fu', 64, lambda x: _float_compare_to_mask(x))
-cmpnleps = vec_vertical_instr('!<=fu', 32, lambda x: _float_compare_to_mask(x))
-cmpnlepd = vec_vertical_instr('!<=fu', 64, lambda x: _float_compare_to_mask(x))
-cmpnless = vec_op_clip('!<=fu', 32, lambda x: _float_compare_to_mask(x))
-cmpnlesd = vec_op_clip('!<=fu', 64, lambda x: _float_compare_to_mask(x))
-cmpordps = vec_vertical_instr('ord', 32, lambda x: _float_compare_to_mask(x))
-cmpordpd = vec_vertical_instr('ord', 64, lambda x: _float_compare_to_mask(x))
-cmpordss = vec_op_clip('ord', 32, lambda x: _float_compare_to_mask(x))
-cmpordsd = vec_op_clip('ord', 64, lambda x: _float_compare_to_mask(x))
-
-# Logical (floating-point)
-#
-
-# MMX/SSE/AVX
-
-
-def pand(_, instr, dst, src):
-    e = []
-    result = dst & src
-    # No flag assigned
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-
-def pandn(_, instr, dst, src):
-    e = []
-    result = (dst ^ dst.mask) & src
-    # No flag assigned
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-
-def por(_, instr, dst, src):
-    e = []
-    result = dst | src
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-
-def cvtdq2pd(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst[:64],
-            m2_expr.ExprOp(
-                'sint_to_fp',
-                src[:32].signExtend(64)
-            )
-        )
-    )
-    e.append(
-        m2_expr.ExprAssign(
-            dst[64:128],
-            m2_expr.ExprOp(
-                'sint_to_fp',
-                src[32:64].signExtend(64)
-            )
-        )
-    )
-    return e, []
-
-
-def cvtdq2ps(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:32], m2_expr.ExprOp('sint_to_fp', src[:32])))
-    e.append(
-        m2_expr.ExprAssign(dst[32:64], m2_expr.ExprOp('sint_to_fp', src[32:64])))
-    e.append(
-        m2_expr.ExprAssign(dst[64:96], m2_expr.ExprOp('sint_to_fp', src[64:96])))
-    e.append(
-        m2_expr.ExprAssign(dst[96:128], m2_expr.ExprOp('sint_to_fp', src[96:128])))
-    return e, []
-
-
-def cvtpd2dq(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:32], m2_expr.ExprOp('fp_to_sint32', src[:64])))
-    e.append(
-        m2_expr.ExprAssign(dst[32:64], m2_expr.ExprOp('fp_to_sint32', src[64:128])))
-    e.append(m2_expr.ExprAssign(dst[64:128], m2_expr.ExprInt(0, 64)))
-    return e, []
-
-
-def cvtpd2pi(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:32], m2_expr.ExprOp('fp_to_sint32', src[:64])))
-    e.append(
-        m2_expr.ExprAssign(dst[32:64], m2_expr.ExprOp('fp_to_sint32', src[64:128])))
-    return e, []
-
-
-def cvtpd2ps(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:32], m2_expr.ExprOp('fpconvert_fp32', src[:64])))
-    e.append(
-        m2_expr.ExprAssign(dst[32:64], m2_expr.ExprOp('fpconvert_fp32', src[64:128])))
-    e.append(m2_expr.ExprAssign(dst[64:128], m2_expr.ExprInt(0, 64)))
-    return e, []
-
-
-def cvtpi2pd(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst[:64],
-            m2_expr.ExprOp(
-                'sint_to_fp',
-                src[:32].signExtend(64)
-            )
-        )
-    )
-    e.append(
-        m2_expr.ExprAssign(
-            dst[64:128],
-            m2_expr.ExprOp(
-                'sint_to_fp',
-                src[32:64].signExtend(64))
-        )
-    )
-    return e, []
-
-
-def cvtpi2ps(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:32], m2_expr.ExprOp('sint_to_fp', src[:32])))
-    e.append(
-        m2_expr.ExprAssign(dst[32:64], m2_expr.ExprOp('sint_to_fp', src[32:64])))
-    return e, []
-
-
-def cvtps2dq(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:32], m2_expr.ExprOp('fp_to_sint32', src[:32])))
-    e.append(
-        m2_expr.ExprAssign(dst[32:64], m2_expr.ExprOp('fp_to_sint32', src[32:64])))
-    e.append(
-        m2_expr.ExprAssign(dst[64:96], m2_expr.ExprOp('fp_to_sint32', src[64:96])))
-    e.append(
-        m2_expr.ExprAssign(dst[96:128], m2_expr.ExprOp('fp_to_sint32', src[96:128])))
-    return e, []
-
-
-def cvtps2pd(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:64], m2_expr.ExprOp('fpconvert_fp64', src[:32])))
-    e.append(
-        m2_expr.ExprAssign(dst[64:128], m2_expr.ExprOp('fpconvert_fp64', src[32:64])))
-    return e, []
-
-
-def cvtps2pi(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:32], m2_expr.ExprOp('fp_to_sint32', src[:32])))
-    e.append(
-        m2_expr.ExprAssign(dst[32:64], m2_expr.ExprOp('fp_to_sint32', src[32:64])))
-    return e, []
-
-
-def cvtsd2si(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:32], m2_expr.ExprOp('fp_to_sint32', src[:64])))
-    return e, []
-
-
-def cvtsd2ss(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:32], m2_expr.ExprOp('fpconvert_fp32', src[:64])))
-    return e, []
-
-
-def cvtsi2sd(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(
-            dst[:64],
-            m2_expr.ExprOp(
-                'sint_to_fp',
-                src[:32].signExtend(64)
-            )
-        )
-    )
-    return e, []
-
-
-def cvtsi2ss(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:32], m2_expr.ExprOp('sint_to_fp', src[:32])))
-    return e, []
-
-
-def cvtss2sd(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:64], m2_expr.ExprOp('fpconvert_fp64', src[:32])))
-    return e, []
-
-
-def cvtss2si(_, instr, dst, src):
-    e = []
-    e.append(
-        m2_expr.ExprAssign(dst[:32], m2_expr.ExprOp('fp_to_sint32', src[:32])))
-    return e, []
-
-
-def _cvtt_tpl(dst, src, numbers, double):
-    e = []
-    for i in numbers:
-        # For CVTT*D2* (Convert with Truncation ... Double-Precision) to work,
-        # a first conversion fp64 -> fp32 is needed
-        if double:
-            tmp_src = m2_expr.ExprOp('fpconvert_fp32', src[i*64:i*64 + 64])
-        else:
-            tmp_src = src[i*32:i*32 + 32]
-
-        e.append(m2_expr.ExprAssign(
-            dst[i*32:i*32 + 32],
-            m2_expr.ExprOp('fp_to_sint32', m2_expr.ExprOp(
-                'fpround_towardszero',
-                tmp_src
-            ))))
-    return e
-
-def cvttpd2pi(_, instr, dst, src):
-    return _cvtt_tpl(dst, src, [0, 1], double=True), []
-
-def cvttpd2dq(_, instr, dst, src):
-    e = _cvtt_tpl(dst, src, [0, 1], double=True)
-    e.append(m2_expr.ExprAssign(dst[64:128], m2_expr.ExprInt(0, 64)))
-    return e, []
-
-def cvttsd2si(_, instr, dst, src):
-    return _cvtt_tpl(dst, src, [0], double=True), []
-
-def cvttps2dq(_, instr, dst, src):
-    return _cvtt_tpl(dst, src, [0, 1, 2, 3], double=False), []
-
-def cvttps2pi(_, instr, dst, src):
-    return _cvtt_tpl(dst, src, [0, 1], double=False), []
-
-def cvttss2si(_, instr, dst, src):
-    return _cvtt_tpl(dst, src, [0], double=False), []
-
-def movss(_, instr, dst, src):
-    e = []
-    if not isinstance(dst, m2_expr.ExprMem) and not isinstance(src, m2_expr.ExprMem):
-        # Source and Destination xmm
-        e.append(m2_expr.ExprAssign(dst[:32], src[:32]))
-    elif not isinstance(src, m2_expr.ExprMem) and isinstance(dst, m2_expr.ExprMem):
-        # Source XMM Destination Mem
-        e.append(m2_expr.ExprAssign(dst, src[:32]))
-    else:
-        # Source Mem Destination XMM
-        e.append(m2_expr.ExprAssign(
-            dst, m2_expr.ExprCompose(src, m2_expr.ExprInt(0, 96))))
-    return e, []
-
-
-def ucomiss(_, instr, src1, src2):
-    e = []
-    e.append(m2_expr.ExprAssign(zf, m2_expr.ExprOp(
-        'ucomiss_zf', src1[:32], src2[:32])))
-    e.append(m2_expr.ExprAssign(pf, m2_expr.ExprOp(
-        'ucomiss_pf', src1[:32], src2[:32])))
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprOp(
-        'ucomiss_cf', src1[:32], src2[:32])))
-
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprInt(0, 1)))
-    e.append(m2_expr.ExprAssign(af, m2_expr.ExprInt(0, 1)))
-    e.append(m2_expr.ExprAssign(nf, m2_expr.ExprInt(0, 1)))
-
-    return e, []
-
-def ucomisd(_, instr, src1, src2):
-    e = []
-    e.append(m2_expr.ExprAssign(zf, m2_expr.ExprOp(
-        'ucomisd_zf', src1[:64], src2[:64])))
-    e.append(m2_expr.ExprAssign(pf, m2_expr.ExprOp(
-        'ucomisd_pf', src1[:64], src2[:64])))
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprOp(
-        'ucomisd_cf', src1[:64], src2[:64])))
-
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprInt(0, 1)))
-    e.append(m2_expr.ExprAssign(af, m2_expr.ExprInt(0, 1)))
-    e.append(m2_expr.ExprAssign(nf, m2_expr.ExprInt(0, 1)))
-
-    return e, []
-
-def blsi(_, instr, dst, src):
-    e = []
-
-    arg1 = m2_expr.ExprInt(0, src.size)
-    neg_src = arg1 - src
-    result = neg_src & src
-
-    e += update_flag_zf(result)
-    e += update_flag_nf(result)
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprInt(0, of.size)))
-
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprCond(src,
-                                                     m2_expr.ExprInt(1, 1),
-                                                     m2_expr.ExprInt(0, 1))))
-
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-def andn(_, instr, dst, src1, src2):
-    e = []
-
-    arg1 = m2_expr.ExprInt(0, src1.size)
-    neg_src1 = arg1 - src1
-    result = neg_src1 & src2
-
-    e += update_flag_zf(result)
-    e += update_flag_nf(result)
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprInt(0, of.size)))
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprInt(0, cf.size)))
-
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-def bextr(_, instr, dst, src1, src2):
-    e = []
-
-    # TODO: change zero extension to 512 bits when AVX is supported
-    start = (src2 & m2_expr.ExprInt(0xFF, src2.size)).zeroExtend(256)
-    length = ((src2 & m2_expr.ExprInt(0xFF00, src2.size)) >> m2_expr.ExprInt(8, src2.size)).zeroExtend(256)
-
-    tmp = src1.zeroExtend(256) >> start
-    mask = m2_expr.ExprInt(0, 256).mask >> (m2_expr.ExprInt(256, 256) - length)
-
-    tmp = tmp & mask
-    result = tmp[:dst.size]
-
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-def blsmsk(_, instr, dst, src):
-    e = []
-
-    tmp = src - m2_expr.ExprInt(1, src.size)
-    result = src ^ tmp
-
-    e += update_flag_nf(result)
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprInt(0, of.size)))
-    e.append(m2_expr.ExprAssign(zf, m2_expr.ExprInt(0, zf.size)))
-
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprCond(src,
-                                                     m2_expr.ExprInt(0, 1),
-                                                     m2_expr.ExprInt(1, 1))))
-
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-def blsr(_, instr, dst, src):
-    e = []
-
-    tmp = src - m2_expr.ExprInt(1, src.size)
-    result = tmp & src
-
-    e += update_flag_zf(result)
-    e += update_flag_nf(result)
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprInt(0, of.size)))
-
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprCond(src,
-                                                     m2_expr.ExprInt(0, 1),
-                                                     m2_expr.ExprInt(1, 1))))
-
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-def tzcnt(ir, instr, dst, src):
-    e = []
-
-    operand_size = m2_expr.ExprInt(dst.size, dst.size)
-
-    result = m2_expr.ExprCond(src, m2_expr.ExprOp("cnttrailzeros", src), operand_size)
-
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprCond(m2_expr.ExprOp("FLAG_EQ_CMP", result, operand_size),
-                                                     m2_expr.ExprInt(1, 1),
-                                                     m2_expr.ExprInt(0, 1))))
-
-    e += update_flag_zf(result)
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-def bzhi(_, instr, dst, src1, src2):
-    e = []
-
-    operand_size = m2_expr.ExprInt(dst.size, dst.size)
-    index = src2[:7].zeroExtend(dst.size)
-    mask = m2_expr.ExprInt(0, dst.size).mask >> (operand_size
-                                                 - index
-                                                 - m2_expr.ExprInt(1, dst.size))
-
-    result = m2_expr.ExprCond(m2_expr.ExprOp("FLAG_SIGN_SUB", index, operand_size),
-                              src1 & mask, src1)
-
-
-    operand_size_dec = operand_size - m2_expr.ExprInt(1, dst.size)
-    e.append(m2_expr.ExprAssign(cf, m2_expr.ExprCond(m2_expr.ExprOp("FLAG_SIGN_SUB", operand_size_dec, index),
-                                                     m2_expr.ExprInt(1, 1),
-                                                     m2_expr.ExprInt(0, 1))))
-
-    e += update_flag_zf(result)
-    e += update_flag_nf(result)
-    e.append(m2_expr.ExprAssign(of, m2_expr.ExprInt(0, of.size)))
-    e.append(m2_expr.ExprAssign(dst, result))
-    return e, []
-
-def pshufb(_, instr, dst, src):
-    e = []
-    if dst.size == 64:
-        bit_l = 3
-    elif dst.size == 128:
-        bit_l = 4
-    else:
-        raise NotImplementedError("bad size")
-    for i in range(0, src.size, 8):
-        index = src[
-            i:i + bit_l].zeroExtend(dst.size) << m2_expr.ExprInt(3, dst.size)
-        value = (dst >> index)[:8]
-        e.append(m2_expr.ExprAssign(dst[i:i + 8],
-                                 m2_expr.ExprCond(src[i + 7:i + 8],
-                                                  m2_expr.ExprInt(0, 8),
-                                                  value)))
-    return e, []
-
-
-def pshufd(_, instr, dst, src, imm):
-    control = int(imm)
-    out = []
-    for i in range(4):
-        shift = ((control >> (i * 2)) & 3) * 32
-        # shift is 2 bits long, expr.size is 128
-        # => shift + 32 <= src.size
-        out.append(src[shift: shift + 32])
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out))], []
-
-
-def pshuflw(_, instr, dst, src, imm):
-    control = int(imm)
-    out = []
-    for i in range(4):
-        shift = ((control >> (i * 2)) & 3) * 16
-        out.append(src[shift: shift + 16])
-    out.append(src[64:])
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out))], []
-
-
-def pshufhw(_, instr, dst, src, imm):
-    control = int(imm)
-    out = [src[:64]]
-    for i in range(4):
-        shift = ((control >> (i * 2)) & 3) * 16
-        out.append(src[shift + 64: shift + 16 + 64])
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out))], []
-
-
-def ps_rl_ll(ir, instr, dst, src, op, size):
-    mask = {16: 0xF,
-            32: 0x1F,
-            64: 0x3F}[size]
-    mask = m2_expr.ExprInt(mask, dst.size)
-
-    # Saturate the counter to 2**size
-    count = src.zeroExtend(dst.size)
-    count = m2_expr.ExprCond(count & expr_simp(~mask),
-                             m2_expr.ExprInt(size, dst.size), # saturation
-                             count, # count < 2**size
-    )
-    count = count[:size]
-    if src.is_int():
-        count = expr_simp(count)
-
-    out = []
-    for i in range(0, dst.size, size):
-        out.append(m2_expr.ExprOp(op, dst[i:i + size], count))
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out))], []
-
-
-def psrlw(ir, instr, dst, src):
-    return ps_rl_ll(ir, instr, dst, src, ">>", 16)
-
-
-def psrld(ir, instr, dst, src):
-    return ps_rl_ll(ir, instr, dst, src, ">>", 32)
-
-
-def psrlq(ir, instr, dst, src):
-    return ps_rl_ll(ir, instr, dst, src, ">>", 64)
-
-
-def psllw(ir, instr, dst, src):
-    return ps_rl_ll(ir, instr, dst, src, "<<", 16)
-
-
-def pslld(ir, instr, dst, src):
-    return ps_rl_ll(ir, instr, dst, src, "<<",  32)
-
-
-def psllq(ir, instr, dst, src):
-    return ps_rl_ll(ir, instr, dst, src, "<<",  64)
-
-
-def psraw(ir, instr, dst, src):
-    return ps_rl_ll(ir, instr, dst, src, "a>>", 16)
-
-
-def psrad(ir, instr, dst, src):
-    return ps_rl_ll(ir, instr, dst, src, "a>>", 32)
-
-
-def pslldq(_, instr, dst, src):
-    assert src.is_int()
-    e = []
-    count = int(src)
-    if count > 15:
-        return [m2_expr.ExprAssign(dst, m2_expr.ExprInt(0, dst.size))], []
-    else:
-        return [m2_expr.ExprAssign(dst, dst << m2_expr.ExprInt(8 * count, dst.size))], []
-
-
-def psrldq(_, instr, dst, src):
-    assert src.is_int()
-    count = int(src)
-    if count > 15:
-        return [m2_expr.ExprAssign(dst, m2_expr.ExprInt(0, dst.size))], []
-    else:
-        return [m2_expr.ExprAssign(dst, dst >> m2_expr.ExprInt(8 * count, dst.size))], []
-
-
-def iret(ir, instr):
-    """IRET implementation
-    XXX: only support "no-privilege change"
-    """
-    size = instr.v_opmode()
-    exprs, _ = retf(ir, instr, m2_expr.ExprInt(size // 8, size=size))
-    tmp = mRSP[instr.mode][:size] + m2_expr.ExprInt((2 * size) // 8, size=size)
-    exprs += _tpl_eflags(tmp)
-    return exprs, []
-
-
-def pcmpeq(_, instr, dst, src, size):
-    e = []
-    for i in range(0, dst.size, size):
-        test = m2_expr.expr_is_equal(dst[i:i + size], src[i:i + size])
-        e.append(m2_expr.ExprAssign(dst[i:i + size],
-                                 m2_expr.ExprCond(test,
-                                                  m2_expr.ExprInt(-1, size),
-                                                  m2_expr.ExprInt(0, size))))
-    return e, []
-
-
-def pcmpgt(_, instr, dst, src, size):
-    e = []
-    for i in range(0, dst.size, size):
-        test = m2_expr.expr_is_signed_greater(dst[i:i + size], src[i:i + size])
-        e.append(m2_expr.ExprAssign(dst[i:i + size],
-                                 m2_expr.ExprCond(test,
-                                                  m2_expr.ExprInt(-1, size),
-                                                  m2_expr.ExprInt(0, size))))
-    return e, []
-
-
-def pcmpeqb(ir, instr, dst, src):
-    return pcmpeq(ir, instr, dst, src, 8)
-
-def pcmpeqw(ir, instr, dst, src):
-    return pcmpeq(ir, instr, dst, src, 16)
-
-def pcmpeqd(ir, instr, dst, src):
-    return pcmpeq(ir, instr, dst, src, 32)
-
-def pcmpeqq(ir, instr, dst, src):
-    return pcmpeq(ir, instr, dst, src, 64)
-
-
-
-
-def pcmpgtb(ir, instr, dst, src):
-    return pcmpgt(ir, instr, dst, src, 8)
-
-def pcmpgtw(ir, instr, dst, src):
-    return pcmpgt(ir, instr, dst, src, 16)
-
-def pcmpgtd(ir, instr, dst, src):
-    return pcmpgt(ir, instr, dst, src, 32)
-
-def pcmpgtq(ir, instr, dst, src):
-    return pcmpgt(ir, instr, dst, src, 64)
-
-
-
-def punpck(_, instr, dst, src, size, off):
-    e = []
-    slices = []
-    for i in range(dst.size // (2 * size)):
-        slices.append(dst[size * i + off: size * i + off + size])
-        slices.append(src[size * i + off: size * i + off + size])
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*slices)))
-    return e, []
-
-
-def punpckhbw(ir, instr, dst, src):
-    return punpck(ir, instr, dst, src, 8, dst.size // 2)
-
-
-def punpckhwd(ir, instr, dst, src):
-    return punpck(ir, instr, dst, src, 16, dst.size // 2)
-
-
-def punpckhdq(ir, instr, dst, src):
-    return punpck(ir, instr, dst, src, 32, dst.size // 2)
-
-
-def punpckhqdq(ir, instr, dst, src):
-    return punpck(ir, instr, dst, src, 64, dst.size // 2)
-
-
-def punpcklbw(ir, instr, dst, src):
-    return punpck(ir, instr, dst, src, 8, 0)
-
-
-def punpcklwd(ir, instr, dst, src):
-    return punpck(ir, instr, dst, src, 16, 0)
-
-
-def punpckldq(ir, instr, dst, src):
-    return punpck(ir, instr, dst, src, 32, 0)
-
-
-def punpcklqdq(ir, instr, dst, src):
-    return punpck(ir, instr, dst, src, 64, 0)
-
-
-def pinsr(_, instr, dst, src, imm, size):
-    e = []
-
-    mask = {8: 0xF,
-            16: 0x7,
-            32: 0x3,
-            64: 0x1}[size]
-
-    sel = (int(imm) & mask) * size
-    e.append(m2_expr.ExprAssign(dst[sel:sel + size], src[:size]))
-
-    return e, []
-
-
-def pinsrb(ir, instr, dst, src, imm):
-    return pinsr(ir, instr, dst, src, imm, 8)
-
-
-def pinsrw(ir, instr, dst, src, imm):
-    return pinsr(ir, instr, dst, src, imm, 16)
-
-
-def pinsrd(ir, instr, dst, src, imm):
-    return pinsr(ir, instr, dst, src, imm, 32)
-
-
-def pinsrq(ir, instr, dst, src, imm):
-    return pinsr(ir, instr, dst, src, imm, 64)
-
-
-def pextr(_, instr, dst, src, imm, size):
-    e = []
-
-    mask = {8: 0xF,
-            16: 0x7,
-            32: 0x3,
-            64: 0x1}[size]
-
-    sel = (int(imm) & mask) * size
-    e.append(m2_expr.ExprAssign(dst, src[sel:sel + size].zeroExtend(dst.size)))
-
-    return e, []
-
-
-def pextrb(ir, instr, dst, src, imm):
-    return pextr(ir, instr, dst, src, imm, 8)
-
-
-def pextrw(ir, instr, dst, src, imm):
-    return pextr(ir, instr, dst, src, imm, 16)
-
-
-def pextrd(ir, instr, dst, src, imm):
-    return pextr(ir, instr, dst, src, imm, 32)
-
-
-def pextrq(ir, instr, dst, src, imm):
-    return pextr(ir, instr, dst, src, imm, 64)
-
-
-def unpckhps(_, instr, dst, src):
-    e = []
-    src = m2_expr.ExprCompose(dst[64:96], src[64:96], dst[96:128], src[96:128])
-    e.append(m2_expr.ExprAssign(dst, src))
-    return e, []
-
-
-def unpckhpd(_, instr, dst, src):
-    e = []
-    src = m2_expr.ExprCompose(dst[64:128], src[64:128])
-    e.append(m2_expr.ExprAssign(dst, src))
-    return e, []
-
-
-def unpcklps(_, instr, dst, src):
-    e = []
-    src = m2_expr.ExprCompose(dst[0:32], src[0:32], dst[32:64], src[32:64])
-    e.append(m2_expr.ExprAssign(dst, src))
-    return e, []
-
-
-def unpcklpd(_, instr, dst, src):
-    e = []
-    src = m2_expr.ExprCompose(dst[0:64], src[0:64])
-    e.append(m2_expr.ExprAssign(dst, src))
-    return e, []
-
-
-def movlpd(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst[:64], src[:64]))
-    return e, []
-
-
-def movlps(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst[:64], src[:64]))
-    return e, []
-
-
-def movhpd(_, instr, dst, src):
-    e = []
-    if src.size == 64:
-        e.append(m2_expr.ExprAssign(dst[64:128], src))
-    elif dst.size == 64:
-        e.append(m2_expr.ExprAssign(dst, src[64:128]))
-    else:
-        raise RuntimeError("bad encoding!")
-    return e, []
-
-
-def movlhps(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst[64:128], src[:64]))
-    return e, []
-
-
-def movhlps(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst[:64], src[64:128]))
-    return e, []
-
-
-def movdq2q(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, src[:64]))
-    return e, []
-
-
-def movq2dq(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst, src[:64].zeroExtend(dst.size)))
-    return e, []
-
-
-def sqrt_gen(_, instr, dst, src, size):
-    e = []
-    out = []
-    for i in range(src.size // size):
-        out.append(m2_expr.ExprOp('fsqrt',
-                                  src[i * size: (i + 1) * size]))
-    src = m2_expr.ExprCompose(*out)
-    e.append(m2_expr.ExprAssign(dst, src))
-    return e, []
-
-
-def sqrtpd(ir, instr, dst, src):
-    return sqrt_gen(ir, instr, dst, src, 64)
-
-
-def sqrtps(ir, instr, dst, src):
-    return sqrt_gen(ir, instr, dst, src, 32)
-
-
-def sqrtsd(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst[:64],
-                             m2_expr.ExprOp('fsqrt',
-                                            src[:64])))
-    return e, []
-
-
-def sqrtss(_, instr, dst, src):
-    e = []
-    e.append(m2_expr.ExprAssign(dst[:32],
-                             m2_expr.ExprOp('fsqrt',
-                                            src[:32])))
-    return e, []
-
-
-def pmovmskb(_, instr, dst, src):
-    e = []
-    out = []
-    for i in range(src.size // 8):
-        out.append(src[8 * i + 7:8 * (i + 1)])
-    src = m2_expr.ExprCompose(*out)
-    e.append(m2_expr.ExprAssign(dst, src.zeroExtend(dst.size)))
-    return e, []
-
-
-def smsw(ir, instr, dst):
-    e = []
-    LOG_X86_SEM.warning("DEFAULT SMSW %s!!", str(dst))
-    e.append(m2_expr.ExprAssign(dst, m2_expr.ExprInt(0x80050033, 32)[:dst.size]))
-    return e, []
-
-
-def bndmov(ir, instr, dst, src):
-    # Implemented as a NOP, because BND side effects are not yet supported
-    return [], []
-
-def palignr(ir, instr, dst, src, imm):
-    # dst.src >> imm * 8 [:dst.size]
-
-    shift = int(imm) * 8
-    if shift == 0:
-        result = src
-    elif shift == src.size:
-        result = dst
-    elif shift > src.size:
-        result = dst >> m2_expr.ExprInt(shift - src.size, dst.size)
-    else:
-        # shift < src.size
-        result = m2_expr.ExprCompose(
-            src[shift:],
-            dst[:shift],
-        )
-
-    return [m2_expr.ExprAssign(dst, result)], []
-
-
-def _signed_to_signed_saturation(expr, dst_size):
-    """Saturate the expr @expr for @dst_size bit
-    Signed saturation return MAX_INT / MIN_INT or value depending on the value
-    """
-    assert expr.size > dst_size
-
-    median = 1 << (dst_size - 1)
-
-    min_int = m2_expr.ExprInt(- median, dst_size)
-    max_int = m2_expr.ExprInt(median - 1, dst_size)
-
-    test_min_int = min_int.signExtend(expr.size)
-    test_max_int = max_int.signExtend(expr.size)
-
-    value = expr[:dst_size]
-
-    return m2_expr.ExprCond(
-        m2_expr.ExprOp(
-            m2_expr.TOK_INF_EQUAL_SIGNED,
-            expr,
-            test_min_int
-        ),
-        min_int,
-        m2_expr.ExprCond(
-            m2_expr.ExprOp(
-                m2_expr.TOK_INF_SIGNED,
-                expr,
-                test_max_int
-            ),
-            value,
-            max_int
-        )
-    )
-
-
-def _signed_to_unsigned_saturation(expr, dst_size):
-    """Saturate the expr @expr for @dst_size bit
-    Unsigned saturation return MAX_INT or value depending on the value
-    """
-    assert expr.size > dst_size
-
-    zero = m2_expr.ExprInt(0, dst_size)
-    test_zero = m2_expr.ExprInt(0, expr.size)
-
-    max_int = m2_expr.ExprInt(-1, dst_size)
-    test_max_int = max_int.zeroExtend(expr.size)
-
-    value = expr[:dst_size]
-
-    return m2_expr.ExprCond(
-        m2_expr.ExprOp(
-            m2_expr.TOK_INF_EQUAL_SIGNED,
-            expr,
-            test_zero
-        ),
-        zero,
-        m2_expr.ExprCond(
-            m2_expr.ExprOp(
-                m2_expr.TOK_INF_SIGNED,
-                expr,
-                test_max_int
-            ),
-            value,
-            max_int
-        )
-    )
-
-
-
-def packsswb(ir, instr, dst, src):
-    out = []
-    for source in [dst, src]:
-        for start in range(0, dst.size, 16):
-            out.append(_signed_to_signed_saturation(source[start:start + 16], 8))
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out))], []
-
-
-def packssdw(ir, instr, dst, src):
-    out = []
-    for source in [dst, src]:
-        for start in range(0, dst.size, 32):
-            out.append(_signed_to_signed_saturation(source[start:start + 32], 16))
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out))], []
-
-
-def packuswb(ir, instr, dst, src):
-    out = []
-    for source in [dst, src]:
-        for start in range(0, dst.size, 16):
-            out.append(_signed_to_unsigned_saturation(source[start:start + 16], 8))
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out))], []
-
-
-def _saturation_sub_unsigned(expr):
-    assert expr.is_op("+") and len(expr.args) == 2 and expr.args[-1].is_op("-")
-
-    # Compute the soustraction on one more bit to be able to distinguish cases:
-    # 0x48 - 0xd7 in 8 bit, should saturate
-    arg1 = expr.args[0].zeroExtend(expr.size + 1)
-    arg2 = expr.args[1].args[0].zeroExtend(expr.size + 1)
-    return _signed_to_unsigned_saturation(arg1 - arg2, expr.size)
-
-def _saturation_sub_signed(expr):
-    assert expr.is_op("+") and len(expr.args) == 2 and expr.args[-1].is_op("-")
-
-    # Compute the subtraction on two more bits, see _saturation_sub_unsigned
-    arg1 = expr.args[0].signExtend(expr.size + 2)
-    arg2 = expr.args[1].args[0].signExtend(expr.size + 2)
-    return _signed_to_signed_saturation(arg1 - arg2, expr.size)
-
-def _saturation_add(expr):
-    assert expr.is_op("+") and len(expr.args) == 2
-
-    # Compute the addition on one more bit to be able to distinguish cases:
-    # 0x48 + 0xd7 in 8 bit, should saturate
-
-    arg1 = expr.args[0].zeroExtend(expr.size + 1)
-    arg2 = expr.args[1].zeroExtend(expr.size + 1)
-
-    # We can also use _signed_to_unsigned_saturation with two additional bits (to
-    # distinguish minus and overflow case)
-    # The resulting expression being more complicated with an impossible case
-    # (signed=True), we rewrite the rule here
-
-    return m2_expr.ExprCond((arg1 + arg2).msb(), m2_expr.ExprInt(-1, expr.size),
-                            expr)
-
-def _saturation_add_signed(expr):
-    assert expr.is_op("+") and len(expr.args) == 2
-
-    # Compute the subtraction on two more bits, see _saturation_add_unsigned
-
-    arg1 = expr.args[0].signExtend(expr.size + 2)
-    arg2 = expr.args[1].signExtend(expr.size + 2)
-
-    return _signed_to_signed_saturation(arg1 + arg2, expr.size)
-
-
-# Saturate SSE operations
-
-psubusb = vec_vertical_instr('-', 8, _saturation_sub_unsigned)
-psubusw = vec_vertical_instr('-', 16, _saturation_sub_unsigned)
-paddusb = vec_vertical_instr('+', 8, _saturation_add)
-paddusw = vec_vertical_instr('+', 16, _saturation_add)
-psubsb = vec_vertical_instr('-', 8, _saturation_sub_signed)
-psubsw = vec_vertical_instr('-', 16, _saturation_sub_signed)
-paddsb = vec_vertical_instr('+', 8, _saturation_add_signed)
-paddsw = vec_vertical_instr('+', 16, _saturation_add_signed)
-
-
-# Others SSE operations
-
-def maskmovq(ir, instr, src, mask):
-    loc_next = ir.get_next_loc_key(instr)
-    loc_next_expr = m2_expr.ExprLoc(loc_next, ir.IRDst.size)
-    blks = []
-
-    # For each possibility, check if a write is necessary
-    check_labels = [m2_expr.ExprLoc(ir.loc_db.add_location(), ir.IRDst.size)
-                    for _ in range(0, mask.size, 8)]
-    # If the write has to be done, do it (otherwise, nothing happen)
-    write_labels = [m2_expr.ExprLoc(ir.loc_db.add_location(), ir.IRDst.size)
-                    for _ in range(0, mask.size, 8)]
-
-    # Build check blocks
-    for i, start in enumerate(range(0, mask.size, 8)):
-        bit = mask[start + 7: start + 8]
-        cur_label = check_labels[i]
-        next_check_label = check_labels[i + 1] if (i + 1) < len(check_labels) else loc_next_expr
-        write_label = write_labels[i]
-        check = m2_expr.ExprAssign(ir.IRDst,
-                                m2_expr.ExprCond(bit,
-                                                 write_label,
-                                                 next_check_label))
-        blks.append(IRBlock(ir.loc_db, cur_label.loc_key, [AssignBlock([check], instr)]))
-
-    # Build write blocks
-    dst_addr = mRDI[instr.mode]
-    for i, start in enumerate(range(0, mask.size, 8)):
-        cur_label = write_labels[i]
-        next_check_label = check_labels[i + 1] if (i + 1) < len(check_labels) else loc_next_expr
-        write_addr = dst_addr + m2_expr.ExprInt(i, dst_addr.size)
-
-        # @8[DI/EDI/RDI + i] = src[byte i]
-        write_mem = m2_expr.ExprAssign(m2_expr.ExprMem(write_addr, 8),
-                                    src[start: start + 8])
-        jump = m2_expr.ExprAssign(ir.IRDst, next_check_label)
-        blks.append(IRBlock(ir.loc_db, cur_label.loc_key, [AssignBlock([write_mem, jump], instr)]))
-
-    # If mask is null, bypass all
-    e = [m2_expr.ExprAssign(ir.IRDst, m2_expr.ExprCond(mask,
-                                                    check_labels[0],
-                                                    loc_next_expr))]
-    return e, blks
-
-
-def emms(ir, instr):
-    # Implemented as a NOP
-    return [], []
-
-def incssp(ir, instr, dst):
-    # Implemented as a NOP
-    return [], []
-
-def rdssp(ir, instr, dst):
-    # Implemented as a NOP
-    return [], []
-
-def saveprevssp(ir, instr):
-    # Implemented as a NOP
-    return [], []
-
-def rstorssp(ir, instr, dst):
-    # Implemented as a NOP
-    return [], []
-
-def wrss(ir, instr, src, dst):
-    # Implemented as a NOP
-    return [], []
-
-def wruss(ir, instr, src, dst):
-    # Implemented as a NOP
-    return [], []
-
-def setssbsy(ir, instr):
-    # Implemented as a NOP
-    return [], []
-
-def clrssbsy(ir, instr, dst):
-    # Implemented as a NOP
-    return [], []
-
-def endbr64(ir, instr):
-    # Implemented as a NOP
-    return [], []
-
-def endbr32(ir, instr):
-    # Implemented as a NOP
-    return [], []
-
-# Common value without too many option, 0x1fa0
-STMXCSR_VALUE = 0x1fa0
-def stmxcsr(ir, instr, dst):
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprInt(STMXCSR_VALUE, dst.size))], []
-
-def ldmxcsr(ir, instr, dst):
-    # Implemented as a NOP
-    return [], []
-
-
-def _select4(src, control):
-    # Implementation inspired from Intel Intrinsics Guide
-    # @control is already resolved (was an immediate)
-
-    if control == 0:
-        return src[:32] # 0
-    elif control == 1:
-        return src[32:64]
-    elif control == 2:
-        return src[64:96]
-    elif control == 3:
-        return src[96:]
-    else:
-        raise ValueError("Control must be on 2 bits")
-
-
-def shufps(ir, instr, dst, src, imm8):
-    out = []
-    control = int(imm8)
-    for i in range(4):
-        if i < 2:
-            source = dst
-        else:
-            source = src
-        out.append(_select4(source, (control >> (i * 2)) & 3))
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out))], []
-
-
-def shufpd(ir, instr, dst, src, imm8):
-    out = []
-    control = int(imm8)
-    out.append(dst[64:] if control & 1 else dst[:64])
-    out.append(src[64:] if control & 2 else src[:64])
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out))], []
-
-def movmskps(ir, instr, dst, src):
-    out = []
-    for i in range(4):
-        out.append(src[(32 * i) + 31:(32 * i) + 32])
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out).zeroExtend(dst.size))], []
-
-def movmskpd(ir, instr, dst, src):
-    out = []
-    for i in range(2):
-        out.append(src[(64 * i) + 63:(64 * i) + 64])
-    return [m2_expr.ExprAssign(dst, m2_expr.ExprCompose(*out).zeroExtend(dst.size))], []
-
-def _roundscalar(ir, inst, dst, src, imm8, double):
-    res = None
-    ctl = int(imm8)
-    dst_expr = dst[:64] if double else dst[:32]
-    src_expr = src[:64] if double else src[:32]
-    if ctl & 0x4 != 0:
-        # Use MXCSR rounding config
-        # TODO: here we assume it's round to nearest, ties to even
-        res = m2_expr.ExprOp('fpround_towardsnearest', src_expr)
-    else:
-        # Use encoded rounding mechanism
-        rounding_mechanism = ctl & 0x3
-        ROUNDING_MODE = {
-            0x0: 'fpround_towardsnearest',
-            0x1: 'fpround_down',
-            0x2: 'fpround_up',
-            0x3: 'fpround_towardszero'
-        }
-        res = m2_expr.ExprOp(ROUNDING_MODE[rounding_mechanism], src_expr)
-    return [m2_expr.ExprAssign(dst_expr, res)], []
-
-def roundss(ir, inst, dst, src, imm8):
-    return _roundscalar(ir, inst, dst, src, imm8, False)
-
-def roundsd(ir, inst, dst, src, imm8):
-    return _roundscalar(ir, inst, dst, src, imm8, True)
-
-def fxsave(_ir, _instr, _src):
-    # Implemented as a NOP for now
-    return [], []
-
-def fxrstor(_ir, _instr, _dst):
-    # Implemented as a NOP for now
-    return [], []
-
-
-mnemo_func = {'mov': mov,
-              'xchg': xchg,
-              'movzx': movzx,
-              'movsx': movsx,
-              'movsxd': movsx,
-              'lea': lea,
-              'add': add,
-              'xadd': xadd,
-              'adc': adc,
-              'sub': sub,
-              'sbb': sbb,
-              'neg': neg,
-              'not': l_not,
-              'cmp': l_cmp,
-              'xor': xor,
-              'pxor': pxor,
-              'or': l_or,
-              'and': l_and,
-              'test': l_test,
-              'rol': l_rol,
-              'ror': l_ror,
-              'rcl': rcl,
-              'rcr': rcr,
-              'sar': sar,
-              'shr': shr,
-              'sal': shl,
-              'shl': shl,
-              'shld': shld,
-              'cmc': cmc,
-              'clc': clc,
-              'stc': stc,
-              'cld': cld,
-              'std': std,
-              'cli': cli,
-              'sti': sti,
-              'bsf': bsf,
-              'bsr': bsr,
-              'inc': inc,
-              'dec': dec,
-              'push': push,
-              'pushw': pushw,
-              'pop': pop,
-              'popw': popw,
-              'sete': sete,
-              'setnz': setnz,
-              'setl': setl,
-              'setg': setg,
-              'setge': setge,
-              'seta': seta,
-              'setae': setae,
-              'setb': setb,
-              'setbe': setbe,
-              'setns': setns,
-              'sets': sets,
-              'seto': seto,
-              'setp': setp,
-              'setpe': setp,
-              'setnp': setnp,
-              'setpo': setnp,
-              'setle': setle,
-              'setng': setle,
-              'setna': setna,
-              'setnbe': setnbe,
-              'setno': setno,
-              'setnc': setnb,
-              'setz': sete,
-              'setne': setnz,
-              'setnb': setae,
-              'setnae': setb,
-              'setc': setb,
-              'setnge': setl,
-              'setnl': setge,
-              'setnle': setg,
-              'setalc': setalc,
-              'bswap': bswap,
-              'cmpsb': lambda ir, instr: cmps(ir, instr, 8),
-              'cmpsw': lambda ir, instr: cmps(ir, instr, 16),
-              'cmpsd': lambda ir, instr: cmps(ir, instr, 32),
-              'cmpsq': lambda ir, instr: cmps(ir, instr, 64),
-              'scasb': lambda ir, instr: scas(ir, instr, 8),
-              'scasw': lambda ir, instr: scas(ir, instr, 16),
-              'scasd': lambda ir, instr: scas(ir, instr, 32),
-              'scasq': lambda ir, instr: scas(ir, instr, 64),
-              'pushfd': pushfd,
-              'pushfq': pushfq,
-              'pushfw': pushfw,
-              'popfd': popfd,
-              'popfq': popfd,
-              'popfw': popfw,
-              'pusha': pusha,
-              'pushad': pushad,
-              'popad': popad,
-              'popa': popa,
-              'call': call,
-              'ret': ret,
-              'retf': retf,
-              'iret': iret,
-              'iretd': iret,
-              'leave': leave,
-              'enter': enter,
-              'jmp': jmp,
-              'jz': jz,
-              'je': jz,
-              'jcxz': jcxz,
-              'jecxz': jecxz,
-              'jrcxz': jrcxz,
-              'jnz': jnz,
-              'jp': jp,
-              'jpe': jp,
-              'jnp': jnp,
-              'ja': ja,
-              'jae': jae,
-              'jb': jb,
-              'jbe': jbe,
-              'jg': jg,
-              'jge': jge,
-              'jl': jl,
-              'jle': jle,
-              'js': js,
-              'jns': jns,
-              'jo': jo,
-              'jno': jno,
-              'loop': loop,
-              'loopne': loopne,
-              'loope': loope,
-              'div': div,
-              'mul': mul,
-              'imul': imul,
-              'idiv': idiv,
-
-              'cbw': cbw,
-              'cwde': cwde,
-              'cdqe': cdqe,
-
-              'cwd': cwd,
-              'cdq': cdq,
-              'cqo': cqo,
-
-              'daa': daa,
-              'das': das,
-              'aam': aam,
-              'aad': aad,
-              'aaa': aaa,
-              'aas': aas,
-              'shrd': shrd,
-              'stosb': lambda ir, instr: stos(ir, instr, 8),
-              'stosw': lambda ir, instr: stos(ir, instr, 16),
-              'stosd': lambda ir, instr: stos(ir, instr, 32),
-              'stosq': lambda ir, instr: stos(ir, instr, 64),
-
-              'lodsb': lambda ir, instr: lods(ir, instr, 8),
-              'lodsw': lambda ir, instr: lods(ir, instr, 16),
-              'lodsd': lambda ir, instr: lods(ir, instr, 32),
-              'lodsq': lambda ir, instr: lods(ir, instr, 64),
-
-              'movsb': lambda ir, instr: movs(ir, instr, 8),
-              'movsw': lambda ir, instr: movs(ir, instr, 16),
-              'movsd': movsd_dispatch,
-              'movsq': lambda ir, instr: movs(ir, instr, 64),
-              'fcomp': fcomp,
-              'fcompp': fcompp,
-              'ficomp': ficomp,
-              'fucom': fucom,
-              'fucomp': fucomp,
-              'fucompp': fucompp,
-              'comiss': comiss,
-              'comisd': comisd,
-              'nop': nop,
-              'ud2': ud2,
-              'prefetch0': prefetch0,
-              'prefetch1': prefetch1,
-              'prefetch2': prefetch2,
-              'prefetchw': prefetchw,
-              'prefetchnta': prefetchnta,
-              'lfence': lfence,
-              'mfence': mfence,
-              'sfence': sfence,
-              'fnop': nop,  # XXX
-              'hlt': hlt,
-              'rdtsc': rdtsc,
-              'fst': fst,
-              'fstp': fstp,
-              'fist': fist,
-              'fistp': fistp,
-              'fisttp': fisttp,
-              'fld': fld,
-              'fldz': fldz,
-              'fld1': fld1,
-              'fldl2t': fldl2t,
-              'fldpi': fldpi,
-              'fldln2': fldln2,
-              'fldl2e': fldl2e,
-              'fldlg2': fldlg2,
-              'fild': fild,
-              'fadd': fadd,
-              'fiadd': fiadd,
-              'fisub': fisub,
-              'fisubr': fisubr,
-              'fpatan': fpatan,
-              'fprem': fprem,
-              'fprem1': fprem1,
-              'fninit': fninit,
-              'fyl2x': fyl2x,
-              'faddp': faddp,
-              'fsub': fsub,
-              'fsubp': fsubp,
-              'fsubr': fsubr,
-              'fsubrp': fsubrp,
-              'fmul': fmul,
-              'fimul': fimul,
-              'fmulp': fmulp,
-              'fdiv': fdiv,
-              'fdivr': fdivr,
-              'fdivrp': fdivrp,
-              'fidiv': fidiv,
-              'fidivr': fidivr,
-              'fdivp': fdivp,
-              'fxch': fxch,
-              'fptan': fptan,
-              'frndint': frndint,
-              'fsin': fsin,
-              'fcos': fcos,
-              'fsincos': fsincos,
-              'fscale': fscale,
-              'f2xm1': f2xm1,
-              'fchs': fchs,
-              'fsqrt': fsqrt,
-              'fabs': fabs,
-              'fnstsw': fnstsw,
-              'fnstcw': fnstcw,
-              'fldcw': fldcw,
-              'fwait': fwait,
-              'fcmovb':   fcmovb,
-              'fcmove':   fcmove,
-              'fcmovbe':  fcmovbe,
-              'fcmovu':   fcmovu,
-              'fcmovnb':  fcmovnb,
-              'fcmovne':  fcmovne,
-              'fcmovnbe': fcmovnbe,
-              'fcmovnu':  fcmovnu,
-              'fnstenv': fnstenv,
-              'fldenv': fldenv,
-              'sidt': sidt,
-              'sldt': sldt,
-              'arpl': arpl,
-              'cmovz': cmovz,
-              'cmove': cmovz,
-              'cmovnz': cmovnz,
-              'cmovpe': cmovpe,
-              'cmovnp': cmovnp,
-              'cmovge': cmovge,
-              'cmovnl': cmovge,
-              'cmovg': cmovg,
-              'cmovl': cmovl,
-              'cmova': cmova,
-              'cmovae': cmovae,
-              'cmovbe': cmovbe,
-              'cmovb': cmovb,
-              'cmovnge': cmovl,
-              'cmovle': cmovle,
-              'cmovng': cmovle,
-              'cmovo': cmovo,
-              'cmovno': cmovno,
-              'cmovs': cmovs,
-              'cmovns': cmovns,
-              'icebp': icebp,
-              'int': l_int,
-              'xlat': xlat,
-              'bt': bt,
-              'cpuid': cpuid,
-              'fcom': fcom,
-              'ftst': ftst,
-              'fxam': fxam,
-              'ficom': ficom,
-              'fcomi': fcomi,
-              'fcomip': fcomip,
-              'fucomi': fucomi,
-              'fucomip': fucomip,
-              'insb': lambda ir, instr: ins(ir, instr, 8),
-              'insw': lambda ir, instr: ins(ir, instr, 16),
-              'insd': lambda ir, instr: ins(ir, instr, 32),
-              'btc': btc,
-              'bts': bts,
-              'btr': btr,
-              'into': into,
-              'in': l_in,
-              'outsb': lambda ir, instr: l_outs(ir, instr, 8),
-              'outsw': lambda ir, instr: l_outs(ir, instr, 16),
-              'outsd': lambda ir, instr: l_outs(ir, instr, 32),
-
-              'out': l_out,
-              "sysenter": l_sysenter,
-              "syscall": l_syscall,
-              "cmpxchg": cmpxchg,
-              "cmpxchg8b": cmpxchg8b,
-              "lds": lds,
-              "les": les,
-              "lss": lss,
-              "lfs": lfs,
-              "lgs": lgs,
-              "lahf": lahf,
-              "sahf": sahf,
-              "lar": lar,
-              "lsl": lsl,
-              "fclex": fclex,
-              "fnclex": fnclex,
-              "str": l_str,
-              "movd": movd,
-              "movdqu": movdqu,
-              "movdqa": movdqu,
-              "movapd": movapd,  # XXX TODO alignment check
-              "movupd": movapd,  # XXX TODO alignment check
-              "movaps": movapd,  # XXX TODO alignment check
-              "movups": movapd,  # XXX TODO alignment check
-              "andps": andps,
-              "andpd": andps,
-              "andnps": andnps,
-              "andnpd": andnps,
-              "orps": orps,
-              "orpd": orps,
-              "xorps": xorps,
-              "xorpd": xorps,
-
-              "movq": movq,
-
-              "pminsw": pminsw,
-              "cvtdq2pd": cvtdq2pd,
-              "cvtdq2ps": cvtdq2ps,
-              "cvtpd2dq": cvtpd2dq,
-              "cvtpd2pi": cvtpd2pi,
-              "cvtpd2ps": cvtpd2ps,
-              "cvtpi2pd": cvtpi2pd,
-              "cvtpi2ps": cvtpi2ps,
-              "cvtps2dq": cvtps2dq,
-              "cvtps2pd": cvtps2pd,
-              "cvtps2pi": cvtps2pi,
-              "cvtsd2si": cvtsd2si,
-              "cvtsd2ss": cvtsd2ss,
-              "cvtsi2sd": cvtsi2sd,
-              "cvtsi2ss": cvtsi2ss,
-              "cvtss2sd": cvtss2sd,
-              "cvtss2si": cvtss2si,
-              "cvttpd2pi": cvttpd2pi,
-              "cvttpd2dq": cvttpd2dq,
-              "cvttps2dq": cvttps2dq,
-              "cvttps2pi": cvttps2pi,
-              "cvttsd2si": cvttsd2si,
-              "cvttss2si": cvttss2si,
-
-
-              "bndmov": bndmov,
-
-
-
-
-              "movss": movss,
-
-              "ucomiss": ucomiss,
-              "ucomisd": ucomisd,
-
-              # BMI operations
-              "blsi": blsi,
-              "andn": andn,
-              "bextr": bextr,
-              "blsmsk": blsmsk,
-              "blsr": blsr,
-              "tzcnt": tzcnt,
-              "bzhi": bzhi,
-
-              #
-              # MMX/AVX/SSE operations
-
-              # Arithmetic (integers)
-              #
-
-              # Additions
-              # SSE
-              "paddb": paddb,
-              "paddw": paddw,
-              "paddd": paddd,
-              "paddq": paddq,
-
-              # Substractions
-              # SSE
-              "psubb": psubb,
-              "psubw": psubw,
-              "psubd": psubd,
-              "psubq": psubq,
-
-              # Multiplications
-              # SSE
-              "pmullb": pmullb,
-              "pmullw": pmullw,
-              "pmulld": pmulld,
-              "pmullq": pmullq,
-              "pmulhub": pmulhub,
-              "pmulhuw": pmulhuw,
-              "pmulhud": pmulhud,
-              "pmulhuq": pmulhuq,
-              "pmulhb": pmulhb,
-              "pmulhw": pmulhw,
-              "pmulhd": pmulhd,
-              "pmulhq": pmulhq,
-              "pmuludq": pmuludq,
-
-              # Mix
-              # SSE
-              "pmaddwd": pmaddwd,
-              "psadbw": psadbw,
-              "pavgb": pavgb,
-              "pavgw": pavgw,
-
-              # Arithmetic (floating-point)
-              #
-
-              # Additions
-              # SSE
-              "addss": addss,
-              "addsd": addsd,
-              "addps": addps,
-              "addpd": addpd,
-
-              # Substractions
-              # SSE
-              "subss": subss,
-              "subsd": subsd,
-              "subps": subps,
-              "subpd": subpd,
-
-              # Multiplications
-              # SSE
-              "mulss": mulss,
-              "mulsd": mulsd,
-              "mulps": mulps,
-              "mulpd": mulpd,
-
-              # Divisions
-              # SSE
-              "divss": divss,
-              "divsd": divsd,
-              "divps": divps,
-              "divpd": divpd,
-
-              # Rounding
-              "roundss": roundss,
-              "roundsd": roundsd,
-
-              # Comparisons (floating-point)
-              #
-              "minps": minps,
-              "minpd": minpd,
-              "minss": minss,
-              "minsd": minsd,
-              "maxps": maxps,
-              "maxpd": maxpd,
-              "maxss": maxss,
-              "maxsd": maxsd,
-              "cmpeqps": cmpeqps,
-              "cmpeqpd": cmpeqpd,
-              "cmpeqss": cmpeqss,
-              "cmpeqsd": cmpeqsd,
-              "cmpltps": cmpltps,
-              "cmpltpd": cmpltpd,
-              "cmpltss": cmpltss,
-              "cmpltsd": cmpltsd,
-              "cmpleps": cmpleps,
-              "cmplepd": cmplepd,
-              "cmpless": cmpless,
-              "cmplesd": cmplesd,
-              "cmpunordps": cmpunordps,
-              "cmpunordpd": cmpunordpd,
-              "cmpunordss": cmpunordss,
-              "cmpunordsd": cmpunordsd,
-              "cmpneqps": cmpneqps,
-              "cmpneqpd": cmpneqpd,
-              "cmpneqss": cmpneqss,
-              "cmpneqsd": cmpneqsd,
-              "cmpnltps": cmpnltps,
-              "cmpnltpd": cmpnltpd,
-              "cmpnltss": cmpnltss,
-              "cmpnltsd": cmpnltsd,
-              "cmpnleps": cmpnleps,
-              "cmpnlepd": cmpnlepd,
-              "cmpnless": cmpnless,
-              "cmpnlesd": cmpnlesd,
-              "cmpordps": cmpordps,
-              "cmpordpd": cmpordpd,
-              "cmpordss": cmpordss,
-              "cmpordsd": cmpordsd,
-
-              # Logical (floating-point)
-              #
-
-              "pand": pand,
-              "pandn": pandn,
-              "por": por,
-
-              "rdmsr": rdmsr,
-              "wrmsr": wrmsr,
-              "pshufb": pshufb,
-              "pshufd": pshufd,
-              "pshuflw": pshuflw,
-              "pshufhw": pshufhw,
-
-              "psrlw": psrlw,
-              "psrld": psrld,
-              "psrlq": psrlq,
-              "psllw": psllw,
-              "pslld": pslld,
-              "psllq": psllq,
-              "pslldq": pslldq,
-              "psrldq": psrldq,
-              "psraw": psraw,
-              "psrad": psrad,
-
-              "palignr": palignr,
-
-              "pmaxub": pmaxub,
-              "pmaxuw": pmaxuw,
-              "pmaxud": pmaxud,
-              "pmaxsw": pmaxsw,
-
-              "pminub": pminub,
-              "pminuw": pminuw,
-              "pminud": pminud,
-
-              "pcmpeqb": pcmpeqb,
-              "pcmpeqw": pcmpeqw,
-              "pcmpeqd": pcmpeqd,
-              "pcmpeqq": pcmpeqq,
-
-              "pcmpgtb": pcmpgtb,
-              "pcmpgtw": pcmpgtw,
-              "pcmpgtd": pcmpgtd,
-              "pcmpgtq": pcmpgtq,
-
-              "punpckhbw": punpckhbw,
-              "punpckhwd": punpckhwd,
-              "punpckhdq": punpckhdq,
-              "punpckhqdq": punpckhqdq,
-
-
-              "punpcklbw": punpcklbw,
-              "punpcklwd": punpcklwd,
-              "punpckldq": punpckldq,
-              "punpcklqdq": punpcklqdq,
-
-              "pinsrb": pinsrb,
-              "pinsrw": pinsrw,
-              "pinsrd": pinsrd,
-              "pinsrq": pinsrq,
-
-              "pextrb": pextrb,
-              "pextrw": pextrw,
-              "pextrd": pextrd,
-              "pextrq": pextrq,
-
-              "unpckhps": unpckhps,
-              "unpckhpd": unpckhpd,
-              "unpcklps": unpcklps,
-              "unpcklpd": unpcklpd,
-
-              "movlpd": movlpd,
-              "movlps": movlps,
-              "movhpd": movhpd,
-              "movhps": movhpd,
-              "movlhps": movlhps,
-              "movhlps": movhlps,
-              "movdq2q": movdq2q,
-              "movq2dq": movq2dq,
-
-              "sqrtpd": sqrtpd,
-              "sqrtps": sqrtps,
-              "sqrtsd": sqrtsd,
-              "sqrtss": sqrtss,
-
-              "pmovmskb": pmovmskb,
-
-              "packsswb": packsswb,
-              "packssdw": packssdw,
-              "packuswb": packuswb,
-
-              "psubusb": psubusb,
-              "psubusw": psubusw,
-              "paddusb": paddusb,
-              "paddusw": paddusw,
-              "psubsb": psubsb,
-              "psubsw": psubsw,
-              "paddsb": paddsb,
-              "paddsw": paddsw,
-
-              "smsw": smsw,
-              "maskmovq": maskmovq,
-              "maskmovdqu": maskmovq,
-              "emms": emms,
-              "shufps": shufps,
-              "shufpd": shufpd,
-              "movmskps": movmskps,
-              "movmskpd": movmskpd,
-              "stmxcsr": stmxcsr,
-              "ldmxcsr": ldmxcsr,
-
-              # CET (Control-flow Enforcement Technology)
-              "incssp": incssp,
-              "rdssp": rdssp,
-              "saveprevssp": saveprevssp,
-              "rstorssp": rstorssp,
-              "wrss": wrss,
-              "wruss": wruss,
-              "setssbsy": setssbsy,
-              "clrssbsy": clrssbsy,
-              "endbr64": endbr64,
-              "endbr32": endbr32,
-              "fxsave": fxsave,
-              "fxrstor": fxrstor,
-              }
-
-
-class Lifter_X86_16(Lifter):
-
-    def __init__(self, loc_db):
-        Lifter.__init__(self, mn_x86, 16, loc_db)
-        self.do_stk_segm = False
-        self.do_ds_segm = False
-        self.do_str_segm = False
-        self.do_all_segm = False
-        self.pc = IP
-        self.sp = SP
-        self.IRDst = m2_expr.ExprId('IRDst', 16)
-        # Size of memory pointer access in IR
-        # 16 bit mode memory accesses may be greater than 16 bits
-        # 32 bit size may be enough
-        self.addrsize = 32
-
-    def mod_pc(self, instr, instr_ir, extra_ir):
-        pass
-
-    def ExprMem(self, ptr, size):
-        """Generate a memory access to @ptr
-        The ptr is resized to a fixed size self.addrsize
-
-        @ptr: Expr instance to the memory address
-        @size: size of the memory"""
-
-        return m2_expr.ExprMem(expraddr(self.addrsize, ptr), size)
-
-    def gen_segm_expr(self, selector, addr):
-        ptr = m2_expr.ExprOp(
-            'segm',
-            selector,
-            addr.zeroExtend(self.addrsize)
-        )
-
-        return ptr
-
-    def get_ir(self, instr):
-        args = instr.args[:]
-        args = [arg.replace_expr(float_replace) for arg in args]
-        args = fix_mem_args_size(instr, *args)
-        my_ss = None
-        if self.do_ds_segm:
-            my_ss = DS
-        if self.do_all_segm and instr.additional_info.g2.value:
-            my_ss = {1: CS, 2: SS, 3: DS, 4: ES, 5: FS, 6: GS}[
-                instr.additional_info.g2.value]
-        if my_ss is not None:
-            for i, a in enumerate(args):
-                if a.is_mem() and not is_mem_segm(a):
-                    args[i] = self.ExprMem(m2_expr.ExprOp('segm', my_ss,
-                                                          a.ptr), a.size)
-
-        if not instr.name.lower() in mnemo_func:
-            raise NotImplementedError(
-                "Mnemonic %s not implemented" % instr.name)
-
-        instr_ir, extra_ir = mnemo_func[
-            instr.name.lower()](self, instr, *args)
-        self.mod_pc(instr, instr_ir, extra_ir)
-        instr.additional_info.except_on_instr = False
-        if instr.additional_info.g1.value & 14 == 0 or \
-                not instr.name in repeat_mn:
-            return instr_ir, extra_ir
-        if instr.name == "MOVSD" and len(instr.args) == 2:
-            return instr_ir, extra_ir
-
-        instr.additional_info.except_on_instr = True
-        admode = instr.v_admode()
-        c_reg = mRCX[instr.mode][:admode]
-
-        zf_val = None
-        # set if zf is tested (cmps, scas)
-        for e in instr_ir:  # +[updt_c]:
-            if e.dst == zf:
-                zf_val = e.src
-
-        cond_dec = m2_expr.ExprCond(c_reg - m2_expr.ExprInt(1, c_reg.size),
-                                    m2_expr.ExprInt(0, 1), m2_expr.ExprInt(1, 1))
-        # end condition
-        if zf_val is None:
-            c_cond = cond_dec
-        elif instr.additional_info.g1.value & 2:  # REPNE and REPNZ
-            c_cond = cond_dec | zf
-        elif instr.additional_info.g1.value & 12:  # REPE, REP and REPZ
-            c_cond = cond_dec | (zf ^ m2_expr.ExprInt(1, 1))
-
-        # gen while
-        loc_do, loc_do_expr = self.gen_loc_key_and_expr(self.IRDst.size)
-        loc_end, loc_end_expr = self.gen_loc_key_and_expr(self.IRDst.size)
-        loc_skip = self.get_next_loc_key(instr)
-        loc_skip_expr = m2_expr.ExprLoc(loc_skip, self.IRDst.size)
-        loc_next = self.get_next_loc_key(instr)
-        loc_next_expr = m2_expr.ExprLoc(loc_next, self.IRDst.size)
-
-        fix_next_loc = {loc_next_expr: loc_end_expr}
-        new_extra_ir = [irblock.modify_exprs(mod_src=lambda expr: expr.replace_expr(fix_next_loc))
-                        for irblock in extra_ir]
-
-        cond_bloc = []
-        cond_bloc.append(m2_expr.ExprAssign(c_reg,
-                                         c_reg - m2_expr.ExprInt(1,
-                                                                 c_reg.size)))
-        cond_bloc.append(m2_expr.ExprAssign(self.IRDst, m2_expr.ExprCond(c_cond,
-                                                                      loc_skip_expr,
-                                                                      loc_do_expr)))
-        cond_bloc = IRBlock(self.loc_db, loc_end, [AssignBlock(cond_bloc, instr)])
-        e_do = instr_ir
-
-        c = IRBlock(self.loc_db, loc_do, [AssignBlock(e_do, instr)])
-        e_n = [m2_expr.ExprAssign(self.IRDst, m2_expr.ExprCond(c_reg, loc_do_expr,
-                                                            loc_skip_expr))]
-        return e_n, [cond_bloc, c] + new_extra_ir
-
-    def expr_fix_regs_for_mode(self, e, mode=64):
-        return e.replace_expr(replace_regs[mode])
-
-    def expraff_fix_regs_for_mode(self, e, mode=64):
-        dst = self.expr_fix_regs_for_mode(e.dst, mode)
-        src = self.expr_fix_regs_for_mode(e.src, mode)
-        return m2_expr.ExprAssign(dst, src)
-
-    def irbloc_fix_regs_for_mode(self, irblock, mode=64):
-        irs = []
-        for assignblk in irblock:
-            new_assignblk = dict(assignblk)
-            for dst, src in viewitems(assignblk):
-                del new_assignblk[dst]
-                # Special case for 64 bits:
-                # If destination is a 32 bit reg, zero extend the 64 bit reg
-                if mode == 64:
-                    if (isinstance(dst, m2_expr.ExprId) and
-                            dst.size == 32 and
-                            dst in replace_regs[64]):
-                        src = src.zeroExtend(64)
-                        dst = replace_regs[64][dst].arg
-                dst = self.expr_fix_regs_for_mode(dst, mode)
-                src = self.expr_fix_regs_for_mode(src, mode)
-                new_assignblk[dst] = src
-            irs.append(AssignBlock(new_assignblk, assignblk.instr))
-        return IRBlock(self.loc_db, irblock.loc_key, irs)
-
-
-class Lifter_X86_32(Lifter_X86_16):
-
-    def __init__(self, loc_db):
-        Lifter.__init__(self, mn_x86, 32, loc_db)
-        self.do_stk_segm = False
-        self.do_ds_segm = False
-        self.do_str_segm = False
-        self.do_all_segm = False
-        self.pc = EIP
-        self.sp = ESP
-        self.IRDst = m2_expr.ExprId('IRDst', 32)
-        self.addrsize = 32
-
-
-class Lifter_X86_64(Lifter_X86_16):
-
-    def __init__(self, loc_db):
-        Lifter.__init__(self, mn_x86, 64, loc_db)
-        self.do_stk_segm = False
-        self.do_ds_segm = False
-        self.do_str_segm = False
-        self.do_all_segm = False
-        self.pc = RIP
-        self.sp = RSP
-        self.IRDst = m2_expr.ExprId('IRDst', 64)
-        self.addrsize = 64
-
-    def mod_pc(self, instr, instr_ir, extra_ir):
-        # fix RIP for 64 bit
-        pc_fixed = {self.pc: m2_expr.ExprInt(instr.offset + instr.l, 64)}
-
-        for i, expr in enumerate(instr_ir):
-            dst, src = expr.dst, expr.src
-            if dst != self.pc:
-                dst = dst.replace_expr(pc_fixed)
-            src = src.replace_expr(pc_fixed)
-            instr_ir[i] = m2_expr.ExprAssign(dst, src)
-
-        for idx, irblock in enumerate(extra_ir):
-            extra_ir[idx] = irblock.modify_exprs(lambda expr: expr.replace_expr(pc_fixed) \
-                                                 if expr != self.pc else expr,
-                                                 lambda expr: expr.replace_expr(pc_fixed))