PyObject *
PyCode_Optimize(PyObject *code, PyObject* consts, PyObject *names,
PyObject *lineno_obj)
{
Py_ssize_t i, j, codelen;
int nops, h, adj;
int tgt, tgttgt, opcode;
unsigned char *codestr = NULL;
unsigned char *lineno;
int *addrmap = NULL;
int new_line, cum_orig_line, last_line, tabsiz;
int cumlc=0, lastlc=0; /* Count runs of consecutive LOAD_CONSTs */
unsigned int *blocks = NULL;
char *name;
/* Bail out if an exception is set */
if (PyErr_Occurred())
goto exitError;
/* Bypass optimization when the lineno table is too complex */
assert(PyBytes_Check(lineno_obj));
lineno = (unsigned char*)PyBytes_AS_STRING(lineno_obj);
tabsiz = PyBytes_GET_SIZE(lineno_obj);
if (memchr(lineno, 255, tabsiz) != NULL)
goto exitUnchanged;
/* Avoid situations where jump retargeting could overflow */
assert(PyBytes_Check(code));
codelen = PyBytes_GET_SIZE(code);
if (codelen > 32700)
goto exitUnchanged;
/* Make a modifiable copy of the code string */
codestr = (unsigned char *)PyMem_Malloc(codelen);
if (codestr == NULL)
goto exitError;
codestr = (unsigned char *)memcpy(codestr,
PyBytes_AS_STRING(code), codelen);
/* Verify that RETURN_VALUE terminates the codestring. This allows
the various transformation patterns to look ahead several
instructions without additional checks to make sure they are not
looking beyond the end of the code string.
*/
if (codestr[codelen-1] != RETURN_VALUE)
goto exitUnchanged;
/* Mapping to new jump targets after NOPs are removed */
addrmap = (int *)PyMem_Malloc(codelen * sizeof(int));
if (addrmap == NULL)
goto exitError;
blocks = markblocks(codestr, codelen);
if (blocks == NULL)
goto exitError;
assert(PyList_Check(consts));
for (i=0 ; i<codelen ; i += CODESIZE(codestr[i])) {
reoptimize_current:
opcode = codestr[i];
lastlc = cumlc;
cumlc = 0;
switch (opcode) {
/* Replace UNARY_NOT POP_JUMP_IF_FALSE
with POP_JUMP_IF_TRUE */
case UNARY_NOT:
if (codestr[i+1] != POP_JUMP_IF_FALSE
|| !ISBASICBLOCK(blocks,i,4))
continue;
j = GETARG(codestr, i+1);
codestr[i] = POP_JUMP_IF_TRUE;
SETARG(codestr, i, j);
codestr[i+3] = NOP;
goto reoptimize_current;
/* not a is b --> a is not b
not a in b --> a not in b
not a is not b --> a is b
not a not in b --> a in b
*/
case COMPARE_OP:
j = GETARG(codestr, i);
if (j < 6 || j > 9 ||
codestr[i+3] != UNARY_NOT ||
!ISBASICBLOCK(blocks,i,4))
continue;
SETARG(codestr, i, (j^1));
codestr[i+3] = NOP;
break;
/* Replace LOAD_GLOBAL/LOAD_NAME None/True/False
with LOAD_CONST None/True/False */
case LOAD_NAME:
case LOAD_GLOBAL:
j = GETARG(codestr, i);
name = _PyUnicode_AsString(PyTuple_GET_ITEM(names, j));
h = load_global(codestr, i, name, consts);
if (h < 0)
goto exitError;
else if (h == 0)
continue;
cumlc = lastlc + 1;
break;
/* Skip over LOAD_CONST trueconst
POP_JUMP_IF_FALSE xx. This improves
"while 1" performance. */
case LOAD_CONST:
cumlc = lastlc + 1;
j = GETARG(codestr, i);
if (codestr[i+3] != POP_JUMP_IF_FALSE ||
!ISBASICBLOCK(blocks,i,6) ||
!PyObject_IsTrue(PyList_GET_ITEM(consts, j)))
continue;
memset(codestr+i, NOP, 6);
cumlc = 0;
break;
/* Try to fold tuples of constants (includes a case for lists and sets
which are only used for "in" and "not in" tests).
Skip over BUILD_SEQN 1 UNPACK_SEQN 1.
Replace BUILD_SEQN 2 UNPACK_SEQN 2 with ROT2.
Replace BUILD_SEQN 3 UNPACK_SEQN 3 with ROT3 ROT2. */
case BUILD_TUPLE:
case BUILD_LIST:
case BUILD_SET:
j = GETARG(codestr, i);
h = i - 3 * j;
if (h >= 0 &&
j <= lastlc &&
((opcode == BUILD_TUPLE &&
ISBASICBLOCK(blocks, h, 3*(j+1))) ||
((opcode == BUILD_LIST || opcode == BUILD_SET) &&
codestr[i+3]==COMPARE_OP &&
ISBASICBLOCK(blocks, h, 3*(j+2)) &&
(GETARG(codestr,i+3)==6 ||
GETARG(codestr,i+3)==7))) &&
tuple_of_constants(&codestr[h], j, consts)) {
assert(codestr[i] == LOAD_CONST);
cumlc = 1;
break;
}
if (codestr[i+3] != UNPACK_SEQUENCE ||
!ISBASICBLOCK(blocks,i,6) ||
j != GETARG(codestr, i+3) ||
opcode == BUILD_SET)
continue;
if (j == 1) {
memset(codestr+i, NOP, 6);
} else if (j == 2) {
codestr[i] = ROT_TWO;
memset(codestr+i+1, NOP, 5);
} else if (j == 3) {
codestr[i] = ROT_THREE;
codestr[i+1] = ROT_TWO;
memset(codestr+i+2, NOP, 4);
}
break;
/* Fold binary ops on constants.
LOAD_CONST c1 LOAD_CONST c2 BINOP --> LOAD_CONST binop(c1,c2) */
case BINARY_POWER:
case BINARY_MULTIPLY:
case BINARY_TRUE_DIVIDE:
case BINARY_FLOOR_DIVIDE:
case BINARY_MODULO:
case BINARY_ADD:
case BINARY_SUBTRACT:
case BINARY_SUBSCR:
case BINARY_LSHIFT:
case BINARY_RSHIFT:
case BINARY_AND:
case BINARY_XOR:
case BINARY_OR:
if (lastlc >= 2 &&
ISBASICBLOCK(blocks, i-6, 7) &&
fold_binops_on_constants(&codestr[i-6], consts)) {
i -= 2;
assert(codestr[i] == LOAD_CONST);
cumlc = 1;
}
break;
/* Fold unary ops on constants.
LOAD_CONST c1 UNARY_OP --> LOAD_CONST unary_op(c) */
case UNARY_NEGATIVE:
case UNARY_INVERT:
case UNARY_POSITIVE:
if (lastlc >= 1 &&
ISBASICBLOCK(blocks, i-3, 4) &&
fold_unaryops_on_constants(&codestr[i-3], consts)) {
i -= 2;
assert(codestr[i] == LOAD_CONST);
cumlc = 1;
}
break;
/* Simplify conditional jump to conditional jump where the
result of the first test implies the success of a similar
test or the failure of the opposite test.
Arises in code like:
"if a and b:"
"if a or b:"
"a and b or c"
"(a and b) and c"
x:JUMP_IF_FALSE_OR_POP y y:JUMP_IF_FALSE_OR_POP z
--> x:JUMP_IF_FALSE_OR_POP z
x:JUMP_IF_FALSE_OR_POP y y:JUMP_IF_TRUE_OR_POP z
--> x:POP_JUMP_IF_FALSE y+3
where y+3 is the instruction following the second test.
*/
case JUMP_IF_FALSE_OR_POP:
case JUMP_IF_TRUE_OR_POP:
tgt = GETJUMPTGT(codestr, i);
j = codestr[tgt];
if (CONDITIONAL_JUMP(j)) {
/* NOTE: all possible jumps here are
absolute! */
if (JUMPS_ON_TRUE(j) == JUMPS_ON_TRUE(opcode)) {
/* The second jump will be
taken iff the first is. */
tgttgt = GETJUMPTGT(codestr, tgt);
/* The current opcode inherits
its target's stack behaviour */
codestr[i] = j;
SETARG(codestr, i, tgttgt);
goto reoptimize_current;
} else {
/* The second jump is not taken
if the first is (so jump past
it), and all conditional
jumps pop their argument when
they're not taken (so change
the first jump to pop its
argument when it's taken). */
if (JUMPS_ON_TRUE(opcode))
codestr[i] = POP_JUMP_IF_TRUE;
else
codestr[i] = POP_JUMP_IF_FALSE;
SETARG(codestr, i, (tgt + 3));
goto reoptimize_current;
}
}
/* Intentional fallthrough */
/* Replace jumps to unconditional jumps */
case POP_JUMP_IF_FALSE:
case POP_JUMP_IF_TRUE:
case FOR_ITER:
case JUMP_FORWARD:
case JUMP_ABSOLUTE:
case CONTINUE_LOOP:
case SETUP_LOOP:
case SETUP_EXCEPT:
case SETUP_FINALLY:
case SETUP_WITH:
tgt = GETJUMPTGT(codestr, i);
/* Replace JUMP_* to a RETURN into just a RETURN */
if (UNCONDITIONAL_JUMP(opcode) &&
codestr[tgt] == RETURN_VALUE) {
codestr[i] = RETURN_VALUE;
memset(codestr+i+1, NOP, 2);
continue;
}
if (!UNCONDITIONAL_JUMP(codestr[tgt]))
continue;
tgttgt = GETJUMPTGT(codestr, tgt);
if (opcode == JUMP_FORWARD) /* JMP_ABS can go backwards */
opcode = JUMP_ABSOLUTE;
if (!ABSOLUTE_JUMP(opcode))
tgttgt -= i + 3; /* Calc relative jump addr */
if (tgttgt < 0) /* No backward relative jumps */
continue;
codestr[i] = opcode;
SETARG(codestr, i, tgttgt);
break;
case EXTENDED_ARG:
if (codestr[i+3] != MAKE_FUNCTION)
goto exitUnchanged;
/* don't visit MAKE_FUNCTION as GETARG will be wrong */
i += 3;
break;
/* Replace RETURN LOAD_CONST None RETURN with just RETURN */
/* Remove unreachable JUMPs after RETURN */
case RETURN_VALUE:
if (i+4 >= codelen)
continue;
if (codestr[i+4] == RETURN_VALUE &&
ISBASICBLOCK(blocks,i,5))
memset(codestr+i+1, NOP, 4);
else if (UNCONDITIONAL_JUMP(codestr[i+1]) &&
ISBASICBLOCK(blocks,i,4))
memset(codestr+i+1, NOP, 3);
break;
}
}
/* Fixup linenotab */
for (i=0, nops=0 ; i<codelen ; i += CODESIZE(codestr[i])) {
addrmap[i] = i - nops;
if (codestr[i] == NOP)
nops++;
}
cum_orig_line = 0;
last_line = 0;
for (i=0 ; i < tabsiz ; i+=2) {
cum_orig_line += lineno[i];
new_line = addrmap[cum_orig_line];
assert (new_line - last_line < 255);
lineno[i] =((unsigned char)(new_line - last_line));
last_line = new_line;
}
/* Remove NOPs and fixup jump targets */
for (i=0, h=0 ; i<codelen ; ) {
opcode = codestr[i];
switch (opcode) {
case NOP:
i++;
continue;
case JUMP_ABSOLUTE:
case CONTINUE_LOOP:
case POP_JUMP_IF_FALSE:
case POP_JUMP_IF_TRUE:
case JUMP_IF_FALSE_OR_POP:
case JUMP_IF_TRUE_OR_POP:
j = addrmap[GETARG(codestr, i)];
SETARG(codestr, i, j);
break;
case FOR_ITER:
case JUMP_FORWARD:
case SETUP_LOOP:
case SETUP_EXCEPT:
case SETUP_FINALLY:
case SETUP_WITH:
j = addrmap[GETARG(codestr, i) + i + 3] - addrmap[i] - 3;
SETARG(codestr, i, j);
break;
}
adj = CODESIZE(opcode);
while (adj--)
codestr[h++] = codestr[i++];
}
assert(h + nops == codelen);
code = PyBytes_FromStringAndSize((char *)codestr, h);
PyMem_Free(addrmap);
PyMem_Free(codestr);
PyMem_Free(blocks);
return code;
exitError:
code = NULL;
exitUnchanged:
if (blocks != NULL)
PyMem_Free(blocks);
if (addrmap != NULL)
PyMem_Free(addrmap);
if (codestr != NULL)
PyMem_Free(codestr);
Py_XINCREF(code);
return code;
}
uint STDCALL ge_load( pbuf in )
{
pubyte cur, end, ptemp;
uint size;
pgehead phead = ( pgehead )buf_ptr( in );
// Проверка заголовка и целостности
// Сравниваем с 'GE' с двумя нулями на конце
if ( *( puint )phead != GE_STRING )//0x00004547 )
msg( MNotGE | MSG_EXIT );
if ( phead->crc != crc( ( pubyte )phead + 12, phead->size - 12, 0xFFFFFFFF ))
msg( MCrcGE | MSG_EXIT );
if ( phead->vermajor != GEVER_MAJOR || phead->verminor > GEVER_MINOR )
msg( MVerGE | MSG_EXIT );
_vm.loadmode = VMLOAD_GE;
_vm.icnv = arr_count( &_vm.objtbl ) - KERNEL_COUNT;
// print("icnv=%i\n", _vm.icnv );
cur = ( pubyte )phead + phead->headsize;
end = ( pubyte )phead + phead->size;
while ( cur < end )
{
ptemp = cur + 5; // type + flag
_vm.ipack = ( *( puint )( cur + 1 )) & GHCOM_PACK ? 1 : 0;
size = load_bwd( &ptemp );
ptemp = cur;
// print("size=%i type=%i flag = %x\n", size, *cur, *( puint )( cur + 1 ) );
switch ( *cur )
{
case OVM_NONE:
load_none();
break;
case OVM_BYTECODE:
load_bytecode( &cur, VMLOAD_GE );
break;
case OVM_EXFUNC:
load_exfunc( &cur, 0 );
_vm.loadmode = VMLOAD_GE;
break;
case OVM_TYPE:
load_type( &cur );
break;
case OVM_GLOBAL:
load_global( &cur );
break;
case OVM_DEFINE:
load_define( &cur );
break;
case OVM_IMPORT:
load_import( &cur );
break;
case OVM_RESOURCE:
load_resource( &cur );
break;
case OVM_ALIAS:
load_alias( &cur );
break;
default:
msg( MUnkGE | MSG_DVAL, cur - ( pubyte )phead );
}
cur = ptemp + size;
}
_vm.loadmode = VMLOAD_G;
_vm.icnv = 0;
return 1;
}