gh-116381: Specialize CONTAINS_OP (GH-116385)

* Specialize CONTAINS_OP

* 📜🤖 Added by blurb_it.

* Add PyAPI_FUNC for JIT

---------

Co-authored-by: blurb-it[bot] <43283697+blurb-it[bot]@users.noreply.github.com>
This commit is contained in:
Ken Jin 2024-03-07 03:30:11 +08:00 committed by GitHub
parent 73807eb634
commit 7114cf20c0
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
21 changed files with 645 additions and 194 deletions

View file

@ -2237,13 +2237,75 @@ dummy_func(
b = res ? Py_True : Py_False;
}
inst(CONTAINS_OP, (left, right -- b)) {
family(CONTAINS_OP, INLINE_CACHE_ENTRIES_CONTAINS_OP) = {
CONTAINS_OP_LIST,
CONTAINS_OP_SET,
CONTAINS_OP_TUPLE,
CONTAINS_OP_DICT,
CONTAINS_OP_STR,
};
op(_CONTAINS_OP, (left, right -- b)) {
int res = PySequence_Contains(right, left);
DECREF_INPUTS();
ERROR_IF(res < 0, error);
b = (res ^ oparg) ? Py_True : Py_False;
}
specializing op(_SPECIALIZE_CONTAINS_OP, (counter/1, left, right -- left, right)) {
#if ENABLE_SPECIALIZATION
if (ADAPTIVE_COUNTER_IS_ZERO(counter)) {
next_instr = this_instr;
_Py_Specialize_ContainsOp(right, next_instr);
DISPATCH_SAME_OPARG();
}
STAT_INC(CONTAINS_OP, deferred);
DECREMENT_ADAPTIVE_COUNTER(this_instr[1].cache);
#endif /* ENABLE_SPECIALIZATION */
}
macro(CONTAINS_OP) = _SPECIALIZE_CONTAINS_OP + _CONTAINS_OP;
inst(CONTAINS_OP_LIST, (unused/1, left, right -- b)) {
DEOPT_IF(!PyList_CheckExact(right));
int res = _PyList_Contains(right, left);
DECREF_INPUTS();
ERROR_IF(res < 0, error);
b = (res ^ oparg) ? Py_True : Py_False;
}
inst(CONTAINS_OP_SET, (unused/1, left, right -- b)) {
DEOPT_IF(!PySet_CheckExact(right));
int res = _PySet_Contains((PySetObject *)right, left);
DECREF_INPUTS();
ERROR_IF(res < 0, error);
b = (res ^ oparg) ? Py_True : Py_False;
}
inst(CONTAINS_OP_TUPLE, (unused/1, left, right -- b)) {
DEOPT_IF(!PyTuple_CheckExact(right));
int res = _PyTuple_Contains((PyTupleObject *)right, left);
DECREF_INPUTS();
ERROR_IF(res < 0, error);
b = (res ^ oparg) ? Py_True : Py_False;
}
inst(CONTAINS_OP_DICT, (unused/1, left, right -- b)) {
DEOPT_IF(!PyDict_CheckExact(right));
int res = PyDict_Contains(right, left);
DECREF_INPUTS();
ERROR_IF(res < 0, error);
b = (res ^ oparg) ? Py_True : Py_False;
}
inst(CONTAINS_OP_STR, (unused/1, left, right -- b)) {
DEOPT_IF(!PyUnicode_CheckExact(right));
int res = PyUnicode_Contains(right, left);
DECREF_INPUTS();
ERROR_IF(res < 0, error);
b = (res ^ oparg) ? Py_True : Py_False;
}
inst(CHECK_EG_MATCH, (exc_value, match_type -- rest, match)) {
if (_PyEval_CheckExceptStarTypeValid(tstate, match_type) < 0) {
DECREF_INPUTS();

View file

@ -2189,6 +2189,96 @@
break;
}
case _CONTAINS_OP_LIST: {
PyObject *right;
PyObject *left;
PyObject *b;
oparg = CURRENT_OPARG();
right = stack_pointer[-1];
left = stack_pointer[-2];
if (!PyList_CheckExact(right)) goto deoptimize;
int res = _PyList_Contains(right, left);
Py_DECREF(left);
Py_DECREF(right);
if (res < 0) goto pop_2_error_tier_two;
b = (res ^ oparg) ? Py_True : Py_False;
stack_pointer[-2] = b;
stack_pointer += -1;
break;
}
case _CONTAINS_OP_SET: {
PyObject *right;
PyObject *left;
PyObject *b;
oparg = CURRENT_OPARG();
right = stack_pointer[-1];
left = stack_pointer[-2];
if (!PySet_CheckExact(right)) goto deoptimize;
int res = _PySet_Contains((PySetObject *)right, left);
Py_DECREF(left);
Py_DECREF(right);
if (res < 0) goto pop_2_error_tier_two;
b = (res ^ oparg) ? Py_True : Py_False;
stack_pointer[-2] = b;
stack_pointer += -1;
break;
}
case _CONTAINS_OP_TUPLE: {
PyObject *right;
PyObject *left;
PyObject *b;
oparg = CURRENT_OPARG();
right = stack_pointer[-1];
left = stack_pointer[-2];
if (!PyTuple_CheckExact(right)) goto deoptimize;
int res = _PyTuple_Contains((PyTupleObject *)right, left);
Py_DECREF(left);
Py_DECREF(right);
if (res < 0) goto pop_2_error_tier_two;
b = (res ^ oparg) ? Py_True : Py_False;
stack_pointer[-2] = b;
stack_pointer += -1;
break;
}
case _CONTAINS_OP_DICT: {
PyObject *right;
PyObject *left;
PyObject *b;
oparg = CURRENT_OPARG();
right = stack_pointer[-1];
left = stack_pointer[-2];
if (!PyDict_CheckExact(right)) goto deoptimize;
int res = PyDict_Contains(right, left);
Py_DECREF(left);
Py_DECREF(right);
if (res < 0) goto pop_2_error_tier_two;
b = (res ^ oparg) ? Py_True : Py_False;
stack_pointer[-2] = b;
stack_pointer += -1;
break;
}
case _CONTAINS_OP_STR: {
PyObject *right;
PyObject *left;
PyObject *b;
oparg = CURRENT_OPARG();
right = stack_pointer[-1];
left = stack_pointer[-2];
if (!PyUnicode_CheckExact(right)) goto deoptimize;
int res = PyUnicode_Contains(right, left);
Py_DECREF(left);
Py_DECREF(right);
if (res < 0) goto pop_2_error_tier_two;
b = (res ^ oparg) ? Py_True : Py_False;
stack_pointer[-2] = b;
stack_pointer += -1;
break;
}
case _CHECK_EG_MATCH: {
PyObject *match_type;
PyObject *exc_value;

View file

@ -2127,14 +2127,144 @@
TARGET(CONTAINS_OP) {
frame->instr_ptr = next_instr;
next_instr += 1;
next_instr += 2;
INSTRUCTION_STATS(CONTAINS_OP);
PREDICTED(CONTAINS_OP);
_Py_CODEUNIT *this_instr = next_instr - 2;
(void)this_instr;
PyObject *right;
PyObject *left;
PyObject *b;
// _SPECIALIZE_CONTAINS_OP
right = stack_pointer[-1];
left = stack_pointer[-2];
int res = PySequence_Contains(right, left);
{
uint16_t counter = read_u16(&this_instr[1].cache);
(void)counter;
#if ENABLE_SPECIALIZATION
if (ADAPTIVE_COUNTER_IS_ZERO(counter)) {
next_instr = this_instr;
_Py_Specialize_ContainsOp(right, next_instr);
DISPATCH_SAME_OPARG();
}
STAT_INC(CONTAINS_OP, deferred);
DECREMENT_ADAPTIVE_COUNTER(this_instr[1].cache);
#endif /* ENABLE_SPECIALIZATION */
}
// _CONTAINS_OP
{
int res = PySequence_Contains(right, left);
Py_DECREF(left);
Py_DECREF(right);
if (res < 0) goto pop_2_error;
b = (res ^ oparg) ? Py_True : Py_False;
}
stack_pointer[-2] = b;
stack_pointer += -1;
DISPATCH();
}
TARGET(CONTAINS_OP_DICT) {
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(CONTAINS_OP_DICT);
static_assert(INLINE_CACHE_ENTRIES_CONTAINS_OP == 1, "incorrect cache size");
PyObject *right;
PyObject *left;
PyObject *b;
/* Skip 1 cache entry */
right = stack_pointer[-1];
left = stack_pointer[-2];
DEOPT_IF(!PyDict_CheckExact(right), CONTAINS_OP);
int res = PyDict_Contains(right, left);
Py_DECREF(left);
Py_DECREF(right);
if (res < 0) goto pop_2_error;
b = (res ^ oparg) ? Py_True : Py_False;
stack_pointer[-2] = b;
stack_pointer += -1;
DISPATCH();
}
TARGET(CONTAINS_OP_LIST) {
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(CONTAINS_OP_LIST);
static_assert(INLINE_CACHE_ENTRIES_CONTAINS_OP == 1, "incorrect cache size");
PyObject *right;
PyObject *left;
PyObject *b;
/* Skip 1 cache entry */
right = stack_pointer[-1];
left = stack_pointer[-2];
DEOPT_IF(!PyList_CheckExact(right), CONTAINS_OP);
int res = _PyList_Contains(right, left);
Py_DECREF(left);
Py_DECREF(right);
if (res < 0) goto pop_2_error;
b = (res ^ oparg) ? Py_True : Py_False;
stack_pointer[-2] = b;
stack_pointer += -1;
DISPATCH();
}
TARGET(CONTAINS_OP_SET) {
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(CONTAINS_OP_SET);
static_assert(INLINE_CACHE_ENTRIES_CONTAINS_OP == 1, "incorrect cache size");
PyObject *right;
PyObject *left;
PyObject *b;
/* Skip 1 cache entry */
right = stack_pointer[-1];
left = stack_pointer[-2];
DEOPT_IF(!PySet_CheckExact(right), CONTAINS_OP);
int res = _PySet_Contains((PySetObject *)right, left);
Py_DECREF(left);
Py_DECREF(right);
if (res < 0) goto pop_2_error;
b = (res ^ oparg) ? Py_True : Py_False;
stack_pointer[-2] = b;
stack_pointer += -1;
DISPATCH();
}
TARGET(CONTAINS_OP_STR) {
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(CONTAINS_OP_STR);
static_assert(INLINE_CACHE_ENTRIES_CONTAINS_OP == 1, "incorrect cache size");
PyObject *right;
PyObject *left;
PyObject *b;
/* Skip 1 cache entry */
right = stack_pointer[-1];
left = stack_pointer[-2];
DEOPT_IF(!PyUnicode_CheckExact(right), CONTAINS_OP);
int res = PyUnicode_Contains(right, left);
Py_DECREF(left);
Py_DECREF(right);
if (res < 0) goto pop_2_error;
b = (res ^ oparg) ? Py_True : Py_False;
stack_pointer[-2] = b;
stack_pointer += -1;
DISPATCH();
}
TARGET(CONTAINS_OP_TUPLE) {
frame->instr_ptr = next_instr;
next_instr += 2;
INSTRUCTION_STATS(CONTAINS_OP_TUPLE);
static_assert(INLINE_CACHE_ENTRIES_CONTAINS_OP == 1, "incorrect cache size");
PyObject *right;
PyObject *left;
PyObject *b;
/* Skip 1 cache entry */
right = stack_pointer[-1];
left = stack_pointer[-2];
DEOPT_IF(!PyTuple_CheckExact(right), CONTAINS_OP);
int res = _PyTuple_Contains((PyTupleObject *)right, left);
Py_DECREF(left);
Py_DECREF(right);
if (res < 0) goto pop_2_error;

View file

@ -182,6 +182,11 @@ static void *opcode_targets[256] = {
&&TARGET_COMPARE_OP_FLOAT,
&&TARGET_COMPARE_OP_INT,
&&TARGET_COMPARE_OP_STR,
&&TARGET_CONTAINS_OP_DICT,
&&TARGET_CONTAINS_OP_LIST,
&&TARGET_CONTAINS_OP_SET,
&&TARGET_CONTAINS_OP_STR,
&&TARGET_CONTAINS_OP_TUPLE,
&&TARGET_FOR_ITER_GEN,
&&TARGET_FOR_ITER_LIST,
&&TARGET_FOR_ITER_RANGE,
@ -230,11 +235,6 @@ static void *opcode_targets[256] = {
&&_unknown_opcode,
&&_unknown_opcode,
&&_unknown_opcode,
&&_unknown_opcode,
&&_unknown_opcode,
&&_unknown_opcode,
&&_unknown_opcode,
&&_unknown_opcode,
&&TARGET_INSTRUMENTED_RESUME,
&&TARGET_INSTRUMENTED_END_FOR,
&&TARGET_INSTRUMENTED_END_SEND,

View file

@ -1206,6 +1206,51 @@
break;
}
case _CONTAINS_OP_LIST: {
_Py_UopsSymbol *b;
b = sym_new_unknown(ctx);
if (b == NULL) goto out_of_space;
stack_pointer[-2] = b;
stack_pointer += -1;
break;
}
case _CONTAINS_OP_SET: {
_Py_UopsSymbol *b;
b = sym_new_unknown(ctx);
if (b == NULL) goto out_of_space;
stack_pointer[-2] = b;
stack_pointer += -1;
break;
}
case _CONTAINS_OP_TUPLE: {
_Py_UopsSymbol *b;
b = sym_new_unknown(ctx);
if (b == NULL) goto out_of_space;
stack_pointer[-2] = b;
stack_pointer += -1;
break;
}
case _CONTAINS_OP_DICT: {
_Py_UopsSymbol *b;
b = sym_new_unknown(ctx);
if (b == NULL) goto out_of_space;
stack_pointer[-2] = b;
stack_pointer += -1;
break;
}
case _CONTAINS_OP_STR: {
_Py_UopsSymbol *b;
b = sym_new_unknown(ctx);
if (b == NULL) goto out_of_space;
stack_pointer[-2] = b;
stack_pointer += -1;
break;
}
case _CHECK_EG_MATCH: {
_Py_UopsSymbol *rest;
_Py_UopsSymbol *match;

View file

@ -103,6 +103,7 @@ _Py_GetSpecializationStats(void) {
return NULL;
}
int err = 0;
err += add_stat_dict(stats, CONTAINS_OP, "contains_op");
err += add_stat_dict(stats, LOAD_SUPER_ATTR, "load_super_attr");
err += add_stat_dict(stats, LOAD_ATTR, "load_attr");
err += add_stat_dict(stats, LOAD_GLOBAL, "load_global");
@ -2561,6 +2562,43 @@ success:
cache->counter = adaptive_counter_cooldown();
}
void
_Py_Specialize_ContainsOp(PyObject *value, _Py_CODEUNIT *instr)
{
assert(ENABLE_SPECIALIZATION);
assert(_PyOpcode_Caches[CONTAINS_OP] == INLINE_CACHE_ENTRIES_COMPARE_OP);
_PyContainsOpCache *cache = (_PyContainsOpCache *)(instr + 1);
if (PyUnicode_CheckExact(value)) {
instr->op.code = CONTAINS_OP_STR;
goto success;
}
if (PyList_CheckExact(value)) {
instr->op.code = CONTAINS_OP_LIST;
goto success;
}
if (PyTuple_CheckExact(value)) {
instr->op.code = CONTAINS_OP_TUPLE;
goto success;
}
if (PyDict_CheckExact(value)) {
instr->op.code = CONTAINS_OP_DICT;
goto success;
}
if (PySet_CheckExact(value)) {
instr->op.code = CONTAINS_OP_SET;
goto success;
}
STAT_INC(CONTAINS_OP, failure);
instr->op.code = CONTAINS_OP;
cache->counter = adaptive_counter_backoff(cache->counter);
return;
success:
STAT_INC(CONTAINS_OP, success);
cache->counter = adaptive_counter_cooldown();
}
/* Code init cleanup.
* CALL_ALLOC_AND_ENTER_INIT will set up
* the frame to execute the EXIT_INIT_CHECK