Skip to content

Experimental variant of BINARY_ADD specialization. #29059

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
markshannon wants to merge 4 commits into from
Closed

Experimental variant of BINARY_ADD specialization. #29059

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
763c297
Experimental variant of BINARY_ADD specialization.
markshannon Oct 19, 2021
a90f27e
Merge branch 'main' into alternative-specialize-binary-add
markshannon Oct 27, 2021
a8128d1
Specialize BINARY_ADD for int+float and float+int.
markshannon Oct 27, 2021
bdb767f
Merge branch 'main' into alternative-specialize-binary-add
markshannon Oct 27, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 8 additions & 1 deletion Include/internal/pycore_code.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,8 @@ typedef struct {
uint8_t original_oparg;
uint8_t counter;
uint16_t index;
uint16_t left_version;
uint16_t right_version;
} _PyAdaptiveEntry;


Expand All @@ -35,6 +37,10 @@ typedef struct {
PyObject *obj;
} _PyObjectCache;

typedef struct {
binaryfunc function;
} _PyFuncPtrCache;

/* Add specialized versions of entries to this union.
*
* Do not break the invariant: sizeof(SpecializedCacheEntry) == 8
Expand All @@ -51,6 +57,7 @@ typedef union {
_PyAttrCache attr;
_PyLoadGlobalCache load_global;
_PyObjectCache obj;
_PyFuncPtrCache func_ptr;
} SpecializedCacheEntry;

#define INSTRUCTIONS_PER_ENTRY (sizeof(SpecializedCacheEntry)/sizeof(_Py_CODEUNIT))
Expand Down Expand Up @@ -307,7 +314,7 @@ int _Py_Specialize_StoreAttr(PyObject *owner, _Py_CODEUNIT *instr, PyObject *nam
int _Py_Specialize_LoadGlobal(PyObject *globals, PyObject *builtins, _Py_CODEUNIT *instr, PyObject *name, SpecializedCacheEntry *cache);
int _Py_Specialize_LoadMethod(PyObject *owner, _Py_CODEUNIT *instr, PyObject *name, SpecializedCacheEntry *cache);
int _Py_Specialize_BinarySubscr(PyObject *sub, PyObject *container, _Py_CODEUNIT *instr);
int _Py_Specialize_BinaryAdd(PyObject *left, PyObject *right, _Py_CODEUNIT *instr);
int _Py_Specialize_BinaryAdd(PyObject *left, PyObject *right, _Py_CODEUNIT *instr, SpecializedCacheEntry *cache);
int _Py_Specialize_BinaryMultiply(PyObject *left, PyObject *right, _Py_CODEUNIT *instr);

#define PRINT_SPECIALIZATION_STATS 0
Expand Down
69 changes: 35 additions & 34 deletions Include/opcode.h
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

1 change: 1 addition & 0 deletions Lib/opcode.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -221,6 +221,7 @@ def jabs_op(name, op):

_specialized_instructions = [
"BINARY_ADD_ADAPTIVE",
"BINARY_ADD_CACHED",
"BINARY_ADD_INT",
"BINARY_ADD_FLOAT",
"BINARY_ADD_UNICODE",
Expand Down
49 changes: 41 additions & 8 deletions Python/ceval.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2080,18 +2080,19 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, InterpreterFrame *frame, int thr
}

TARGET(BINARY_ADD_ADAPTIVE) {
if (oparg == 0) {
SpecializedCacheEntry *cache = GET_CACHE();
if (cache->adaptive.counter == 0) {
PyObject *left = SECOND();
PyObject *right = TOP();
next_instr--;
if (_Py_Specialize_BinaryAdd(left, right, next_instr) < 0) {
if (_Py_Specialize_BinaryAdd(left, right, next_instr, cache) < 0) {
goto error;
}
DISPATCH();
}
else {
STAT_INC(BINARY_ADD, deferred);
UPDATE_PREV_INSTR_OPARG(next_instr, oparg - 1);
cache->adaptive.counter--;
STAT_DEC(BINARY_ADD, unquickened);
JUMP_TO_INSTRUCTION(BINARY_ADD);
}
Expand All @@ -2103,7 +2104,9 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, InterpreterFrame *frame, int thr
DEOPT_IF(!PyUnicode_CheckExact(left), BINARY_ADD);
DEOPT_IF(Py_TYPE(right) != Py_TYPE(left), BINARY_ADD);
STAT_INC(BINARY_ADD, hit);
record_hit_inline(next_instr, oparg);
SpecializedCacheEntry *caches = GET_CACHE();
_PyAdaptiveEntry *cache0 = &caches[0].adaptive;
record_cache_hit(cache0);
PyObject *res = PyUnicode_Concat(left, right);
STACK_SHRINK(1);
SET_TOP(res);
Expand Down Expand Up @@ -2132,7 +2135,9 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, InterpreterFrame *frame, int thr
PyObject *var = GETLOCAL(next_oparg);
DEOPT_IF(var != left, BINARY_ADD);
STAT_INC(BINARY_ADD, hit);
record_hit_inline(next_instr, oparg);
SpecializedCacheEntry *caches = GET_CACHE();
_PyAdaptiveEntry *cache0 = &caches[0].adaptive;
record_cache_hit(cache0);
GETLOCAL(next_oparg) = NULL;
Py_DECREF(left);
STACK_SHRINK(1);
Expand All @@ -2150,7 +2155,9 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, InterpreterFrame *frame, int thr
DEOPT_IF(!PyFloat_CheckExact(left), BINARY_ADD);
DEOPT_IF(Py_TYPE(right) != Py_TYPE(left), BINARY_ADD);
STAT_INC(BINARY_ADD, hit);
record_hit_inline(next_instr, oparg);
SpecializedCacheEntry *caches = GET_CACHE();
_PyAdaptiveEntry *cache0 = &caches[0].adaptive;
record_cache_hit(cache0);
double dsum = ((PyFloatObject *)left)->ob_fval +
((PyFloatObject *)right)->ob_fval;
PyObject *sum = PyFloat_FromDouble(dsum);
Expand All @@ -2170,7 +2177,9 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, InterpreterFrame *frame, int thr
DEOPT_IF(!PyLong_CheckExact(left), BINARY_ADD);
DEOPT_IF(Py_TYPE(right) != Py_TYPE(left), BINARY_ADD);
STAT_INC(BINARY_ADD, hit);
record_hit_inline(next_instr, oparg);
SpecializedCacheEntry *caches = GET_CACHE();
_PyAdaptiveEntry *cache0 = &caches[0].adaptive;
record_cache_hit(cache0);
PyObject *sum = _PyLong_Add((PyLongObject *)left, (PyLongObject *)right);
SET_SECOND(sum);
Py_DECREF(right);
Expand All @@ -2182,6 +2191,30 @@ _PyEval_EvalFrameDefault(PyThreadState *tstate, InterpreterFrame *frame, int thr
DISPATCH();
}

TARGET(BINARY_ADD_CACHED) {
SpecializedCacheEntry *caches = GET_CACHE();
_PyAdaptiveEntry *cache0 = &caches[0].adaptive;
PyObject *left = SECOND();
PyObject *right = TOP();
DEOPT_IF(Py_TYPE(left)->tp_version_tag != cache0->left_version, BINARY_ADD);
DEOPT_IF(Py_TYPE(right)->tp_version_tag != cache0->right_version, BINARY_ADD);
record_cache_hit(cache0);
binaryfunc function = caches[-1].func_ptr.function;
PyObject *sum = function(left, right);
if (sum == Py_NotImplemented) {
Py_DECREF(sum);
sum = NULL;
Comment on lines +2184 to +2185

@Fidget-Spinner Fidget-Spinner Oct 19, 2021

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You may want to consider replacing this with

Suggested change
Py_DECREF(sum);
sum = NULL;
Py_SETREF(sum, NULL);

@brandtbucher brandtbucher Oct 27, 2021

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Or:

Suggested change
Py_DECREF(sum);
sum = NULL;
Py_CLEAR(sum);

}
SET_SECOND(sum);
Py_DECREF(right);
Py_DECREF(left);
STACK_SHRINK(1);
if (sum == NULL) {
goto error;
}
DISPATCH();
}

TARGET(BINARY_SUBTRACT) {
PyObject *right = POP();
PyObject *left = TOP();
Expand Down Expand Up @@ -4983,7 +5016,7 @@ MISS_WITH_CACHE(STORE_ATTR)
MISS_WITH_CACHE(LOAD_GLOBAL)
MISS_WITH_CACHE(LOAD_METHOD)
MISS_WITH_OPARG_COUNTER(BINARY_SUBSCR)
MISS_WITH_OPARG_COUNTER(BINARY_ADD)
MISS_WITH_CACHE(BINARY_ADD)
MISS_WITH_OPARG_COUNTER(BINARY_MULTIPLY)

binary_subscr_dict_error:
Expand Down
34 changes: 17 additions & 17 deletions Python/opcode_targets.h
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

Loading