980 Commits

Author SHA1 Message Date
Victor Stinner
865a0f621f Optimize _PyFunction_FastCallDict() when kwargs is {}
Issue #28839: Optimize _PyFunction_FastCallDict() when kwargs is an empty
dictionary, avoid the creation of an useless empty tuple.
2017-01-03 02:01:42 +01:00
INADA Naoki
9c15776fac Issue #29049: Remove unnecessary Py_DECREF 2016-12-26 18:52:46 +09:00
INADA Naoki
6a3cedf8dd Issue #29049: Fix refleak introduced by f5eb0c4f5d37. 2016-12-26 18:01:46 +09:00
INADA Naoki
5a625d0aa6 Issue #29049: Call _PyObject_GC_TRACK() lazily when calling Python function.
Calling function is up to 5% faster.
2016-12-24 20:19:08 +09:00
Serhiy Storchaka
5bb8b9134b Issue #18896: Python function can now have more than 255 parameters.
collections.namedtuple() now supports tuples with more than 255 elements.
2016-12-16 19:19:02 +02:00
Serhiy Storchaka
5ab81d787f Issue #28959: Added private macro PyDict_GET_SIZE for retrieving the size of dict. 2016-12-16 16:18:57 +02:00
Yury Selivanov
f2392133eb Issue #26110: Add LOAD_METHOD/CALL_METHOD opcodes.
Special thanks to INADA Naoki for pushing the patch through
the last mile, Serhiy Storchaka for reviewing the code, and to
Victor Stinner for suggesting the idea (originally implemented
in the PyPy project).
2016-12-13 19:03:51 -05:00
Victor Stinner
59a73276e9 Backed out changeset 99c34e47348b
The change broke test_gdb.
2016-12-09 18:51:13 +01:00
Victor Stinner
0ca246c5e7 Inline PyEval_EvalFrameEx() in callers
The PEP 523 modified PyEval_EvalFrameEx(): it's now an indirection to
interp->eval_frame().

Inline the call in performance critical code. Leave PyEval_EvalFrame()
unchanged, this function is only kept for backward compatibility.
2016-12-09 17:12:17 +01:00
Victor Stinner
f17c3de263 Use _PyObject_CallNoArg()
Replace:
    PyObject_CallFunctionObjArgs(callable, NULL)
with:
    _PyObject_CallNoArg(callable)
2016-12-06 18:46:19 +01:00
Victor Stinner
a5ed5f000a Use _PyObject_CallNoArg()
Replace:
    PyObject_CallObject(callable, NULL)
with:
    _PyObject_CallNoArg(callable)
2016-12-06 18:45:50 +01:00
Victor Stinner
2d0eb65f45 Uniformize argument names of "call" functions
Issue #28838: Rename parameters of the "calls" functions of the Python C API.

* Rename 'callable_object' and 'func' to 'callable': any Python callable object
  is accepted, not only Python functions
* Rename 'method' and 'nameid' to 'name' (method name)
* Rename 'o' to 'obj'
* Move, fix and update documentation of PyObject_CallXXX() functions
  in abstract.h
* Update also the documentaton of the C API (update parameter names)
2016-12-06 16:27:24 +01:00
Victor Stinner
de4ae3d486 Backed out changeset b9c9691c72c5
Issue #28858: The change b9c9691c72c5 introduced a regression. It seems like
_PyObject_CallArg1() uses more stack memory than
PyObject_CallFunctionObjArgs().
2016-12-04 22:59:09 +01:00
Victor Stinner
842cfff321 WITH_CLEANUP_START uses fastcall
Modify WITH_CLEANUP_START bytecode: replace PyObject_CallFunctionObjArgs() with
_PyObject_FastCall().
2016-12-01 14:45:31 +01:00
Victor Stinner
27580c1fb5 Replace PyObject_CallFunctionObjArgs() with fastcall
* PyObject_CallFunctionObjArgs(func, NULL) => _PyObject_CallNoArg(func)
* PyObject_CallFunctionObjArgs(func, arg, NULL) => _PyObject_CallArg1(func, arg)

PyObject_CallFunctionObjArgs() allocates 40 bytes on the C stack and requires
extra work to "parse" C arguments to build a C array of PyObject*.

_PyObject_CallNoArg() and _PyObject_CallArg1() are simpler and don't allocate
memory on the C stack.

This change is part of the fastcall project. The change on listsort() is
related to the issue #23507.
2016-12-01 14:43:22 +01:00
Victor Stinner
b69ee8c386 call_function(): document PyMethod optimization 2016-11-28 18:32:31 +01:00
Victor Stinner
048afd98b3 Remove CALL_PROFILE special build
Issue #28799:

* Remove the PyEval_GetCallStats() function.
* Deprecate the untested and undocumented sys.callstats() function.
* Remove the CALL_PROFILE special build

Use the sys.setprofile() function, cProfile or profile module to profile
function calls.
2016-11-28 11:59:04 +01:00
Victor Stinner
ed6de7345e Merge 3.6 2016-11-24 22:33:49 +01:00
Victor Stinner
f7d199ff32 Fix _PyGen_yf()
Issue #28782: Fix a bug in the implementation ``yield from`` when checking
if the next instruction is YIELD_FROM. Regression introduced by WORDCODE
(issue #26647).

Reviewed by Serhiy Storchaka and Yury Selivanov.
2016-11-24 22:33:01 +01:00
Raymond Hettinger
3fa28fd075 merge 2016-11-22 11:50:56 -08:00
Raymond Hettinger
64e2f9ac86 Issue #27100: Fix ref leak 2016-11-22 11:50:40 -08:00
Raymond Hettinger
9af740b99a merge 2016-11-21 17:24:58 -08:00
Raymond Hettinger
a3fec1543d Issue #27100: With statement reports missing __enter__ before __exit__. (Contributed by Jonathan Ellington.) 2016-11-21 17:24:23 -08:00
Serhiy Storchaka
85b0f5beb1 Added the const qualifier to char* variables that refer to readonly internal
UTF-8 represenatation of Unicode objects.
2016-11-20 10:16:47 +02:00
Serhiy Storchaka
a98c4a984b Replaced outdated macros _PyUnicode_AsString and _PyUnicode_AsStringAndSize
with PyUnicode_AsUTF8 and PyUnicode_AsUTF8AndSize.
2016-11-20 09:13:40 +02:00
Serhiy Storchaka
06515833fe Replaced outdated macros _PyUnicode_AsString and _PyUnicode_AsStringAndSize
with PyUnicode_AsUTF8 and PyUnicode_AsUTF8AndSize.
2016-11-20 09:13:07 +02:00
Raymond Hettinger
c32f9db846 Issue #28665: Use macro form of PyCell_GET/SET 2016-11-12 04:10:35 -05:00
Raymond Hettinger
b2b154374d merge 2016-11-11 04:32:11 -08:00
Raymond Hettinger
13527123a1 Issue #28665: Harmonize STORE_DEREF with STORE_FAST and LOAD_DEREF giving a 40% speedup. 2016-11-11 04:31:18 -08:00
Victor Stinner
c6944e7edc Issue #28618: Make hot functions using __attribute__((hot))
When Python is not compiled with PGO, the performance of Python on call_simple
and call_method microbenchmarks depend highly on the code placement. In the
worst case, the performance slowdown can be up to 70%.

The GCC __attribute__((hot)) attribute helps to keep hot code close to reduce
the risk of such major slowdown. This attribute is ignored when Python is
compiled with PGO.

The following functions are considered as hot according to statistics collected
by perf record/perf report:

* _PyEval_EvalFrameDefault()
* call_function()
* _PyFunction_FastCall()
* PyFrame_New()
* frame_dealloc()
* PyErr_Occurred()
2016-11-11 02:13:35 +01:00
Serhiy Storchaka
818b5cc6db Fixed possible abort in ceval loop if _PyUnicode_FromId() fails.
Every opcode should end with DISPATCH() or goto error.
2016-11-08 23:14:00 +02:00
Serhiy Storchaka
4678b2f448 Fixed possible abort in ceval loop if _PyUnicode_FromId() fails.
Every opcode should end with DISPATCH() or goto error.
2016-11-08 23:13:36 +02:00
Serhiy Storchaka
70b72f0f96 Fixed possible abort in ceval loop if _PyUnicode_FromId() fails.
Every opcode should end with DISPATCH() or goto error.
2016-11-08 23:12:46 +02:00
Yury Selivanov
0ee446c894 Merge 3.6 (issue #27243) 2016-11-08 15:15:42 -05:00
Yury Selivanov
2edd8a1e2c Issue #27243: Change PendingDeprecationWarning -> DeprecationWarning.
As it was agreed in the issue, __aiter__ returning an awaitable
should result in PendingDeprecationWarning in 3.5 and in
DeprecationWarning in 3.6.
2016-11-08 15:13:07 -05:00
Victor Stinner
d65f42a132 Issue #21955: Please don't try to optimize int+int 2016-10-20 12:18:10 +02:00
Raymond Hettinger
4186222e63 Minor fix-up to apply the stack adjustment macros consistent with the other opcodes 2016-10-15 19:03:06 -07:00
Serhiy Storchaka
5665301bae Issue #28257: Improved error message when pass a non-mapping as a var-keyword
argument.
2016-10-07 23:32:41 +03:00
Berker Peksag
419968c235 Issue #27358: Merge from 3.6 2016-10-02 13:08:47 +03:00
Berker Peksag
8e9045d0d8 Issue #27358: Fix typo in error message 2016-10-02 13:08:25 +03:00
Serhiy Storchaka
2e84de3638 Issue #27358: Optimized merging var-keyword arguments and improved error
message when pass a non-mapping as a var-keyword argument.
2016-10-02 11:07:29 +03:00
Serhiy Storchaka
e036ef8fa2 Issue #27358: Optimized merging var-keyword arguments and improved error
message when pass a non-mapping as a var-keyword argument.
2016-10-02 11:06:43 +03:00
Serhiy Storchaka
775a0ea0da Issue #28257: Improved error message when pass a non-iterable as
a var-positional argument.  Added opcode BUILD_TUPLE_UNPACK_WITH_CALL.
2016-10-02 10:34:46 +03:00
Serhiy Storchaka
7344285c19 Issue #28257: Improved error message when pass a non-iterable as
a var-positional argument.  Added opcode BUILD_TUPLE_UNPACK_WITH_CALL.
2016-10-02 10:33:46 +03:00
Serhiy Storchaka
c019158a4c Issue #27703: Got rid of unnecessary NULL checks in do_raise() in release mode.
Patch by Xiang Zhang.
2016-09-27 11:37:10 +03:00
Serhiy Storchaka
208bbd29d3 Silence GCC warning.
The code was correct, but GCC is not enough clever.
2016-09-22 19:59:46 +03:00
Serhiy Storchaka
63dc548109 Issue #28086: Single var-positional argument of tuple subtype was passed
unscathed to the C-defined function.  Now it is converted to exact tuple.
2016-09-22 19:41:20 +03:00
Victor Stinner
57f91ac95a Document kwnames in _PyObject_FastCallKeywords() and _PyStack_AsDict()
Issue #27213.
2016-09-12 13:37:07 +02:00
Victor Stinner
eece2229e2 Issue #27213: Fix reference leaks 2016-09-12 11:16:37 +02:00
Serhiy Storchaka
b72810583e Issue #27213: Fixed different issues with reworked CALL_FUNCTION* opcodes.
* BUILD_TUPLE_UNPACK and BUILD_MAP_UNPACK_WITH_CALL no longer generated with
  single tuple or dict.
* Restored more informative error messages for incorrect var-positional and
  var-keyword arguments.
* Removed code duplications in _PyEval_EvalCodeWithName().
* Removed redundant runtime checks and parameters in _PyStack_AsDict().
* Added a workaround and enabled previously disabled test in test_traceback.
* Removed dead code from the dis module.
2016-09-12 00:52:40 +03:00