New submission from Sergey Fedoseev :
When compiled with default NSMALLPOSINTS, _PyLong_FromUnsignedChar() is
significantly faster than other PyLong_From*():
$ python -m perf timeit -s "from collections import deque; consume =
deque(maxlen=0).extend; b = bytes(2**20)" "consume
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +14971
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/15251
___
Python tracker
<https://bugs.python.org/issu
New submission from Sergey Fedoseev :
bytearray_getitem() adjusts negative index, though that's already done by
PySequence_GetItem().
This makes PySequence_GetItem(bytearray(1), -2) to return 0 instead of raise
IndexError.
--
components: Interpreter Core
messages: 349545
nosy
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +14972
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/15250
___
Python tracker
<https://bugs.python.org/issu
New submission from Sergey Fedoseev :
Argument Clinic generates `{NULL, NULL}` initializer for Py_buffer variables.
Such initializer zeroes all Py_buffer members, but as I understand only `obj`
and `buf` members are really had to be initialized. Avoiding unneeded
initialization provides tiny
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +14975
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/15254
___
Python tracker
<https://bugs.python.org/issu
New submission from Sergey Fedoseev :
PyLong_As*() functions computes result for large longs like this:
size_t x, prev;
x = 0;
while (--i >= 0) {
prev = x;
x = (x << PyLong_SHIFT) | v->ob_digit[i];
if ((x >> PyLong_SHIFT) != prev) {
*overflow = sign;
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +15074
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/15363
___
Python tracker
<https://bugs.python.org/issu
Change by Sergey Fedoseev :
--
pull_requests: +15094
pull_request: https://github.com/python/cpython/pull/15385
___
Python tracker
<https://bugs.python.org/issue27
Change by Sergey Fedoseev :
--
pull_requests: +15095
pull_request: https://github.com/python/cpython/pull/15386
___
Python tracker
<https://bugs.python.org/issue27
Change by Sergey Fedoseev :
--
pull_requests: +15098
pull_request: https://github.com/python/cpython/pull/15388
___
Python tracker
<https://bugs.python.org/issue27
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +15152
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/15457
___
Python tracker
<https://bugs.python.org/issu
New submission from Sergey Fedoseev :
PyLong_As*() functions have a lot of duplicated code, creating draft PR with
attempt to deduplicate them.
--
components: Interpreter Core
messages: 350367
nosy: sir-sigurd
priority: normal
severity: normal
status: open
title: refactor PyLong_As
Sergey Fedoseev added the comment:
$ gcc -v 2>&1 | grep 'gcc version'
gcc version 8.3.0 (Debian 8.3.0-19)
using ./configure --enable-optimizations --with-lto
$ python -m perf timeit -s "from collections import deque; consume =
deque(maxlen=0).extend; b = bytes(2**20)&qu
Sergey Fedoseev added the comment:
Previous benchmarks results were obtained with non-LTO build.
Here are results for LTO build:
$ python -m perf timeit -s "from itertools import repeat; _len = repeat(None,
0).__length_hint__" "_len()" --compare-to=../cpython-ma
Sergey Fedoseev added the comment:
These last results are invalid :-)
I thought that I was checking _PyLong_FromUnsignedChar() on top of GH-15192,
but that wasn't true. So the correct results for LTO build are:
$ python -m perf timeit -s "from collections import deque; consum
New submission from Sergey Fedoseev :
In [8]: help(sys.float_info)
...
|
| dig
| DBL_DIG -- digits
|
This is not very helpful,
https://docs.python.org/3/library/sys.html#sys.float_info is more verbose, so
probably docstrings should be updated from where.
--
assignee
New submission from Sergey Fedoseev :
In [3]: help(zip)
class zip(object)
| zip(*iterables) --> zip object
|
| Return a zip object whose .__next__() method returns a tuple where
| the i-th element comes from the i-th iterable argument. The .__next__()
| method continues until
New submission from Sergey Fedoseev :
zip() shadows TypeError raised in __iter__() of source iterable:
In [21]: class Iterable:
...: def __init__(self, n):
...: self.n = n
...: def __iter__(self):
...: return iter(range(self.n))
...:
In [22
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +15268
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/15592
___
Python tracker
<https://bugs.python.org/issu
Sergey Fedoseev added the comment:
Maybe it's not clear from description, but traceback only show the line with
zip(), so it doesn't help at localizing the source of exception at all.
You only see that 'argument #N must support iteration', but that argument has
__iter_
Sergey Fedoseev added the comment:
Also using this example class:
In [5]: iter(Iterable('one'))
---
TypeError Traceback (most recent call last)
in ()
> 1 iter(Iterable
Sergey Fedoseev added the comment:
> map() does not have anything special.
Just for the reference, in Python 2.7 map() has the same behavior and it caused
many problems for me and other people working with/developing Django.
--
___
Python trac
New submission from Sergey Fedoseev :
This patch simplifies fast path for floats that fit into C long and moves it
from float.__trunc__ to PyLong_FromDouble().
+-+-+--+
| Benchmark | long-from-float-ref | long-from
Change by Sergey Fedoseev :
--
pull_requests: +15372
pull_request: https://github.com/python/cpython/pull/15718
___
Python tracker
<https://bugs.python.org/issue38
Sergey Fedoseev added the comment:
I added similar patch that replaces get_small_int() with macro version, since
it also induces unnecessary casts and makes machine code less efficient.
Example assembly can be checked at https://godbolt.org/z/1SjG3E.
This change produces tiny, but
Sergey Fedoseev added the comment:
I use GCC 9.2.
--
___
Python tracker
<https://bugs.python.org/issue38015>
___
___
Python-bugs-list mailing list
Unsubscribe:
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +15471
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/14838
___
Python tracker
<https://bugs.python.org/issu
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +15517
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/15274
___
Python tracker
<https://bugs.python.org/issu
New submission from Sergey Fedoseev :
GCC (along with Clang and ICC) has __builtin_unreachable() and MSVC has
__assume() builtins, that can be used to optimize out unreachable conditions.
So we could add macro like this:
#ifdef Py_DEBUG
# define Py_ASSUME(cond) (assert(cond))
#else
# if
Sergey Fedoseev added the comment:
I believe that the problem is caused by the change in Py_UNREACHABLE()
(https://github.com/python/cpython/commit/3ab61473ba7f3dca32d779ec2766a4faa0657923).
Before the mentioned commit Py_UNREACHABLE() was an expression, now it's a
block. Py_UNREAC
Sergey Fedoseev added the comment:
Also quote from Py_UNREACHABLE() doc:
> Use this in places where you might be tempted to put an assert(0) or abort()
> call.
https://github.com/python/cpython/commit/6b519985d23bd0f0bd072b5d5d5f2c60a81a19f2
does exactly that, it replaces assert(0
New submission from Sergey Fedoseev :
I wrote patch that cleans up type_init():
1. Removes conditions already checked by assert()
2. Removes object_init() call that effectively creates an empty tuple and
checks that this tuple is empty
--
messages: 352710
nosy: sir-sigurd
priority
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +15852
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/16257
___
Python tracker
<https://bugs.python.org/issu
Sergey Fedoseev added the comment:
FWIW I proposed to add Py_ASSUME() macro that uses __builtin_unreachable() in
bpo-38147.
--
___
Python tracker
<https://bugs.python.org/issue38
Sergey Fedoseev added the comment:
> If you care of _PyLong_Copy() performance, you should somehow manually inline
> _PyLong_New() inside _PyLong_Copy().
It doesn't solve this:
> We could add a function that bypass that check, but in LTO build
> PyObject_MALLOC(
Sergey Fedoseev added the comment:
These warnings are caused by
https://github.com/python/cpython/commit/c6734ee7c55add5fdc2c821729ed5f67e237a096.
I'd fix them, but I'm not sure if we are going to restore CHECK_SMALL_INT()
¯\_(ツ)_/¯
--
nosy: +
Sergey Fedoseev added the comment:
issue38524 is related.
--
nosy: +sir-sigurd
___
Python tracker
<https://bugs.python.org/issue38517>
___
___
Python-bugs-list m
Change by Sergey Fedoseev :
--
nosy: +sir-sigurd
___
Python tracker
<https://bugs.python.org/issue38524>
___
___
Python-bugs-list mailing list
Unsubscribe:
Change by Sergey Fedoseev :
--
nosy: +sir-sigurd
___
Python tracker
<https://bugs.python.org/issue39759>
___
___
Python-bugs-list mailing list
Unsubscribe:
Change by Sergey Fedoseev :
--
nosy: +sir-sigurd
nosy_count: 8.0 -> 9.0
pull_requests: +25593
pull_request: https://github.com/python/cpython/pull/27033
___
Python tracker
<https://bugs.python.org/issu
Change by Sergey Fedoseev :
--
pull_requests: +16894
pull_request: https://github.com/python/cpython/pull/17413
___
Python tracker
<https://bugs.python.org/issue37
Change by Sergey Fedoseev :
--
pull_requests: +17021
pull_request: https://github.com/python/cpython/pull/17539
___
Python tracker
<https://bugs.python.org/issue27
Change by Sergey Fedoseev :
--
nosy: +sir-sigurd
nosy_count: 3.0 -> 4.0
pull_requests: +19338
pull_request: https://github.com/python/cpython/pull/15659
___
Python tracker
<https://bugs.python.org/issu
New submission from Sergey Fedoseev :
Currently patlib.Path.iterdir() filters out '.' and '..'.
It's unneeded since patlib.Path.iterdir() uses os.listdir() under the hood,
which returns neither '.' nor '..'.
https://docs.python.org/3/library/os.ht
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +20336
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/21179
___
Python tracker
<https://bugs.python.org/issu
New submission from Sergey Fedoseev :
In [191]: import dataclasses, pydoc
In [192]: @dataclass
...: class C:
...: pass
...:
In [193]: print(pydoc.render_doc(C))
Python Library Documentation: class C in module __main__
class C(builtins.object)
| C() -> None
|
|
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +20793
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/21652
___
Python tracker
<https://bugs.python.org/issu
Change by Sergey Fedoseev :
--
nosy: +sir-sigurd
___
Python tracker
<https://bugs.python.org/issue26834>
___
___
Python-bugs-list mailing list
Unsubscribe:
Change by Sergey Fedoseev :
--
nosy: +sir-sigurd
___
Python tracker
<https://bugs.python.org/issue35276>
___
___
Python-bugs-list mailing list
Unsubscribe:
Change by Sergey Fedoseev :
--
nosy: +sir-sigurd
nosy_count: 4.0 -> 5.0
pull_requests: +21279
pull_request: https://github.com/python/cpython/pull/21763
___
Python tracker
<https://bugs.python.org/issu
Change by Sergey Fedoseev :
--
pull_requests: +13420
___
Python tracker
<https://bugs.python.org/issue31862>
___
___
Python-bugs-list mailing list
Unsubscribe:
Sergey Fedoseev added the comment:
`BytesIO.write()` and `BytesIO.writelines()` are independent of each other.
--
___
Python tracker
<https://bugs.python.org/issue34
New submission from Sergey Fedoseev :
Currently PyLong_FromSize_t() uses PyLong_FromLong() for values < PyLong_BASE.
It's suboptimal because PyLong_FromLong() needs to handle the sign. Removing
PyLong_FromLong() call and handling small ints directly in PyLong_FromSize_t()
makes i
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +14924
stage: -> patch review
pull_request: https://github.com/python/cpython/pull/15192
___
Python tracker
<https://bugs.python.org/issu
New submission from Sergey Fedoseev:
Currently TestResult.addSubTest() is called just before TestResult.stopTest(),
but docs says that addSubTest is "Called when a subtest finishes". IMO that
means that it will be called immediately after subtest finishes, but not after
indefinite t
Changes by Sergey Fedoseev :
--
components: +Library (Lib)
___
Python tracker
<http://bugs.python.org/issue30856>
___
___
Python-bugs-list mailing list
Unsub
Changes by Sergey Fedoseev :
--
pull_requests: +3025
___
Python tracker
<http://bugs.python.org/issue31108>
___
___
Python-bugs-list mailing list
Unsubscribe:
New submission from Sergey Fedoseev:
> python -mtimeit -s "l = list(range(10))" "l[-1] in l"
1000 loops, best of 3: 1.34 msec per loop
> python -mtimeit -s "l = list(range(10))" "l[-1] in iter(l)"
>
Change by Sergey Fedoseev :
--
pull_requests: +10490
___
Python tracker
<https://bugs.python.org/issue33234>
___
___
Python-bugs-list mailing list
Unsubscribe:
New submission from Sergey Fedoseev :
PyTuple_New() fills items array with NULLs to make usage of Py_DECREF() safe
even when array is not fully filled with real items.
There are multiple cases when this initialization step can be avoided to
improve performance. For example it gives such speed
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +11952
stage: -> patch review
___
Python tracker
<https://bugs.python.org/issue36030>
___
_
New submission from Sergey Fedoseev :
There are several cases in CPython sources when PyList_AsTuple() is used with
just created list and immediately after that this list is Py_DECREFed. This
operation can be performed more effectively since refcount of items is not
changed. For example it
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +11953
stage: -> patch review
___
Python tracker
<https://bugs.python.org/issue36031>
___
_
Sergey Fedoseev added the comment:
Does this look like more real world example?
Before:
$ python -m perf timeit -s "t = (1, 2, 3)" "(*t, 4, 5)"
.
Mean +- std dev: 95.7 ns +- 2.3 ns
After:
$ python -m perf timeit -s "t = (1, 2, 3)" "(*t
Change by Sergey Fedoseev :
--
pull_requests: +11979
___
Python tracker
<https://bugs.python.org/issue36030>
___
___
Python-bugs-list mailing list
Unsubscribe:
Sergey Fedoseev added the comment:
Here's benchmarks for PyTuple_FromArray() PR:
$ python -m perf timeit -s "l = [None]*0; tuple_ = tuple" "tuple_(l)"
--duplicate=100 --compare-to=../cpython-master/venv/bin/python
/home/sergey/tmp/cpython-master/venv/bin/python: .
New submission from Sergey Fedoseev :
list_slice() is used by PyList_GetSlice(), list_subscript() and for list
copying. In list_subscript() slice indices are already normalized with
PySlice_AdjustIndices(), so slice normalization currently performed in
list_slice() is only needed for
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +11993
stage: -> patch review
___
Python tracker
<https://bugs.python.org/issue36062>
___
_
New submission from Sergey Fedoseev :
This change produces minor speed-up:
$ python-other -m perf timeit -s "divmod_ = divmod" "divmod_(1, 1)"
--duplicate=1000 --compare-to=../cpython-master/venv/bin/python
python: . 64.6 ns +-
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +11998
stage: -> patch review
___
Python tracker
<https://bugs.python.org/issue36063>
___
_
New submission from Sergey Fedoseev :
In [186]: from itertools import cycle
In [187]: class ContainerLike:
...: def __init__(self):
...: self.chars = cycle('12')
...: def __getitem__(self, key):
...: return next(self.chars)
...:
Change by Sergey Fedoseev :
--
title: str.translate() behave differently for ASCII-only and other strings ->
str.translate() behaves differently for ASCII-only and other strings
___
Python tracker
<https://bugs.python.org/issu
New submission from Sergey Fedoseev :
It's somewhat similar to bpo-10811, but for converter function:
In [197]: import sqlite3 as sqlite
...: con = sqlite.connect(':memory:', detect_types=sqlite.PARSE_COLNAMES)
...: cur = con.cursor()
...: sqlite.converte
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +12008
stage: -> patch review
___
Python tracker
<https://bugs.python.org/issue36073>
___
_
Change by Sergey Fedoseev :
--
pull_requests: +12062
___
Python tracker
<https://bugs.python.org/issue36030>
___
___
Python-bugs-list mailing list
Unsubscribe:
Change by Sergey Fedoseev :
--
pull_requests: +12078
___
Python tracker
<https://bugs.python.org/issue36030>
___
___
Python-bugs-list mailing list
Unsubscribe:
Sergey Fedoseev added the comment:
I added WIP PR with discussed micro-optimization, here are benchmark results:
$ python -m perf compare_to --min-speed 1 -G master.json tuple-untracked.json
Slower (1):
- sqlite_synth: 5.16 us +- 0.10 us -> 5.22 us +- 0.08 us: 1.01x slower (+1%)
Faster
Sergey Fedoseev added the comment:
This optimization also can be used for BUILD_TUPLE opcode and in pickle module,
if it's OK to add _PyTuple_StealFromArray() function :-)
--
___
Python tracker
<https://bugs.python.org/is
Sergey Fedoseev added the comment:
> Can you please convert msg336142 into a perf script?
> And then run again these benchmarks on PR 12052.
++-+--+
| Benchmark | ref | unt
Sergey Fedoseev added the comment:
>> This optimization also can be used for BUILD_TUPLE opcode and in pickle
>> module, if it's OK to add _PyTuple_StealFromArray() function :-)
> I would like to see a micro-benchmark show
Change by Sergey Fedoseev :
Added file: https://bugs.python.org/file48174/list_as_tuple_bench.py
___
Python tracker
<https://bugs.python.org/issue36030>
___
___
Pytho
Change by Sergey Fedoseev :
Added file: https://bugs.python.org/file48173/build_tuple_bench.py
___
Python tracker
<https://bugs.python.org/issue36030>
___
___
Python-bug
New submission from Sergey Fedoseev :
Before:
$ ./python -m timeit -s "from binascii import unhexlify; b = b'aa'*2**20"
"unhexlify(b)"
50 loops, best of 5: 5.68 msec per loop
After:
$ ./python -m timeit -s "from binascii import unhexlify; b = b'aa'
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +4508
stage: -> patch review
___
Python tracker
<https://bugs.python.org/issue32147>
___
___
Py
Change by Sergey Fedoseev :
--
components: +Library (Lib)
___
Python tracker
<https://bugs.python.org/issue32147>
___
___
Python-bugs-list mailing list
Unsub
Sergey Fedoseev added the comment:
Serhiy, did you use the same benchmark as mentioned here?
--
___
Python tracker
<https://bugs.python.org/issue32147>
___
___
Sergey Fedoseev added the comment:
> OS
x86_64 GNU/Linux
> compiler
gcc version 7.2.0 (Debian 7.2.0-16)
> CPU
Architecture:x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
CPU(s): 4
On-line CPU(s) list: 0-3
Thread(s) per core: 2
Co
Sergey Fedoseev added the comment:
Is there anything I can do to push this forward?
BTW, Serhiy, what are your OS, compiler and CPU?
--
___
Python tracker
<https://bugs.python.org/issue32
Change by Sergey Fedoseev :
--
nosy: +sir-sigurd
___
Python tracker
<https://bugs.python.org/issue8488>
___
___
Python-bugs-list mailing list
Unsubscribe:
Change by Sergey Fedoseev :
--
nosy: +sir-sigurd
___
Python tracker
<https://bugs.python.org/issue33576>
___
___
Python-bugs-list mailing list
Unsubscribe:
Change by Sergey Fedoseev :
--
nosy: +sir-sigurd
___
Python tracker
<https://bugs.python.org/issue21145>
___
___
Python-bugs-list mailing list
Unsubscribe:
New submission from Sergey Fedoseev :
SQLite converters are documented to be sensitive to the case of type names, but
they're not.
In [50]: import sqlite3
...:
...: sqlite3.converters.clear()
...: sqlite3.register_converter('T', lambda x: 'UPPE
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +7651
stage: -> patch review
___
Python tracker
<https://bugs.python.org/issue34018>
___
___
Py
New submission from Sergey Fedoseev :
SQLiter 3.8.3 and higher allows to mark created functions as deterministic to
allow additional optimizations. There isn't currently a way to it from Python.
https://sqlite.org/c3ref/create_function.html
--
components: Extension Modules
mes
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +7687
stage: -> patch review
___
Python tracker
<https://bugs.python.org/issue34041>
___
___
Py
New submission from Sergey Fedoseev :
In [1]: import sqlite3
In [2]: con = sqlite3.connect(':memory:')
In [3]: con.execute('SELECT f()')
---
OperationalError Traceback (most r
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +7702
stage: -> patch review
___
Python tracker
<https://bugs.python.org/issue34052>
___
___
Py
New submission from Sergey Fedoseev :
Currently list concatenation, slicing and repeating operations are using
PyList_New() which allocates memory for the items by calloc(). malloc() could
be used instead, since the allocated memory is overwritten by mentioned
operations.
I made benchmarks
Change by Sergey Fedoseev :
--
keywords: +patch
pull_requests: +7869
stage: -> patch review
___
Python tracker
<https://bugs.python.org/issue34151>
___
___
Py
1 - 100 of 136 matches
Mail list logo
