Microbenchmarks show 2-2.5x improvement. Built-in 'divmod' function
is now also ~10% faster.
-m timeit -s "x=22331" "x//2;x//-3;x//4;x//5;x//-6;x//7;x//8;x//-99;x//100;"
with patch: 0.321 without patch: 0.633
-m timeit -s "x=22331" "x%2;x%3;x%-4;x%5;x%6;x%-7;x%8;x%99;x%-100;"
with patch: 0.224 without patch: 0.66
Big thanks to Serhiy Storchaka, Mark Dickinson and Victor Stinner for
thorow code reviews and algorithms improvements.
Optimize bytes.__mod__(args) for integere formats: %d (%i, %u), %o, %x and %X.
_PyBytesWriter is now used to format directly the integer into the writer
buffer, instead of using a temporary bytes object.
Formatting is between 30% and 50% faster on a microbenchmark.
Some time ago we changed the docs to consistently use the term 'bytes-like
object' in all the contexts where bytes, bytearray, memoryview, etc are used.
This patch (by Ezio Melotti) completes that work by changing the error
messages that previously reported that certain types did "not support the
buffer interface" to instead say that a bytes-like object is required. (The
glossary entry for bytes-like object references the discussion of the buffer
protocol in the docs.)
Reduce the base by the modulus when the base is larger than
the modulus. This can unboundedly speed the "startup costs"
of doing modular exponentiation, particularly in cases where
the base is much larger than the modulus. Original patch
by Armin Rigo, inspired by https://github.com/pyca/ed25519.
Merged from 3.3.
Reduce the base by the modulus when the base is larger than
the modulus. This can unboundedly speed the "startup costs"
of doing modular exponentiation, particularly in cases where
the base is much larger than the modulus. Original patch
by Armin Rigo, inspired by https://github.com/pyca/ed25519.
- replace 'long int' / 'long' by 'int'
- fix capitalization of "Python" in PyLong_AsUnsignedLong
- "is too large" -> "too large", for consistency with other messages.
PyStructSequence_InitType() except that it has a return value (0 on success,
-1 on error).
* PyStructSequence_InitType2() now raises MemoryError on memory allocation failure
* Fix also some calls to PyDict_SetItemString(): handle error
* Add also min_char attribute to _PyUnicodeWriter structure (currently unused)
* _PyUnicodeWriter_Init() has no more argument (except the writer itself):
min_length and overallocate must be set explicitly
* In error handlers, only enable overallocation if the replacement string
is longer than 1 character
* CJK decoders don't use overallocation anymore
* Set min_length, instead of preallocating memory using
_PyUnicodeWriter_Prepare(), in many decoders
* _PyUnicode_DecodeUnicodeInternal() checks for integer overflow
I've left a couple of them in: zlib (third-party lib), getaddrinfo.c
(doesn't include Python.h, and probably obsolete), _sre.c (legitimate
use for the re.LOCALE flag), mpdecimal (needs to build without Python.h).
I've left a couple of them in: zlib (third-party lib), getaddrinfo.c
(doesn't include Python.h, and probably obsolete), _sre.c (legitimate
use for the re.LOCALE flag), mpdecimal (needs to build without Python.h).
I've left a couple of them in: zlib (third-party lib), getaddrinfo.c
(doesn't include Python.h, and probably obsolete), _sre.c (legitimate
use for the re.LOCALE flag).
This commit rewrites the docstring for int() to incorporate the documentation
changes made in issue #16036. It also switches the docstrings for int(),
str(), range(), and slice() to use multi-line signatures.
* Formatting string, int, float and complex use the _PyUnicodeWriter API. It
avoids a temporary buffer in most cases.
* Add _PyUnicodeWriter_WriteStr() to restore the PyAccu optimization: just
keep a reference to the string if the output is only composed of one string
* Disable overallocation when formatting the last argument of str%args and
str.format(args)
* Overallocation allocates at least 100 characters: add min_length attribute
to the _PyUnicodeWriter structure
* Add new private functions: _PyUnicode_FastCopyCharacters(),
_PyUnicode_FastFill() and _PyUnicode_FromASCII()
The speed up is around 20% in average.
PyLong_AsDouble() and PyLong_AsUnsignedLongLong() now raise TypeError (rather
than SystemError) when passed a non-integer argument, matching the behavior of
all the other PyLong_As*() functions.
types. Added a new API function, PyUnicode_TransformDecimalToASCII(),
which transforms non-ASCII decimal digits in a Unicode string to their
ASCII equivalents.
(instances of int, float, complex, decimal.Decimal and
fractions.Fraction) that makes it easy to maintain the invariant that
hash(x) == hash(y) whenever x and y have equal value.
svn+ssh://pythondev@svn.python.org/python/trunk
........
r79843 | mark.dickinson | 2010-04-06 17:46:09 +0100 (Tue, 06 Apr 2010) | 4 lines
Issue #8259: Get rid of 'outrageous left shift count' error when
left-shifting an integer by more than 2**31 on a 64-bit machine. Also
convert shift counts to a Py_ssize_t instead of a C long.
........
r79844 | mark.dickinson | 2010-04-06 17:47:55 +0100 (Tue, 06 Apr 2010) | 1 line
Misc/NEWS entry for r79843.
........
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r78918 | mark.dickinson | 2010-03-13 11:34:40 +0000 (Sat, 13 Mar 2010) | 4 lines
Issue #8014: Fix PyLong_As<c-integer-type> methods not to produce an
internal error on non-integer input: they now raise TypeError instead.
This is needed for attributes declared via PyMemberDefs.
........
r78920 | mark.dickinson | 2010-03-13 13:23:05 +0000 (Sat, 13 Mar 2010) | 3 lines
Issue #8014: Fix incorrect error checks in structmember.c, and re-enable
previously failing test_structmember.py tests.
........
svn+ssh://pythondev@svn.python.org/python/trunk
........
r77842 | mark.dickinson | 2010-01-30 10:08:33 +0000 (Sat, 30 Jan 2010) | 4 lines
Issue #7767: Add new C-API function PyLong_AsLongLongAndOverflow, a
long long variant of PyLong_AsLongAndOverflow. Patch by Case Van
Horsen.
........
svn+ssh://pythondev@svn.python.org/python/trunk
........
r77234 | mark.dickinson | 2010-01-02 14:45:40 +0000 (Sat, 02 Jan 2010) | 7 lines
Refactor some longobject internals: PyLong_AsDouble and _PyLong_AsScaledDouble
(the latter renamed to _PyLong_Frexp) now use the same core code. The
exponent produced by _PyLong_Frexp now has type Py_ssize_t instead of the
previously used int, and no longer needs scaling by PyLong_SHIFT. This
frees the math module from having to know anything about the PyLong
implementation. This closes issue #5576.
........
svn+ssh://pythondev@svn.python.org/python/trunk
........
r75110 | mark.dickinson | 2009-09-28 17:52:40 +0100 (Mon, 28 Sep 2009) | 9 lines
Style/consistency/nano-optimization nit: replace occurrences of
(high_bits << PyLong_SHIFT) + low_bits with
(high_bits << PyLong_SHIFT) | low_bits
in Objects/longobject.c. Motivation:
- shouldn't unnecessarily mix bit ops with arithmetic ops (style)
- this pattern should be spelt the same way thoughout (consistency)
- it's very very very slightly faster: no need to worry about
carries to the high digit (nano-optimization).
........
svn+ssh://pythondev@svn.python.org/python/trunk
........
r72202 | mark.dickinson | 2009-05-02 18:55:01 +0100 (Sat, 02 May 2009) | 3 lines
Remove unnecessary use of context for long getters.
(Related to issue #5880).
........
svn+ssh://pythondev@svn.python.org/python/trunk
........
r71772 | mark.dickinson | 2009-04-20 22:13:33 +0100 (Mon, 20 Apr 2009) | 5 lines
Issue #3166: Make long -> float (and int -> float) conversions
correctly rounded, using round-half-to-even. This ensures that the
value of float(n) doesn't depend on whether we're using 15-bit digits
or 30-bit digits for Python longs.
........
svn+ssh://pythondev@svn.python.org/python/trunk
........
r70542 | mark.dickinson | 2009-03-23 18:25:13 +0000 (Mon, 23 Mar 2009) | 14 lines
Issue #5512: speed up the long division algorithm for Python longs.
The basic algorithm remains the same; the most significant speedups
come from the following three changes:
(1) normalize by shifting instead of multiplying and dividing
(2) the old algorithm usually did an unnecessary extra iteration of
the outer loop; remove this. As a special case, this means that
long divisions with a single-digit result run twice as fast as
before.
(3) make inner loop much tighter.
Various benchmarks show speedups of between 50% and 150% for long
integer divisions and modulo operations.
........
- new configure option --enable-big-digits
- new structseq sys.int_info giving information about the internal format
By default, 30-bit digits are enabled on 64-bit machines but
disabled on 32-bit machines.
Benjamin Peterson
Raymond Hettinger
Mark Dickinson
Victor Stinner
Martin Panter
Serhiy Storchaka
Yury Selivanov
R David Murray
Christian Heimes
Tim Peters
Antoine Pitrou
Gregory P. Smith
Mark Dickinson
Chris Jerdonek
Victor Stinner
Martin v. Löwis
Stefan Krah
Nadeem Vawda
Brian Curtin
Ezio Melotti
Alexander Belopolsky
Georg Brandl
Stefan Krah
Alexandre Vassalotti