The `used` field must be written using atomic stores because `set_len`
and iterators may access the field concurrently without holding the
per-object lock.
The `_PySeqLock_EndRead` function needs an acquire fence to ensure that
the load of the sequence happens after any loads within the read side
critical section. The missing fence can trigger bugs on macOS arm64.
Additionally, we need a release fence in `_PySeqLock_LockWrite` to
ensure that the sequence update is visible before any modifications to
the cache entry.
Make error message for index() methods consistent
Remove the repr of the searched value (which can be arbitrary large)
from ValueError messages for list.index(), range.index(), deque.index(),
deque.remove() and ShareableList.index(). Make the error messages
consistent with error messages for other index() and remove()
methods.
Refactor the fast Unicode hash check into `_PyObject_HashFast` and use relaxed
atomic loads in the free-threaded build.
After this change, the TSAN doesn't report data races for this method.
In some cases, previously computed as (nan+nanj), we could
recover meaningful component values in the result, see
e.g. the C11, Annex G.5.2, routine _Cdivd().
Fix warnings when using -Wimplicit-fallthrough compiler flag.
Annotate explicitly "fall through" switch cases with a new
_Py_FALLTHROUGH macro which uses __attribute__((fallthrough)) if
available. Replace "fall through" comments with _Py_FALLTHROUGH.
Add _Py__has_attribute() macro. No longer define __has_attribute()
macro if it's not defined. Move also _Py__has_builtin() at the top
of pyport.h.
Co-Authored-By: Nikita Sobolev <mail@sobolevn.me>
This PR sets up tagged pointers for CPython.
The general idea is to create a separate struct _PyStackRef for everything on the evaluation stack to store the bits. This forces the C compiler to warn us if we try to cast things or pull things out of the struct directly.
Only for free threading: We tag the low bit if something is deferred - that means we skip incref and decref operations on it. This behavior may change in the future if Mark's plans to defer all objects in the interpreter loop pans out.
This implies a strict stack reference discipline is required. ALL incref and decref operations on stackrefs must use the stackref variants. It is unsafe to untag something then do normal incref/decref ops on it.
The new incref and decref variants are called dup and close. They mimic a "handle" API operating on these stackrefs.
Please read Include/internal/pycore_stackref.h for more information!
---------
Co-authored-by: Mark Shannon <9448417+markshannon@users.noreply.github.com>
Remove the const qualifier of the argument of functions:
* _PyLong_IsCompact()
* _PyLong_CompactValue()
Py_TYPE() argument is not const.
Fix the compiler warning:
Include/cpython/longintrepr.h: In function ‘_PyLong_CompactValue’:
Include/pyport.h:19:31: error: cast discards ‘const’ qualifier from
pointer target type [-Werror=cast-qual]
(...)
Include/cpython/longintrepr.h:133:30: note: in expansion of macro
‘Py_TYPE’
assert(PyType_HasFeature(Py_TYPE(op), Py_TPFLAGS_LONG_SUBCLASS));
PyDict_Next no longer locks the dictionary in the free-threaded build. Locking
around individual PyDict_Next calls is not sufficient because the function
returns borrowed references and because it allows concurrent modifications
during the iteraiton loop.
The internal locking also interferes with correct external synchronization
because it may suspend outer critical sections created by the caller.
Moves the logic to update the type's dictionary to its own function in order
to make the lock scoping more clear.
Also, ensure that `name` is decref'd on the error path.
PyUnicode_FromFormat() no longer produces the ending \ufffd
character for truncated C string when use precision with %s and %V.
It now truncates the string before the start of truncated multibyte sequences.
This makes the following macros public as part of the non-limited C-API for
locking a single object or two objects at once.
* `Py_BEGIN_CRITICAL_SECTION(op)` / `Py_END_CRITICAL_SECTION()`
* `Py_BEGIN_CRITICAL_SECTION2(a, b)` / `Py_END_CRITICAL_SECTION2()`
The supporting functions and structs used by the macros are also exposed for
cases where C macros are not available.
* Add an InternalDocs file describing how interning should work and how to use it.
* Add internal functions to *explicitly* request what kind of interning is done:
- `_PyUnicode_InternMortal`
- `_PyUnicode_InternImmortal`
- `_PyUnicode_InternStatic`
* Switch uses of `PyUnicode_InternInPlace` to those.
* Disallow using `_Py_SetImmortal` on strings directly.
You should use `_PyUnicode_InternImmortal` instead:
- Strings should be interned before immortalization, otherwise you're possibly
interning a immortalizing copy.
- `_Py_SetImmortal` doesn't handle the `SSTATE_INTERNED_MORTAL` to
`SSTATE_INTERNED_IMMORTAL` update, and those flags can't be changed in
backports, as they are now part of public API and version-specific ABI.
* Add private `_only_immortal` argument for `sys.getunicodeinternedsize`, used in refleak test machinery.
* Make sure the statically allocated string singletons are unique. This means these sets are now disjoint:
- `_Py_ID`
- `_Py_STR` (including the empty string)
- one-character latin-1 singletons
Now, when you intern a singleton, that exact singleton will be interned.
* Add a `_Py_LATIN1_CHR` macro, use it instead of `_Py_ID`/`_Py_STR` for one-character latin-1 singletons everywhere (including Clinic).
* Intern `_Py_STR` singletons at startup.
* For free-threaded builds, intern `_Py_LATIN1_CHR` singletons at startup.
* Beef up the tests. Cover internal details (marked with `@cpython_only`).
* Add lots of assertions
Co-Authored-By: Eric Snow <ericsnowcurrently@gmail.com>
The public PyUnicodeWriter API enables overallocation by default and
so is more efficient.
Benchmark:
python -m pyperf timeit \
-s 't = int | float | complex | str | bytes | bytearray' \
' | memoryview | list | dict' \
'str(t)'
Result:
1.29 us +- 0.02 us -> 1.00 us +- 0.02 us: 1.29x faster
The public PyUnicodeWriter API enables overallocation by default and
so is more efficient.
Benchmark:
python -m pyperf timeit \
-s 't = list[int, float, complex, str, bytes, bytearray, ' \
'memoryview, list, dict]' \
'str(t)'
Result:
1.49 us +- 0.03 us -> 1.10 us +- 0.02 us: 1.35x faster
This exposes `PyUnstable_Object_ClearWeakRefsNoCallbacks` as an unstable
C-API function to provide a thread-safe mechanism for clearing weakrefs
without executing callbacks.
Some C-API extensions need to clear weakrefs without calling callbacks,
such as after running finalizers like we do in subtype_dealloc.
Previously they could use `_PyWeakref_ClearRef` on each weakref, but
that's not thread-safe in the free-threaded build.
Co-authored-by: Petr Viktorin <encukou@gmail.com>
In gh-120009 I used an atexit hook to finalize the _datetime module's static types at interpreter shutdown. However, atexit hooks are executed very early in finalization, which is a problem in the few cases where a subclass of one of those static types is still alive until the final GC collection. The static builtin types don't have this probably because they are finalized toward the end, after the final GC collection. To avoid the problem for _datetime, I have applied a similar approach here.
Also, credit goes to @mgorny and @neonene for the new tests.
FYI, I would have liked to take a slightly cleaner approach with managed static types, but wanted to get a smaller fix in first for the sake of backporting. I'll circle back to the cleaner approach with a future change on the main branch.
We need to write to `ob_ref_local` and `ob_tid` before `ob_ref_shared`.
Once we mark `ob_ref_shared` as merged, some other thread may free the
object because the caller also passes in `-1` as `extra` to give up its
only reference.
We make use of the same mechanism that we use for the static builtin types. This required a few tweaks.
The relevant code could use some cleanup but I opted to avoid the significant churn in this change. I'll tackle that separately.
This change is the final piece needed to make _datetime support multiple interpreters. I've updated the module slot accordingly.
The free-threaded build currently immortalizes objects that use deferred
reference counting (see gh-117783). This typically happens once the
first non-main thread is created, but the behavior can be suppressed for
tests, in subinterpreters, or during a compile() call.
This fixes a race condition involving the tracking of whether the
behavior is suppressed.
Remove the delegation of `int` to the `__trunc__` special method: `int` will now only delegate to `__int__` and `__index__` (in that order). `__trunc__` continues to exist, but its sole purpose is to support `math.trunc`.
---------
Co-authored-by: Bénédikt Tran <10796600+picnixz@users.noreply.github.com>
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
* Add docs for new APIs
* Add soft-deprecation notices
* Add What's New porting entries
* Update comments referencing `PyFrame_LocalsToFast()` to mention the proxy instead
* Other related cleanups found when looking for refs to the deprecated APIs
* Passing a string as the "real" keyword argument is now an error;
it should only be passed as a single positional argument.
* Passing a complex number as the "real" or "imag" argument is now deprecated;
it should only be passed as a single positional argument.
* Remove the equivalence with real+imag*1j which can be incorrect in corner
cases (non-finite numbers, the sign of zeroes).
* Separately document the three roles of the constructor: parsing a string,
converting a number, and constructing a complex from components.
* Document positional-only parameters of complex(), float(), int() and bool()
as positional-only.
* Add examples for complex() and int().
* Specify the grammar of the string for complex().
* Improve the grammar of the string for float().
* Describe more explicitly the behavior when real and/or imag arguments are
complex numbers. (This will be deprecated in future.)
The deadlock only affected the free-threaded build and only occurred
when the GIL was enabled at runtime. The `Py_DECREF(old_name)` call
might temporarily release the GIL while holding the type seqlock.
Another thread may spin trying to acquire the seqlock while holding the
GIL.
The deadlock occurred roughly 1 in ~1,000 runs of `pool_in_threads.py`
from `test_multiprocessing_pool_circular_import`.
Add unicode_decode_utf8_writer() to write directly characters into a
_PyUnicodeWriter writer: avoid the creation of a temporary string.
Optimize PyUnicode_FromFormat() by using the new
unicode_decode_utf8_writer().
Rename unicode_fromformat_write_cstr() to
unicode_fromformat_write_utf8().
Microbenchmark on the code:
return PyUnicode_FromFormat(
"%s %s %s %s %s.",
"format", "multiple", "utf8", "short", "strings");
Result: 620 ns +- 8 ns -> 382 ns +- 2 ns: 1.62x faster.
Add `Py_BEGIN_CRITICAL_SECTION_SEQUENCE_FAST` and
`Py_END_CRITICAL_SECTION_SEQUENCE_FAST` macros and update `str.join` to use
them. Also add a regression test that would crash reliably without this
patch.
As reported in #117847 and #115366, an unpaired backtick in a docstring
tends to confuse e.g. Sphinx running on subclasses of standard library
objects, and the typographic style of using a backtick as an opening
quote is no longer in favor. Convert almost all uses of the form
The variable `foo' should do xyz
to
The variable 'foo' should do xyz
and also fix up miscellaneous other unpaired backticks (extraneous /
missing characters).
No functional change is intended here other than in human-readable
docstrings.
The PEP 649 implementation will require a way to load NotImplementedError
from the bytecode. @markshannon suggested implementing this by converting
LOAD_ASSERTION_ERROR into a more general mechanism for loading constants.
This PR adds this new opcode. I will work on the rest of the implementation
of the PEP separately.
Co-authored-by: Irit Katriel <1055913+iritkatriel@users.noreply.github.com>
* BaseException_vectorcall() now creates a tuple from 'args' array.
* Creation an exception using BaseException_vectorcall() is now a
single function call, rather than having to call
BaseException_new() and then BaseException_init().
Calling BaseException_init() is inefficient since it overrides
the 'args' attribute.
* _PyErr_SetKeyError() now uses PyObject_CallOneArg() to create the
KeyError instance to use BaseException_vectorcall().
The `list_preallocate_exact` function did not zero initialize array
contents. In the free-threaded build, this could expose uninitialized
memory to concurrent readers between the call to
`list_preallocate_exact` and the filling of the array contents with
items.
Use relaxed atomics when reading / writing to the field. There are still a
few places in the GC where we do not use atomics. Those should be safe as
the world is stopped.
We already intern and immortalize most string constants. In the
free-threaded build, other constants can be a source of reference count
contention because they are shared by all threads running the same code
objects.
Add _PyType_LookupRef and use incref before setting attribute on type
Makes setting an attribute on a class and signaling type modified atomic
Avoid adding re-entrancy exposing the type cache in an inconsistent state by decrefing after type is updated
This interns the strings for `co_filename`, `co_name`, and `co_qualname`
on codeobjects in the free-threaded build. This partially addresses a
reference counting bottleneck when creating closures concurrently. The
closures take the name and qualified name from the code object.
This PR adds the ability to enable the GIL if it was disabled at
interpreter startup, and modifies the multi-phase module initialization
path to enable the GIL when loading a module, unless that module's spec
includes a slot indicating it can run safely without the GIL.
PEP 703 called the constant for the slot `Py_mod_gil_not_used`; I went
with `Py_MOD_GIL_NOT_USED` for consistency with gh-104148.
A warning will be issued up to once per interpreter for the first
GIL-using module that is loaded. If `-v` is given, a shorter message
will be printed to stderr every time a GIL-using module is loaded
(including the first one that issues a warning).
The module itself is a thin wrapper around calls to functions in
`Python/codecs.c`, so that's where the meaningful changes happened:
- Move codecs-related state that lives on `PyInterpreterState` to a
struct declared in `pycore_codecs.h`.
- In free-threaded builds, add a mutex to `codecs_state` to synchronize
operations on `search_path`. Because `search_path_mutex` is used as a
normal mutex and not a critical section, we must be extremely careful
with operations called while holding it.
- The codec registry is explicitly initialized as part of
`_PyUnicode_InitEncodings` to simplify thread-safety.
The code for Tier 2 is now only compiled when configured
with `--enable-experimental-jit[=yes|interpreter]`.
We drop support for `PYTHON_UOPS` and -`Xuops`,
but you can disable the interpreter or JIT
at runtime by setting `PYTHON_JIT=0`.
You can also build it without enabling it by default
using `--enable-experimental-jit=yes-off`;
enable with `PYTHON_JIT=1`.
On Windows, the `build.bat` script supports
`--experimental-jit`, `--experimental-jit-off`,
`--experimental-interpreter`.
In the C code, `_Py_JIT` is defined as before
when the JIT is enabled; the new variable
`_Py_TIER2` is defined when the JIT *or* the
interpreter is enabled. It is actually a bitmask:
1: JIT; 2: default-off; 4: interpreter.
Deferred reference counting is not fully implemented yet. As a temporary
measure, we immortalize objects that would use deferred reference
counting to avoid multi-threaded scaling bottlenecks.
This is only performed in the free-threaded build once the first
non-main thread is started. Additionally, some tests, including refleak
tests, suppress this behavior.
It's not safe to raise an exception in `PyObject_ClearWeakRefs()` if one
is not already set, since it may be called by `_Py_Dealloc()`, which
requires that the active exception does not change.
Additionally, make sure we clear the weakrefs even when tuple allocation
fails.
Fix data races in the method cache in free-threaded builds
These are technically data races, but I think they're benign (to
the extent that that is actually possible). We update cache entries
non-atomically but read them atomically from another thread, and there's
nothing that establishes a happens-before relationship between the
reads and writes that I can see.
Stefano Rivera
satori1995
Sam Gross
Serhiy Storchaka
byundojin
Bénédikt Tran
AN Long
Sergey B Kirpichev
Steve Dower
Irit Katriel
Victor Stinner
Ken Jin
Xie Yanbo
Pablo Galindo Salgado
Petr Viktorin
Nikita Sobolev
Mark Shannon
Eric Snow
Steele Farnsworth
Ruben Vorderman
E. M. Bray
Saul Shanabrook
Kirill Podoprigora
d.grigonis
Donghee Na
Mark Dickinson
Alyssa Coghlan
Jelle Zijlstra
Germán Méndez Bravo
Ayato Hayashi
Josh {*()} Rosenberg
Dino Viehland
Geoffrey Thomas
Tian Gao
mpage
James Hilton-Balfe
Thomas Grainger
Brett Simmers
Guido van Rossum
neonene
Shantanu
Noah Kim
Rostyslav Lobov
Jens Hedegaard Nielsen
Erlend E. Aasland