Add special cases for classmethod and staticmethod descriptors in
_PyObject_GetMethodStackRef() to avoid calling tp_descr_get, which
avoids reference count contention on the bound method and underlying
callable. This improves scaling when calling classmethods and
staticmethods from multiple threads.
Also refactor method_vectorcall in classobject.c into a new _PyObject_VectorcallPrepend() helper so that it can be used by
PyObject_VectorcallMethod as well.
Add `_Py_type_getattro_stackref`, a variant of type attribute lookup
that returns `_PyStackRef` instead of `PyObject*`. This allows returning
deferred references in the free-threaded build, reducing reference count
contention when accessing type attributes.
This significantly improves scaling of namedtuple instantiation across
multiple threads.
* Add blurb
* Rename PyObject_GetAttrStackRef to _PyObject_GetAttrStackRef
* Apply suggestion from @vstinner
Co-authored-by: Victor Stinner <vstinner@python.org>
* Apply suggestion from @vstinner
Co-authored-by: Victor Stinner <vstinner@python.org>
* format
* Update Include/internal/pycore_function.h
Co-authored-by: Victor Stinner <vstinner@python.org>
---------
Co-authored-by: Victor Stinner <vstinner@python.org>
This combines most _PyStackRef functions and macros between the free
threaded and default builds.
- Remove Py_TAG_DEFERRED (same as Py_TAG_REFCNT)
- Remove PyStackRef_IsDeferred (same as !PyStackRef_RefcountOnObject)
Replace code that directly accesses PyASCIIObject.hash with
PyUnstable_Unicode_GET_CACHED_HASH().
Remove redundant "assert(PyUnicode_Check(op))" from
PyUnstable_Unicode_GET_CACHED_HASH(), _PyASCIIObject_CAST() already
implements the check.
In the free-threaded build, avoid data races caused by updating type
slots or type flags after the type was initially created. For those
(typically rare) cases, use the stop-the-world mechanism. Remove the
use of atomics when reading or writing type flags.
Adds `_PyObject_GetMethodStackRef` which uses stackrefs and takes advantage of deferred reference counting in free-threading while calling method objects in vectorcall.
In the free-threaded build, avoid data races caused by updating type slots
or type flags after the type was initially created. For those (typically
rare) cases, use the stop-the-world mechanism. Remove the use of atomics
when reading or writing type flags. The use of atomics is not sufficient to
avoid races (since flags are sometimes read without a lock and without
atomics) and are no longer required.
In the free threaded build, the `_PyObject_LookupSpecial()` call can lead to
reference count contention on the returned function object becuase it
doesn't use stackrefs. Refactor some of the callers to use
`_PyObject_MaybeCallSpecialNoArgs`, which uses stackrefs internally.
This fixes the scaling bottleneck in the "lookup_special" microbenchmark
in `ftscalingbench.py`. However, the are still some uses of
`_PyObject_LookupSpecial()` that need to be addressed in future PRs.
This fixes a fairly subtle bug involving finalizers and resurrection in
debug free threaded builds: if `_PyObject_ResurrectEnd` returns `1`
(i.e., the object was resurrected by a finalizer), it's not safe to
access the object because it might still be deallocated. For example:
* The finalizer may have exposed the object to another thread. That
thread may hold the last reference and concurrently deallocate it any
time after `_PyObject_ResurrectEnd()` returns `1`.
* `_PyObject_ResurrectEnd()` may call `_Py_brc_queue_object()`, which
may internally deallocate the object immediately if the owning thread
is dead.
Therefore, it's important not to access the object after it's
resurrected. We only violate this in two cases, and only in debug
builds:
* We assert that the object is tracked appropriately. This is now moved
up betewen the finalizer and the `_PyObject_ResurrectEnd()` call.
* The `--with-trace-refs` builds may need to remember the object if
it's resurrected. This is now handled by `_PyObject_ResurrectStart()`
and `_PyObject_ResurrectEnd()`.
Note that `--with-trace-refs` is currently disabled in `--disable-gil`
builds because the refchain hash table isn't thread-safe, but this
refactoring avoids an additional thread-safety issue.
In the free threading build, the per thread reference counting uses a
unique id for some objects to index into the local reference count
table. Use 0 instead of -1 to indicate that the id is not assigned. This
avoids bugs where zero-initialized heap type objects look like they have
a unique id assigned.
* Use TABLES_LOCK() to protect 'tracemalloc_config.tracing'.
* Hold TABLES_LOCK() longer while accessing tables.
* tracemalloc_realloc() and tracemalloc_free() no longer
remove the trace on reentrant call.
* _PyTraceMalloc_Stop() unregisters _PyTraceMalloc_TraceRef().
* _PyTraceMalloc_GetTraces() sets the reentrant flag.
* tracemalloc_clear_traces_unlocked() sets the reentrant flag.
Objects may be temporarily "resurrected" in destructors when calling
finalizers or watcher callbacks. We previously undid the resurrection
by decrementing the reference count using `Py_SET_REFCNT`. This was not
thread-safe because other threads might be accessing the object
(modifying its reference count) if it was exposed by the finalizer,
watcher callback, or temporarily accessed by a racy dictionary or list
access.
This adds internal-only thread-safe functions for temporary object
resurrection during destructors.
The CALL family of instructions were mostly thread-safe already and only required a small number of changes, which are documented below.
A few changes were needed to make CALL_ALLOC_AND_ENTER_INIT thread-safe:
Added _PyType_LookupRefAndVersion, which returns the type version corresponding to the returned ref.
Added _PyType_CacheInitForSpecialization, which takes an init method and the corresponding type version and only populates the specialization cache if the current type version matches the supplied version. This prevents potentially caching a stale value in free-threaded builds if we race with an update to __init__.
Only cache __init__ functions that are deferred in free-threaded builds. This ensures that the reference to __init__ that is stored in the specialization cache is valid if the type version guard in _CHECK_AND_ALLOCATE_OBJECT passes.
Fix a bug in _CREATE_INIT_FRAME where the frame is pushed to the stack on failure.
A few other miscellaneous changes were also needed:
Use {LOCK,UNLOCK}_OBJECT in LIST_APPEND. This ensures that the list's per-object lock is held while we are appending to it.
Add missing co_tlbc for _Py_InitCleanup.
Stop/start the world around setting the eval frame hook. This allows us to read interp->eval_frame non-atomically and preserves the behavior of _CHECK_PEP_523 documented below.
* Mark almost all reachable objects before doing collection phase
* Add stats for objects marked
* Visit new frames before each increment
* Update docs
* Clearer calculation of work to do.
Enable specialization of LOAD_GLOBAL in free-threaded builds.
Thread-safety of specialization in free-threaded builds is provided by the following:
A critical section is held on both the globals and builtins objects during specialization. This ensures we get an atomic view of both builtins and globals during specialization.
Generation of new keys versions is made atomic in free-threaded builds.
Existing helpers are used to atomically modify the opcode.
Thread-safety of specialized instructions in free-threaded builds is provided by the following:
Relaxed atomics are used when loading and storing dict keys versions. This avoids potential data races as the dict keys versions are read without holding the dictionary's per-object lock in version guards.
Dicts keys objects are passed from keys version guards to the downstream uops. This ensures that we are loading from the correct offset in the keys object. Once a unicode key has been stored in a keys object for a combined dictionary in free-threaded builds, the offset that it is stored in will never be reused for a different key. Once the version guard passes, we know that we are reading from the correct offset.
The dictionary read fast-path is used to read values from the dictionary once we know the correct offset.
* Mark almost all reachable objects before doing collection phase
* Add stats for objects marked
* Visit new frames before each increment
* Remove lazy dict tracking
* Update docs
* Clearer calculation of work to do.
Use per-thread refcounting for the reference from function objects to
their corresponding code object. This can be a source of contention when
frequently creating nested functions. Deferred refcounting alone isn't a
great fit here because these references are on the heap and may be
modified by other libraries.
Currently, we only use per-thread reference counting for heap type objects and
the naming reflects that. We will extend it to a few additional types in an
upcoming change to avoid scaling bottlenecks when creating nested functions.
Rename some of the files and functions in preparation for this change.
Petr Viktorin