Title: gh-115999: Specialize `CALL_KW` in free-threaded builds by mpage · Pull Request #127713 · python/cpython · GitHub
Open Graph Title: gh-115999: Specialize `CALL_KW` in free-threaded builds by mpage · Pull Request #127713 · python/cpython
X Title: gh-115999: Specialize `CALL_KW` in free-threaded builds by mpage · Pull Request #127713 · python/cpython
Description: The CALL_KW family was already thread-safe. Only one change to fix a bug in _PY_FRAME_KW was needed: _PY_FRAME_KW pushes a pointer to the new frame onto the stack for consumption by the next uop. When pushing the frame fails, we do not want to push the result, NULL, to the stack because it is not a valid stackref. This works in the default build because PyStackRef_NULL and NULL are the same value, so the PyStackRef_XCLOSE() in the error handler ignores it. In the free-threaded build the values are not the same, causing PyStackRef_XCLOSE() to decref a null pointer. Single-threaded performance Performance looks neutral on both builds. I think this is still worth merging to keep the two builds consistent, however. Default build Free-threaded build Scalability Scalability looks roughly unchanged: Base PR object_cfunction 1.5x slower 1.4x slower cmodule_function 1.6x slower 1.6x slower mult_constant 12.8x faster 13.2x faster generator 13.5x faster 10.1x faster pymethod 1.9x slower 1.9x slower pyfunction 10.7x faster 13.4x faster module_function 2.1x slower 2.0x slower load_string_const 14.2x faster 13.1x faster load_tuple_const 13.6x faster 13.2x faster create_pyobject 15.5x faster 15.7x faster create_closure 12.3x faster 13.5x faster create_dict 13.3x faster 12.9x faster thread_local_read 4.0x slower 3.8x slower Thread safety The specialized instructions are composed of the follow uops, whose thread-safety is documented below. _CALL_KW_NON_PY - Tuples are immutable; it's fine to read their size non-atomically. _CHECK_FUNCTION_VERSION_KW - This uop guards that the top of the stack is a function and that its version matches the version stored in the inline cache. Instructions assume that if the guard passes, the version, and any properties verified by the version, will not change for the remainder of the instruction execution, assuming there are no escaping calls in between the guard and the code that relies on the guard. This property is preserved in free-threaded builds: the world is stopped whenever a function's version changes. _CHECK_IS_NOT_PY_CALLABLE_KW - Performs type checks. _CHECK_METHOD_VERSION_KW - This loads a function from a PyMethodObject and guards that its version matches what is stored in the cache. PyMethodObjects are immutable; their fields can be accessed non-atomically. The thread safety of function version guards was already documented above. _CHECK_PEP_523 - Thread safety was addressed as part of CALL. _CHECK_PERIODIC - Thread safety was previously addressed as part of the 3.13 release. _EXPAND_METHOD_KW - Only loads from PyMethodObjects and performs type checks. _PUSH_FRAME - Only manipulates tstate->current_frame and fields that are not read by other threads. _PY_FRAME_KW - Loads immutable fields (tuple size, code flags) or fields that require stopping the world to modify (function code) and uses thread-safe APIs. _SAVE_RETURN_OFFSET - Stores only to the frame's return_offset which is not read by other threads. Specialization only looks at properties that require stopping the world to update (i.e. instance type, eval_frame, func_code, and the function version) and contains no escaping calls. Issue: gh-115999
Open Graph Description: The CALL_KW family was already thread-safe. Only one change to fix a bug in _PY_FRAME_KW was needed: _PY_FRAME_KW pushes a pointer to the new frame onto the stack for consumption by the next uop. W...
X Description: The CALL_KW family was already thread-safe. Only one change to fix a bug in _PY_FRAME_KW was needed: _PY_FRAME_KW pushes a pointer to the new frame onto the stack for consumption by the next uop. W...
Opengraph URL: https://github.com/python/cpython/pull/127713
X: @github
Domain: github.com
| route-pattern | /:user_id/:repository/pull/:id/files(.:format) |
| route-controller | pull_requests |
| route-action | files |
| fetch-nonce | v2:72674e33-7c30-cc9d-b5c1-35795f49f60f |
| current-catalog-service-hash | ae870bc5e265a340912cde392f23dad3671a0a881730ffdadd82f2f57d81641b |
| request-id | D3DE:37ED2:8DBD1:BDFD7:6A5676AD |
| html-safe-nonce | 3fd261eaf23e9a1efd27dc71061cd6967007ec1f1be758657d00e5deef790eba |
| visitor-payload | eyJyZWZlcnJlciI6IiIsInJlcXVlc3RfaWQiOiJEM0RFOjM3RUQyOjhEQkQxOkJERkQ3OjZBNTY3NkFEIiwidmlzaXRvcl9pZCI6IjYxMTY4MDg5NjQ5NDI3NTU1MDEiLCJyZWdpb25fZWRnZSI6ImlhZCIsInJlZ2lvbl9yZW5kZXIiOiJpYWQifQ== |
| visitor-hmac | 1c39bf82527f4229744e0c1cb0a43e063b40db493db874c7b8693c1b5dc09a78 |
| hovercard-subject-tag | pull_request:2221031920 |
| github-keyboard-shortcuts | repository,pull-request-list,pull-request-conversation,pull-request-files-changed,copilot |
| google-site-verification | Apib7-x98H0j5cPqHWwSMm6dNU4GmODRoqxLiDzdx9I |
| octolytics-url | https://collector.github.com/github/collect |
| analytics-location | / |
| fb:app_id | 1401488693436528 |
| apple-itunes-app | app-id=1477376905, app-argument=https://github.com/python/cpython/pull/127713/files |
| twitter:image | https://avatars.githubusercontent.com/u/577841?s=400&v=4 |
| twitter:card | summary_large_image |
| og:image | https://avatars.githubusercontent.com/u/577841?s=400&v=4 |
| og:image:alt | The CALL_KW family was already thread-safe. Only one change to fix a bug in _PY_FRAME_KW was needed: _PY_FRAME_KW pushes a pointer to the new frame onto the stack for consumption by the next uop. W... |
| og:site_name | GitHub |
| og:type | object |
| hostname | github.com |
| expected-hostname | github.com |
| None | 19b5ee9431dcfce15ab4b001d626e4e3a6ba1acb7be814813a52bbebd9a86ef5 |
| turbo-cache-control | no-preview |
| diff-view | unified |
| go-import | github.com/python/cpython git https://github.com/python/cpython.git |
| octolytics-dimension-user_id | 1525981 |
| octolytics-dimension-user_login | python |
| octolytics-dimension-repository_id | 81598961 |
| octolytics-dimension-repository_nwo | python/cpython |
| octolytics-dimension-repository_public | true |
| octolytics-dimension-repository_is_fork | false |
| octolytics-dimension-repository_network_root_id | 81598961 |
| octolytics-dimension-repository_network_root_nwo | python/cpython |
| turbo-body-classes | logged-out env-production page-responsive full-width |
| disable-turbo | true |
| browser-stats-url | https://api.github.com/_private/browser/stats |
| browser-errors-url | https://api.github.com/_private/browser/errors |
| release | afc01ffdb045f4ff81c6122091366ade7b6166b7 |
| ui-target | full |
| theme-color | #1e2327 |
| color-scheme | light dark |
Links:
Viewport: width=device-width