René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Make `dict` objects thread-safe in `--disable-gil` builds","articleBody":"We are going to need a variety of techniques to make dictionaries thread-safe in `--disable-gil` builds. I expect this to be implemented across multiple PRs.\r\n\r\nFor context, here is the change from the `nogil-3.12` fork, but things might be done a bit differently in CPython 3.13: https://github.com/colesbury/nogil-3.12/commit/d896dfc8db\r\n\r\n1) Most operations should acquire the dictionary object lock using the critical section API.\r\n2) Dictionary global version counter should use atomic increments \r\n3) `PyDictKeysObject` needs special handling (see below)\r\n4) `PyDictOrValues` needs special handling (see below)\r\n5) Accessing a single element should optimistically avoid locking for performance\r\n\r\nIn general, we want to avoid making changes which hurt the performance of the default build. The \"critical sections\" are no-ops in the default build, but we will need to take care with other changes. Some may be behind `#ifdef` guards.\r\n\r\n### `PyDictKeysObject`\r\n\r\nNote that `PyDictKeysObject` is *NOT* a `PyObject` subclass.\r\n\r\nWe need a mutex in `PyDictKeysObject` because multiple threads may concurrently insert keys into shared `PyDictKeysObject`. The mutex is only used for modifications to \"shared\" keys; non-shared keys rely on the locking for the dict object.\r\n\r\nIn `--disable-gil` builds, we want to avoid refcounting shared `PyDictKeysObject` for performance and thread-safety reasons. Instead shared keys objects should be freed during cyclic GC. (Non-shared keys don't need reference counting because they \"owned\" by a single dict object.) We may want to consider making this change for both the default build and `--disable-gil` builds to make maintenance easier if there is not a negative perf impact.\r\n\r\n### `PyDictOrValues`\r\n\r\n`PyDictOrValues` is the \"managed\" dict in some PyObject pre-headers. We need some locking/atomic operations to handle the transition between `PyDictValues*` and storing a `PyDictObject*`\r\n\r\n### Optimistically Avoid Locking\r\n\r\nSee https://peps.python.org/pep-0703/#optimistically-avoiding-locking.\n\n\u003c!-- gh-linked-prs --\u003e\n### Linked PRs\n* gh-112247\n* gh-114256\n* gh-114508\n* gh-114512\n* gh-114543\n* gh-114568\n* gh-114629\n* gh-114741\n* gh-114742\n* gh-114768\n* gh-114778\n* gh-114931\n* gh-115108\n* gh-115109\n* gh-115786\n* gh-115855\n* gh-116087\n* gh-116336\n* gh-116344\n* gh-116620\n* gh-117528\n* gh-117529\n* gh-117824\n* gh-118499\n* gh-118676\n* gh-119288\n* gh-121545\n* gh-129336\n* gh-139737\n\u003c!-- /gh-linked-prs --\u003e\n","author":{"url":"https://github.com/colesbury","@type":"Person","name":"colesbury"},"datePublished":"2023-11-14T19:14:52.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":9},"url":"https://github.com/112075/cpython/issues/112075"}