{"id":160720,"date":"2026-02-12T17:11:59","date_gmt":"2026-02-12T15:11:59","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/v-yazyke-nim-predstavlen-yrc-potokobezopasnyj-sborshhik-cziklicheskih-ssylok"},"modified":"2026-02-12T17:11:59","modified_gmt":"2026-02-12T15:11:59","slug":"v-yazyke-nim-predstavlen-yrc-potokobezopasnyj-sborshhik-cziklicheskih-ssylok","status":"publish","type":"post","link":"https:\/\/prohoster.info\/en\/blog\/news\/v-yazyke-nim-predstavlen-yrc-potokobezopasnyj-sborshhik-cziklicheskih-ssylok","title":{"rendered":"The Nim language introduces YRC \u2014 a thread-safe cyclic reference collector.","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
Andreas Rumpf (Araq), the creator of the Nim programming language, has announced a new memory management algorithm called YRC (pronounced \u201cUrk\u201d), which addresses one of the key issues with existing mechanisms in Nim: the inability to correctly handle cyclic references that cross thread boundaries. <\/p>\n
Before the introduction of YRC, the proposed memory management algorithms in Nim had the following limitations: ARC \u2014 did not support multi-threading or cycle handling; Atomic ARC \u2014 was thread-safe but did not handle cyclic references; ORC \u2014 handled cyclic references but could do so correctly only within a single thread (using shared references between threads resulted in memory leaks). <\/p>\n
The proposed YRC combines thread safety with cycle handling across threads by using a hybrid approach: for acyclic data, it employs atomic reference counting, while for cyclic data \u2014 a write barrier that activates only when pointer assignments occur. In the proposed implementation, the collector runs as needed (no stop-the-world pauses); root RC objects are explicitly defined and collected once (no requirement to scan thread stacks); multi-threaded deletion is avoided during iteration (there's no global sweep phase); mutators can freely read data; any thread can trigger garbage collection as needed (without a dedicated GC thread). <\/p>\n
YRC utilizes complete information about incRef\/decRef events, which traditional tracing garbage collectors discard and then must restore through stack scanning and graph traversal. The implementation is only 550 lines of code and has formal verification for safety and freedom from deadlocks through specifications in TLA+ and proofs in the Lean toolkit. YRC is positioned as 'the almost last cycle collector based on reference counting' (the letter Y precedes Z in the alphabet) and as the simplest thread-safe garbage collector \u2014 according to the author, it does not require the multitude of complex mechanisms inherent in traditional tracing garbage collectors. <\/p>\n
YRC provides the same API as ORC, with destructors executed during garbage collection. The collector only processes those object subgraphs that threads are working with, leaving unrelated data structures (caches, long-lived objects) untouched \u2014 a method similar to a 'perfect' generational garbage collector without using generations. The main drawback is performance: YRC shows a slowdown of 1.5-2.0 times compared to ORC in the orcbench performance test. The author considers this an acceptable trade-off for fully thread-safe handling of cyclic references. <\/p>\n
YRC is already available in the development version of Nim and is enabled with the flag '\u2014mm:yrc'. However, in later messages, the author admitted that the initial implementation contained serious bugs and did not collect cycles correctly. At the time of publication, a set of fixes has been prepared that addresses the major issues. The author continues to refine the collection heuristics and fix remaining bugs, while the basic algorithm and its formal verification remain correct. \n Source: opennet.ru<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"