Parallelization of Dynamic Languages - Synchronizing Built-in Collections

on YouTube ↗23m

" summary

An implementation in TruffleRuby adds a local vs shared heap tag and applies gradual synchronization to Ruby Array and Hash instances, switching from no locks to a lightweight layout lock and lock-free bucket/list structures. Benchmarks up to 44 CPUs show near-zero single-thread overhead, linear scaling for Array/Hash accesses, and scalability comparable to Fortran/Java on NAS parallel kernels; the system is implemented and the measurements support the main claims.

AI-written summary · about

" chapters (7)

1. 0:00Start
2. 0:17Contribution Parallelization of Dynamic Languages
3. 1:23Background Concurrency in Dynamic Languages
4. 2:44Challenges
5. 5:12Concurrent Strategies
6. 17:30Lightweight Layout Lock Concurrent Accesses
7. 18:32Summary Dynamic Languages Built-in Collections

" related talks

• 38m"Parallel and Thread-Safe Ruby at High-Speed with TruffleRuby"Benoit Daloze · RubyC 2019
• 43mParallel and Thread-Safe Ruby at High-Speed with TruffleRubyBenoit Daloze · Ruby on Ice 2019
• 25mMaking Hash Parallel, Thread-Safe and Fast!Benoit Daloze · RubyKaigi 2026
• 28mConcurrency for RubyistsFederico Carrone · RubyConf Argentina 2014
• 23mParallel Processing With RubyDilum Navanjana · RubyConf MY 2017
• 34mWriting concurrent libraries for all Ruby runtimesPetr Chalupa · RubyConf 2015
• 40mRUBY 3.0 - SINGLE PROCESS, MULIPLE POSSIBILITIESMaciej Mensfeld · Grill.rb 2017
• 39mParallel programming in Ruby3 with GuildKoichi Sasada · RubyConf 2018