Methods of Memory Management in MRI

on YouTube ↗49m

" summary

Covers how MRI actually allocates and reclaims memory: Ruby heap layout with 16KB pages, 40B object slots; tag‑pointer fixnum encoding and the 2^62 limit; mark‑and‑sweep, generational + tri‑color incremental tracing with write‑barriers and remembered sets; allocation slabs, bump pointer, eden/tomb behavior; introspection via GC.stat, ObjectSpace.dump and GC::Profiler; experimental heap‑splitting and a GitHub patch that groups classes/modules and yields a ~17% heap reduction on a Rails app — includes code, numbers, and measured results.

AI-written summary · about

" chapters (19)

1. 0:00Start
2. 5:35Bear Metal
3. 16:30Mark Phase
4. 16:57Sweep Phase
5. 17:12Mark & Sweep
6. 18:37Generational
7. 22:32Object Colors
8. 22:48Algorithm
9. 24:36Tri-Color Marking
10. 29:38Page Layout
11. 30:22Full Pages
12. 34:37Encode Number 2
13. 35:14Biggest Fixnum
14. 36:40Fixnums are singletons
15. 37:35Tagged Pointers
16. 37:59Poor Reclamation
17. 40:14Two Page Types
18. 46:01GC::Profiler
19. 46:44ObjectSpace.dump

" related talks

• 48mMethods of Memory Management in MRIAaron Patterson · RubyConf Portugal 2016
• 44mKeynote: Taking Out The TrashAaron Patterson · RedDotRubyConf 2016
• 01h 01mKeynote: Mark/Compact GC in MRIAaron Patterson · Red Dot Ruby Conference 2017
• 54mKeynote: Exploring Memory in Ruby - Building a Compacting GCAaron Patterson · RubyConf MY 2017
• 51mCompacting GC for MRIAaron Patterson · Balkan Ruby 2019
• 42mThe Mechanics of RubyAdam Keys · LoneStarRuby Conf 2013
• 46mBuilding a Compacting GC for MRIAaron Patterson · RubyConf 2017
• 24mMRI Magic TricksCharlie Somerville · Wicked Good Ruby 2013