Methods of Memory Management in MRI

on YouTube ↗48m

" summary

MRI uses a mark-and-sweep collector that is generational and incremental, with write barriers and a remembered set to avoid missing newly-referenced objects. Ruby allocates 16KB slabs of 40-byte slots, uses tagged pointers for small integers/true/false/nil, and promotes objects after three GCs to promote to older generations. A page-segregation patch groups long-lived objects and reduces Rails heap by 17% smaller heap in tests. Useful tools: GC.stat, GC::Profiler, ObjectSpace.dump_all and undocumented RUBY_GC_* environment tunables.

AI-written summary · about

" chapters (9)

1. 0:00Introduction
2. 5:03Memory basics in Ruby
3. 10:35Garbage collection algorithms
4. 27:40Allocation algorithms
5. 36:57Tagged pointers and hacks
6. 38:21Copy-on-write issues
7. 40:32Optimizing heap usage
8. 43:03GC introspection APIs
9. 46:14GC tuning in production

" related talks

• 49mMethods of Memory Management in MRIAaron Patterson · RubyConf 2016
• 44mKeynote: Taking Out The TrashAaron Patterson · RedDotRubyConf 2016
• 01h 01mKeynote: Mark/Compact GC in MRIAaron Patterson · Red Dot Ruby Conference 2017
• 51mCompacting GC for MRIAaron Patterson · Balkan Ruby 2019
• 54mKeynote: Exploring Memory in Ruby - Building a Compacting GCAaron Patterson · RubyConf MY 2017
• 46mBuilding a Compacting GC for MRIAaron Patterson · RubyConf 2017
• 51mCompacting GC in MRIAaron Patterson · RubyConf Colombia 2017
• 48mDefragging RubyAaron Patterson · RubyConf AU 2017