Optimizing Optimizations, Declaratively: Optimizing the Higher-Order Functions in(…)
summary
Two concrete uses of egglog inside JijModeling 2: resugaring comprehensions reconstructs Python comprehension syntax from map/flat_map/filter and then applies a custom cost model to minimize temporary let rebindings, typically finishing within 100 ms; declarative constraint detection encodes parameterized constraints with Henkin-like constants, propagates side conditions inside egglog, and turns a previously slow/nonterminating detection case from minutes or nontermination down to a few seconds. Practical performance requires decomposing and optimizing egglog rule premises.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:46What is JuJu Modeling
- 4:15Desugaring comprehensions
- 8:57Resugaring comprehensions
- 9:19Constraint detection with Egglog
- 17:14Optimizing Egglog rules
- 19:00Conclusion
- 20:04Q&A
related talks
41mNo stats, no problem: building feedback-driven optimizers for lakehouses
14mAchieving High-Performance the Functional Way - A Functional Pearl on Expressing High...
44mMachine Learning for Faster Optimization
43mPost Functional
24mCost and Performance Optimization of LLM Inferencing.
39mPractical Optimization for Stats Nerds
57mGood Practices for Applied Machine Learning- 57mGood Practices for Applied Machine Learning