Explore a 21-minute conference talk from Dyalog '24 where Aaron Hsu presents recent developments and future plans for the Co-dfns compiler, focusing on improved integration with traditional APL code. Learn about the compiler's current performance capabilities, implementation features, and practical use cases including GPU utilization, scalar operations optimization, static analysis, and system integration. Discover insights into traditional APL programming techniques, parsing mechanisms, scoping considerations, and memory management improvements. Gain valuable recommendations for compiler usage, understand its current limitations, and explore unique APL programming possibilities enabled by Co-dfns. Access accompanying presentation materials in both PPTX and PDF formats for detailed reference of concepts covered in this comprehensive technical discussion.
Co-dfns Compiler Roadmap and Updates - Implementation and Integration Features
Dyalog User Meetings via YouTube
Overview
Syllabus
A compiled implementation of APL
A guide to APL programming techniques
A tool for scaling APL
Use case #1: GPUs
Use case #2: Scalar Hot Spots
Use case #3: Static Analysis
Use case #4: Integration
Traditional APL
Traditional APL: Features
Traditional APL: Parse vs. Exec
Parser-only supported features
Trad-fns: Scoping concerns
Trad-fns: Scoping approach
Lending a helping hand
Current efforts
Progress: Parser
Progress: Memory overheads
Benchmarking
Recommendations/Limits
Unique features
Taught by
Dyalog User Meetings
