EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Lobsters·Aug 23·5 min read
Optimizing memory usage in a markdown parser
TL;DR
A C++ port of the Rust markdown parser markdown-rs was developed, focusing on optimizing the Abstract Syntax Tree (AST) node size. The largest node in Rust is 152 bytes, and the goal is to minimize this in the C++ version.
✦ Why It Matters
Consider reviewing your data structures to minimize memory usage in your C++ applications.
Key Takeaways
- Implement an arena allocator for efficient memory management.
- Reorder struct fields to minimize padding and reduce size.
- Compress pointers to save memory in child node arrays.
- Optimize string handling to prevent unnecessary allocations.
How It Works
The arena allocator allows for all nodes to be allocated in a single memory block, which simplifies memory management and enables optimizations like pointer compression. By reordering struct fields and packing boolean values, the overall size of the Node struct was minimized, leading to significant memory savings.