Context
Following the recent work, the next optimization pass should focus on reworking the core data structures. A significant share of the performance loss also comes from image processing, which is a prime candidate for acceleration.
Goals
- Refactor by failure class
- Re-run the test suite, identify the largest class of failures, and refactor it. Iterate from there.
- Optimize image processing
- Image processing accounts for a large part of the performance loss. Investigate:
Parallelism / vectorization
Other general-purpose acceleration techniques
3. Improve remaining patterns
Address the dyn patterns. There is likely a crate that handles this more cleanly, worth researching.
Trait segmentation was pushed a bit too far. We should be able to remove all impl blocks that just return Some(self).
Coding guidelines
Per agents.md:
Write semantic Rust
Keep abstraction levels distinct
Code should read like documentation
Context
Following the recent work, the next optimization pass should focus on reworking the core data structures. A significant share of the performance loss also comes from image processing, which is a prime candidate for acceleration.
Goals
Parallelism / vectorization
Other general-purpose acceleration techniques
3. Improve remaining patterns
Address the dyn patterns. There is likely a crate that handles this more cleanly, worth researching.
Trait segmentation was pushed a bit too far. We should be able to remove all impl blocks that just return Some(self).
Coding guidelines
Per agents.md:
Write semantic Rust
Keep abstraction levels distinct
Code should read like documentation