Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request addresses a critical memory issue in ONNX-based GPU feature extraction by preventing excessive thread allocation, which previously led to unbounded memory growth. It also enhances the flexibility and correctness of feature extraction by making the GPU/CPU dispatch mechanism aware of the specific feature type being processed, rather than relying on SIFT-specific configurations. These changes improve stability and resource utilization during feature extraction.

Highlights

• Memory Management: Fixed unbounded memory growth in ONNX-based GPU feature extraction by preventing each extractor thread from inheriting the full hardware_concurrency() thread count for its internal ONNX session.
• Feature Extraction Logic: Refactored the dispatch logic for GPU vs. CPU extraction to be feature-type-aware, moving away from SIFT-specific options for all extractor types.
• Threading Optimization: Implemented measures to prevent nested threading within feature extractors, ensuring that num_threads is appropriately set for both GPU and CPU workers to avoid over-subscription.

• src/colmap/controllers/feature_extraction.cc
  • Refactored the condition for GPU extraction to consider feature type and SIFT-specific options (domain_size_pooling, estimate_affine_shape).
  • Introduced a worker_extraction_options variable to manage extraction settings for individual threads.
  • Adjusted the num_threads for GPU extractors to prevent nested threading, dividing the total threads by the number of GPU indices.
  • Removed redundant explicit use_gpu = false setting for CPU extractors, as the worker_extraction_options now correctly handles GPU enablement.

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

ahojnnes deleted the user/jsch/nested-threading-onnx-extractor branch