Optimize LBVH Construction using Bottom-up Build

[zfergus]

Description

• Replace top-down split search with a single bottom-up pass (Apetrei 2014) to build the hierarchy and bounding boxes simultaneously
• Update ConstructionInfo to track range endpoints instead of parent indices
• Add Catch2 benchmarks for LBVH::build performance testing

Fixes # (issue)

The single-pass bottom-up construction of Apetrei [2014] is up to 10% faster than the two-pass construction of Karras [2012]. Benchmarked on Apple M3 Pro with 11 cores.

Type of change

• Enhancement (non-breaking change which improves existing functionality)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.84%. Comparing base (e400bb0) to head (abaf7f6).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #230      +/-   ##
==========================================
- Coverage   95.84%   95.84%   -0.01%     
==========================================
  Files         160      160              
  Lines       16611    16589      -22     
  Branches      923      918       -5     
==========================================
- Hits        15921    15899      -22     
  Misses        690      690              

Flag	Coverage Δ
unittests	95.84% <100.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR optimizes LBVH construction by switching from the prior top-down split search / two-pass approach to a single-pass bottom-up build (Apetrei 2014), while also adding benchmarks to measure LBVH::build performance.

Changes:

• Replaced hierarchy construction with a bottom-up, single-pass algorithm that builds internal nodes and bounding boxes simultaneously.
• Updated ConstructionInfo to track subtree key-range endpoints (left/right) instead of parent indices.
• Added Catch2 benchmark coverage for LBVH::build across several meshes/scenes.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 6 comments.

File	Description
tests/src/tests/broad_phase/test_lbvh.cpp	Adds Catch2 benchmarks to measure LBVH::build performance on multiple scenes.
src/ipc/broad_phase/lbvh.hpp	Updates construction metadata to store subtree range endpoints used by bottom-up build.
src/ipc/broad_phase/lbvh.cpp	Implements the Apetrei 2014 bottom-up build, including root swapping to preserve traversal assumptions.

Comments suppressed due to low confidence (1)

src/ipc/broad_phase/lbvh.cpp:269

• std::atomic::operator++ defaults to memory_order_seq_cst, which is typically more expensive than needed for this arrival gate in a performance-critical build. Consider switching to fetch_add(1, std::memory_order_acq_rel) (or equivalent) so (1) the memory ordering is explicit for correctness of publishing the child pointer/range/AABB data, and (2) you avoid the stronger global ordering constraints of seq_cst.

                if (info.visitation_count++ == 0) {
                    // this is the first thread that arrived at this
                    // node -> finished
                    break;
                }

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.