All pull requests

Pull requests

• Performance/: probes, a host/device harness and a per-PR comparison

  #207·rmrsk opened Oct 19, 2026Owner·9 tasks·

  ·
  2

  

• [DO NOT MERGE] DCEL: reconcile a mesh with GPU::parallelFor

  do not mergetriangles
  #203·rmrsk opened Oct 17, 2026Owner·8 tasks·

  ·
  7

• [DO NOT MERGE] Batched traversal: a Morton query order and pruneTraverseBatch

  bvhdo not merge
  #202·rmrsk opened Oct 17, 2026Owner·8 tasks·

  ·
  6

• [DO NOT MERGE] Build a PointCloudHashGrid with GPU::parallelFor

  do not mergepoint clouds
  #201·rmrsk opened Oct 17, 2026Owner·8 tasks·

  ·
  3

• [DO NOT MERGE] Tape: Add, Displace and a point cloud as a leaf

  csgdo not mergepoint clouds
  #200·rmrsk opened Oct 17, 2026Owner·8 tasks·

  ·
  6

• [DO NOT MERGE] PC-6: build a PointCloudBVH with GPU::parallelFor

  bvhdo not mergepoint clouds
  #199·rmrsk opened Oct 17, 2026Owner·8 tasks·

  ·
  5

• [DO NOT MERGE] PackedBVH: refit on a device, one tree level at a time (BVH-11, device part)

  bvhdo not merge
  #198·rmrsk opened Oct 17, 2026Owner·8 tasks·

  ·
  5

• Triangles: one signed-distance kernel for every SIMD width and the scalar paths (MESH-8)

  #197·rmrsk opened Oct 17, 2026Owner·8 tasks·

  ·
  4

• PackedBVH: named static builds, pack() by value (BVH-10)

  #196·rmrsk opened Oct 17, 2026Owner·8 tasks·

  ·
  4

• TreeBVH: fold interior boxes with merged(), no vector per union (BVH-11, host part)

  #195·rmrsk opened Oct 17, 2026Owner·8 tasks·

  ·
  5

• PackedBVH: store every box once (BVH-8)

  #194·rmrsk opened Oct 17, 2026Owner·8 tasks·

  ·
  4