Skip to content

v0.9.2

Choose a tag to compare

View all tags
@manuelkNVDA manuelkNVDA released this 11 Apr 22:50
· 34 commits to main since this release
v0.9.2
aa17d1d

Performance Improvements

Test Scene: amy_kitchenset.scene.json (default camera: 79M microtriangles)
Hardware: RTX 5090 @ 4K Render Resolution (r572.83)

  • Compute Cluster tiling: 4.0ms -> 1.0ms (300% speedup)
    • Coalesced UAV writes for structs, unaligned members were causing UAV readbacks
    • Coalesce per wave atomics into a groupshared atomic to reduce pressure on global/UAV atomics by 4x
  • Fill clusters: 4.9ms to -> 1.0ms (390% speedup)
    • Fixed cases where the compiler was unable to unroll loops due to dynamic loop counts
    • Specialized shaders by subdivision surface type, with a special path for Pure BSpline surfaces. Prefetch all control points into shared memory to be used wave wide.

Minor Fixes

  • Fix SpecularHitT guide buffer to DLSS-RR to improve coherence of specular rays.

Full Changelog: v0.9.1...v0.9.2

Assets 3
Loading