@plasius/gpu-renderer 0.2.6 → 0.2.8

package/CHANGELOG.md

@@ -23,7 +23,7 @@ All notable changes to this project will be documented in this file.
 - **Security**
   - (placeholder)
 
-## [0.2.6] - 2026-06-14
+## [0.2.8] - 2026-06-20
 
 - **Added**
   - Added `updateCamera(...)` support for wavefront renderers so validation
@@ -60,6 +60,10 @@ All notable changes to this project will be documented in this file.
     highlight proxy.
   - Added ADR and design coverage for prefiltered HDRI, BRDF LUT, and MIS-based
     environment lighting in the wavefront display-quality path.
+  - Added ADR/design coverage and public contract support for authored volume
+    transport so mesh materials can derive Beer-Lambert media from glTF-style
+    attenuation inputs while preserving shell thickness in GPU material
+    packing.
   - Added adaptive `renderFrame(...)` sampling controls so callers can cap a
     frame with `frameTimeBudgetMs`, guarantee a `minimumSamplesPerPixel`, and
     inspect actual `renderedSamplesPerPixel` separately from the configured SPP
@@ -68,10 +72,31 @@ All notable changes to this project will be documented in this file.
     wavefront SPP adaptation to `@plasius/gpu-performance` without duplicating
     renderer-specific ladder construction in app or demo code.
 
+- **Fixed**
+  - Awaited wavefront frame waits now scale their submitted-work timeout by
+    actual triangle load as well as pass count, preventing mesh-heavy
+    validation frames from failing early while the GPU is still legitimately
+    finishing the submitted work.
+  - Awaited high-SPP wavefront frames now break first-frame GPU work into
+    progressive 1-SPP submission slices with bounded wait windows, which avoids
+    queueing an entire heavy validation frame ahead of a single completion wait
+    and materially reduces browser-side device-loss risk on mesh renders.
+  - Fixed the display-quality `cpu-upload` mesh path so uploaded triangle
+    records preserve raw material factors, atlas rects, and texture settings
+    for GPU hit-time sampling instead of baking CPU-side averages that could
+    corrupt leather, wood, and chrome shading with the wrong atlas region.
+  - Fixed awaited `>= 8 spp` wavefront scheduling to stay tile-major while the
+    accumulation buffer is tile-local, preventing repeated-tile/striped image
+    corruption and stalled validation captures on higher-SPP frames.
+
 - **Changed**
   - Changed internal wavefront frame batching and dispatch-diagnostics plumbing
     to live in a dedicated runtime helper module so scheduling concerns stay
     separated from shader and pipeline assembly.
+  - Changed display-quality mesh tracing to default to `accelerationBuildMode:
+    "cpu-upload"` so stable CPU-built BVH uploads are used for validation and
+    demo rendering unless callers explicitly opt into the experimental GPU-side
+    BVH construction path.
   - Changed wavefront frame dispatch to split large tile/sample workloads into
     bounded command submissions instead of encoding an entire high-resolution
     frame into one command buffer.
@@ -413,4 +438,4 @@ All notable changes to this project will be documented in this file.
 [0.1.12]: https://github.com/Plasius-LTD/gpu-renderer/releases/tag/v0.1.12
 [0.1.14]: https://github.com/Plasius-LTD/gpu-renderer/releases/tag/v0.1.14
 [0.2.1]: https://github.com/Plasius-LTD/gpu-renderer/releases/tag/v0.2.1
-[0.2.6]: https://github.com/Plasius-LTD/gpu-renderer/releases/tag/v0.2.6
+[0.2.8]: https://github.com/Plasius-LTD/gpu-renderer/releases/tag/v0.2.8

package/README.md

@@ -180,15 +180,20 @@ project-wide display-quality baseline for path-traced rendering, not a
 Product-Studio-only requirement.
 
 Mesh inputs are normalized into triangle records, packed into GPU buffers, and
-uploaded as source buffers for GPU triangle assembly and GPU BVH construction.
-Vertex normals are preserved for smooth shading; when normals are absent the
-triangle geometric normal is used. The display-quality path uses
-`accelerationBuildMode: "gpu"` and rejects CPU-built acceleration. The
-`createWavefrontMeshAcceleration(...)` helper remains available only for debug
-fixtures and deterministic layout tests. GPU BVH construction now uses
-Morton-style centroid keys to sort leaf references before sorted leaves and
-level-concurrent internal nodes are materialized. The current mesh path is the
-GPU runtime baseline under active hardening.
+uploaded as source buffers for tracing. Vertex normals are preserved for smooth
+shading; when normals are absent the triangle geometric normal is used. The
+display-quality path now defaults to `accelerationBuildMode: "cpu-upload"` so
+CPU-built BVH nodes and triangle records are uploaded once and then reused by
+the GPU tracing passes. Set `accelerationBuildMode: "gpu"` only when you
+explicitly want the experimental GPU-side BVH assembly path for validation or
+development. The `createWavefrontMeshAcceleration(...)` helper is therefore no
+longer debug-only; it is the stable display-quality acceleration builder,
+whereas GPU BVH construction remains available behind the explicit mode switch.
+CPU-upload records still preserve the raw material factors plus atlas rects and
+texture settings expected by the GPU hit-time samplers; they are not meant to
+replace exact UV-driven sampling with CPU-baked triangle averages.
+The GPU BVH path still uses Morton-style centroid keys to sort leaf references
+before sorted leaves and level-concurrent internal nodes are materialized.
 When mesh inputs also carry UVs plus decoded base-colour,
 metallic-roughness, normal, occlusion, or emissive maps, the display-quality
 path now packs them into GPU texture atlases and samples them at the resolved
@@ -201,6 +206,15 @@ response now also samples reflection-aligned environment radiance so leather,
 chrome, and other polished authored materials can read from the active
 environment map or procedural sky instead of relying mostly on a sun-direction
 proxy.
+Authored participating-media inputs can also enter through the same mesh path.
+A mesh may provide an explicit `medium` block or a glTF-style `material.volume`
+block with `thickness`, `attenuationColor`, and `attenuationDistance`; the
+renderer derives the medium-table entry, carries the active medium id on the
+GPU ray, and applies Beer-Lambert transmittance to travelled segments on the
+GPU. Thickness is now preserved in material packing for later shell-volume
+work, but the current renderer still tracks only one active medium id per ray
+and does not yet implement nested media, spectral dispersion, or a branching
+reflection/refraction tree.
 
 `samplesPerPixel` controls how many GPU primary-ray samples are accumulated per
 screen pixel within a single render. This multiplies dispatch work but does not
@@ -272,7 +286,9 @@ with active continuation rays instead of the maximum tile capacity, while still
 avoiding CPU readback between bounces. WebGPU
 still preserves ordering between dependent bounce passes, but the renderer
 keeps CPU queue submissions bounded rather than forcing one submission per
-tile/sample.
+tile/sample. Awaited higher-SPP submission slicing remains tile-major because
+the accumulation buffer is tile-local; changing that order to sample-major
+would mix samples across tiles and corrupt the resolved image.
 Environment-light portals can additionally guide and gate sky/HDRI contribution
 through rectangular openings such as windows. `environmentPortalMode: "guide"`
 biases diffuse continuation rays toward configured openings, while