DiskLight / RectLight with spread < 1: intersect() reports a different pdf than sample()
Nadie ha tomado este issue todavía.
Evaluación
- Dificultad
- 3/5
- Tiempo estimado
- 1-2 días
- Aptitud para principiantes
- 68/100
- Tipo de issue
- Error
- Claridad
- Bien especificado
- Estado de actividad
- Activo
- Stack tecnológico
- cpp
- Área
- computer-graphics
Línea de trabajo
Read lib/rendering/pbr/light/DiskLight.cc and RectLight.cc, then locate the corresponding ISPC implementations and compare how sample() and intersect() calculate overlap bounds and PDFs. Reproduce the listed spread and reverse-sidedness cases, and check for relevant tests. Done means intersection PDFs match sampled PDFs for disk and rectangular lights across these cases, with the reported MIS discrepancy resolved.
Escrito por el modelo de indexación a partir del texto del issue.
Descripción
Two related problems in lib/rendering/pbr/light/DiskLight.cc and RectLight.cc (same in the ISPC versions), version 2026.29.1.
1. DiskLight: bounding box instead of the disk
Reproduce. Glossy floor (roughness 0.3) under a disk light, max_depth 1:
| spread | light samples only | BSDF samples only | both |
|---|---|---|---|
| 1.0 | 0.04509 | 0.04510 | 0.04509 |
| 0.95 | 0.04114 | 0.04116 | 0.04139 |
| 0.8 | 0.02979 | 0.02980 | 0.02997 |
Cause. DiskLight::sample() uses mInvArea (1 / (pi R^2)) when the whole disk lies inside the spread region (lightIsInsideSpread), and the overlap rectangle otherwise. DiskLight::intersect() always calls getOverlapBounds, which in the first case returns the 2R x 2R bounding box, so the hit pdf is 1 / (4 R^2). The MIS weights of light and BSDF samples then do not sum to one.
Fix (untested). Use the same case distinction in intersect() as in sample().
2. Reverse and two-sided lights: signed distance
Reproduce. The same floor under a light with sidedness reverse and spread 0.8:
| light | light samples only | BSDF samples only | both |
|---|---|---|---|
| DiskLight | 0.02980 | 0 | 0.02942 |
| RectLight | 0.03794 | 0 | 0.03725 |
Cause. intersect() computes float spreadLength = localP.z * mTanSpreadTheta; with the signed localP.z, which is negative on the reverse side. getOverlapBounds then finds no overlap and returns false without writing width and height; the return value is not checked and the pdf is computed from them. sample() uses the sign-corrected plane distance and is fine.
Fix (untested). Use abs(localP.z), and check the return value of getOverlapBounds.
- Lenguaje dominante
- CMake
- Estrellas
- 4.7k
- Forks
- 321
- Merge medio
- 13 min
- PR fusionados (30 d)
- 3
Preparar el entorno
- Sin Dockerfile ni archivo de Docker Compose
- Tiene una plantilla de pull request
- Leer la guía de contribución
Primeros pasos
- Lee el issue completo y luego la guía de contribución del proyecto.
- Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
- Haz un fork del repositorio y trabaja en una rama.
- Abre un pull request que haga referencia al número del issue.
Más de OpenMoonRay/openmoonray
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 76/100
OpenMoonRay/openmoonray#332 ·
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 64/100
OpenMoonRay/openmoonray#320 ·
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 64/100
OpenMoonRay/openmoonray#233 ·
-
Dificultad 3/5 1-2 días Aptitud para principiantes 57/100
OpenMoonRay/openmoonray#331 ·
-
Dificultad 4/5 3-5 días Aptitud para principiantes 52/100
OpenMoonRay/openmoonray#329 ·
Todos los issues de OpenMoonRay/openmoonray
Issues similares
-
needs-triage
Dificultad 2/5 1-3 horas Aptitud para principiantes 85/100
bherbruck/solvecraft#13447 ·
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 83/100
facebookresearch/pytorch3d#2067 ·
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 78/100
Los mantenedores suelen responder en 2 días
-
Dificultad 2/5 1-3 horas Aptitud para principiantes 78/100
BabylonJS/Babylon.js#19127 ·
Los mantenedores suelen responder en 1 día