Skip to content

QBnd: sample the boundary texture with POINT filtering#344

Draft
ggalgoczi wants to merge 1 commit into
mainfrom
qbnd-point-filter
Draft

QBnd: sample the boundary texture with POINT filtering#344
ggalgoczi wants to merge 1 commit into
mainfrom
qbnd-point-filter

Conversation

@ggalgoczi
Copy link
Copy Markdown
Contributor

cudaFilterModeLinear blends ~0.4% from the adjacent (categorically distinct) material/surface row on large (>5000-row) bnd tables, corrupting every per-photon material/surface lookup. Add a POINT-mode path selected by QBND_FILTER_POINT (run with QBND_FILTER_POINT=1).

@ggalgoczi ggalgoczi self-assigned this May 29, 2026
@ggalgoczi ggalgoczi added the bug Something isn't working label May 29, 2026
@ggalgoczi ggalgoczi marked this pull request as draft May 29, 2026 13:45
@ggalgoczi ggalgoczi force-pushed the qbnd-point-filter branch from 760b0cb to d4311d7 Compare May 29, 2026 14:03
The boundary QTex was sampled with cudaFilterModeLinear, blending ~0.4% from the adjacent (categorically distinct) material/surface row on large (>5000-row) bnd tables, corrupting every per-photon material/surface lookup. Switch to POINT filtering (QBND_FILTER_POINT; run with =1). POINT also makes the wavelength axis nearest-bin, so qbnd::boundary_lookup recovers G4-style per-wavelength linear interpolation via a manual 2-tap lerp (material axis stays POINT). The two go together -- nearest-bin wavelength sampling alone is ~2.3% low on the dRICH count.
@ggalgoczi ggalgoczi force-pushed the qbnd-point-filter branch from d4311d7 to 1961ad1 Compare May 29, 2026 14:15
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

bug Something isn't working

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant