TY - JOUR
KW - cache
KW - daylight simulation
KW - Morton code
KW - out-of-core
KW - paging
KW - photon map
AU - Roland Schregle
AU - Lars Grobe
AU - Stephen Wittkopf
AB - Climate-based daylight modelling (CBDM) is an effective means of assessing the performance of daylight redirecting components (DRCs) with highly directional scattering to determine their impact on daylight availability and visual comfort. Such a simulation imposes significant computational demands on commodity hardware as it requires high density luminance samples obtained by forward raytracing. We propose an out-of-core photon mapping method within the Radiance framework to compute high quality daylight coefficients as a basis for CBDM. The method is particularly suited to angularly selective DRCs exhibiting strong redirection in conjunction with non-uniform sky luminance distributions with high resolution subdivisions. Our implementation is a work in progress and currently accommodates up to 4.3G photons on disk, while optimizing the in-core memory footprint by loading only photons which actually contribute flux to sensor points. We also leverage the fact that photon paths are independent through parallelization.
BT - Journal of Building Performance Simulation
DA - 05/2016
DO - 10.1080/19401493.2016.1177116
IS - 6
LA - eng
N2 - Climate-based daylight modelling (CBDM) is an effective means of assessing the performance of daylight redirecting components (DRCs) with highly directional scattering to determine their impact on daylight availability and visual comfort. Such a simulation imposes significant computational demands on commodity hardware as it requires high density luminance samples obtained by forward raytracing. We propose an out-of-core photon mapping method within the Radiance framework to compute high quality daylight coefficients as a basis for CBDM. The method is particularly suited to angularly selective DRCs exhibiting strong redirection in conjunction with non-uniform sky luminance distributions with high resolution subdivisions. Our implementation is a work in progress and currently accommodates up to 4.3G photons on disk, while optimizing the in-core memory footprint by loading only photons which actually contribute flux to sensor points. We also leverage the fact that photon paths are independent through parallelization.
PY - 2016
SP - 620
EP - 632
T2 - Journal of Building Performance Simulation
TI - An out-of-core photon mapping approach to daylight coefficients
UR - http://dx.doi.org/10.1080/19401493.2016.1177116
VL - 9
ER -