Just to let you know that if the number of current viewing blocks is less than a certain amount(256 in your configurations, which is the number of threads in a block), the cuda version of prefix sum won't work cause it only adds the 256th element of the buffer to the total number of rendering blocks. That does happen sometimes if too few blocks are within the current camera viewing frustum(or set voxel size too large). Maybe should consider comparing total viewable blocks with number thread blocks?
FYI, it is on the 61st line of file "ITMCUDAUtils.h"
Btw, big fan of yours, awesome job!
Just to let you know that if the number of current viewing blocks is less than a certain amount(256 in your configurations, which is the number of threads in a block), the cuda version of prefix sum won't work cause it only adds the 256th element of the buffer to the total number of rendering blocks. That does happen sometimes if too few blocks are within the current camera viewing frustum(or set voxel size too large). Maybe should consider comparing total viewable blocks with number thread blocks? FYI, it is on the 61st line of file "ITMCUDAUtils.h" Btw, big fan of yours, awesome job!