fangq
commented
7 years ago your tab setting seems still different from mine (8 space per tab), but I accepted it anyways.
with the way you calculated photons_per_blk, will mcxcl still run exactly the specified total photon numbers if it is not multiple of thread numbers? (previously it was take cared by oddphotons)
Use local memory to dynamically distribute workloads to each thread in the work-group. The OpenCL lib directory for cuda is added to the Makefile.