Configure local (shared) memory for OpenCL using Nvidia platforms
I want to optimize my local memory access pattern in my OpenCL core. I read somewhere about configurable local memory. For example. we should be able to customize what amount is used for local mem and how much is used for automatic caching.
Also I read that the pot size can be selected for the latest (Kepler) Nvidia hardware here: http://www.acceleware.com/blog/maximizing-shared-memory-bandwidth-nvidia-kepler-gpus . This point seems to be very important for storing double precision values in local memory.
Does Nvidia provide local memory customization functionality exclusively for CUDA users? I cannot find similar methods for OpenCL. So maybe it could be called differently or it really doesn't exist?
+3
source to share