Maximum shared memory can be used

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi All,

My kernel is dealing with a fairly big data structure, “Ring”, which has sizeof(Ring)==41932. It’s about 42KB, and barely fit into the shared memory for one block. Now, I am thinking to launch a kernel will 1000 blocks, will cuda-3.2 be capable of having a 42KB data structure in shared memory for each block? My GPU is “Tesla C2050”.

Thanks,

Kai