摘要:
integer :: istat, nDevices,ierrSync,ierrAsync ,ierr istat = cudaGetDeviceCount(nDevices) if (istat /= cudaSuccess) write(*,*) cudaGetErrorString(ierr) 阅读全文
摘要:
MAX_REGISTERS_PRE_BLOCK(或者是MAX_REGISTERS_PRE_MULTIPROCESSOR?),假设为65535。那么,如果调用核函数时,blocksize为(x,y,z),每个线程调用的寄存器个数为r,则x*y*z*r应该小于65535。 From Stack Over 阅读全文