Sign Up to our social questions and Answers Engine to ask questions, answer people’s questions, and connect with other people.
Login to our social questions & Answers Engine to ask questions answer people’s questions & connect with other people.
Lost your password? Please enter your email address. You will receive a link and will create a new password via email.
Please briefly explain why you feel this question should be reported.
Please briefly explain why you feel this answer should be reported.
Please briefly explain why you feel this user should be reported.
How are constant memory and texture memory implemented in GPU? I know that device memory is off the streaming multiprocessors and is a DRAM. Shared memory is on-chip, and probably an SRAM.
I also read somewhere that parameters are passed to the kernel via constant memory (i.e., parameters to the kernel function are copied from the CPU to the GPU constant memory). Is this correct?
Texture memory and constant memory are allocated in the off-chip memory used by global memory, but accessed via dedicated hardware. Both memories have their own cache space (quite small) and special features like 2D pre-fetching for 2D texture memory or broadcasting for the constant memory.
Kernel parameters are stored in constant memory in Fermi and Kepler GPUs, on older G80/90/200 GPUs they are stored in shared memory.
See the CUDA C Programming Guide for more details about the caches sizes and those special features.