0

I am learning to work with GPUs using C++ CUDA, and I could not find a clear answer to a question. Could you please explain briefly the difference between:

  1. Constant Memory
  2. Texture Memory
  3. Pinned Memory

I need a small amount of memory that is accessible by all kernels. Is the best approach is using constant memory?

halfer
  • 19,824
  • 17
  • 99
  • 186
Mojtaba Valizadeh
  • 746
  • 4
  • 12
  • 2
    This covers 1) and 2): https://stackoverflow.com/questions/8306967/constant-memory-vs-texture-memory-vs-global-memory-in-cuda?rq=1 – R2RT Mar 04 '22 at 09:36
  • 1
    This somewhat covers 3): [Pinned memory in CUDA](https://stackoverflow.com/q/12439807/10107454). If the data is not modified by the kernels, constant memory seems like a good idea. – paleonix Mar 04 '22 at 12:02

0 Answers0