Hello, nvidia:
I was trying to convert YUV to RGB using npp API.
This function named “cudaMemcpy” seems to call “memcpy” which caused CPU performance issue.
I’m wondering whether there is something I can do to avoid the calling of “memcpy” in the “cudaMemcpy”. thanks!
Hi,
Could you share which NPP function you use?
Or a simple reproducible source for us to check.
In general, if you allocate a buffer with unified or zero-copy memory.
You don’t manually copy the data between CPU and GPU.
Below is the different memory types that supported by Jetson for your reference:
Thanks.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.