Performance between elgImg map and cudaMalloc?

I want know there is different of the two methods. Because I need to save a picture, which is faster?
The source is nvsurfacefd and map the eglimg, the dst is cudaMalloc or nvsurfacefd is better?
At last I will encoder the dst picture to jpg.
Thanks!

Hi,
For using hardware JPEG encoder, we would suggest have frame data in NvBufSurface in YUV420/NV12. So that the loading is on VIC(hardware converter) and NVJPG engines.

I mean, if I always malloc memory using NvBufSurf::NvAllocate and elgimg to cuda, does it let the cuda’s memory performance weaker?

Hi,

No. We would expect identical performance for allocating buffer through cudaMalloc() or
NvBufSurfaceAllocate()

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.