Multiple NVMM buffers to CUDA

[Nicolas Dufresne]

Le mardi 19 septembre 2017 à 04:05 -0700, Andreaaa a écrit :
> I have two images allocated to the *NVMM memory* using "nvvidconv !
> 'video/x-raw(memory:NVMM)' and I would like to retrieve and use them
> with
> CUDA. A possible solution would be to use *nvivafilter* but it takes
> only
> one input at a time. 
> 
> I created a custom plugin that retrieves the buffer from the pipeline
> and
> processes it inside the chain function, but if I print the size of
> the
> buffer I get a value of *808 bytes* that is much smaller than the
> image size
> it should contain.
> 
> To sum up, my questions are:
> - how can I correctly receive NVMM buffers directly allocated in CUDA
> memory? 
> - If so, how can I create a custom plugin that can manage multiple
> sinks?

All the NVMM memory code is proprietary to NVidia. As this code is not
Open Source, it will be impossible for us to provide you with support.
Please contact you vendor for more support.

Nicolas