I wish to redirect the AGX GPU output to an PCIe / Ethernet / MIPI interface (to an external device, for example FPGA). instead of display port.
How can I modify the GPU output destination?
For the PCIe case, do I need to use GPUDirect?
I asked a similar question here but i opened this thread to rephrase my question:
I wish to output the GPU generated frames to a remote FPGA memory via PCIe instead of display port output.
I saw there is a GPUdirect mechanism but couldn’t find any guides walkthorugh regarding this subject.
Where can I find some getting started help?
Is it possible to redirect only part of the gpu generated frames? I mean that some of them will output to the display port and others to the FPGA memory
Who controls the transfers? the CPU/GPU/FPGA? I mean who tells the GPU to redir…
The existing solution is to pass the data through PCIe. Please refer to
Video over PCIe - #8 by AastaLLL
thank you for you answer.
Is PCIe my only possible choice?
I wish to benchmark the BW/performance for my specific use case when redirecting through PCIe/ETH/MIPI
The last two not possible?
For using Ethernet, a general use-case is to encode frame data into h264/h265 stream and send out through UDP or RTSP. We don’t know whether there’s frameworks to send RAW frame data.
For sending MIPI signals, it is not supported on Jetson devices. There are CSI ports for receiving MIPI signals, but cannot run as a source to generate MIPI signals.
Thank you for your answer.
Just to clarify:
If I wish to send data via udp from the GPU memory, do I have to copy it into RAM first by the CPU and then use the network driver? or can I use the network driver to the the encoded frame directly from the GPU memory?
The possible solution is like:
GPU buffer → NvBufSurface → hardware encoder → h264stream → UDP or RTSP
We support gstreamer and jetson_multimedia_api. You can choose one for developing the use-case.
Thanks I’ll try and update
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.