Jetson Orin Nano Super Developer Kit 8GB
Hardware - Jetson JP 6.2
Deepstream Version - 7.1
Issue Type: Bug
Hi. It’s well known that on Jetson Orin Nano Super we don’t have accelerated encoders.
I have been experiencing issues with the software encoder and I found out that the problem spreads with most of my pipelines.
gst-launch-1.0 videotestsrc ! "video/x-raw, format=NV12, width=1920, height=1080" ! nvvideoconvert nvbuf-memory-type=4 ! "video/x-raw(memory:NVMM), format=RGBA, width=(int)1920, height=(int)1080" ! mux.sink_0 nvstreammux name=mux batch-size=1 width=1920 height=1080 sync-inputs=true live-source=1 batched-push-timeout=40000 ! nvinfer name=Detector0 batch-size=1 gpu-id=0 config-file-path=/app/assets/configs/primary_detector_config.txt ! nvtracker name=Tracker tracker-width=1920 tracker-height=1088 gpu-id=0 compute-hw=0 user-meta-pool-size=64 display-tracking-id=1 ll-lib-file="/opt/nvidia/deepstream/deepstream/lib/libnvds_nvmultiobjecttracker.so" ll-config-file="/app/assets/configs/config_tracker_NvDCF_perf.yml" ! nvvideoconvert ! x264enc ! mpegtsmux ! udpsink host=0.0.0.0 port=5005 sync=true async=true
gst-launch-1.0 videotestsrc ! "video/x-raw, format=NV12, width=1920, height=1080" ! nvvideoconvert nvbuf-memory-type=4 ! "video/x-raw(memory:NVMM), format=RGBA, width=(int)1920, height=(int)1080" ! mux.sink_0 nvstreammux name=mux batch-size=1 width=1920 height=1080 ! nvinfer name=Detector0 batch-size=1 gpu-id=0 config-file-path=/app/assets/configs/primary_detector_config.txt ! nvtracker name=Tracker tracker-width=1920 tracker-height=1088 gpu-id=0 compute-hw=0 ll-lib-file="/opt/nvidia/deepstream/deepstream/lib/libnvds_nvmultiobjecttracker.so" ll-config-file="/app/assets/configs/config_tracker_NvDCF_perf.yml" ! nvvideoconvert ! video/x-raw,format=I420 ! jpegenc quality=95 ! multifilesink name=JPEGSink location=\"/app/build/snapshotTesting/frame_%06d.jpg\" max-files=10
Those two with those two pipelines you can see the issue clearly.
After running for N seconds the pipeline stops with the following error
/dvs/git/dirty/git-master_linux/nvutils/nvbufsurftransform/nvbufsurftransform_copy.cpp:438: => Failed in mem copy
terminate called after throwing an instance of ‘thrust::THRUST_200500_520_NS::system::system_error’
what(): trivial_device_copy D->H failed: cudaErrorIllegalAddress: an illegal memory access was encountered
I have researched similar issues and I know that the current solution is provided in the following forum: https://forums.developer.nvidia.com/t/failed-in-mem-copy/320358/23
But the pipelines do not work even if you change the nvvideoconvert to be computing on the GPU with UNIFIED memory.
The models used are yolov11m with Deepsteam-Yolo.
Both pipelines run on dGPU.
My questions are:
What is the current status of this issue?
Can you provide me with more info why does it occur and how to workaround it.