Multi GPU pipeline Deesptream

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) :- GPU
• DeepStream Version :- 8.0/7.1

So, If my pipeline looks like following

nvurisrcbin_0 → 

nvurisrcbin_1 →

                streammux -> nvinfer1 -> nvtrakcer -> nvinfer2 -> sink 

nvurisrcbin_2 →

nvurisrcbin_3 →

And my machine has two GPUs A4000 and PRO 2000.
can I use ‘gpu-id’ property of nvinfer and set gpu-id of nvinfer1 as 0 and gpu-id of nvinfer2 as 1 ?

will these nvinfers start using these two GPUs ? or do I have to implement some mechanism that copies buffers between GPUs

2

contnuing the question,
can I run all nvurisrcbins on one GPU and all nvinfers on another ? is it just a config change ?

5

With your pipeline, you must make sure the “nvbuf-memory-type” property of nvstreammux is set to 3(Unified CUDA memory).

“cudadec-memtype” of nvurisrcbins should be set to 2(memtype_unified)

6

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks.