CGFDemoapp.json Mutex explaination

Autonomous Vehicles DRIVE AGX Orin DRIVE AGX Orin General
1

Software Version
DRIVE OS 6.0.8.1

Target Operating System
Linux

Hardware Platform
DRIVE AGX Orin Developer Kit (940-63710-0010-300)

Host Machine Version

native Ubuntu Linux 20.04 Host installed with DRIVE OS Docker Containers

Issue Description
this is not an issue, but needs an explanation,

In the CGFDemo.app.json file, section “camerahyperepoch” in the “resources” subsection, how do you explain
below mutex locks ?

                    "camera_master.CUDA_STREAM0:dGPU": [],
                    "camera_master.CUDA_MUTEX_LOCK": [],
                    "camera_pipeline0.CUDA_STREAM0:dGPU": [],
                    "camera_pipeline0.CUDA_MUTEX_LOCK": [
                        "cgfDemo.arender.renderingNode",
                        "cgfDemo.cameraPipelineFront0.cameraSensor.cameraNode.PROCESSED_OUTPUT_CPU_SYNC",
                        "cgfDemo.cameraPipelineFront0.cameraSensor.cameraNode.RAW_OUTPUT_CPU_SYNC"

PROCESSED_OUTPUT_CPU_SYNC or RAW_OUTPUT_CPU_SYNC where are they defined, is there any node documentation for the same? how they are used? If possible point me to the right documentation.

Thank you,
Sushant

3

in System Task Manager SDK Reference: Compute Graph And Constraints (nvidia.com) you will see the topic Scheduling Mutex Resource Type

Any resource type not known by the compiler is modeled as a scheduling mutex. There are no naming conventions associated with either the resource type or the resource ID for a scheduling mutex. Interfering runnables can specify a scheduling mutex as a resource requirement to prevent the compiler from scheduling them concurrently.

4

Hi @lizhensheng ,
do you mean the names are just representative of a resource access only?

PROCESSED_OUTPUT_CPU_SYNC

RAW_OUTPUT_CPU_SYNC

just represents camera nodes GPU access?
I was curious if Mutexs which are implemented inside camera node for GPU access are specifically tagged to STM compile? to be liked with other nodes? Need to understand it more clearly.
Since there are no example GPU/VPI nodes provided with cgf examples. If you have any, please share with me, if you can that will be great.

Thanks,
Sushant

5

Sorry, all I can access is the public official docs, I cannot answer the question based on the docs.

The official forum service is @SivaRamaKrishnaNV
Looking for help

1 Like
6

Hi @VickNV @SivaRamaKrishnaNV
If you give help on this topic?

Dear @lizhensheng, Thank you for your suggestion, specially the Scheduling Mutex Resource Type helped.
I was trying to ask if you were able to build GPU/VPI CGF node, I had always been stuck with it.
As contradictory to claimed CGF documentation, is not so clear about implementing any GPU/VPI nodes.

7

Dear @sushant.bahadure ,
I will check with engineering team on your query and update you.
@lizhensheng Thank you for support to developer community.

1 Like