NvSciBufAttrListReconcileAndObjAlloc/NvSciCommonMemcpyS consume too much CPU

Please provide the following info (tick the boxes after creating this topic):
Software Version
DRIVE OS 6.0.9.0
DRIVE OS 6.0.8.1
DRIVE OS 6.0.6
DRIVE OS 6.0.5
DRIVE OS 6.0.4 (rev. 1)
DRIVE OS 6.0.4 SDK
other

Target Operating System
Linux
QNX
other

Hardware Platform
DRIVE AGX Orin Developer Kit (940-63710-0010-300)
DRIVE AGX Orin Developer Kit (940-63710-0010-200)
DRIVE AGX Orin Developer Kit (940-63710-0010-100)
DRIVE AGX Orin Developer Kit (940-63710-0010-D00)
DRIVE AGX Orin Developer Kit (940-63710-0010-C00)
DRIVE AGX Orin Developer Kit (not sure its number)
other

SDK Manager Version
2.1.0
other

Host Machine Version
native Ubuntu Linux 20.04 Host installed with SDK Manager
native Ubuntu Linux 20.04 Host installed with DRIVE OS Docker Containers
native Ubuntu Linux 18.04 Host installed with DRIVE OS Docker Containers
other

Issue Description
When encoding one stream with hardware acceleration, CPU usage reaches ~17%. Flame graph shows excessive CPU time spent in:
• NvSciBufAttrListReconcileAndObjAlloc()
• NvSciCommonMemcpyS()
This issue does NOT occur in SDK 6040. Perreference demos, we’ve created and reconciled the attribute list before encoder initialization.
Questions:

  1. Was the internal logic of the encoder modified in SDK 6090?
  2. Could the reconciled attribute list passed to the encoder fail to meet its requirements?
  3. How should we troubleshoot this?

flameGraph

Dear @yangsiliang ,
I notice you are using DRIVE OS 6.0.9 which is not a devzone release and you must have got it via channel. Could you please reach out to your NVIDIA representative to know right support channel.

This issue is likely a chronic problem across multiple SDK versions. Vendors probably won’t optimize the encoder unless they receive substantial funding and can surpass NVIDIA’s video encoding capabilities. For solutions, refer to the image_2d module ref Seeking Alternative to NvSciBufObjPutPixels for Lower Latency and CPU Usage - #6 by huangweijia.alex

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.