Delay nvvidconv element in TX2i 4 channels

[ealali]

Hi

We use nvvidconv element to convert from YUY2 to I420 format

Each channel resolution is 1920x1080 (Full HD)

It works fine when we use one or two channels BUT when we use 4 channels we get big latency

What can cause such problem?

What can cause this problem?

[DaneLLL]

Hi,
We enable dynamic frequency scaling to hardware converter(VIC) in latest releases. Please check the steps in

Nvvideoconvert issue, nvvideoconvert in DS4 is better than Ds5?

Hi, Please excute the steps to run VIC in max clock and try again: Disable runtime suspend of VIC $ echo on > /sys/devices/13e10000.host1x/15340000.vic/power/control Set userspace governor $ echo userspace > /sys/devices/13e10000.host1x/15340000.vic/devfreq/15340000.vic/governor Set max_freq $ cat /sys/devices/13e10000.host1x/15340000.vic/devfreq/15340000.vic/available_frequencies $ echo [max_freq_val] > /sys/devices/13e10000.host1x/15340000.vic/devfreq/15340000.vic/max_freq Set targ…

And run the engine in max clock and try again.

[ealali]

Hi

We are using L4T version 28.2.1

Can you tell me how to control VIC using this version of L4T?

Thanks

[ealali]

1. Do you measure benchmark with convert 4 channels from YUY2->I420 on TX2\TX2i?
2. What version of L4T do you use?

[DaneLLL]

Hi,
If you run like

v4l2src ! video/x-raw,format=YUY2,width=1920,height=1080 ! nvvidconv ! 'video/x-raw(memory:NVMM),format=I420' ! ...

In nvvidconv plugin, it copies YUY2 data from CPU buffers to NVMM buffers and convert YUY2 to I420 NVMM buffers. In multi channels, copying CPU buffers to NVMM buffers may be the bottleneck. You need to execute sudo nvpmodel -m 0 and sudo jetson_clocks, to get max CPU clocks.
VIC engine is at max clock on r28 release. DFS is applied on r32 releases.