Nvcsi max_rate not supported quad imx390?

[gaoming.song]

orin nano 4gb R35.1
nvcsi max_rate is 214M,
imx390c sensor pix_clk_hz is 148 500 000 :(2l ane dphy)
pix_clk_hz * 12(pix_bit ) = 1782 000 000
pix_clk_hz * 12(pix_bit ) /2lane = 891 000 000
pix_clk_hz * 12(pix_bit ) /2lane /2(ddr) = 445 500 000
nvcsi_rate (byte) = pix_clk_hz * 12(pix_bit ) /2lane /2(ddr) /8 = 55 687 500

I don’t know if that’s the right calculation.
As above, a 4-way camera requires： 55 687 500 * 4 = 222 750 000
but nvcsi_max_rate is 214M，is nvcsi not supported 4 camera(imx390 raw@12)？
Tested to support 3 cameras 1920108030fps raw@12bit * 3camera

Later changed to 4lane (dphy), is it necessary to configure pix_clk_hz = 148500000/2

[gaoming.song]

@gaoming.song

nvcsi_max_rate is 214M，is nvcsi not supported 4 camera(imx390 raw@12)？
Tested to support 3 cameras 19201080*30fps raw@12bit * 3camera

Tested 4 cameras video will jam frames

[ShaneCCC]

Suppose Jetson Orin support 4 cameras. Any problem for 4 imx390?

[gaoming.song]

@gaoming.song

nvcsi max_rate is 214M,
imx390c sensor pix_clk_hz is 148 500 000 :(2l ane dphy)
pix_clk_hz * 12(pix_bit ) = 1782 000 000
pix_clk_hz * 12(pix_bit ) /2lane = 891 000 000
pix_clk_hz * 12(pix_bit ) /2lane /2(ddr) = 445 500 000
nvcsi_rate (byte) = pix_clk_hz * 12(pix_bit ) /2lane /2(ddr) /8 = 55 687 500

I don’t know if that’s the right calculation.

use:
nvgstcapture --sensord-id 0~3
hdmi monitor display lag , is orin support quad imx390 display.

[gaoming.song]

nvgstcapture video decoding shows lack of performance, 1 camera is normal, 2 drop frames, nvgstcapture can specify cpu core?

[gaoming.song]

Use:
gst-launch-1.0 nvarguscamerasrc sensor-id=0 ! 'video/x-raw(memory:NVMM),width=1920, height=1080, framerate=30/1, format=NV12' ! nvvidconv ! 'video/x-raw(memory:NVMM),format=I420' ! fpsdisplaysink text-overlay=0 video-sink=fakesink sync=0 -v
Or
nvgstcapture --sensor-id=[0-3]

[ShaneCCC]

Boost the clocks to try.

sudo jetson_clocks
sudo su
echo 1 > /sys/kernel/debug/bpmp/debug/clk/vi/mrq_rate_locked
echo 1 > /sys/kernel/debug/bpmp/debug/clk/isp/mrq_rate_locked
echo 1 > /sys/kernel/debug/bpmp/debug/clk/nvcsi/mrq_rate_locked
echo 1 > /sys/kernel/debug/bpmp/debug/clk/emc/mrq_rate_locked
cat /sys/kernel/debug/bpmp/debug/clk/vi/max_rate |tee /sys/kernel/debug/bpmp/debug/clk/vi/rate
cat /sys/kernel/debug/bpmp/debug/clk/isp/max_rate | tee  /sys/kernel/debug/bpmp/debug/clk/isp/rate
cat /sys/kernel/debug/bpmp/debug/clk/nvcsi/max_rate | tee /sys/kernel/debug/bpmp/debug/clk/nvcsi/rate
cat /sys/kernel/debug/bpmp/debug/clk/emc/max_rate | tee /sys/kernel/debug/bpmp/debug/clk/emc/rate

[gaoming.song]

use v4l2 dump data is 30fps*2.
It looks like it’s a single core that can play 30fps, playing multiple cameras running on a single core, top shows 105% cpu usage on a single camera

[gaoming.song]

Either one is 30fps when tested individually, but not at the same time，even with these added configurations.

[gaoming.song]

Currently basically sure is nvgstcapture single core running problem, now we use v4l2 +opencv display can 30fps *3cam(cpu 300%), just cv2.cvtColor(img, cv2.COLOR_BayerRG2RGB) after converting the image green chroma is very dark, not as good as nvgstcapture normal

[ShaneCCC]

v4l2 should must have much lower CPU/GPU usage due to didn’t involve ISP pipeline and 3A and lot of software/HW feature.

v4l2-ctl only capture the frame to memory a very simple pipeline.

[gaoming.song]

When trying to turn on the second camera of the test, both fps drop to very low, and can see that a certain cpu core idel that is not fixed is very low

[gaoming.song]

It seems that kernel dynamic_debug=y is causing the problem