Hello,
Is there a way to monitor the general processing load on the gpu?
Cheers!
-
- Posts: 501
- Joined: Mon Apr 06, 2020 3:49 am
- DougieLawson
- Posts: 42636
- Joined: Sun Jun 16, 2013 11:19 pm
- Location: A small cave in deepest darkest Basingstoke, UK
Re: Monitoring load on GPU?
There isn't for the GPU. It's a closed black box with limited external controls.
Languages using left-hand whitespace for syntax are ridiculous
DMs sent on https://twitter.com/DougieLawson or LinkedIn will be answered next month.
Fake doctors - are all on my foes list.
The use of crystal balls and mind reading is prohibited.
DMs sent on https://twitter.com/DougieLawson or LinkedIn will be answered next month.
Fake doctors - are all on my foes list.
The use of crystal balls and mind reading is prohibited.
-
- Posts: 7755
- Joined: Sat Aug 18, 2012 2:33 pm
Re: Monitoring load on GPU?
from my notes filesviewtopic.php?f=29&t=195178&start=100
> What are the comparative loadings on the 2 VPU cores? "sudo vcdbg hist gnuplot" should give you two files called task.gpt and task.gpt, and "gnuplot task.gpt" should then display the three graphs of VPU0, VPU1, and combined.
`vcdbg hist gnuplot` shows VPU usage??
but that is only for the VPU usage,not the V3D usage (where shaders/opengl runs)
i am curious aout v3d usage, but i heard it needs linux tracing functions to measure
-
- Posts: 501
- Joined: Mon Apr 06, 2020 3:49 am
Re: Monitoring load on GPU?
Ok thanks for the clarification.DougieLawson wrote: ↑Thu Feb 18, 2021 7:38 amThere isn't for the GPU. It's a closed black box with limited external controls.
If that is because of lack of resources from the Raspberry organization I would personally love to see some diverted that direction for future hardware. Not that I am using the Pi professionally but the gpu is a strong part of the hardware and could be used better if it was easier to get some insight into how it runs. I'm currently attempting to optimize a program to run on the Pi4b by moving compute to gl shaders at the same time as realtime streaming in video and I'd love to know how much I can shift over to the gpu and how much load each 'part' adds.
Re: Monitoring load on GPU?
On Pi4 all hardware for Open GL and Vulcan is controlled ARM side in the open source Mesa drivers plus fairly small Kernel drivers.
The RTOS in the VPU doesn’t get involved at all. That’s just for ISP, H264, power and clock management these days if you are running the KMS display driver.
The RTOS in the VPU doesn’t get involved at all. That’s just for ISP, H264, power and clock management these days if you are running the KMS display driver.
- DougieLawson
- Posts: 42636
- Joined: Sun Jun 16, 2013 11:19 pm
- Location: A small cave in deepest darkest Basingstoke, UK
Re: Monitoring load on GPU?
It's because the RPF don't own the hardware. It's Broadcom's proprietary IP.RedMarsBlueMoon wrote: ↑Thu Feb 18, 2021 4:59 pmOk thanks for the clarification.
If that is because of lack of resources from the Raspberry organization I would personally love to see some diverted that direction for future hardware.
As the ARM processors get faster more of the stuff done by the GPU will be moved off the GPU.
Languages using left-hand whitespace for syntax are ridiculous
DMs sent on https://twitter.com/DougieLawson or LinkedIn will be answered next month.
Fake doctors - are all on my foes list.
The use of crystal balls and mind reading is prohibited.
DMs sent on https://twitter.com/DougieLawson or LinkedIn will be answered next month.
Fake doctors - are all on my foes list.
The use of crystal balls and mind reading is prohibited.
-
- Posts: 501
- Joined: Mon Apr 06, 2020 3:49 am
Re: Monitoring load on GPU?
Thanks for the comments.
To clarify, not that it matters to the discussion, I'm moving compute off the CPU onto the GPU (through glsl shaders), but also doing video streaming stuff feeding pixels/data in.
To clarify, not that it matters to the discussion, I'm moving compute off the CPU onto the GPU (through glsl shaders), but also doing video streaming stuff feeding pixels/data in.