Send Motion-user mailing list submissions to
motion-user@lists.sourceforge.net
To subscribe or unsubscribe via the World Wide Web, visit
https://lists.sourceforge.net/lists/listinfo/motion-user
or, via email, send a message with subject or body 'help' to
motion-user-requ...@lists.sourceforge.net
You can reach the person managing the list at
motion-user-ow...@lists.sourceforge.net
When replying, please edit your Subject line so it is more specific
than "Re: Contents of Motion-user digest..."
Today's Topics:
1. Re: Motion 4.2.2 vs 4.1.1 CPU usage (tosiara)
----------------------------------------------------------------------
Message: 1
Date: Fri, 19 Jul 2019 22:01:56 +0300
From: tosiara <tosi...@gmail.com>
To: Motion discussion list <motion-user@lists.sourceforge.net>
Subject: Re: [Motion-user] Motion 4.2.2 vs 4.1.1 CPU usage
Message-ID:
<CACHTdwQ2rAZyA_H2UDHy0K0gx8S-a7u6jKwee=b6ernbrhp...@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"
4.2 should not use less CPU in your case. It should be the same as 4.1
NEON optimizations have not been released. The patch must be merged manually
On Fri, Jul 19, 2019 at 9:12 PM rmbusy <rmbusy+mot...@gmail.com> wrote:
> Thanks for this info. It sounds like 4.2.2 should be less CPU usage than
> 4.1.1, so I'll have to do some experimenting here to see if I can figure
> out why it's gone up (the profiling link should help with that). I have
> all the notes from when I built 4.1.1. initially, and the source zip files,
> so I should be able to reproduce that if need be.
>
> I would not expect going from Ubuntu 16.04 to 18.04 to be the cause, but I
> won't rule it out either. I have a spare drive, so I could install 16.04
> on it, and in theory, completely recreate my original setup. If only I had
> the free time to do everything I want to do right now.
>
> I'll keep you posted.
>
>
> --
> Rob.
>
>
>
> On 7/19/19 10:35 AM, tosiara wrote:
>
> This is my ARM board CPU usage over the previous year:
>
> [image: image.png]
>
> It is using USB webcam. In 2018 I changed pixel format from MJPEG to RAW
> and it decreased the CPU usage from ~65% to ~35%. In the middle on 2018 I
> merged NEON optimizations which saved even more CPU - to ~23%. I constantly
> update my motion to master branch, but since 2018 CPU usage on this ARM
> device has not changed
>
> So I doubt your issue is in motion's code. It could be compile flags. If
> want to be sure, try profiling motion and compare:
> http://www.lavrsen.dk/foswiki/bin/view/Motion/MotionProfiling
>
>
>
>
> On Fri, Jul 19, 2019 at 7:16 PM rmbusy <rmbusy+mot...@gmail.com> wrote:
>
>> One other thing, in an attempt to reduce CPU usage, I've disabled
>> auto_tune, and added mask files to 5 of the cameras to mask out foliage
>> and areas that could cause excess triggering, and bumped the threshold
>> from 5,000 to 10,000 on all of the 2MP cameras. These changes were not
>> needed with 4.1.1.
>>
>> I've also noticed a lot more motion detection of changing shadows,
>> compared to what I used to see.
>>
>>
>> --
>> Rob.
>>
>>
>> On 7/19/19 9:08 AM, rmbusy wrote:
>> > OS is Linux (before, Ubuntu Xenial 16.04 LTS, now Ubuntu Bionic
>> > 18.04.2 LTS). In both cases, I built motion from source. I started
>> > by installing the package with apt-get install motion to get all the
>> > dependencies, then built from source and switched to it.
>> >
>> > Since there are really only 3 types / configurations of cameras, I
>> > only tested the 3 types. I watched the output of 'top' for about a
>> > minute for each test. I also waited for the initial startup to settle
>> > (time for camera to respond with images) before recording the
>> > numbers. I don't have 4.1.1 building yet, so these numbers are for
>> > 4.2.2.
>> >
>> > For the 2MP snapshot camera, after startup, before connecting to the
>> > web page, it hovered between 25% and 45% CPU usage.
>> > After starting the web page, the numbers hovered between 78.5% and
>> 81.5%.
>> >
>> > For the 1MP streaming camera (5 frames / sec), it hovered between 23%
>> > and 33%.
>> > After starting the web page, 33.7% to 45.9%.
>> >
>> > For the 2MP streaming camera (5 frames / sec), it hovered between 42%
>> > and 53.6%.
>> > After starting the web page, 63.6% and 70.3%.
>> >
>> > I'm a little surprised no one had usage numbers between the two
>> > versions. Considering I'm using the same configuration files (with
>> > updates for the new parameter names in 4.2), I was expecting others
>> > would see this problem. It's an issue for me, because I have more
>> > cameras I want to add, but at this point with 4.2.2, the CPU is maxed
>> > out.
>> >
>> >
>> > --
>> > Rob.
>> >
>> >
>> > On 7/18/19 11:12 PM, tosiara wrote:
>> >> Could you compare CPU numbers running only one camera? The best to
>> >> make 8 separate tests with each camera. This could help to understand
>> >> the issue better
>> >> Also, which OS are were you using before and now, and which motion
>> >> version - binary package or compiled yourself?
>> >>
>> >> On Thu, Jul 18, 2019 at 9:33 PM rmbusy <rmbusy+mot...@gmail.com>
>> wrote:
>> >>> I upgraded my system to the latest version of Linux, and at the same
>> >>> time, upgraded Motion from 4.1.1 to 4.2.2.
>> >>>
>> >>> I have 8 RTSP cameras set up, 5 cameras configured for snapshot mode
>> at
>> >>> 1 frame per second (all 2MP), and 3 streaming at 5 frames per second
>> (2
>> >>> 2MP, 1 1MP). All of this is running on a quad core Rock64 w/ 4GB
>> DRAM,
>> >>> at the default CPU frequency (1.3GHz).
>> >>>
>> >>> With 4.1.1, I was seeing around 150% CPU utilization (via top), and
>> >>> when
>> >>> connecting to the web page, around 300% CPU utilization. Since the
>> >>> Rock64 is a quad core ARM processor, max is 400%.
>> >>>
>> >>> Now with 4.2.2 and the same camera configuration files (updated with
>> >>> new
>> >>> option names), I'm seeing CPU utilization of 250%, and when connecting
>> >>> to the web page, the numbers are maxed out, and the web page is
>> >>> basically non-responsive.
>> >>>
>> >>> Since I upgraded the underlying OS, I can't rule out that it's
>> >>> requiring
>> >>> more CPU cycles for Motion to do the same thing, but that seems
>> >>> unlikely. Running top doesn't show anything else using CPU cycles
>> >>> (other than top).
>> >>>
>> >>> Can anyone confirm if 4.2.2 is using significantly more CPU cycles
>> than
>> >>> 4.1.1? If I can't find a way to reduce the usage, I may have to go
>> >>> back
>> >>> to 4.1.1.
>> >>>
>> >>> Any suggestions would be appreciated.
>> >>>
>> >>>
>> >>> --
>> >>> Rob.
>> >>>
>> >>>
>> >>>
>> >>>
>> >>> _______________________________________________
>> >>> Motion-user mailing list
>> >>> Motion-user@lists.sourceforge.net
>> >>> https://lists.sourceforge.net/lists/listinfo/motion-user
>> >>> https://motion-project.github.io/
>> >>>
>> >>> Unsubscribe: https://lists.sourceforge.net/lists/options/motion-user
>> >>
>> >> _______________________________________________
>> >> Motion-user mailing list
>> >> Motion-user@lists.sourceforge.net
>> >> https://lists.sourceforge.net/lists/listinfo/motion-user
>> >> https://motion-project.github.io/
>> >>
>> >> Unsubscribe: https://lists.sourceforge.net/lists/options/motion-user
>> >
>>
>>
>>
>> _______________________________________________
>> Motion-user mailing list
>> Motion-user@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/motion-user
>> https://motion-project.github.io/
>>
>> Unsubscribe: https://lists.sourceforge.net/lists/options/motion-user
>
>
>
> _______________________________________________
> Motion-user mailing
> listMotion-user@lists.sourceforge.nethttps://lists.sourceforge.net/lists/listinfo/motion-userhttps://motion-project.github.io/
>
> Unsubscribe: https://lists.sourceforge.net/lists/options/motion-user
>
>
> _______________________________________________
> Motion-user mailing list
> Motion-user@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/motion-user
> https://motion-project.github.io/
>
> Unsubscribe: https://lists.sourceforge.net/lists/options/motion-user
-------------- next part --------------
An HTML attachment was scrubbed...
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 17360 bytes
Desc: not available
------------------------------
------------------------------
Subject: Digest Footer
_______________________________________________
Motion-user mailing list
Motion-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/motion-user
------------------------------
End of Motion-user Digest, Vol 157, Issue 26
********************************************
