Firefox and CPU cycles

classic Classic list List threaded Threaded
13 messages Options
Reply | Threaded
Open this post in threaded view
|

Firefox and CPU cycles

Bob-351
I have installed Firefox 10.0.7 with a new profile and no addons on an
ecs 2.1 SMP system. When I start Firefox to a blank page it loads, then
continues to use cycles (as shown in TOP) in bursts varying from 25% to
40%. What could be causing this activity.
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Dave Yeo-3
On 09/05/12 07:54 pm, Bob wrote:
> I have installed Firefox 10.0.7 with a new profile and no addons on an
> ecs 2.1 SMP system. When I start Firefox to a blank page it loads, then
> continues to use cycles (as shown in TOP) in bursts varying from 25% to
> 40%. What could be causing this activity.

What does the XCenter CPU monitor show? I started getting high CPU with
10.0.6 and delayed releasing while trying to figure it out, then
released a test version to see if anyone else had the problem. You seem
to be the first besides me. See thread 10.06 and high CPU on eCs started
Jul 31. First I thought SET NSPR_OS2_NO_HIRES_TIMER=1 fixed it but I was
mistaken. You might try setting it in your environment.
Strange thing is it comes and goes, is the worst with Thunderbird
10.0.6+ and doesn't show up at all on non-SMP Warp v4.
Dave

_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Paul Ratcliffe
In reply to this post by Bob-351
On Thu, 06 Sep 2012 14:54:33 +1200, Bob <[hidden email]> wrote:

> I have installed Firefox 10.0.7 with a new profile and no addons on an
> ecs 2.1 SMP system. When I start Firefox to a blank page it loads, then
> continues to use cycles (as shown in TOP) in bursts varying from 25% to
> 40%. What could be causing this activity.

It does it with any version. Been doing it for years. Who the f* knows
what this increasing pile of bloatware is constantly doing in the
background. Why not ask the real big-boy developers? I'm sure they'll
tell you.
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Bob-351
I've got this one running on a P4 with HT, so I disabled HT in the BIOS.
Firefox now bursts into life periodically but uses only about 5% of the
CPU. It could be that the OS/2 SMP doesn't play well with HT. I didn't
revert to the single thread kernel. Interestingly the Google home screen
uses cycles like mad. It's full of Javascript and CSS for really simple
screen.

Paul Ratcliffe wrote:

> On Thu, 06 Sep 2012 14:54:33 +1200, Bob <[hidden email]> wrote:
>
>> I have installed Firefox 10.0.7 with a new profile and no addons on an
>> ecs 2.1 SMP system. When I start Firefox to a blank page it loads, then
>> continues to use cycles (as shown in TOP) in bursts varying from 25% to
>> 40%. What could be causing this activity.
>
> It does it with any version. Been doing it for years. Who the f* knows
> what this increasing pile of bloatware is constantly doing in the
> background. Why not ask the real big-boy developers? I'm sure they'll
> tell you.
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Robert Blair-2
On Fri, 7 Sep 2012 05:27:27 UTC, Bob <[hidden email]> wrote:

> I've got this one running on a P4 with HT, so I disabled HT in the BIOS.
> Firefox now bursts into life periodically but uses only about 5% of the
> CPU. It could be that the OS/2 SMP doesn't play well with HT. I didn't
> revert to the single thread kernel. Interestingly the Google home screen
> uses cycles like mad. It's full of Javascript and CSS for really simple
> screen.

The ACPI documentation states that HT should be disabled because the OS/2
kernel does not have support for HT.  Some systems running OS/2 do not work
properly with HT turned on.  My machine appears to work properly with HT
enabled but I have it disabled in the BIOS since I updated ACPI.


--
Robert Blair
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Mark Dodel
On Fri, 7 Sep 2012 06:56:23 UTC, "Robert Blair" <[hidden email]>
wrote:

-> On Fri, 7 Sep 2012 05:27:27 UTC, Bob <[hidden email]> wrote:
->
-> > I've got this one running on a P4 with HT, so I disabled HT in the BIOS.
-> > Firefox now bursts into life periodically but uses only about 5% of the
-> > CPU. It could be that the OS/2 SMP doesn't play well with HT. I didn't
-> > revert to the single thread kernel. Interestingly the Google home screen
-> > uses cycles like mad. It's full of Javascript and CSS for really simple
-> > screen.
->
-> The ACPI documentation states that HT should be disabled because the OS/2
-> kernel does not have support for HT.  Some systems running OS/2 do not work
-> properly with HT turned on.  My machine appears to work properly with HT
-> enabled but I have it disabled in the BIOS since I updated ACPI.
->

Thanks for that tip Robert.  I turned off HT in my Lenovo x201 and
after a few hours at least FF 10.0.7 is still fairly useable.  Now I
will see how it is tomorrow morning after being on all night with FF
running.

Mark


--
From the eComStation  2.1 GA desktop of Mark Dodel

Warpstock 2013 - http://www.warpstock.org
Warpstock Europe 2013 - http://www.warpstock.eu 
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Dave Yeo-3
In reply to this post by Robert Blair-2
Robert Blair wrote:

> On Fri, 7 Sep 2012 05:27:27 UTC, Bob<[hidden email]>  wrote:
>
>> I've got this one running on a P4 with HT, so I disabled HT in the BIOS.
>> Firefox now bursts into life periodically but uses only about 5% of the
>> CPU. It could be that the OS/2 SMP doesn't play well with HT. I didn't
>> revert to the single thread kernel. Interestingly the Google home screen
>> uses cycles like mad. It's full of Javascript and CSS for really simple
>> screen.
>
> The ACPI documentation states that HT should be disabled because the OS/2
> kernel does not have support for HT.  Some systems running OS/2 do not work
> properly with HT turned on.  My machine appears to work properly with HT
> enabled but I have it disabled in the BIOS since I updated ACPI.
>
>

Intel actually recommends that HT be turned off on any operating system
that does not support HT. HT seems to work best with 2 threads from the
same program as it is basically 2 pipelines (with parts of the pipelines
shared) going to one execution unit. If one pipeline stalls then the
other can run. Works best with a scheduler that is smart enough to
schedule similar threads together. They're also using shared cache.
I gather where it really breaks down is with 2+ cores with HT. The
scheduler needs to know enough that if 2 threads are running, each
should go to a different core.\
Dave
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Dave Yeo-3
In reply to this post by Bob-351
Bob wrote:
> I've got this one running on a P4 with HT, so I disabled HT in the BIOS.
> Firefox now bursts into life periodically but uses only about 5% of the
> CPU. It could be that the OS/2 SMP doesn't play well with HT. I didn't
> revert to the single thread kernel. Interestingly the Google home screen
> uses cycles like mad. It's full of Javascript and CSS for really simple
> screen.

I have a core2duo so truly 2 CPU's. BIOS (old) doesn't seem to have a
setting to disable one core so I remmed out ACPI.PSD. Behaviour of
Mozilla didn't change except it brought the system to its knees with the
single CPU running at 99% often. Just closing SeaMonkey took minutes.
Yet here I am on Warp V4 with a SMP kernel but no PSD file to enable SMP
and typing this I'm using about 3% CPU consistently with the odd peak at
7%. Loading google.ca doesn't change it much either.
Like to know what it is about my eCS install that uses so much CPU. I'll
have to do another install and test.
Dave


>
> Paul Ratcliffe wrote:
>> On Thu, 06 Sep 2012 14:54:33 +1200, Bob <[hidden email]>
>> wrote:
>>
>>> I have installed Firefox 10.0.7 with a new profile and no addons on
>>> an ecs 2.1 SMP system. When I start Firefox to a blank page it loads,
>>> then continues to use cycles (as shown in TOP) in bursts varying from
>>> 25% to 40%. What could be causing this activity.
>>
>> It does it with any version. Been doing it for years. Who the f* knows
>> what this increasing pile of bloatware is constantly doing in the
>> background. Why not ask the real big-boy developers? I'm sure they'll
>> tell you.

_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Dave Saville-6
On Sun, 9 Sep 2012 05:46:12 UTC, Dave Yeo <[hidden email]>
wrote:

snip>

> I have a core2duo so truly 2 CPU's. BIOS (old) doesn't seem to have a
> setting to disable one core so I remmed out ACPI.PSD. Behaviour of
> Mozilla didn't change except it brought the system to its knees with the
> single CPU running at 99% often. Just closing SeaMonkey took minutes.
> Yet here I am on Warp V4 with a SMP kernel but no PSD file to enable SMP
> and typing this I'm using about 3% CPU consistently with the odd peak at
> 7%. Loading google.ca doesn't change it much either.
> Like to know what it is about my eCS install that uses so much CPU. I'll
> have to do another install and test.
>

My server is an ancient twin CPU box. It is running both CPU's on ECS
2.0 with

PSD=OS2APIC.PSD

and APM.

HTH
--
Regards
Dave Saville
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Barry Landy
On Sun, 9 Sep 2012, Dave Saville wrote:

:>On Sun, 9 Sep 2012 05:46:12 UTC, Dave Yeo <[hidden email]>
:>wrote:
:>
:>snip>
:>> I have a core2duo so truly 2 CPU's. BIOS (old) doesn't seem to have a
:>> setting to disable one core so I remmed out ACPI.PSD. Behaviour of
:>> Mozilla didn't change except it brought the system to its knees with the
:>> single CPU running at 99% often. Just closing SeaMonkey took minutes.
:>> Yet here I am on Warp V4 with a SMP kernel but no PSD file to enable SMP
:>> and typing this I'm using about 3% CPU consistently with the odd peak at
:>> 7%. Loading google.ca doesn't change it much either.
:>> Like to know what it is about my eCS install that uses so much CPU. I'll
:>> have to do another install and test.
:>>
:>
:>My server is an ancient twin CPU box. It is running both CPU's on ECS
:>2.0 with
:>
:>PSD=OS2APIC.PSD
:>
:>and APM.
:>
:>HTH
:>

I have just realised that my T61 shows the 100% COU cycles as well; both
with two cores and maxcpu=1. version 10.0.5

Modern ACPI (now 3.21.02)

--
Barry Landy                        Email: Remove nospam in from address
192, Gilbert Road, Cambridge CB4 3PB









_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Dariusz Piatkowski
On Mon, 10 Sep 2012 22:00:05 UTC, Barry Landy <[hidden email]> wrote:

> On Sun, 9 Sep 2012, Dave Saville wrote:
>
> :>On Sun, 9 Sep 2012 05:46:12 UTC, Dave Yeo <[hidden email]>
> :>wrote:
> :>
> :>snip>
> :>> I have a core2duo so truly 2 CPU's. BIOS (old) doesn't seem to have a
> :>> setting to disable one core so I remmed out ACPI.PSD. Behaviour of
> :>> Mozilla didn't change except it brought the system to its knees with the
> :>> single CPU running at 99% often. Just closing SeaMonkey took minutes.
> :>> Yet here I am on Warp V4 with a SMP kernel but no PSD file to enable SMP
> :>> and typing this I'm using about 3% CPU consistently with the odd peak at
> :>> 7%. Loading google.ca doesn't change it much either.
> :>> Like to know what it is about my eCS install that uses so much CPU. I'll
> :>> have to do another install and test.
> :>>
> :>
> :>My server is an ancient twin CPU box. It is running both CPU's on ECS
> :>2.0 with
> :>
> :>PSD=OS2APIC.PSD
> :>
> :>and APM.
> :>
> :>HTH
> :>
>
> I have just realised that my T61 shows the 100% COU cycles as well; both
> with two cores and maxcpu=1. version 10.0.5
>
> Modern ACPI (now 3.21.02)

I have been able to avoid this with the previously mentioned (and posted about)
high-res timer OFF option for the Firefox app. This is on an AMD Phenom2 X4
machine.

However...given the amount of problems this causes, is there an OS2 tool which
allows one to actually detect what thread/process is causing the CPU to spike
and better yet, track this into a log of some sorts?

I am not looking for your normal TOP style process monitor...instead it's
something more along the lines of a CPU/process monitor which will identify the
processor affinity (is that possible in OS2??? I think/hope so...).
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Barry Landy
On Mon, 10 Sep 2012, Dariusz Piatkowski wrote:

:>On Mon, 10 Sep 2012 22:00:05 UTC, Barry Landy <[hidden email]> wrote:
:>
:>> On Sun, 9 Sep 2012, Dave Saville wrote:
:>>
:>> :>On Sun, 9 Sep 2012 05:46:12 UTC, Dave Yeo <[hidden email]>
:>> :>wrote:
:>> :>
:>> :>snip>
:>> :>> I have a core2duo so truly 2 CPU's. BIOS (old) doesn't seem to have a
:>> :>> setting to disable one core so I remmed out ACPI.PSD. Behaviour of
:>> :>> Mozilla didn't change except it brought the system to its knees with the
:>> :>> single CPU running at 99% often. Just closing SeaMonkey took minutes.
:>> :>> Yet here I am on Warp V4 with a SMP kernel but no PSD file to enable SMP
:>> :>> and typing this I'm using about 3% CPU consistently with the odd peak at
:>> :>> 7%. Loading google.ca doesn't change it much either.
:>> :>> Like to know what it is about my eCS install that uses so much CPU. I'll
:>> :>> have to do another install and test.
:>> :>>
:>> :>
:>> :>My server is an ancient twin CPU box. It is running both CPU's on ECS
:>> :>2.0 with
:>> :>
:>> :>PSD=OS2APIC.PSD
:>> :>
:>> :>and APM.
:>> :>
:>> :>HTH
:>> :>
:>>
:>> I have just realised that my T61 shows the 100% COU cycles as well; both
:>> with two cores and maxcpu=1. version 10.0.5
:>>
:>> Modern ACPI (now 3.21.02)
:>
:>I have been able to avoid this with the previously mentioned (and posted about)
:>high-res timer OFF option for the Firefox app. This is on an AMD Phenom2 X4
:>machine.
:>
:>However...given the amount of problems this causes, is there an OS2 tool which
:>allows one to actually detect what thread/process is causing the CPU to spike
:>and better yet, track this into a log of some sorts?
:>
:>I am not looking for your normal TOP style process monitor...instead it's
:>something more along the lines of a CPU/process monitor which will identify the
:>processor affinity (is that possible in OS2??? I think/hope so...).
:>

I already have the NO_HIRES_TIMER=1

--
Barry Landy                        Email: Remove nospam in from address
192, Gilbert Road, Cambridge CB4 3PB









_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox and CPU cycles

Bob-351
In reply to this post by Dariusz Piatkowski
There are a bunch of process managers at
http://hobbes.nmsu.edu/h-browse.php?dir=/pub/os2/util/process. I do
recall having a few problems with some of them on newer kernels. I don't
recall one that logs, even amongst the older commercial products.

Dariusz Piatkowski wrote:

> On Mon, 10 Sep 2012 22:00:05 UTC, Barry Landy <[hidden email]> wrote:
>
>> On Sun, 9 Sep 2012, Dave Saville wrote:
>>
>> :>On Sun, 9 Sep 2012 05:46:12 UTC, Dave Yeo <[hidden email]>
>> :>wrote:
>> :>
>> :>snip>
>> :>> I have a core2duo so truly 2 CPU's. BIOS (old) doesn't seem to have a
>> :>> setting to disable one core so I remmed out ACPI.PSD. Behaviour of
>> :>> Mozilla didn't change except it brought the system to its knees with the
>> :>> single CPU running at 99% often. Just closing SeaMonkey took minutes.
>> :>> Yet here I am on Warp V4 with a SMP kernel but no PSD file to enable SMP
>> :>> and typing this I'm using about 3% CPU consistently with the odd peak at
>> :>> 7%. Loading google.ca doesn't change it much either.
>> :>> Like to know what it is about my eCS install that uses so much CPU. I'll
>> :>> have to do another install and test.
>> :>>
>> :>
>> :>My server is an ancient twin CPU box. It is running both CPU's on ECS
>> :>2.0 with
>> :>
>> :>PSD=OS2APIC.PSD
>> :>
>> :>and APM.
>> :>
>> :>HTH
>> :>
>>
>> I have just realised that my T61 shows the 100% COU cycles as well; both
>> with two cores and maxcpu=1. version 10.0.5
>>
>> Modern ACPI (now 3.21.02)
>
> I have been able to avoid this with the previously mentioned (and posted about)
> high-res timer OFF option for the Firefox app. This is on an AMD Phenom2 X4
> machine.
>
> However...given the amount of problems this causes, is there an OS2 tool which
> allows one to actually detect what thread/process is causing the CPU to spike
> and better yet, track this into a log of some sorts?
>
> I am not looking for your normal TOP style process monitor...instead it's
> something more along the lines of a CPU/process monitor which will identify the
> processor affinity (is that possible in OS2??? I think/hope so...).
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2