Tom (Dec 23 2008)


log in

Advanced search

Message boards : Technical News : Tom (Dec 23 2008)

Previous · 1 · 2 · 3 · Next
Author Message
OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13542
Credit: 29,433,366
RAC: 16,011
United States
Message 845857 - Posted: 28 Dec 2008, 4:06:25 UTC - in response to Message 845854.

thanks just like to help as much as possible, omly dual core PC so two tasks it is

Cheers


No problem. Running more than one task per CPU would actually be counter-productive with all the internal CPU task switching from app to app, including cache flushes and reloads. What would be great is to do a single WU with multiple CPUs or cores, but an efficient way hasn't been found to do that yet as you run into other issues such as keeping each CPU/core updated as to what the other is doing, and not having each L2 cache per CPU/core duplicate the same data and trying to manipulate it at the same time.
____________

Profile littlegreenmanfrommars
Volunteer tester
Avatar
Send message
Joined: 28 Jan 06
Posts: 1410
Credit: 934,158
RAC: 0
Australia
Message 845959 - Posted: 28 Dec 2008, 11:57:39 UTC

What I'd like to have explained is how I managed to have THREE WU's running, when I have a dual core CPU.

One of the three has recently completed, and there are now only two WU's running, as I was led to expect.

Anyone out there with an idea how this happened?
____________

OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13542
Credit: 29,433,366
RAC: 16,011
United States
Message 845989 - Posted: 28 Dec 2008, 14:26:44 UTC - in response to Message 845959.
Last modified: 28 Dec 2008, 14:28:15 UTC

I heard about a similar issues a few years ago reported here on the forums. As far as I know, it seems to be an anomaly that is not reproducible or explainable. Perhaps some strange branch in the BOINC code that is only encountered in rare situations, or some code messed up in memory due to a RAM glitch (which is actually caused by solar flares and why ECC memory was designed to prevent, as these errors happen roughly once a month per GB of RAM) causing a third app to be executed.

Plenty of possibilities, but not many actual facts to work with.
____________

Josef W. Segur
Volunteer developer
Volunteer tester
Send message
Joined: 30 Oct 99
Posts: 4207
Credit: 1,031,237
RAC: 263
United States
Message 846037 - Posted: 28 Dec 2008, 16:45:43 UTC - in response to Message 845959.

What I'd like to have explained is how I managed to have THREE WU's running, when I have a dual core CPU.

One of the three has recently completed, and there are now only two WU's running, as I was led to expect.

Anyone out there with an idea how this happened?

Three tasks is exactly right, one on each of the CPUs, one on the GPU. The GPU task uses a very small amount of CPU time, but so does the OS in managing display, file system, etc., etc.; that's the nature of multitasking operating systems. If you check Task Manager you'll probably find there are 40 or so processes getting some parts of the CPU time.

The fact that BOINC fails to have three tasks running if there's GPU work available should be what you're wondering about.
Joe

Profile littlegreenmanfrommars
Volunteer tester
Avatar
Send message
Joined: 28 Jan 06
Posts: 1410
Credit: 934,158
RAC: 0
Australia
Message 846326 - Posted: 29 Dec 2008, 11:15:18 UTC - in response to Message 846037.

What I'd like to have explained is how I managed to have THREE WU's running, when I have a dual core CPU.

One of the three has recently completed, and there are now only two WU's running, as I was led to expect.

Anyone out there with an idea how this happened?

Three tasks is exactly right, one on each of the CPUs, one on the GPU. The GPU task uses a very small amount of CPU time, but so does the OS in managing display, file system, etc., etc.; that's the nature of multitasking operating systems. If you check Task Manager you'll probably find there are 40 or so processes getting some parts of the CPU time.

The fact that BOINC fails to have three tasks running if there's GPU work available should be what you're wondering about.
Joe


Hi Josef

I am running S@h and SETI Beta on the rig in question.
I had to re-attach it to both projects when I downloaded and installed BOINC 6.4.5

Initially, I attached to the main project, for a few WU's, to test things out.

One WU would run on one CPU core, the other on the GPU. "(0.05 CPU's 1 CUDA)" indicating the WU running on the GPU.

Having satisfied myself all was as it should be, I re-attached to the Beta project.

This produced a situation where the rig was crunching three WU's.

Once one of the three had finished, the situation changed, with one WU running on a CPU core, and 1 on the CUDA.

Reading other threads, I got the impression there should only be two WU's running on my rig. Please don't tell me it has installed wrongly! *sigh*


____________

Profile littlegreenmanfrommars
Volunteer tester
Avatar
Send message
Joined: 28 Jan 06
Posts: 1410
Credit: 934,158
RAC: 0
Australia
Message 846327 - Posted: 29 Dec 2008, 11:19:30 UTC - in response to Message 845989.

I heard about a similar issues a few years ago reported here on the forums. As far as I know, it seems to be an anomaly that is not reproducible or explainable. Perhaps some strange branch in the BOINC code that is only encountered in rare situations, or some code messed up in memory due to a RAM glitch (which is actually caused by solar flares and why ECC memory was designed to prevent, as these errors happen roughly once a month per GB of RAM) causing a third app to be executed.

Plenty of possibilities, but not many actual facts to work with.


Hi Ozz

I like your explanation! Thoughtful and written in non-technical English, it is succinct yet sufficiently ethereal to prevent the reader from understanding a word.

I shall, therefore translate.

There are two possible translations:

Option 1: "Blowed if I know"

Option 2: "It happens"

lol
____________

Profile Virtual Boss*
Volunteer tester
Avatar
Send message
Joined: 4 May 08
Posts: 417
Credit: 6,171,561
RAC: 787
Australia
Message 846348 - Posted: 29 Dec 2008, 13:19:23 UTC
Last modified: 29 Dec 2008, 13:20:36 UTC

There are two possible translations:

Option 1: "Blowed if I know"

Option 2: "It happens"

lol


Of course Option 2 is normally assoiated with option 3: Sh happens! ...lol...


Initially, I attached to the main project, for a few WU's, to test things out.

One WU would run on one CPU core, the other on the GPU. "(0.05 CPU's 1 CUDA)" indicating the WU running on the GPU.

Having satisfied myself all was as it should be, I re-attached to the Beta project.

This produced a situation where the rig was crunching three WU's.

Once one of the three had finished, the situation changed, with one WU running on a CPU core, and 1 on the CUDA.


I have not used cuda but after reading a lot of posts about it my interpretation is this:

I believe it will not run two seperate tasks (for same project) on 1 core,
as in Task A = CUDA (using GPU + part CPU X to control GPU) and Task B = MB (using CPU X), only one task (A or B) at the same time.

However on your duo it will run 2 tasks for same project, as you have noted, 1 on each core, or 1 on core 1 + 1 on GPU (using core 2 to control GPU).

A second project is still able to use core 2 to run a task, giving a total of three tasks running.

You will only be able to get the third task running if the first two are from the same project.(EDIT... And 1 is on the GPU)

I do not have a CUDA compatible card so I canot confirm this, so if you are able to confirm/deny this hypothesis please post the result.

Josef W. Segur
Volunteer developer
Volunteer tester
Send message
Joined: 30 Oct 99
Posts: 4207
Credit: 1,031,237
RAC: 263
United States
Message 846400 - Posted: 29 Dec 2008, 15:43:35 UTC - in response to Message 846348.

Initially, I attached to the main project, for a few WU's, to test things out.

One WU would run on one CPU core, the other on the GPU. "(0.05 CPU's 1 CUDA)" indicating the WU running on the GPU.

Having satisfied myself all was as it should be, I re-attached to the Beta project.

This produced a situation where the rig was crunching three WU's.

Once one of the three had finished, the situation changed, with one WU running on a CPU core, and 1 on the CUDA.


I have not used cuda but after reading a lot of posts about it my interpretation is this:

I believe it will not run two seperate tasks (for same project) on 1 core,
as in Task A = CUDA (using GPU + part CPU X to control GPU) and Task B = MB (using CPU X), only one task (A or B) at the same time.

However on your duo it will run 2 tasks for same project, as you have noted, 1 on each core, or 1 on core 1 + 1 on GPU (using core 2 to control GPU).

A second project is still able to use core 2 to run a task, giving a total of three tasks running.

You will only be able to get the third task running if the first two are from the same project.(EDIT... And 1 is on the GPU)

I do not have a CUDA compatible card so I canot confirm this, so if you are able to confirm/deny this hypothesis please post the result.

Substitute "application" for "project". Two AP and one MB CUDA should be supported, though BOINC 6.4.5 may leave a CPU essentially idle unless a cc_config.xml is used to encourage sensible usage.
Joe

Profile KWSN THE Holy Hand Grenade!
Volunteer tester
Avatar
Send message
Joined: 20 Dec 05
Posts: 1903
Credit: 9,243,631
RAC: 14,331
United States
Message 846410 - Posted: 29 Dec 2008, 16:14:01 UTC

It is possible to run 2 different WU's on one core: provided that they start on different cores... You go into Win Task Manager and "set affinity" for both WU's to the same core. I do this when backing up my computer, as (for reasons unknown to me) this speeds up the backup process by about 10-15x. On my quad-core, I run 4 WU's on 3 cores by excluding one core for the same reason.
____________
.

Profile littlegreenmanfrommars
Volunteer tester
Avatar
Send message
Joined: 28 Jan 06
Posts: 1410
Credit: 934,158
RAC: 0
Australia
Message 846762 - Posted: 30 Dec 2008, 5:12:05 UTC

OK, I am s-l-o-w-l-y getting a picture here...

Am I right in saying I need a third project (Say Einstein)so I can run that on one core, while the present two projects (applications) run on the second core and the GPU?

Thanks for your patience, guys. I can be a litle dense at times, but once I understand the concept, I shall never forget it!
____________

Fred W
Volunteer tester
Send message
Joined: 13 Jun 99
Posts: 2524
Credit: 11,954,210
RAC: 0
United Kingdom
Message 846811 - Posted: 30 Dec 2008, 8:55:57 UTC - in response to Message 846762.



Am I right in saying I need a third project (Say Einstein)so I can run that on one core, while the present two projects (applications) run on the second core and the GPU?


Errrr... No - you don't *NEED* a third project to use all cores. AstroPulse can keep all your cores busy while the GPU will do all the MB (i.e. Seti Enhanced) if you set up your machine as described in other threads in NC.

F.
____________

Profile littlegreenmanfrommars
Volunteer tester
Avatar
Send message
Joined: 28 Jan 06
Posts: 1410
Credit: 934,158
RAC: 0
Australia
Message 846822 - Posted: 30 Dec 2008, 10:16:19 UTC

Hi Fred

Thanks for your input.

HOWEVER:

WHICH THREADS???????????????????????

I am going nuts, trying to make sense of the (largely bollox) posts all over NC.

I have not been able to find one single post in which it is laid out how the hell one can configure BOINC to utilise all three processors (or 5, 9, etc).

Maybe I am stupid, blind, or worse.

If there is a post showing this, could you please be so kind as to give this slow-witted alien a link to it?

Cheers

lgm
____________

Fred W
Volunteer tester
Send message
Joined: 13 Jun 99
Posts: 2524
Credit: 11,954,210
RAC: 0
United Kingdom
Message 846826 - Posted: 30 Dec 2008, 10:51:48 UTC - in response to Message 846822.
Last modified: 30 Dec 2008, 10:53:33 UTC

WHICH THREADS???????????????????????


Try this for starters. It includes the following:

It is possible to run 5 wu at a time on a quaddie with a gpu (ie 4 cpu + 1 gpu), but you need to have a cc_config.xml with a <ncpus>5</ncpus> statement in it. By default it will run as 3+1. There is a "feeder" app that passes the data to the gpu and BOINC treats it as a task too, even though it only requires a small percentage of cpu time, which is why most of us use a cc_config.xml.

I believe there is more info on cc_config.xml in the FAQ's.

Hope this helps.

F.

[edit]Using the "Search forums" at the top of this thread for cc_config will turn up a lot more posts that may be useful to you. [/edit]
____________

Profile littlegreenmanfrommars
Volunteer tester
Avatar
Send message
Joined: 28 Jan 06
Posts: 1410
Credit: 934,158
RAC: 0
Australia
Message 846833 - Posted: 30 Dec 2008, 11:40:25 UTC

Hi again Fred

I have been trawling through the threads you indicated.

Much of it is above my head, as information I require seems to be assumed as obvious by the posters. This is very frustrating, as I'm sure you will understand. Having been away from the boards for several months, even the common abbreviation "MB" took a half hour of research to translate. (Multi-beam)

The default directories mentioned for locating the cc_config file do not exist on this PC. Is there something wrong with my BOINC installation? Is there something missing in the information I am reading?

What I am looking at now is:
2 x AP WU's. One is "Waiting to run"
The other is running.
A long list of CUDA WU's. One is running.
The rest are "Ready to start"

I understand why the CUDA apps are waiting to start, since there is only 1 GPU to process them.

What is stopping the second AP app from running?

I am continuing to search threads for a post written in English. *sigh*



____________

Claggy
Volunteer tester
Send message
Joined: 5 Jul 99
Posts: 4048
Credit: 32,693,513
RAC: 428
United Kingdom
Message 846835 - Posted: 30 Dec 2008, 11:56:39 UTC - in response to Message 846833.

Have you unhidden your hidden folders?

Claggy
____________


Profile littlegreenmanfrommars
Volunteer tester
Avatar
Send message
Joined: 28 Jan 06
Posts: 1410
Credit: 934,158
RAC: 0
Australia
Message 846838 - Posted: 30 Dec 2008, 12:04:02 UTC - in response to Message 846835.

Hi Claggy

Yes.

I have found the directory to place the cc_config.xml file. NOT where the posts I read said, but if you look in the initial messages in BOINC Advanced view, the correct directory is there.

Copying and pasting a sample cc_config.xml file did nothing, so I restarted BOINC, with the result my machine was detached from both projects.

*ggggggggggrrrrrrrr, snarrrrrrrrrrllllllllll*
____________

Claggy
Volunteer tester
Send message
Joined: 5 Jul 99
Posts: 4048
Credit: 32,693,513
RAC: 428
United Kingdom
Message 846857 - Posted: 30 Dec 2008, 13:29:20 UTC - in response to Message 846838.

I hope you didn't overwrite an existing file, not sure if you had to edit it and put that line in.
(I'm posting from a PSP, and haven't got a PC here with Boinc on to check).
Post in Number Crunching and someone will help.

Claggy

Profile Dr. C.E.T.I.
Avatar
Send message
Joined: 29 Feb 00
Posts: 15993
Credit: 690,597
RAC: 6
United States
Message 846864 - Posted: 30 Dec 2008, 13:49:01 UTC - in response to Message 846838.

Hi Claggy

Yes.

I have found the directory to place the cc_config.xml file. NOT where the posts I read said, but if you look in the initial messages in BOINC Advanced view, the correct directory is there.

Copying and pasting a sample cc_config.xml file did nothing, so I restarted BOINC, with the result my machine was detached from both projects.

*ggggggggggrrrrrrrr, snarrrrrrrrrrllllllllll*


. . . LGM - you may also see Eric K's Forum here: Questions and Answers : CUDA <--- click me

and, as Well - Paul D Buck's: Message boards : Number crunching : CUDA and Resource Share <--- click me

> PS - Have a Good New Year LGM and i Hope this helped Sir!




____________
BOINC Wiki . . .

Science Status Page . . .

Profile Byron S Goodgame
Volunteer tester
Avatar
Send message
Joined: 16 Jan 06
Posts: 1151
Credit: 3,936,993
RAC: 0
United States
Message 846867 - Posted: 30 Dec 2008, 13:59:03 UTC - in response to Message 846838.
Last modified: 30 Dec 2008, 14:17:33 UTC

Hi Claggy

Yes.

I have found the directory to place the cc_config.xml file. NOT where the posts I read said, but if you look in the initial messages in BOINC Advanced view, the correct directory is there.

Copying and pasting a sample cc_config.xml file did nothing, so I restarted BOINC, with the result my machine was detached from both projects.

*ggggggggggrrrrrrrr, snarrrrrrrrrrllllllllll*

Hi,

Here's an explanation of how to edit the cc_config.xml.

note* When saving with a text editor, make sure you save as "All types". That way the editor won't auto-add a TXT extension.

For your XP OS the default location to place it would be in C:\Documents and Setting\All Users\Application Data\BOINC and is probably where the Boinc Manager message log says it is.

The Application Data folder is a hidden folder that you would have to un-hide in folder options from the Windows Control Panel.

Hope this helps get all your cpu's crunching.
____________

Profile littlegreenmanfrommars
Volunteer tester
Avatar
Send message
Joined: 28 Jan 06
Posts: 1410
Credit: 934,158
RAC: 0
Australia
Message 846999 - Posted: 31 Dec 2008, 0:51:20 UTC

Hi Guys

I'm at work atm, so will try this after I get home.

I think I was a little tired yesterday, and SOME of the threads are am\\making more sense today.

After what happened with the cut and paste cc_config.xml last night, I must say I'm a little wary of doing it again, but "Nothing ventured, nothing gained".
____________

Previous · 1 · 2 · 3 · Next

Message boards : Technical News : Tom (Dec 23 2008)

Copyright © 2014 University of California