Welcome to MilkyWay@home

Posts by Conan

1) Message boards : Number crunching : Why is this project using all 8 cores when another project is trying to run (Message 76957)
Posted 23 days ago by Profile Conan
Post:
This project is using all eight available cores while another project is trying to run. Why is that?


G'Day Richard,

The default for N-Body work units is to use up to 16 cores in multi-thread.

You can change this to what you want (2,3,4,5,6,7 whatever) using an app_config.xml file (placed in your MilkyWay project directory).
Speed of processing may change you would have to try it out to see what works best.

I currently use 8 cores so I can fit 3 work units at once on my 12 core 24 thread Ryzen CPU.

See also This thread

Conan
2) Message boards : Number crunching : Thread to report issues after server migration (Message 76688)
Posted 7 Dec 2023 by Profile Conan
Post:
I rejoined after long absence and uploaded the new link. Now the work unit keeps getting stuck and the only solution is to reboot which is not always convenient, currently stuck for 3rd time after c.30% progress. WCG has the same issue after changing servers, but usually one reboot is enough


Are you using an app_config.xml file to limit the number of cpu cores the app can use? If not the one Link has is very helpful, the problem seems to be that if MilkyWay is using all of your cpu cores to crunch with then when the pc wants to do something else, ie check you mail, open a webpage etc, it stops crunching and then doesn't resume crunching again.


Yes I limit them to 8 threads, otherwise it will use 16 by default. I can run 3 at a time. Have not had any issue with the stopping of the work unit and then not restarting.

I will do the detach and re-attach maybe tomorrow or when I get round to it, same with SRBase and Einstein, who also want me to use a different url, works with the old one.

Conan
3) Message boards : Number crunching : Thread to report issues after server migration (Message 76684)
Posted 7 Dec 2023 by Profile Conan
Post:
If I detach and re-attach to this project to correct the "milkyway-new" issue now that all my work units have been returned, do I use "http" as stated in the join link or do I use "https" as shown on the web search panel.
I am guessing "https".

Conan
4) Message boards : Number crunching : Thread to report issues after server migration (Message 76674)
Posted 1 Dec 2023 by Profile Conan
Post:
And these are some of the errors I get when trying to check my tasks (either from main account page or from directly from the Host page)
This is just some as it goes for a few pages


Warning: Attempt to read property "appid" on null in /home/boinc/boinc/milkyway/html/inc/result.inc on line 83

Warning: Attempt to read property "platformid" on null in /home/boinc/boinc/milkyway/html/inc/result.inc on line 84

Warning: Attempt to read property "user_friendly_name" on null in /home/boinc/boinc/milkyway/html/inc/result.inc on line 85

Warning: Attempt to read property "version_num" on null in /home/boinc/boinc/milkyway/html/inc/result.inc on line 86

Warning: Attempt to read property "version_num" on null in /home/boinc/boinc/milkyway/html/inc/result.inc on line 86

Warning: Attempt to read property "plan_class" on null in /home/boinc/boinc/milkyway/html/inc/result.inc on line 87

Warning: Attempt to read property "name" on null in /home/boinc/boinc/milkyway/html/inc/result.inc on line 92

Warning: Attempt to read property "appid" on null in /home/boinc/boinc/milkyway/html/inc/result.inc on line 83

If you keep scrolling past all the warnings the tasks do show

Conan
5) Message boards : Number crunching : Thread to report issues after server migration (Message 76646)
Posted 29 Nov 2023 by Profile Conan
Post:
Conan,

As suggested by MIchael Setzer II in the News forum, you could add the following to your /etc/hosts file so that your client(s) can still get at the server using the now-withdrawn name. You should be able to run down your work to a point where you can take action to restore the original site name!

128.113.126.54 milkyway-new.cs.rpi.edu


This was working for me until the various BOINC scheduler services were taken offline some time on th eevening of 28th November (confirmed by the server status page); uploads are still getting through but then sit at "Ready to report" and, of course, downloads aren't possible at present (no scheduler...)

Hope this helps in some way.

Cheers - Al


Thanks Al,
I managed to get this to work and thanks to you and Michael It worked so I have allowed the remaining work units to finish before I reset the project.

As root I used echo 'text' | tee -a /file.txt > /dev/null

to make the change (don't leave the 'nnn' around the text and put the file name in that you need)

Conan
6) Message boards : Number crunching : Thread to report issues after server migration (Message 76642)
Posted 28 Nov 2023 by Profile Conan
Post:
Marcin wrote:
another issue to report!
I am getting hostname resolution errors now:
[code]34852: 19-Nov-2023 11:13:37 (low) [Milkyway@home] Sending scheduler request: To report completed tasks.
34853: 19-Nov-2023 11:13:37 (low) [Milkyway@home] Reporting 1 completed tasks
34854: 19-Nov-2023 11:13:37 (low) [Milkyway@home] Not requesting tasks: don't need (job cache full)
34855: 19-Nov-2023 11:13:42 (low) [] Project communication failed: attempting access to reference site
34856: 19-Nov-2023 11:13:42 (low) [Milkyway@home] Scheduler request failed: Couldn't resolve host name



[quote]It seems that I can eliminate the issue locally in the client by shutting down the client, removing the <scheduler_url>url</scheduler_url> line from client_state.xml, and restarting the client. But that's just a workaround of course, the real fix needs to happen on the server.


I have 17 pc's running here at my home and get that at every Project over the course of a day as I manually update them, I REALLY dislike seeing a whole page of completed tasks sitting there waiting to be returned and also dislike seeing tasks that are stuck uploading so I will keep clicking until they get returned or uploaded as the case may be. Now if your 'fix' really does fix it then I wonder if it's the change-over to https from http that is causing the problems I'm seeing at all the other projects?


I, like a lot of others (or everyone), received the message that I had to change my URL from https://milkyway.cs.rpi.edu/milkyway/ to http://milkyway-new.cs.rpi.edu/milkyway/.
I could not get work until I did this change.
I was able to download work on both my computers but now I can't upload any results due the server no longer accepting my HOST NAME.
So I was told to change from https to http not the other way round.
Now I see a forum message that I will need to change it back again.

There goes a heap of work that will benefit no one.

Just what is the correct thing to do, or do I just drop this hot potato, lose the work and move on.

In the NEWS section it states that the DNS has been changed to make milkyway-new to no longer exist.
I noticed that my Master file in the BOINC directory has nothing in it at all.

So BOINC no longer knows what to do with my results, URL no longer exists, my work is wasted, I can't return the results as it trying to send to a non existent URL, that was in existence just a day ago.

Is there any editing I can do in the ACCOUNT file in BOINC that will allow me to return the work or do I now have to RESET the project AGAIN and lose all my work just to go back to where I was in the first place?

I have suspended the remaining work units pending an answer here as I don't want to waste any more time on work that may be lost anyway.

Conan
7) Questions and Answers : Unix/Linux : 15 CPUs cause running tasks to stop running (Message 76299)
Posted 20 Jul 2023 by Profile Conan
Post:
mikey - for information:

"PERF" as in Ithena Measurements Perf tasks? If so it's like any other Boinc project you choose what kind of tasks you want to run, Perf is for Windows only while Ooni tasks are Linux only and the Cnode tasks are for both I think. There is also a Project Ithena computation that has Hex tasks https://comp.ithena.net/usr/
Nope - we're talking Linux, and perf is one of a set of kernel-specific system tools; it offers various different ways of looking at system performance...

Cheers - Al.


Just for information, 'Perf' work units can run on both Linux and Windows.

I run them on my Linux computers along with CNode and OONI Probe.
The only problem is I have to often hit update for OONI and Perf to download as CNode is the default. Project Admin has not worked out how to download all at once unfortunately.

Conan
8) Message boards : News : Separation Application Shutting Down on Tuesday, Jun 20th (Message 76216)
Posted 10 Jul 2023 by Profile Conan
Post:
The thing with Bench marking, in relation to BOINC, as that is what we are doing, it serves less purpose than when originally designed.
With so many projects using fixed credit systems it is becoming redundant for use as a measure for granting credit.
However for the projects that still use them, bench marks are used for the granting of credits and in some cases with a quorum of two required then it is often the lower claim that is awarded.
This really comes to a head when you don't run bench marks as you get the default BOINC bench mark of 1 billion floating points and 1 billion drystones even if you have never run a bench mark in BOINC.

This will always give you a very low credit claim and your so called wingman will (even if they have run bench marks) get a low awarded credit score as the lowest is awarded.

So running a bench mark on the projects that use them does serve a purpose (for the granting of credit), for other BOINC projects that don't use them they serve no purpose and are not needed at all.

Bench marks are supposed to be part of the BOINC software for the use of the CreditNew system to award credit.
It does not work very well,
YAFU only uses the default unmodified BOINC software and CreditNew.
I recently ran a 16t work unit on my faster higher bench marking 7900X Ryzen system and it used all 24 threads for 9 hours (equal to about 7 days of processing) and got 1,450 odd points.
I then ran an 8t work unit on my lower bench marked Ryzen 5900X and it ran for 8 hours using up to 24 threads (equal to about 5 days processing) and it was awarded 54,200 odd points.

I can't figure that out, bench marks and CreditNew just don't work.

Bench marks are not the be all and end all for the testing of a CPU, as no single bench mark does every thing (at least as far as I have seen), it there was a single bench mark that could test every thing then they would not have to run a "Test Suite" to see how a CPU preforms they could just run one programme.

I don't need a bench mark to tell me that my Ryzen 5900X runs a programme at a slower pace than my Ryzen 7900X, it is obvious with the times taken.

My computers are not hidden you can see the Bench Mark difference but they don't tell the whole story.

Bench Marks, don't believe them. CreditNew, don't use it.

Rant over, thanks for your time.

Conan
9) Message boards : News : Separation Application Shutting Down on Tuesday, Jun 20th (Message 76102)
Posted 29 Jun 2023 by Profile Conan
Post:
It can work though, depending on the project.

I told BOINC I had 128 cores on my 12 core 24 thread Ryzen CPU so I could run YAFU work units of 32, 64 and 128 threads.
Worked well, but I was a bit slow changing it back once my goal on YAFU had been reached and BOINC flooded my computer with work from other projects thinking I had 128 cores but I only had 24.

The lucky thing about both cc_config.xml and app_config.xml files is that you can change it any time with BOINC still running then in BOINC Manager hit "Read Config Files" and it updates straight away.

Conan
10) Message boards : News : Server Maintenance 6/28/2023 12 PM EDT (4 PM UTC) (Message 76099)
Posted 29 Jun 2023 by Profile Conan
Post:
Thank you Tom,

You have done a great job and I appreciate what you have done. I just got a heap of credit for work that I had written off as not going to get credit for and then I have it. Great work.

People like you are the ones that make people stay at projects, active on the forums, responding to help requests, and actually doing something about it.

Thanks again
Conan
11) Message boards : News : Separation Application Shutting Down on Tuesday, Jun 20th (Message 75886)
Posted 23 Jun 2023 by Profile Conan
Post:
Could the Admins just run the credit granting scrip to all pending Separation work units, and once done then flush all the Separation work units from the Data Base?

I have over 1200 just sitting there and others would have many more.

It would tidy things up a bit and then they could concentrate on the N-Body work.

I will be back doing N-Body work units when some other commitments are met, wish they paid the same as the Separation work did.

Conan
12) Message boards : News : Separation Project Coming To An End (Message 75673)
Posted 18 Jun 2023 by Profile Conan
Post:
P.S. the ... post you made and I assume you wanted to delete it, changing to two spaces makes it go away. Weird, but it works.


Thanks Peter,

This tip is great and works well. Just deleted everything in a double posting that I had made then hit the space bar twice hit OK and the one I didn't want disappeared.

Conan
13) Message boards : Number crunching : One WU using only 16 threads. (Message 75672)
Posted 18 Jun 2023 by Profile Conan
Post:
I have a 12 core 24 thread Ryzen CPU, whenever I ran N-body work units only 1 would run and it would take 16 cores, when that one finished then another would start.

Buggered up the work throughput as most would not be completed before the deadline.

I added an app_config.xml file to make it use only 8 cores and now 3 will happily run and use all threads.

I think 16 is the default, at least for me it was.

In theory if you have 32 threads then it should run 2 work units but not sure about the default nature of the settings if it will then start another WU if one is already running.

Conan
14) Message boards : News : Separation Project Coming To An End (Message 75623)
Posted 16 Jun 2023 by Profile Conan
Post:
I only have CPU anyway so not a biggy for me, however I have now lost another project that runs on 32 bit windows which is really starting to limit options.

However my 64 bit Linux machines can easily do N-Body work,
it is just a pity that the credit is so low when 8 cores at a time are locked up on one work unit,
and often give less points that a single Separation work unit would, and it uses just the one core.

Conan
15) Message boards : Number crunching : New team member lots of pendings or invalids. (Message 71097)
Posted 3 Sep 2021 by Profile Conan
Post:
G'Day Rick,

From what I can see, most of the Validaion Inconclusive work units are just waiting for a wingman to return their work so it can be compared to your work and then validated.
They are marked as pending as they are still waiting on the other person.

The invalids I am not sure about, possibly just not agreeing with the other person's returned result. Other people who know more might be able to expand on that.

Conan
16) Message boards : News : New Separation Runs 6/9/2021 (Message 71049)
Posted 3 Aug 2021 by Profile Conan
Post:
Thanks Jim and Al. That seems to make sense.

Upgrading may be an issue as I have the last version output by BOINC themselves.

The later versions seem to have a lot to do with the distributors of the various Linux versions.
I have tried some in the past but I don't like the way they store my data all over the place and I lose control as it all goes to root control.

But there is a way I think to take back that control and install a later version my way, I will have to think about it and have a go at a later date.

Conan
17) Message boards : News : New Separation Runs 6/9/2021 (Message 71045)
Posted 2 Aug 2021 by Profile Conan
Post:
All Linux Separation work units are failing with the following error

Task 276551426
Name de_modfit_85_bundle4_4s_south4s_gapfix_1627399316_3938308_0
Workunit 145671322
Created 1 Aug 2021, 16:42:15 UTC
Sent 1 Aug 2021, 16:53:15 UTC
Report deadline 13 Aug 2021, 16:53:15 UTC
Received 2 Aug 2021, 22:00:35 UTC
Server state Over
Outcome Computation error
Client state Compute error
Exit status 1 (0x00000001) Unknown error code
Computer ID 448629
Run time 2 sec
CPU time
Validate state Invalid
Credit 0.00
Device peak FLOPS 4.33 GFLOPS
Application version Milkyway@home Separation v1.46
x86_64-pc-linux-gnu
Peak disk usage 0.01 MB
Stderr output

<core_client_version>7.4.25</core_client_version>
<![CDATA[
<message>
process exited with code 1 (0x1, -255)
</message>
<stderr_txt>
<search_application> milkyway_separation 1.46 Linux x86_64 double </search_application>
Reading preferences ended prematurely
Error loading Lua script 'astronomy_parameters.txt': [string "number_parameters: 4..."]:1: '<name>' expected near '4'
Switching to Parameter File 'astronomy_parameters.txt'
<number_WUs> 4 </number_WUs>
<number_params_per_WU> 26 </number_params_per_WU>
stream sigma 0.0 is invalid
Failed to get stream constants
07:58:52 (32011): called boinc_finish(1)

</stderr_txt

Windows is working OK on Separation WUs, N-Body works fine on same Linux computer, using just CPU for both WU types.

Conan
18) Message boards : Number crunching : Running N-Body work units (Message 65979)
Posted 30 Nov 2016 by Profile Conan
Post:
After a bit of an absence from the project I have come back to run some of the N-Body simulations after reading about it in an astronomy magazine.

I recently also started running an older 8 core/16 thread xeon server picked up cheaply.

I have been noticing a few things about these Multi-Threaded work units (mt).

The first thing you of course notice is the very low credit per work unit. Searching other results as well as mine I have seen as low as 1.48 credits for a WU and as high as 41 credits, with most around the 15 to 26 credit mark.

Second thing is this credit to some degree seems related to the benchmark figures.

Third thing is the more cores running the WU the less credit per hour you will get.
This is due to the extra CPU time run up by the extra cores.

A 4 core machine may take a RUN time of 800 seconds for a CPU time of 3,200 seconds (this is approximate).
My 16 core machine drops the RUN time to 500 seconds but the CPU time goes up to 8,000 seconds in doing so.

This then means that if 26 credits are awarded then the 4 core computer gets 29.25 cr/h and is doing OK, however my 16 core only gets 11.7 cr/h.

The WU's only run a short time, however a number get downloaded at once and take some time to work through, stopping my computer from running any other projects.

With credit per hour measured against CPU time not RUN time I don't get much reward for these work units.

Yes I could limit the number of cores being used but that is a lot of mucking around and will affect other projects that I run (which is over 2 dozen).

I was surprised that a Multi Threaded aplication is not awarding the effort of extra cores being used to solve the problem, less cores more reward.

So does this application become less efficient after a certain number of cores are reached? If so that should be stated to anyone wanting to run the application.

Thanks
Conan
19) Message boards : News : Nbody Status Update (Message 63848)
Posted 3 Aug 2015 by Profile Conan
Post:
You mentioned about compiling nbody on an XP machine. Last I recalled, XP was only 32 bit. You seem to be doing a lot of floating point arithmetic. Are you doing double-precision? If so, wouldn't a 64-bit compiled version have to use different code and at least a 64-bit compiler? Am I missing something?


I understood that Windows XP came in both 32 and 64 Bit flavours, not just 32 Bit.
But as Sidd took down the Windows app for N-Body due to a bug that appears to yet be fixed then this is probably a mute point.

I just came back to run a few N-Body work units on my Linux computers, but as none are currently available I will have to wait.

Conan
20) Message boards : Number crunching : Listed Application GigaFLOP estimations (Message 63765)
Posted 24 Jun 2015 by Profile Conan
Post:
I was looking through the currently available Applications for Milkyway and found that there must be thousands of Macs running this project.

When I say thousands I mean many thousands as the GigaFLOP output for the
Mac OS X 10.5 running Intel 64 bit 1.36(opencl_nvidia_101) shows 31,422,013,947,084 GigaFLOPS

Nvidia must be very happy as well.

I suppose it could be a computing cluster supplying the processing power?

Very impressive output.

Conan


Next 20

©2024 Astroinformatics Group