Posts by Toby Broom

1) Message boards : Questions and problems : Editing RAM usage for project (Message 92809)
Posted 15 Sep 2019 by Toby Broom
Post:
I think I found the problem another config file was limiting the max number of tasks.
2) Message boards : Questions and problems : Editing RAM usage for project (Message 92808)
Posted 15 Sep 2019 by Toby Broom
Post:
Hi,

For my LHC@Home project the project team do not calculate the amount of RAM usage correctly so I would like to fix it client side.

I am able to edit the rsc_memory_bound & working_set_size_smoothed to the correct values but the scheduler doesn't seems to see that the memory load is reduced.

How does the calculation of the free memory work?
75			09/15/19 15:20:54	[mem_usage] enforce: available RAM 128359.19MB swap 75217.38MB	
76	LHC@home	09/15/19 15:20:57	[mem_usage]  (not running)Theory_2417912_1568071912.492574_0: WS 84.60MB, smoothed 5120.00MB, swap 103.84MB, 0.00 page faults/sec, user CPU 16.063, kernel CPU 1923.344	
77	LHC@home	09/15/19 15:20:57	[mem_usage]  (not running)Theory_2475718_1568080319.011424_0: WS 85.05MB, smoothed 5120.00MB, swap 103.87MB, 0.00 page faults/sec, user CPU 22.547, kernel CPU 3868.422	
78	LHC@home	09/15/19 15:20:57	[mem_usage]  (not running)Theory_2505470_1568084822.594092_0: WS 85.52MB, smoothed 5120.00MB, swap 100.29MB, 0.00 page faults/sec, user CPU 19.781, kernel CPU 3618.375	
79	Radioactive@Home	09/15/19 15:20:57	[mem_usage] sample_10775727_0: WS 4.75MB, smoothed 4.75MB, swap 6.39MB, 0.00 page faults/sec, user CPU 0.000, kernel CPU 0.063	
80	LHC@home	09/15/19 15:20:57	[mem_usage]  (not running)Theory_2471900_1568079718.688719_0: WS 84.43MB, smoothed 5120.00MB, swap 94.43MB, 0.00 page faults/sec, user CPU 3.313, kernel CPU 44.984	
81	LHC@home	09/15/19 15:20:57	[mem_usage] Theory_2386456_1568067408.805761_0: WS 86.16MB, smoothed 5120.00MB, swap 96.68MB, 0.00 page faults/sec, user CPU 7.016, kernel CPU 188.047	
82	LHC@home	09/15/19 15:20:57	[mem_usage] Theory_2513686_1568086023.457098_0: WS 89.50MB, smoothed 5120.00MB, swap 101.08MB, 0.00 page faults/sec, user CPU 6.484, kernel CPU 176.688	
83	LHC@home	09/15/19 15:20:57	[mem_usage] Theory_2520971_1568087224.201573_0: WS 89.66MB, smoothed 5120.00MB, swap 101.04MB, 0.00 page faults/sec, user CPU 5.813, kernel CPU 137.344	
84			09/15/19 15:20:57	[mem_usage] BOINC totals: WS 609.68MB, smoothed 35844.75MB, swap 707.62MB, 0.00 page faults/sec	
85			09/15/19 15:20:57	[mem_usage] All others: WS 3381.68MB, swap 3284.45MB, user 6516.125s, kernel 6734.047s	
86			09/15/19 15:20:57	[mem_usage] non-BOINC CPU usage: 0.00%	

3) Message boards : Questions and problems : Simulate an amount of memory (Message 90768)
Posted 21 Mar 2019 by Toby Broom
Post:
Is there an setting similar to <ncpus> but for memory?

thanks
4) Message boards : Questions and problems : Maximum CPUS (threads) supported by BOINC client (Message 90767)
Posted 21 Mar 2019 by Toby Broom
Post:
I think these limits are server side. See these config options, for a project.

<max_ncpus>N</max_ncpus>
An upper bound on NCPUS (default: 64)

I can imagine that most projects didn't increase this from default so you are correct.

I do what you propose which is run 2 clients on the same PC, you have to take care about the memory usage, since both clients think they have all the ram, so if your pushing the ram limits then you can end up with very slow computer.

I'm not sure if the limit is 1000?, it should be 64x<max_wus_in_progress> N </max_wus_in_progress>, whatever the project configures N to be. Each project should balance this, as you say if they set it low then the users of project would be hammering the server for WU's so they need to increase it to lower server load.

I assume the client it's self has no limits. If you need more than 64 then you can ask the project admins to increase the limits to 128.
5) Message boards : Questions and problems : Adjust Working Set Client-Side (Message 87227)
Posted 22 Jul 2018 by Toby Broom
Post:
Can I adjust the working set size of tasks from the client?

The project I'm connected to allows between 1-8 work units at a time or an unlimited option, when the option is set to 1 for example then the working set is correctly calculated, however your are limited to 1 WU at a time which is wasteful given my computers have many more than 8 cores.

If I choose unlimited then I can use the work buffers as set in the BOINC config but the working set is not calculated correctly anymore so once it thinks it's used all the memory then it will not start any further work.
6) Message boards : Questions and problems : Suspend CPU when there is GPU work (Message 85742)
Posted 5 Apr 2018 by Toby Broom
Post:
Hi is it possiable to suspend CPU work when there is GPU work?

It seems like the suspend option stops everything and the GPU works the other way around.
7) Message boards : Projects : Which BOINC projects use the most memory? (Message 84182)
Posted 3 Jan 2018 by Toby Broom
Post:
I'm able to use about 100GB of ram on ATLAS project, the 1.4GB number is not correct anymore it's at least 2.6GB
8) Message boards : Projects : Radioactive@Home domain is dead. (Message 84181)
Posted 3 Jan 2018 by Toby Broom
Post:
It was working up until a few weeks ago, I think we have to wait for a while to see if it's really dead.
9) Message boards : Projects : For the betterment of BOINC (Message 79816)
Posted 27 Jul 2017 by Toby Broom
Post:
1. I don't find the Disk and Statistics much use, I find that the statists are better servered on the website like boincstats etc. I always use the advanced view. BoincTasks is super valuable tool. the concepts of jobs, tasks, wu's are hard to configure on these virtualbox projects. I still have strange effects like some settings cause my PC to run 1 task when I set to unlimited for example. I think this is becasuse it too hard for the admins to make correct configurations. The projects like this are beyond a normal user to setup. I don't critise this help as its excellent but a avergae person wouldn't have a clue what to do.

https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4161#29359

editing client_state.xml etc...
Boinc knows the specs of your PC so it could help you?

I like davids concept that you would optin to an area of science, this makes it more likly that people would join projects as now you have to discover them so how. Currently project are optin mabybe optout would be better, athough maybe an invasion, athough you optin to BOINC

I it would be nice if BOINC could post to socialmedia that I achived something for science, found new prime or computed a low energy confimation.

Since we give our resources to the projects I have to wonder what is the return, from davids comments it seems like the main users are people that liek to tinker with computer so they do it for those reasons. Gridcon maybe to make money. I imagine for many people these aren't such motivating reasons.

2. I'm not sure what I could do?
10) Message boards : BOINC client : My Wish List (Message 77263)
Posted 10 Apr 2017 by Toby Broom
Post:
Some project are non cpu intensive and the multithreaded ones can be limited to a specificed number of cores so I would like to know how it's all working.
11) Message boards : BOINC client : My Wish List (Message 77249)
Posted 9 Apr 2017 by Toby Broom
Post:
Can the boinccmd --get_tasks report the number of CPU/GPU cores used by the task?
12) Message boards : Questions and problems : Controling blend of mutlicore and single core VM's (Message 77096)
Posted 1 Apr 2017 by Toby Broom
Post:
I finally got it to work

The project_max_concurrent is jobs, so it will always run that many tasks regardless of the number of cores per task.

I forced the number of CPU's as per your suggestion.

Now it respects the number of cores correctly. I can also limit the number of cores in use with the Use at most N% or the <ncpus>.

I think the concept of a job is quite confusing, not sure why it would be useful, max cores per project would be better?
13) Message boards : Questions and problems : Controling blend of mutlicore and single core VM's (Message 76708)
Posted 22 Mar 2017 by Toby Broom
Post:
I expect still does not work correctly? This would be my expected configuration

<app_config>
<project_max_concurrent>8</project_max_concurrent>
<app>
<name>Dual Core</name>
<max_concurrent>4</max_concurrent>
</app>
<app>
<name>Single Core</name>
<max_concurrent>8</max_concurrent>
</app>
</app_config>

My example run time of 4 Dual & 4 Single is fine with this config but use 12 cores.

BOINC knows that the dual core task core take 2 core as the tasks have (2 cores) after the name, it just doesn't seem to respect the limits.
14) Message boards : Questions and problems : Controling blend of mutlicore and single core VM's (Message 76543)
Posted 19 Mar 2017 by Toby Broom
Post:
I have some tasks that are mutli core and some single core, I can create an app_config like so:

<app_config>
<project_max_concurrent>8</project_max_concurrent>
</app_config>

This will limit the number of tasks to 8 total, however if I have 8 dual core then this is actually 16 cores of usage. So I can create a limit like so:

<app_config>
<app>
<name>App2</name>
<max_concurrent>4</max_concurrent>
</app>
</app_config>

This limits the app to 4 so 8 cores.

However, if I have 4 of App1 & 4 of App2 then both conditions are met 8 total and 4 of app2 but the total number of cores used is 8+4 so again over used.

Is there a way to configure BOINC so it will respect the number of cores?

BOINC 7.6.33, Windows 10, LHC@Home, not installed as service
15) Message boards : Questions and problems : cc_config, appnames (Message 52074)
Posted 21 Jan 2014 by Toby Broom
Post:
It's not bad for one computer but kind of a pain for 8 of them.

Thanks for feedback, I'll write up a script to make my life easier.
16) Message boards : Questions and problems : cc_config, appnames (Message 52069)
Posted 21 Jan 2014 by Toby Broom
Post:
Hi,

I have in my cc_config to stop GPU work when the CPU is fully loaded so I added the name of the app e.g.:

<exclusive_gpu_app>sixtrack_win32_4463_pni.exe</exclusive_gpu_app>

However if the project release a new app e.g.:

sixtrack_win64_4463_pni.exe

Now I need a new line in the cc_config.

Is there a way to wild card it to say sixtrack*

then all apps from that project would be exclusive?
17) Message boards : News : BOINC 7.2.28 released to the public (Message 51260)
Posted 11 Nov 2013 by Toby Broom
Post:
Could you include the VirtualBox version, included in the package?
18) Message boards : GPUs : OpenCL but no CAL? (Message 50765)
Posted 4 Oct 2013 by Toby Broom
Post:
I guess I'd like to use a driver later than 13.5 (12.5 was a typo).

for some reason all the later ones don't detect CAL?
19) Message boards : GPUs : OpenCL but no CAL? (Message 50763)
Posted 3 Oct 2013 by Toby Broom
Post:
I have a 7790 GPU in my computer if I install a driver after the 12.5 beta I see the following when I load up BOINC:
03-Oct-13 18:52:45 | | Starting BOINC client version 7.0.64 for windows_x86_64
03-Oct-13 18:52:45 | | log flags: file_xfer, sched_ops, task
03-Oct-13 18:52:45 | | Libraries: libcurl/7.25.0 OpenSSL/1.0.1 zlib/1.2.6
03-Oct-13 18:52:45 | | Data directory: C:\ProgramData\BOINC
03-Oct-13 18:52:45 | | Running under account Administrator
03-Oct-13 18:52:45 | | Processor: 24 GenuineIntel Intel(R) Xeon(R) CPU L5640 @ 2.27GHz [Family 6 Model 44 Stepping 2]
03-Oct-13 18:52:45 | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 popcnt aes syscall nx lm vmx smx tm2 dca pbe
03-Oct-13 18:52:45 | | OS: Microsoft Windows Server 2008 "R2": Standard x64 Edition, Service Pack 1, (06.01.7601.00)
03-Oct-13 18:52:45 | | Memory: 23.98 GB physical, 47.97 GB virtual
03-Oct-13 18:52:45 | | Disk: 148.95 GB total, 71.29 GB free
03-Oct-13 18:52:45 | | Local time is UTC -4 hours
03-Oct-13 18:52:45 | | VirtualBox version: 4.2.16
03-Oct-13 18:52:45 | | OpenCL: AMD/ATI GPU 0: AMD Radeon HD 7790 series (Bonaire) (driver version 1124.2 (VM), device version OpenCL 1.2 AMD-APP (1124.2), 1024MB, 966MB available, 2408 GFLOPS peak)

If I have 13.5 then I get this extra line:
03-Oct-13 18:52:45 | | CAL: ATI GPU 0: AMD Radeon HD 7790 series (Bonaire) (CAL version 1.4.1741, 1024MB, 966MB available, 2408 GFLOPS peak)

Any idea's how I can use a later driver?

I have another computer with a 7870XT and it seems to show CAL all the time?
20) Message boards : Questions and problems : Scheduling with T4t and CPDN (Message 43592)
Posted 18 Apr 2012 by Toby Broom
Post:
I'll just nurse the work units though then until I clear all my CPND work units.

The scheduler doesn't seem to work the way I want in 7.0.25

Thanks for the sugestions.


Next 20

Copyright © 2019 University of California. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.