Coming back to folding@home / Intel i3 & nVidia Gefore GT 52

[codysluder]

All projects expire, though the dates depend on the project. If it has been that long, the WU would have been assigned to someone else and it would have been completed long ago. The data from your old laptop can be discarded.

Please download a new client. Use the same UserName/Team Number/Passkey that you used before.

For a list of the active projects and the deadlines associated with them, see the Project Summary

[Slash_2CPU]

You can run 2, 3, or 4 uniprocessor cores under a single v7 client if you want to fully use your i3.

Under v7, "FAH Control" provides a single interface to monitor and configure all cores.

You can run FAH and BOINC at the same time if that is what your earlier question was. Run BOINC on the GPU, and run FAH on the CPU cores. There is an option you need to enable on FAH for "slightly higher" thread scheduling priority for it all to work right.

[brainvision]

OK, finally I start folding..

I have to say that for a Slackware user thing are really not easy.. not easy at all! I'm going to ask to the beta client releasers if they share the source code or at least the x86 binaries so that I can build a Slackware package or so that a Linux user could compile it by himself.

Edit by Mod: The issue of installing FAH on other distros of Linux is really a new topic. I'm separating that part of your discussion and moving it here so that it becomes part of the discussion of the V7 beta development effort.


going on, I would like to ask a little question, again.. just for curiosity, I choose into the config file to set the SMP donor.. How could I be sure that it is not working? I think that maybe, if SMP is working, it would download more than one FahCore_a4 binary.. that's right? Actually under cores dir I got only one dir pointing to FahCore_a4.. if I could use SMP they would be more than one, that's right?

You can run 2, 3, or 4 uniprocessor cores under a single v7 client if you want to fully use your i3.

Under v7, "FAH Control" provides a single interface to monitor and configure all cores.

You can run FAH and BOINC at the same time if that is what your earlier question was. Run BOINC on the GPU, and run FAH on the CPU cores. There is an option you need to enable on FAH for "slightly higher" thread scheduling priority for it all to work right.

Thanks Slash_2CPU.. but, how could do I do this? How can I choose which CPU use for folding? Maybe with the expert option into the FAHControl Preferences? Sincerely it's still a little bit confusing for me that interface..
That's my log..

Code: Select all

*********************** Log Started 2011-12-07T19:49:23 ************************
19:49:23:************************* Folding@home Client *************************
19:49:23:    Website: http://folding.stanford.edu/
19:49:23:  Copyright: (c) 2009-2011 Stanford University
19:49:23:     Author: Joseph Coffland <[email protected]>
19:49:23:       Args: --chdir=/home/brainvision/.FAHClient/ --lifeline 20714
19:49:23:             --command-port=36330
19:49:23:     Config: /home/brainvision/.FAHClient/config.xml
19:49:23:******************************** Build ********************************
19:49:23:    Version: 7.1.38
19:49:23:       Date: Oct 6 2011
19:49:23:       Time: 18:39:13
19:49:23:    SVN Rev: 3080
19:49:23:     Branch: fah/trunk/client
19:49:23:   Compiler: Intel(R) C++ g++ 4.1 mode 1110
19:49:23:    Options: -std=gnu++98 -diag-disable 279 -O3 -funroll-loops -msse
19:49:23:             -axSSE2,SSE3,SSSE3,SSE4.1,SSE4.2 -openmp -restrict
19:49:23:   Platform: linux2 2.4.27-3-386
19:49:23:       Bits: 32
19:49:23:       Mode: Release
19:49:23:******************************* System ********************************
19:49:23:        CPU: Intel(R) Core(TM) i3-2310M CPU @ 2.10GHz
19:49:23:     CPU ID: GenuineIntel Family 6 Model 42 Stepping 7
19:49:23:       CPUs: 4
19:49:23:     Memory: 3.87GiB
19:49:23:Free Memory: 2.63GiB
19:49:23:    Threads: POSIX_THREADS
19:49:23: On Battery: false
19:49:23: UTC offset: 1
19:49:23:        PID: 20902
19:49:23:        CWD: /home/brainvision/.FAHClient
19:49:23:         OS: Linux 2.6.38.7-smp i686
19:49:23:    OS Arch: X86
19:49:23:       GPUs: 1
19:49:23:      GPU 0: NVIDIA:1 GT520M Nvidia GeForce GT520M
19:49:23:       CUDA: 2.1
19:49:23:CUDA Driver: 4010
19:49:23:***********************************************************************
19:49:23:<config>
19:49:23:  <!-- Folding Slot Configuration -->
19:49:23:  <smp v='true'/>
19:49:23:
19:49:23:  <!-- User Information -->
19:49:23:  <passkey v='********************************'/>
19:49:23:  <team v='51670'/>
19:49:23:  <user v='brainvision'/>
19:49:23:
19:49:23:  <!-- Folding Slots -->
19:49:23:</config>
19:49:23:Trying to access database...
19:49:23:Successfully acquired database lock
19:49:23:Enabled folding slot 00: READY smp:4
19:49:23:Starting Unit 00
19:49:23:Running core: /home/brainvision/.FAHClient/cores/www.stanford.edu/~pande/Linux/x86/Core_a4.fah/FahCore_a4 -dir 00 -suffix 01 -lifeline 20902 -version 701 -checkpoint 15 -np 4
19:49:23:Started core on PID 20910
19:49:23:FahCore 0xa4 started
19:49:24:Unit 00:
19:49:24:Unit 00:*------------------------------*
19:49:24:Unit 00:Folding@Home Gromacs GB Core
19:49:24:Unit 00:Version 2.27 (Dec. 15, 2010)
19:49:24:Unit 00:
19:49:24:Unit 00:Preparing to commence simulation
19:49:24:Unit 00:- Ensuring status. Please wait.
19:49:26:Server connection id=1 on 0.0.0.0:36330 from 127.0.0.1
19:49:33:Unit 00:- Looking at optimizations...
19:49:33:Unit 00:- Working with standard loops on this execution.
19:49:33:Unit 00:- Previous termination of core was improper.
19:49:33:Unit 00:- Going to use standard loops.
19:49:33:Unit 00:- Files status OK
19:49:33:Unit 00:- Expanded 661065 -> 1008860 (decompressed 152.6 percent)
19:49:33:Unit 00:Called DecompressByteArray: compressed_data_size=661065 data_size=1008860, decompressed_data_size=1008860 diff=0
19:49:33:Unit 00:- Digital signature verified
19:49:33:Unit 00:
19:49:33:Unit 00:Project: 7903 (Run 3, Clone 9, Gen 3)
19:49:33:Unit 00:
19:49:33:Unit 00:Entering M.D.
19:49:39:Unit 00:Using Gromacs checkpoints
19:49:39:Unit 00:Mapping NT from 4 to 4 
19:49:40:Unit 00:Resuming from checkpoint
19:49:40:Unit 00:Verified 00/wudata_01.log
19:49:40:Unit 00:Verified 00/wudata_01.trr
19:49:40:Unit 00:Verified 00/wudata_01.edr
19:49:40:Unit 00:Completed 222180 out of 2500000 steps  (8%)
19:52:26:Unit 00:Completed 225000 out of 2500000 steps  (9%)

Do I donor with only 1 cpu? So, do I unset the SMP option?

Thanks!! As always you are very nice with me..

[bruce]

Another big problem is that launching the program without no option but the username, the passkey and group ID

This is the subject of an open ticket for V7.1.38 but it will probably be fixed in the instal script so it might not be easy for you to adapt to Slackware.

going on, I would like to ask a little question, again.. just for curiosity, I choose into the config file to set the SMP donor.. How cxould I be sure that it is not working? I think that maybe, if SMP is working, it would downlaod more than one FahCore_a4 binary.. that's right? Actually under cores dir I got only one dir pointing to FahCore_a4.. if I could use SMP they would be more than one, that's right?

By default, a single copy of a SMP core creates as many concurrent threads as as you have CPUs. You can reduce that number if you need to leave resources free, but since it runs at a suitable nice level, that shouldn't be required. If your configuration is not SMP but instead is configured as one or more Uniprocessor slots, you'll either get an assignment for FahCore_78 or FahCore_a4 but they'll be using only one CPU. In effect, FahCore_a4 understands smp:N for any value of N from the number of processors you have all the way down to 1 (even though there's actually no such setting as smp:1).

How many CPU cores do you have, and what percent of them are devoted to FahCore_a4?

All settings can be managed from FAHControl and the current status is updated regularly.

[brainvision]

mmm.. I'm not sure that I well understand the last part of your words, sorry.. I need to read it again, I'll try that..
anyway, did you see my log r3eported? Do you think that FAH is using SMP or uniprocessor folding?
So, it's not true that, for example, if I choose a SMP folding (naturally with a supported SO, it's an example) I would find more "cores" directories or FAHCore_XX processes? Actually I have only one, even if I configured it for SMP.. but I have a Slackware 32 bit, so it is supposed to not work..
Last question: if I'm using only one CPU because SMP doesn't work for my 32 bit OS, why I see that ALL 4 CPU are always 100% even if the laptop is only folding at that moment?

[Joe_H]

Looking at your log, you are folding with SMP with the A4 core and using 4 threads. See these parts of the log:

Code: Select all

19:49:23:<config>
19:49:23:  <!-- Folding Slot Configuration -->
19:49:23:  <smp v='true'/>

19:49:23:Enabled folding slot 00: READY smp:4
19:49:23:Starting Unit 00

19:49:33:Unit 00:Project: 7903 (Run 3, Clone 9, Gen 3)
19:49:33:Unit 00:
19:49:33:Unit 00:Entering M.D.
19:49:39:Unit 00:Using Gromacs checkpoints
19:49:39:Unit 00:Mapping NT from 4 to 4

Now since I do not do folding under Linux, I have not been following all the progress in the Version 7 Beta forum on the Linux client. It appears that the 32-bit version will do SMP folding, at least with the A4 core. The A4 core is the newest and will process as either a single-threaded core or as a multi-threaded core when that is available. It does not need multiple copies in your directories or loaded into separate processes to do that. Thread synchronization is done within the process unlike with the older A1 and A2 cores. You might not be able to do SMP folding with other cores such as the A3, that depends on whether a 32-bit version is available for Linux. Someone who knows more about the Linux support and its current state compared to the latest information update on the Folding web pages would have to comment on that.

[brainvision]

Looking at your log, you are folding with SMP with the A4 core and using 4 threads. See these parts of the log:

Code: Select all

19:49:23:<config>
19:49:23:  <!-- Folding Slot Configuration -->
19:49:23:  <smp v='true'/>

19:49:23:Enabled folding slot 00: READY smp:4
19:49:23:Starting Unit 00

19:49:33:Unit 00:Project: 7903 (Run 3, Clone 9, Gen 3)
19:49:33:Unit 00:
19:49:33:Unit 00:Entering M.D.
19:49:39:Unit 00:Using Gromacs checkpoints
19:49:39:Unit 00:Mapping NT from 4 to 4

Now since I do not do folding under Linux, I have not been following all the progress in the Version 7 Beta forum on the Linux client. It appears that the 32-bit version will do SMP folding, at least with the A4 core. The A4 core is the newest and will process as either a single-threaded core or as a multi-threaded core when that is available. It does not need multiple copies in your directories or loaded into separate processes to do that. Thread synchronization is done within the process unlike with the older A1 and A2 cores. You might not be able to do SMP folding with other cores such as the A3, that depends on whether a 32-bit version is available for Linux. Someone who knows more about the Linux support and its current state compared to the latest information update on the Folding web pages would have to comment on that.

WOOOW! That's absolutely confusing!! It's not your fault, obviously: I'm talking about the all situation..

[bruce]

WOOOW! That's absolutely confusing!! It's not your fault, obviously: I'm talking about the all situation..

I think you're talking about the default being to use 4 cores out of 4. If not, please help me understand what you mean.

FAH has always been designed to use UNUSED resources. It is intentionally designed to run at the lowest possible priority so that it can use any processor resources which would otherwise be idle time. There's an assumption that everything else that you run will run at a more normal priority and therefore will preempt the FAH tasks. Whatever resources are needed by other tasks should instantly be available to you.

You also have the option to reduce that default setting to allow resources to be wasted if that's what you want to do.

[brainvision]

I think you're talking about the default being to use 4 cores out of 4. If not, please help me understand what you mean.

yea, that's exactly the question which confused me.. as Joe_H said, it seems that I'm using SMP folding, even if I have a 32bit OS! Or maybe we are both right.. what do you if you take a look at my log reported? I have to advice that I launched FAHClient with --configure option and I set "true" at the SMP choice, so it's not a configuration settled by the client itself..

And please: what's A1, A2, A3 or A4??

So do you think it's normal that I have all the 4 CPU at 100%..

And is there a way to assign a folding to CPU 1, another to CPU 2 and so on? how can I tell the client that a project must use only 1 CPU,, even at 100%?

[bruce]

what's A1, A2, A3 or A4??

http://fahwiki.net/index.php/Cores

So do you think it's normal that I have all the 4 CPU at 100%.

Yes.

And is there a way to assign a folding to CPU 1, another to CPU 2 and so on? how can I tell the client that a project must use only 1 CPU,, even at 100%?

Yes, it's possible, but the scientific results will be worth less. You can configure separate uniprocessor slots instead of the single SMP slot.

Are you using FAHControl to configure your client or are you manually editing config.xml?

Did you look at the file sample-config.xml?

[brainvision]

mmm.. but if you think the result would be worst I don't want to do it.. maybe just 2, surely not 4!! do you think 2 project would be good?
Actually I'm just becoming familiar with the GUI and its option.. and yes, naturally I noticed the sample config file, but you must admit that for a novice donor it's not easy to understand.. and it's still in english!

please, bruce, can you take a look at the log I posted here http://foldingforum.org/viewtopic.php?f ... 88#p200934
it seems that I'm using SMP folding.. don't you agree? but I have a 32 bit!!

thanks!!

[bruce]

Yes, that's what it looks like. The client is 32-bit. FahCore_a4 is 32-bit. Some of the other SMP cores are 64-bit.

What I do not know is whether the Assignment Server is smart enough to only give you 32-bit assignments or not. I'll check with somebody who knows.