Welcome to the MacNN Forums.

If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below.

You are here: MacNN Forums > Community > Team MacNN > New Altivec-enhanced Seti worker in need of testing

New Altivec-enhanced Seti worker in need of testing (Page 7)
Thread Tools
Junior Member
Join Date: Oct 2004
Location: Germany
Status: Offline
Oct 26, 2005, 09:18 AM
 
Originally Posted by darcybaston
I have a Dual Core 2.0GHz, stock config (http://www.apple.com/powermac/specs.html) and Alex walked me through the process of doing a benchmark. Because of the 2x1MB caches, you're going to love these results:
Code:
~/Downloads/seti@home-G5-a5 >time ./seti@home-G5-a5 real 37m11.570s user 36m50.564s sys 0m19.551s init_data.xml contains: <wu_cpu_time>3098.689364</wu_cpu_time>
One question, is this result from the reference WU? It's not or?

Karl
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 26, 2005, 09:21 AM
 
If reference unit, you mean a file called "work_unit.sah", then yes. I used whatever Alex gave me.
     
alexkan  (op)
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status: Offline
Oct 26, 2005, 06:17 PM
 
Originally Posted by Karl Schimanek
One question, is this result from the reference WU? It's not or?

Karl
Yes, it's the reference work unit.

Darcy, are you planning on joining Team MacNN, or were you just in the mood to run the BOINC benchmarks as well?
     
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Oct 26, 2005, 07:08 PM
 
Going by these results, we can assume that a Quad 2.5 GHz G5 would churn out 4 units every half hour. I'm stomped!

Here's the Math:

37 min 12 sec = 2232 sec
2232 sec * 2.0 GHz / 2.5 GHz = 1776 sec (assuming linear performance increase)
1776 sec = 29.6 min
Since seti workers are single-threaded (or so I thought), each dual core processor can crunch two of these in parallel. Hence, a quad can crunch four.

If I went wrong anywhere, please correct me...
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 26, 2005, 07:29 PM
 
Originally Posted by alexkan
Darcy, are you planning on joining Team MacNN, or were you just in the mood to run the BOINC benchmarks as well?
I'm signed up in Team MacNN and joined the Evangelist RC-72 team. Is that what you mean?
     
alexkan  (op)
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status: Offline
Oct 26, 2005, 08:41 PM
 
Originally Posted by halimedia
Going by these results, we can assume that a Quad 2.5 GHz G5 would churn out 4 units every half hour. I'm stomped!

Here's the Math:

37 min 12 sec = 2232 sec
2232 sec * 2.0 GHz / 2.5 GHz = 1776 sec (assuming linear performance increase)
1776 sec = 29.6 min
Since seti workers are single-threaded (or so I thought), each dual core processor can crunch two of these in parallel. Hence, a quad can crunch four.

If I went wrong anywhere, please correct me...
Well, whether or not that holds true depends on whether or not the dual-core machines really scale linearly with clockspeed and number of cores. There are good reasons why they might not, both pertaining to memory.

Specifically, there's still a limit to how quickly we can stream the work unit out of memory. The larger caches have definitely cut down the amount of traffic to and from RAM, though. I haven't thought carefully enough about the implications of putting two cores on one FSB, but I can only assume that 4 cores worth of processing is going to saturate a bus somewhere.

Won't stop the quad-core G5 from being a beast, however. Anyone going to buy one?

So Darcy, I was kind of hoping that you would give SETI a try for a bit. I'm pretty sure that your machine could easily hit 1000 RAC, and likely more, if you ran it 24/7. (Incidentally, this would land you a pretty good spot on SETI's Top Computers list.) You seem like a RC5-72 kind of guy, though, so I don't know if I could turn you away from that. Also, any luck with those Shark traces?
     
Junior Member
Join Date: Oct 2004
Location: Germany
Status: Offline
Oct 27, 2005, 12:07 PM
 
Now, they have only to solve the "BOINC Manager timing problem".

Originally Posted by alexkan
Yes, it's the reference work unit.
WOW! What a speed up!
Can't be true. To good to be true
The SPEC numbers impressive, too

Something new on the alpha6 front?
And yes, i have ordered one, On or before 30/11/2005, cross fingers
Karl
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 27, 2005, 05:01 PM
 
Originally Posted by alexkan
So Darcy, I was kind of hoping that you would give SETI a try for a bit. I'm pretty sure that your machine could easily hit 1000 RAC, and likely more, if you ran it 24/7. (Incidentally, this would land you a pretty good spot on SETI's Top Computers list.) You seem like a RC5-72 kind of guy, though, so I don't know if I could turn you away from that. Also, any luck with those Shark traces?
I did some 24/7 SETI some years back and had a fun time. I don't mind switching. I'll do just that.

I sent you the Shark data last night at 07:47:30 PM CDT to your berkeley account. You haven't received it? It was an 8 meg (5.9 megs un-mimed) email because of the size of the data file. Do you suppose it got rejected? Or have you received it by now?

For conversation's sake, the timing while doing a Shark capture was:
Code:
<wu_cpu_time>3090.702161</wu_cpu_time> real 38m0.907s user 37m29.592s sys 0m20.616s
     
alexkan  (op)
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status: Offline
Oct 27, 2005, 05:16 PM
 
Originally Posted by darcybaston
I did some 24/7 SETI some years back and had a fun time. I don't mind switching. I'll do just that.

I sent you the Shark data last night at 07:47:30 PM CDT to your berkeley account. You haven't received it? It was an 8 meg (5.9 megs un-mimed) email because of the size of the data file. Do you suppose it got rejected? Or have you received it by now?

For conversation's sake, the timing while doing a Shark capture was:
Code:
<wu_cpu_time>3090.702161</wu_cpu_time> real 38m0.907s user 37m29.592s sys 0m20.616s
Oh, don't worry, I got it. (I probably should have sent an acknowledgment or something.) I've forwarded it to Rick, so we can figure out what steps we want to take next. Our code is still taking about 25-30% of the total run time (the rest is just FFTs, which are vDSP's job), so we might still be able to speed things up a bit if we're clever enough. If you get a chance, could you try timing two copies of the client running at the same time? We'd like to see how performance scales when you're utilizing both cores simultaneously. (You might have to make a copy of the directory.)

Nice to hear that you're joining the SETI team. Were you able to switch your SETI client to the optimized client in BOINC? If not, there should be plenty of instructions around the forums regarding how to do this. Do post a link to your computer stats so we can watch its RAC rise.

Also, Karl--I'm stoked about the prospect of seeing quad-G5 numbers. Let's hope that machine gets here sooner rather than later.
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 27, 2005, 07:50 PM
 
I ran two processes simultaneously, or no more than half a second apart; just enough time to click on another shell window and press enter on the prepped command.

1:
Code:
real 41m41.362s user 40m30.539s sys 0m28.482s 3359.437197
2:
Code:
real 42m4.284s user 40m52.837s sys 0m28.471s 3358.865312
     
alexkan  (op)
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status: Offline
Oct 27, 2005, 08:05 PM
 
Well, this looks promising, at least for those of you with the single dual-core G5s. At least from these statistics (~10% performance hit in processing time when going from one core to two), it looks like there's enough FSB bandwidth to keep both cores fed when doing SETI processing. Unfortunately, it's hard to extrapolate as to whether or not the quad-core G5s will have enough memory bandwidth to keep four cores fed. Only time (and Karl) will tell, in that regard. In any case, three work units per hour (or likely more, depending on how the reference unit stacks up to the average work unit) is nothing to scoff at.
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 28, 2005, 01:26 PM
 
Originally Posted by alexkan
Nice to hear that you're joining the SETI team. Were you able to switch your SETI client to the optimized client in BOINC? If not, there should be plenty of instructions around the forums regarding how to do this. Do post a link to your computer stats so we can watch its RAC rise.
Well, I created an account and downloaded the full client witih GUI and screen saver. I attached to the project and ran the screen saver all not long. I got to the computer the next day and launched the full client to look at stats or whatever, and everything was still at 0, and the gui client reported only being at 30%, which is where it was before I quit it the night before to let the screen saver version run.

I'm confused now. My stats page then showed 2 new processes, both at zero. There was only one of those in there before I ran the screen saver.

I tried putting the core you gave me Alex, into /library/application support/boinc/data/0 or whatever, and when I launched the boinc client, it kept emptying out the directory and replacing the files. So I tried to trick it by renaming the alpha5 core to the same filename as the default 4.x core, and then all the files disappeared from that directory. And I broke the client .

I'll reinstall when I get home tonight.
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 28, 2005, 01:33 PM
 
Ok, last message on this topic as it's off-topic. I found better instructions:
Make sure BOINC is not running.

Find MenuBar. Control-Click on it to bring up the little pop-up menu window and select Show Package Contents. Another window will open up with the Icon for the "folder" MenuBar. Open this, remove the "boinc" file and put it into the trash. Now rename the optimized client to "boinc", making sure it's all lower-case (and no quotes), and put it into the MenuBar folder in place of the other one you just took out. Close the window, and go to your Applications folder and double-click on MenuBar to start it up. All the other files for MenuBar, including your SETI and other clients should be in a folder yourusername/Library/Application Support/Boinc Data/

To use the optimized SETI client, first set your account preferences to .01 contacts per day to run your cache way down, then make sure BOINC is not running. Open the yourusername/Library/Application Support/Boinc Data/Projects/Seti/setiathome.berkeley.edu folder, remove the setiathome... file and trash it. Repace it with the optimized SETI client, and the app_info.xml file that comes with it, and close everything. Restart MenuBar.
     
alexkan  (op)
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status: Offline
Oct 28, 2005, 07:45 PM
 
Originally Posted by darcybaston
Ok, last message on this topic as it's off-topic. I found better instructions:
If your BOINC/SETI installation is working now, can you post a link to your computer's stats?
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 28, 2005, 09:41 PM
 
Still not working. I'll have to use whatever default installation is available. I can't get any alphas to work. I even tried the commandline and ended up having to kill 5 instances of 'boinc' before a seti@home4.x thread would go away. It kept getting respawned.
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 28, 2005, 09:42 PM
 
I'll start doing whatever's the default early tomorrow. I'd like to use my computer for gaming tonight.
     
Anon
Guest
Status:
Oct 29, 2005, 03:46 AM
 
     
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Oct 29, 2005, 09:16 AM
 
Darcy: since you seem to have trouble getting SETI/BOINC to work, I thought I'd provide you (and anyone else interested) with step-by-step instructions on how to get Alex and Rick's optimized SETI workers to run on your machine. And trust me, it's absolutely worth it! The optimized clients are as much as three times faster than the stock one!!!

Follow these steps, and you won't look back:

(Please note that these instructions are particularly geared towards BOINC Menubar. They should also largely apply to BOINC Manager, but since I don't use that version of the client, I don't really know. If anyone has any corrections, please feel free to post them! Setting-up the command line version of BOINC is a different ball of wax alltogether...)


1.) If you haven't done so already, register for SETI/BOINC here:
http://setiathome.berkeley.edu/create_account_form.php
Be sure to keep your registration info! I would recommend saving it in some form (text, web archive or pdf) instead of printing it, since you will have to enter your rather lengthy account key on every machine you want to run SETI/BOINC on.


2.) Delete everything relating to previous BOINC-installs from your system, particularly the following directory:
~/Library/Application Support/BOINC Data/
Also delete anything that starts with edu.berkeley.boinc... in ~/Library/Preferences

Note: I'm recommending this only because you've mentioned that you had problems with previous installation attempts. Newcomers can safely ignore this step.


3.) Download BOINC client software. I recommend you use BOINC Menubar, since this version of the client uses less cycles than BOINC Manager (full graphic client), and it works just as well. But use whatever suits you best. You can download different optimized versions of both BONC Menubar and BOINC Manager here:
http://members.dslextreme.com/~reade...eam/boinc.html
Use the superbench versions if you're planning to crunch 24/7 or near there. Otherwise, use the standard version. If you use the superbench client and don't crunch nearly full time, your machine will over-commit, and you'll likely end up with expiring work units.


4.) Uncompress and install your BOINC software of choice. I would recommend installing it in either /Applications or /Applications/Utilities. Once done, start up your BOINC client.


5.) Add SETI as a project, using the registration information you obtained as the result of step 1. When using BONC Menubar, use the menu option 'Manage Projects...' to get started. The rest is pretty much self-explanatory.


6.) Start computation on your BOINC client (that's not the same as launching the BOINC client). In BOINC Menubar, that's done by clicking on the 'Start' menu item.


7.) As soon as you have done so, open the log window of your BOINC client (with BOINC Menubar, you can have the log window open all the time while starting and stopping computation, etc.), and watch what's being written to the log. Wait until you see something that looks similar to this in the log:
2005-10-29 16:05:06 [SETI@home] Started download of 23my04aa.20947.17553.129826.155
2005-10-29 16:05:12 [SETI@home] Finished download of 23my04aa.20947.17553.129826.155
2005-10-29 16:05:12 [SETI@home] Throughput 67271 bytes/sec
What's happening here is that a work unit has been requested and successfully downloaded. The name/numbers of your work unit will be different, but you'll get the idea, I'm sure...


8.) Immediately stop computation or quit your BOINC client once you see the above lines in your log! Thist will prevent you from starting computation on the downloaded work unit with the stock worker.


9.) Download Alex and Rick's optimized SETI-worker for Tiger/G5 here:
http://inst.eecs.berkeley.edu/~alexk...home-G5-a5.zip
This is Alpha-5, the latest version as of this writing. Others running Tiger on G4s should use this worker:
http://inst.eecs.berkeley.edu/~alexk...home-G4-a5.zip
Other (currently older) versions of their optimized workers can be obtained here:
http://writhe.org.uk/seti@home/

10.) Once you've downloaded it, decompress it and copy the contents of the resulting directory (two files: seti@home-G5-a5 and app_info.xml) into this directory:
~/Library/Application Support/BOINC Data/projects/setiathome.berkeley.edu/

11.) From that same directory, remove the stock SETI worker by deleting the following file:
setiathome_4.18_powerpc-apple-darwin

12.) Start your BOINC client again, and you're good to go!


Hope this helps! Happy crunching!

Ron

(To Alex & Rick: feel free to use/alter these instructions and stick them on your site!)
( Last edited by halimedia; Oct 29, 2005 at 02:23 PM. )
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 29, 2005, 10:00 AM
 
I don't think step 5 and 6 went well. I may not have stopped the client in time. The log window filled up with this before I could click on 'stop':
Code:
2005-10-29 07:54:17 [---] Starting BOINC client version 4.44 for powerpc-apple-darwin 2005-10-29 07:54:17 [---] Data directory: /Library/Application Support/BOINC Data 2005-10-29 07:54:17 [http://setiathome.berkeley.edu/] Computer ID: not assigned yet; location: ; project prefs: default 2005-10-29 07:54:17 [---] No general preferences found - using BOINC defaults 2005-10-29 07:54:19 [---] Running CPU benchmarks 2005-10-29 07:55:17 [---] Benchmark results: 2005-10-29 07:55:17 [---] Number of CPUs: 2 2005-10-29 07:55:17 [---] 1819 double precision MIPS (Whetstone) per CPU 2005-10-29 07:55:17 [---] 4721 integer MIPS (Dhrystone) per CPU 2005-10-29 07:55:17 [---] Finished CPU benchmarks 2005-10-29 07:55:17 [---] Resuming computation and network activity 2005-10-29 07:55:17 [---] Insufficient work; requesting more 2005-10-29 07:55:19 [http://setiathome.berkeley.edu/] Requesting 8640.00 seconds of work 2005-10-29 07:55:19 [http://setiathome.berkeley.edu/] Sending scheduler request to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi 2005-10-29 07:55:20 [http://setiathome.berkeley.edu/] Scheduler request to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi succeeded 2005-10-29 07:55:20 [SETI@home] New host venue: home 2005-10-29 07:55:20 [SETI@home] Deferring communication with project for 10 minutes and 5 seconds 2005-10-29 07:55:20 [---] schedule_cpus: must schedule 2005-10-29 07:55:20 [---] Received signal 15 2005-10-29 07:55:20 [---] Exit requested by user 2005-10-29 07:55:20 [---] request_reschedule_cpus: exit_tasks
The benchmarks are probably lower since I had iTunes playing music.

Anyway, after that, I noticed the client was already stopped. So, I sent into ~/library/application support and found NO "Boinc" related folders.

However, off the root of the hard drive in /library/application support there IS a Boinc Data folder. Is that what you meant? I'm going to try to follow the rest of your steps using that folder.

*** update: Step 10 can't be completed as there was no client in folder setiathome.berkeley.edu. It was completely empty.

*** update 2: I copied alpha 5 and its xml file into that folder just the same, and started the client. The client then made the 4.x client your instructions asked me to remove appear. So I stopped the client and deleted it and its progress file to try to complete step 10. But each time I run the client, the 4.x executable shows up in there. It's as if it keeps "repairing" itself. I think these line in the log shows what keeps happening:
Code:
2005-10-29 08:03:55 [SETI@home] Started download of setiathome_4.18_powerpc-apple-darwin
( Last edited by darcybaston; Oct 29, 2005 at 10:07 AM. )
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 29, 2005, 10:10 AM
 
Where the heck is it getting the 4.4 core from? The domain name folder only has the alpha 5 and its XML in it, and yet when I start the client, I get this:
Code:
2005-10-29 08:09:41 [---] Starting BOINC client version 4.44 for powerpc-apple-darwin
So now there are different versions that run whenever they choose? LOL! 4.18, 4.44, what gives?
     
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Oct 29, 2005, 10:40 AM
 
Weird stuff going on!

You definitely have no BOINC Data folder in Users/yourname/Library/Application Support?

The 4.44 BOINC Client has nothing to do with SETI per se. There's an important distinction here to be aware of. While BOINC ist the framework that handles the whole distribution, tracking and consolidation of new, running and finished work units, the code that's doing the actual computation is separate from BOINC. In this case, it's the customized SETI worker you installed. I hope that makes sense...
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 29, 2005, 10:43 AM
 
That does make sense. Yeah, my /users/myname/library/application support was consistently devoid of any BOINC related folder.

I'm going to leave the app support folder open in list view, and repeat all the steps while watching it just to make sure a goblin doesn't run in and see as shiny and removeable.

I'm baffled by the difficulty I'm having. I've installed more complex things than this before!
     
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Oct 29, 2005, 10:48 AM
 
Thanks for the reply! I will now proceed with following my own instructions to the letter on a test machine, to see where things go wonky. I was going from memory, and I might have missed something. I'll let you know what's happening...
( Last edited by halimedia; Oct 30, 2005 at 08:13 AM. )
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 29, 2005, 11:07 AM
 
I "think" I got it finally.

I went through your steps again. Things that differed:

-I didn't get to stop the client when that cgi message appears, because it flies through to "request_reschedule_cpus: exit tasks" in the blink of an eye after the benchmarks are done. I can't click stop fast enough.
Code:
2005-10-29 08:55:15 [---] Starting BOINC client version 4.44 for powerpc-apple-darwin 2005-10-29 08:55:15 [---] Data directory: /Users/darcy/Library/Application Support/BOINC Data 2005-10-29 08:55:15 [http://setiathome.berkeley.edu/] Computer ID: not assigned yet; location: ; project prefs: default 2005-10-29 08:55:15 [---] No general preferences found - using BOINC defaults 2005-10-29 08:55:17 [---] Running CPU benchmarks 2005-10-29 08:56:15 [---] Benchmark results: 2005-10-29 08:56:15 [---] Number of CPUs: 2 2005-10-29 08:56:15 [---] 1821 double precision MIPS (Whetstone) per CPU 2005-10-29 08:56:15 [---] 4729 integer MIPS (Dhrystone) per CPU 2005-10-29 08:56:15 [---] Finished CPU benchmarks 2005-10-29 08:56:15 [---] Resuming computation and network activity 2005-10-29 08:56:15 [---] Insufficient work; requesting more 2005-10-29 08:56:17 [http://setiathome.berkeley.edu/] Requesting 8640.00 seconds of work 2005-10-29 08:56:17 [http://setiathome.berkeley.edu/] Sending scheduler request to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi 2005-10-29 08:56:18 [http://setiathome.berkeley.edu/] Scheduler request to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi succeeded 2005-10-29 08:56:18 [SETI@home] General preferences have been updated 2005-10-29 08:56:18 [SETI@home] New host venue: home 2005-10-29 08:56:18 [---] General prefs: from SETI@home (last modified 2005-10-29 08:50:50) 2005-10-29 08:56:18 [---] General prefs: using your defaults 2005-10-29 08:56:18 [SETI@home] Deferring communication with project for 10 minutes and 5 seconds 2005-10-29 08:56:18 [---] schedule_cpus: must schedule 2005-10-29 08:56:19 [SETI@home] Started download of setiathome_4.18_powerpc-apple-darwin 2005-10-29 08:56:19 [SETI@home] Started download of better_banner.jpg 2005-10-29 08:56:20 [SETI@home] Finished download of better_banner.jpg 2005-10-29 08:56:20 [SETI@home] Throughput 17015 bytes/sec 2005-10-29 08:56:20 [SETI@home] Started download of 23my04ab.17965.7777.997154.143 2005-10-29 08:56:20 [---] Received signal 15 2005-10-29 08:56:20 [---] Exit requested by user 2005-10-29 08:56:20 [---] request_reschedule_cpus: exit_tasks
I had both my application data folders open, one was my personal account and the second was off the root. THIS time my user account got the BOINC Data folder. Yay!

I then put the new set@home-G5-a5 file in the right folder, and deleted t he 4.18 named one.

I started the client. And here I did something different. The last time I did this, the 4.18 file showed up again, so I stopped the client. I shouldn't have. Because within a second or two, the 4.18 file that had reappeared, disappeared again, on its own. I must have halted that process. I really wasn't expecting all this "magic" as it wasn't mentioned anywhere.

Anyway, I think things are running fine now. The menu icon has changed from a B to a satelite image with a black speck in the bottom right corner of itself. Oh, actually, there are two satelite images up there now. What does that mean? My log window is displaying things like:
Code:
2005-10-29 09:06:28 [SETI@home] Scheduler request to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi succeeded 2005-10-29 09:06:28 [SETI@home] Computation for result 23my04ab.17965.7777.997154.143_0 finished 2005-10-29 09:06:28 [---] schedule_cpus: must schedule 2005-10-29 09:06:29 [SETI@home] Started upload of 23my04ab.17965.7777.997154.143_0_0 2005-10-29 09:06:29 [SETI@home] Started download of 14my04aa.8313.15568.290886.113 2005-10-29 09:06:32 [SETI@home] Finished download of 14my04aa.8313.15568.290886.113 2005-10-29 09:06:32 [SETI@home] Throughput 127799 bytes/sec 2005-10-29 09:06:32 [SETI@home] Started download of 13my04aa.5608.29617.98570.87 2005-10-29 09:06:32 [---] request_reschedule_cpus: files downloaded 2005-10-29 09:06:32 [---] schedule_cpus: must schedule 2005-10-29 09:06:32 [SETI@home] Starting result 14my04aa.8313.15568.290886.113_0 using setiathome version 4.18 2005-10-29 09:06:33 [SETI@home] Finished upload of 23my04ab.17965.7777.997154.143_0_0 2005-10-29 09:06:33 [SETI@home] Throughput 114166 bytes/sec 2005-10-29 09:06:35 [SETI@home] Finished download of 13my04aa.5608.29617.98570.87 2005-10-29 09:06:35 [SETI@home] Throughput 127051 bytes/sec 2005-10-29 09:06:35 [---] request_reschedule_cpus: files downloaded 2005-10-29 09:06:35 [---] schedule_cpus: must schedule 2005-10-29 09:06:35 [SETI@home] Starting result 13my04aa.5608.29617.98570.87_2 using setiathome version 4.18 2005-10-29 09:07:30 [SETI@home] Result 13my04aa.5608.29617.98570.87_2 exited with zero status but no 'finished' file 2005-10-29 09:07:30 [SETI@home] If this happens repeatedly you may need to reset the project. 2005-10-29 09:07:30 [---] request_reschedule_cpus: process exited 2005-10-29 09:07:30 [---] schedule_cpus: must schedule 2005-10-29 09:07:30 [SETI@home] Restarting result 13my04aa.5608.29617.98570.87_2 using setiathome version 4.18
     
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Oct 29, 2005, 11:38 AM
 
Great! I'm glad it's working! I worked through my instructions, and found a few boo-boos. I changed the following: the main thing that I had forgotten was that you first have to start computation in the BOINC cilent before the 4.18 worker is actually downloaded. Therefore, I also changed the piece of information to look for in the log, now being the download of the first work unit. That way your BOINC Data dir is definitely in place. Now the instructions should be correct, particularly for BOINC Menubar.

The satelite dish is BOINC Menubar's icon indicating that it's currently crunching a SETI unit. The little black dot to the left of it is a progress indicator that will grow vertically as you crunch through your work unit.

The reason you had your BOINC Data directory in /Library/Application Support rather than ~/Library/Application Support was undoubtedly you installing it using the root account. Incidentally, I would recommend against enabling root access on OS X client. Idon't know of any reason why it would need to be enabled ever since Panther came around. All you need to do can be done as an admin user, either in the Finder or in Terminal. If I were you, I would never log-in as root. Just puts wonky things in /Library that are not usually expected there.

You don't need to worry about the following line, as long as it doesn't appear all the time:

Result 13my04aa.5608.29617.98570.87_2 exited with zero status but no 'finished' file

It simply means that you started computing your work unit with the stock 4.18 worker, then resumed computation with the alpha-5 worker. BOINC doesn't like this, and therefore aborts computation on that work unit, starting with the next one instead. This will only happen this once, and you haven't lost much, since you've only been crunching it for a very short time.

OK, that's about that. I hope all is well now! You can check by opening Terminal and typing top -u followed by return. If something starting with seti@home- is at the top of your list of active processes, all is well.

Happy crunching! And don't forget to watch the top machine list at SETI/BOINC:

http://setiathome.berkeley.edu/top_hosts.php

You'll be climbing really fast if you keep BOINC running 24/7. And maybe another important nugget: you really can keep crunching all the time, since the scheduling priority of the SETI worker is very low. This means that just about any other process coming along requesting CPU time will get it at the expense of SETI. This works really well (thanks to the excellent MACH kernel architecture), and it's actually very well possible to run very processor intensive applications concurrently with SETI without any ill effect (e.g. Final Cut Pro, Motion, Soundtrack, etc.). That's common practice around my world, and there are no dropped frames or any other problems at all.

Happy crunching!

Ron
( Last edited by halimedia; Oct 29, 2005 at 02:58 PM. )
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 29, 2005, 11:55 AM
 
Originally Posted by halimedia
The reason you had your BOINC Data directory in /Library/Application Support rather than ~/Library/Application Support was undoubtedly you installing it using the root account.
I regret friend, but the root account is not enabled and I've never used it. I have no idea how the BOINC folder got into there, except either my tests with screen saver or full client acted funny.

This client IS as fast as the work unit test I think. I'm already at about 70% complete on each thread. I only had enough time to make toast and coffee and have a few bites and sips before it got there. Nice!
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 29, 2005, 11:56 AM
 
The URL for my stats:

http://setiweb.ssl.berkeley.edu/show...hostid=1585039

(updated url)
( Last edited by darcybaston; Oct 29, 2005 at 01:56 PM. Reason: updated url)
     
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Oct 29, 2005, 12:35 PM
 
Sorry, I assumed incorrectly that you had used BOINC Menubar in all your efforts to get BOINC running. I suppose it makes sense that a screen saver version or BOINC Manager in 'service' mode installs the BOINC Data directory in /Library, since BOINC can then keep crunching the same units regardless of what user is logged-in (or even when noone is logged in).

The URL you provided is that of your account, and only accessible by you. Could you post the stat URL of your G5 instead? I will look something like this:

http://setiweb.ssl.berkeley.edu/show...hostid=1116838

Incidentally, this is my dual 2.5 GHz G5, running BOINC Menubar superbench and the alpha-5 worker. Might be interesting to compare over time...

Cheers,

Ron
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 29, 2005, 01:57 PM
 
Ron, how come your claimed credits are in the 40s, and mine are in the teens? How is that calculated?
     
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Oct 29, 2005, 02:30 PM
 
I don't know the specifics. AFAIK, the claimed credit is calculated based on the local CPU benchmarks and the final computation time of each work unit. Possibly other parameters are involved as well; I don't know for sure. The superbench, standard optimized and stock clients will probably each claim different amounts of credit.

Which version of the BOINC client are you running?

The whole system of credit claiming/granting in SETI is a bit suspect to me, and I haven't found much explanation of the process anywhere, either. But what counts is that we crunch lots of units and that the optimized workers produce the same results as the stock one but muuuuuuch faster!

This thread at BOINC Synergy has some interesting insights on the topic (first post):

http://www.boincsynergy.com/forums/viewtopic.php?t=2307

HTH,

Ron
     
Gecko_r7
Guest
Status:
Oct 29, 2005, 06:22 PM
 
Darcy,

I may be mistaken and please forgive me if I am, but your benchmarks look lower than I'd expect them to be on 4.44 superbench menubar +Alpha 5 w/ the quad. Granted, you're crunching 4 WU at a time to my 1 and spitting them out a heck of alot quicker vs. my single 1.25 G4, but the benchmarks seem low. By the way, I'm averaging @ 5350 secs per WU (short WUs removed) w/ 4.44 menubar+ Alpha 5. Don't get me wrong, your Quad absolutely flies and I'm already working on my sell-in story for the Mrs!!! I suspect there's still more performance to be gained in fine-tuning your optimizing. Only a perception. It will be fun to watch your system get up to full potential. Your RAC should be amazing!
Welcome aboard.

http://setiweb.ssl.berkeley.edu/show...hostid=1162773

CPU type Power Macintosh PowerMac3,6
Number of CPUs 1
Operating System Darwin 8.2.0
Memory 1536 MB
Cache 976.56 KB
Swap space 0 MB
Measured floating point speed 1855.04 million ops/sec
Measured integer speed 5222.23 million ops/sec
     
Dedicated MacNNer
Join Date: Nov 2004
Location: Virginia
Status: Offline
Oct 29, 2005, 06:33 PM
 
Darcy:
I suspect your claimed credits are so low because you installed the "regular" optimized BOINC 4.44, instead of the "superbench" optimized BOINC 4.44. You can easily replace the "regular" one by going back and downloading the "superbench".

Quit BOINC (both CPU's). Then find the MenuBar application you already installed, drag it to the trash, then put the "superbench" version in the folder where the first one was located. That's all you have to do. To restart everything, just double-click on the MenuBar application.

beadman
     
Junior Member
Join Date: Oct 2004
Location: Germany
Status: Offline
Oct 29, 2005, 06:38 PM
 
Gecko_r7, he didn't say anything about a quad or i missed the connection ...

Karl
     
Junior Member
Join Date: Jul 2005
Location: USA, Virginia
Status: Offline
Oct 29, 2005, 07:15 PM
 
Originally Posted by beadman
Darcy:
I suspect your claimed credits are so low because you installed the "regular" optimized BOINC 4.44, instead of the "superbench" optimized BOINC 4.44. You can easily replace the "regular" one by going back and downloading the "superbench".

Quit BOINC (both CPU's). Then find the MenuBar application you already installed, drag it to the trash, then put the "superbench" version in the folder where the first one was located. That's all you have to do. To restart everything, just double-click on the MenuBar application.

beadman

All true except one point. Because the BOINC version number will not change you will have to rerun the benchmarks manually. Unless the version changes BOINC will not re-benchmark and as such you will not claim the proper credit. After you re-benchmark manually this will fix it.

Please try to stay in the left lane at all times so you do not frignten the Windoze folks that are in the slow lanes on your way to the top.

Regards
Phil
We must seek intelligent life on other planets as it is increasingly apparent we will not find any on our own.

Link: http://www.boincsynergy.com/images/stats/comb-2033.jpg
     
Gecko_r7
Guest
Status:
Oct 29, 2005, 07:55 PM
 
Originally Posted by Karl Schimanek
Gecko_r7, he didn't say anything about a quad or i missed the connection ...

Karl
Karl, you're correct. My apologies. I misread from one of the earlier comments.
Indeed, it's a dual core 2.0GHZ. Sorry.
We'll, that certainly explains part of my question.
Curious to see w/ 1MB of L2 each core, how it compares to previous gen 512KB L2 2.5 dual once it's sorted out.
Regards.
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 29, 2005, 10:01 PM
 
I was using the regular client. It was explained on the download page that the regular client was better for multitasking, and I wanted that so I could still play games like Halo while it was running. I did play Halo today and didn't notice any slowdown, because it relinquished one core of the cpu to the game.

I'll try the superbench one and see what happens.

** addendum: I use this computer for all kinds of things from game playing, to media conversion to software development. Please don't expect my scores to consistently reflect "what a 2GHz core can REALLY do when all it's doing is SETI". I will have the client turned off from time to time. ***
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 30, 2005, 12:02 AM
 
Ok, I've got the superbench going. I'm looking at my computer stats trying to look for any differences. The time in seconds is longer but the credit count is higher. Is that good news? The computer's been dedicated to the task during this new superbench phase.
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 30, 2005, 12:05 AM
 
With the superbench, my benchmark speed was:
Code:
2005-10-29 20:07:36 [---] Benchmark results: 2005-10-29 20:07:36 [---] Number of CPUs: 2 2005-10-29 20:07:36 [---] 3641 double precision MIPS (Whetstone) per CPU 2005-10-29 20:07:36 [---] 9562 integer MIPS (Dhrystone) per CPU
I wasn't playing any music on iTunes this time around either.
     
Dedicated MacNNer
Join Date: Nov 2004
Location: Virginia
Status: Offline
Oct 30, 2005, 01:09 AM
 
Looks pretty good, Darcy - your claimed credit has gone from around 16 per WU to around 32.

As you experiment, you may find you can leave BOINC running all the time - I frequently have a lot of apps open on my PowerBook, and I haven't noticed any slowdowns at all. The only times I turn off BOINC is when I have to run Disk Utility. BTW, if you haven't done so, you really ought to go to your Applications/Utilities folder and run Disk Utility to repair disk permissions each time you install some new software (one of Apple's recommendations). It takes a while, depending on how many files you have on your machine, but everything runs smoother and faster after repairing the permissions.

beadman
     
Junior Member
Join Date: Jul 2005
Location: USA, Virginia
Status: Offline
Oct 30, 2005, 01:51 AM
 
Originally Posted by beadman
Looks pretty good, Darcy - your claimed credit has gone from around 16 per WU to around 32.

As you experiment, you may find you can leave BOINC running all the time - I frequently have a lot of apps open on my PowerBook, and I haven't noticed any slowdowns at all. The only times I turn off BOINC is when I have to run Disk Utility. BTW, if you haven't done so, you really ought to go to your Applications/Utilities folder and run Disk Utility to repair disk permissions each time you install some new software (one of Apple's recommendations). It takes a while, depending on how many files you have on your machine, but everything runs smoother and faster after repairing the permissions.

beadman
As a matter of fact I use my machines for all kind of stuff. I leave the BOINC projects running all the time unless I am digitizing video or rendering a time line. he machine crunches all the time (while my wife plays games, e-mail, posting to these lists, browsing) the BOINC system just keeps chucking. If you try something and the system bogs down, turn off BOINC while you do that function, and start it up after. On some apps you will pay a time penalty for turning it off at the wrong time, but with SETI that is not a real issue.

And yes the function of the Superbench BOINC is to give a proper reading of your systems actual speed, and provide credit claims accordingly. So the improvements in that area are what you want to see.

Regards
Phil
We must seek intelligent life on other planets as it is increasingly apparent we will not find any on our own.

Link: http://www.boincsynergy.com/images/stats/comb-2033.jpg
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 30, 2005, 10:10 AM
 
Thanks for the wise words. One point of contention is that my system is swapping from virtual ram TONS more than it was before with this SETI stuff running. I only have 512MB and OS X can get sluggish already if I open Photoshop Elements and GoLive together. With SETI it makes the computer useage experience feels quite similar to my iBook 933MHz days.

I'll have to bump this system up to a gig or something to get the comfort back.
     
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Oct 30, 2005, 11:03 AM
 
Definitely! 512 MB is marginal for a Panther/Tiger-box doing heavy lifting. More than 2.5 GB is overkill, however (unless you have apps/processes that can adress 64-bit memory space).

BTW, I consider MenuMeters highly valuable when trying to get a feel of resource usage of an OS X box in different situations. It can display CPU usage, network traffic, disk access and memory usage in a customizable fashion right in the menu bar. It has a very low memory and CPU footprint (much lower than Activity Monitor or top/BigTop), has never interfered with anything in my useage over the last two years, and - maybe best of all - it's free!

More information here: http://www.versiontracker.com/dyn/moreinfo/macosx/17713

Disclaimer: I have no affiliation with Raging Menace or VersionTracker. I'm just a very happy user!
( Last edited by halimedia; Oct 30, 2005 at 11:22 AM. )
     
Dedicated MacNNer
Join Date: Nov 2004
Location: Virginia
Status: Offline
Oct 30, 2005, 11:16 AM
 
Only 512MB? I agree that's probably the cause of the slow down. My PowerBook (work machine) has a single processor and 1 GB of RAM; my iBook has a single with 768MB. Your dual processor is probably "memory starved". One thing you can do to see how things are going is to go to Applications/Utilities folder and drag the Activity Monitor app to your Dock. Launch the app, then close the monitor (without quitting it) and the icon in the dock will display your usage - green is user (your open apps), red is system, and blue is BOINC. Since it does steal CPU seconds to run AM, I normally only open it if I want to monitor something - otherwise, I quit it and the Dock icon shows shows an icon that does nothing. I suspect you'll see the red (system) being fairly large.

By the way, I have noticed every few days that MenuBar unexplicably seems to stop doing anything. I look in AM and see no blue, then open the AM fully and see that SETI, Einstein, and MenuBar are all present, but doing nothing. I have to force quit all three (using the stop sign button in AM) and then restart MenuBar (or just reboot the system). Based on log entries, it appears that this happens when swiching between SETI and Einstein... almost as if BOINC loses track of what it's doing and then just sits there. This may be a part of the "reporting of short times" problem discussed extensively in another thread.

beadman
     
Grizzled Veteran
Join Date: May 2000
Location: ON, Canada
Status: Offline
Oct 30, 2005, 12:30 PM
 
I often use activity monitor and already have menu meters installed. Thanks!
     
alexkan  (op)
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status: Offline
Oct 30, 2005, 02:29 PM
 
Originally Posted by darcybaston
I often use activity monitor and already have menu meters installed. Thanks!
Interesting...your work unit times don't seem to reflect the performance increase relative to the old G5s, at least from what I can tell. Older G5s tend to report (in init_data.xml) between 4000-4500 seconds per work unit as well, and your machine was reporting 3350 or so on the reference work unit when you were running it twice. Perhaps I should have asked you to run two different workunits simultaneously.

Maybe it's because I'm starting to feel the squeeze from the x86 clients (after all, Altivec and number of cores can only make up for so much for a deficit in raw clock speed), so forgive me for asking what may seem like an obvious question. Changing to the superbench didn't revert the clients, did it? Also, how many instances of SETI are running? (It might be worth running only one at a time just for a little while to see if we really are having memory bandwidth problems with feeding both cores with a single bus.)
     
Forum Regular
Join Date: Oct 2005
Location: Las Vegas, NV
Status: Offline
Oct 30, 2005, 02:48 PM
 
Originally Posted by alexkan
Interesting...your work unit times don't seem to reflect the performance increase relative to the old G5s, at least from what I can tell. Older G5s tend to report (in init_data.xml) between 4000-4500 seconds per work unit as well, and your machine was reporting 3350 or so on the reference work unit when you were running it twice. Perhaps I should have asked you to run two different workunits simultaneously.

Maybe it's because I'm starting to feel the squeeze from the x86 clients (after all, Altivec and number of cores can only make up for so much for a deficit in raw clock speed), so forgive me for asking what may seem like an obvious question. Changing to the superbench didn't revert the clients, did it? Also, how many instances of SETI are running? (It might be worth running only one at a time just for a little while to see if we really are having memory bandwidth problems with feeding both cores with a single bus.)
Hi Alex,
Have you been following this thread the past few days?

http://setiathome.ssl.berkeley.edu/f...d.php?id=21660
     
Junior Member
Join Date: Jul 2005
Location: USA, Virginia
Status: Offline
Oct 30, 2005, 03:03 PM
 
Originally Posted by darcybaston
Thanks for the wise words. One point of contention is that my system is swapping from virtual ram TONS more than it was before with this SETI stuff running. I only have 512MB and OS X can get sluggish already if I open Photoshop Elements and GoLive together. With SETI it makes the computer useage experience feels quite similar to my iBook 933MHz days.

I'll have to bump this system up to a gig or something to get the comfort back.
You definatly need more memory. Even if you just added a single 512 module, you would see a dramatic improvement for the applications you described. With only 512 the system will have to page for almost anything it does. While it looks at paged memory as just more memory, the system will run a lot faster without all the page faults. Remember OS 10 is a big girl, and she has an appitite for memory too. With two processers you really need more.

Look at it this way, speed of light in a copper wire verses the speed of a read cycle on a hard drive, moderated by the bus the drive is on. The more memory you can add, the better, up to about 2 GB. after that the applications would have to be written to to take advantage of it for it to be useful.

Regards
Phil
We must seek intelligent life on other planets as it is increasingly apparent we will not find any on our own.

Link: http://www.boincsynergy.com/images/stats/comb-2033.jpg
     
alexkan  (op)
Forum Regular
Join Date: Aug 2005
Location: Cupertino, CA
Status: Offline
Oct 30, 2005, 03:37 PM
 
Originally Posted by Gecko_r7
Hi Alex,
Have you been following this thread the past few days?

http://setiathome.ssl.berkeley.edu/f...d.php?id=21660
Yes, of course I have. I follow other clients pretty closely--in fact, Howard Naparst and I have exchanged a couple emails on occasion. Not all of their tweaks help us--for one thing, caching helps them and not us because their source is based off a different version of the SETI worker, and also, we're generally better off calculating things instead of loading them from memory because G4s have such slow memory and G5s have high memory latency...except for the kind of caching in t_funct that they did, which we were actually doing since late August.

Rest assured, though, that we won't take this sort of encroachment lying down.
     
Dedicated MacNNer
Join Date: Oct 2005
Location: Switzerland
Status: Offline
Oct 30, 2005, 04:00 PM
 
At a going average rate of approx. 4200 s/wu, Darcy's box is just about exactly as fast as my dual-processor 2.5 GHz G5, and his is 'only' a dual-core 2.0 GHz. That's still an extrapolated performance increase of 20 % compared to the previous G5 generation! Nothing to sneeze at, I'd say!

Alex, if you like you can send me the ref wu and I'll crunch it real quick for comparison...

Happy crunching!

Ron
     
Forum Regular
Join Date: Oct 2005
Location: Las Vegas, NV
Status: Offline
Oct 30, 2005, 06:49 PM
 
Originally Posted by alexkan
Yes, of course I have. I follow other clients pretty closely--in fact, Howard Naparst and I have exchanged a couple emails on occasion. Not all of their tweaks help us--for one thing, caching helps them and not us because their source is based off a different version of the SETI worker, and also, we're generally better off calculating things instead of loading them from memory because G4s have such slow memory and G5s have high memory latency...except for the kind of caching in t_funct that they did, which we were actually doing since late August.

Rest assured, though, that we won't take this sort of encroachment lying down.
You, Rick and everyone else testing & debugging have done one heck of a great job! I've been crunching Seti on one machine or another since 99' (started w/P2-200mmx) and have certainly had the most fun since "goin' back to Mac" and especially since running your improvements these last few months. Pretty proud of what you guys have been able to wring-out of my humble G4....Lookin forward to given' em' heck w/ Alpha-6. Keep up the great work and Thanks again!
     
 
Thread Tools
Forum Links
Forum Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On
Top
Privacy Policy
All times are GMT -4. The time now is 07:34 PM.
All contents of these forums © 1995-2015 MacNN. All rights reserved.
Branding + Design: www.gesamtbild.com
vBulletin v.3.8.8 © 2000-2015, Jelsoft Enterprises Ltd., Content Relevant URLs by vBSEO 3.3.2