GNU/Linux S5R3 "power users" App 4.27 available

Conan
Conan
Joined: 19 Jun 05
Posts: 172
Credit: 7099171
RAC: 2794

RE: RE: This App

Message 77429 in response to message 77391

Quote:
Quote:

This App incorporates SSE vector code, but doesn't have a CPU feature detection. It will badly crash on non-SSE machines.

Other calculations are done in x87 FPU math, including the linear SIN/COS approximation, which should make it run (and run fast!) on older CPUs (i.e. PIII).

Thank you, thank you, thank you!!!

As the owner of a large number of SSE capable PIII tualatins and Athlon XPs which are running Linux, I'll be in hog's heaven with this app. I've just converted a tualatin now which was running 4.14. It has restarted with 4.27 after I added some lines to app_info.xml to allow the old 4.14 version to be handled by 4.27. There were no problems with the transition but it'll be a while before the new speed becomes apparent.

As a general warning to unsuspecting users who might be tempted to use this app, please be aware that if you are currently running a version of the app that is older than 4.20 (eg 4.14 like me) and want to switch midstream to 4.27, you will need to make a small addition to app_info.xml to make the transition happen smoothly.

Otherwise any tasks in your cache that are "branded" with a version older than 4.20 will be trashed.

You are so right Gary, I forgot to edit app_info (I was also using 4.14) and it killed two that were running and 6 that had yet to start, no idea where they went.
New 4.27 WU's downloaded ok so we will see how they run.
I have only done on one computer to see the results, if ok will add the other two Linux machines as well.

It's funny how the Windows 4.26 app did not kill the older WU's but that maybe because I was running the standard app not another Beta version.

Never mind, you live and learn, and because I forget, I then learn again.

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282700
RAC: 0

RE: It's funny how the

Message 77430 in response to message 77429

Quote:

It's funny how the Windows 4.26 app did not kill the older WU's but that maybe because I was running the standard app not another Beta version.

I went from 4.25 to 4.26... As I said in that thread, "no pain no gain". I only had one result that I could've trashed and it was only 4% into it...

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109384656135
RAC: 35939664

RE: RE: As a general

Message 77431 in response to message 77429

Quote:
Quote:


As a general warning to unsuspecting users who might be tempted to use this app, please be aware that if you are currently running a version of the app that is older than 4.20 (eg 4.14 like me) and want to switch midstream to 4.27, you will need to make a small addition to app_info.xml to make the transition happen smoothly.

Otherwise any tasks in your cache that are "branded" with a version older than 4.20 will be trashed.

You are so right Gary, I forgot to edit app_info (I was also using 4.14) and it killed two that were running and 6 that had yet to start, no idea where they went.

Unfortunately, they were just unceremoniously dumped :(.
If an app version is not included in the app_info.xml file then any tasks "branded" with that particular version (eg 4.14) cannot be processed (even if already started).

I decided to post the previous warning because I felt sure there would be people like me still running the 4.14 version as later versions didn't seem to solve the problems and were slower as well. 4.14 was working fine for me, so I was just waiting for a new app that did solve the issues and was likely to be faster. 4.27 seems to be a winner on both counts.

So here is a more explicit repeat of the warning for any person currently running a Linux beta older than 4.20. Take a look inside the new app_info.xml file and notice the several blocks of code between ... tags that allow earlier version data to be crunched by the latest 4.27 version app. You will need to add a similar block of code for the 4.14 -> 4.27 transition. Here is a copy of that added block of code:-


einstein_S5R3
414
6.1.0

einstein_S5R3_4.27_i686-pc-linux-gnu



einstein_S5R3_4.27_graphics_i686-pc-linux-gnu
graphics_app

Quote:

It's funny how the Windows 4.26 app did not kill the older WU's but that maybe because I was running the standard app not another Beta version.

That's because the old windows 4.15 beta (which had been made official anyway) was already included in the app_info.xml file. So no changes were needed there.

If anybody is still puzzled by how the app_info.xml file works, here is a message I posted some time ago which explained things, piece by piece. The examples refer to much older Windows beta apps and don't include the newer tags like and but the version handling mechanism is what is important and that is explained.

Please ask if you are still puzzled.

Cheers,
Gary.

Knorr
Knorr
Joined: 18 Feb 06
Posts: 16
Credit: 3129905
RAC: 0

RE: RE: RE: I'm running

Message 77432 in response to message 77428

Quote:
Quote:
Quote:

I'm running Ubuntu 8.04 64 bit with the 5.10.30 boinc-client.

The "Show Graphics" button in boinc manager doesn't produce any output.
But if I open a terminal and run the einstein_S5R3_4.27_graphics_i686-pc-linux-gnu program with a workunit from slots/0 as parameter I get a perfect graphics window.

From stderr:

No protocol specified
GLUT: Fatal Error in BOINC: could not open display: :1.0

http://einsteinathome.org/task/90982448


Open a terminal and type "echo $DISPLAY". If you get something different than ":1" or ":1.0", terminate the Manager and the client ("killall boinc") and run "BOINC/run_manager" again. Having a DISPLAY :1 is slightly unusual if you have only one screen and are not logged in remotely.

BM

The terminal echoed :1.0
The graphics on this system is a bit odd sometimes, but the fact that I can manually run the graphics app is strange I think.

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494410
RAC: 0

Well, my latest results do

Well, my latest results do certainly seem to have been faster, at least that's my impression... hard to give exact numbers due to runtime variation but if I had to guess I'd say roughly 10% on my Core CPU...
Oh, I noticed that most of the time my box now covers two sky positions between checkpoints instead of just one. Has the checkpointing intervall changed or does this also indicate a speed increase?

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 686042851
RAC: 592401

RE: Well, my latest results

Message 77434 in response to message 77433

Quote:
Well, my latest results do certainly seem to have been faster, at least that's my impression... hard to give exact numbers due to runtime variation but if I had to guess I'd say roughly 10% on my Core CPU...
Oh, I noticed that most of the time my box now covers two sky positions between checkpoints instead of just one. Has the checkpointing intervall changed or does this also indicate a speed increase?

Tht would indeed indicate a speed increase!

CU
H-B

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494410
RAC: 0

I just finished and uploaded

I just finished and uploaded another WU and it gave the same impression: A runtime much at the lower end of what I'm used to and again the larger intervall between checkpoints... I mean I know this is not scientific or anything ;-) but I'd definitely say there is a noticeable speed increase on this CPU aswell, not only Core 2 or so.

Wedge009
Wedge009
Joined: 5 Mar 05
Posts: 117
Credit: 15649833783
RAC: 7422122

First 4.27 application WU

First 4.27 application WU completed. Even taking the completion time variation into account, there seems to be a substantial speed increase against the 4.24 application on my Barton Athlon XP. Thanks for the work involved in optimising the application.

Soli Deo Gloria

Conan
Conan
Joined: 19 Jun 05
Posts: 172
Credit: 7099171
RAC: 2794

Can't say I have noticed much

Can't say I have noticed much in the way of a speed increase. My first did go under 30,000 seconds for the first time but the next two are back the same as before over 41,000 seconds. So after 3 results it is no faster on an AMD Opteron 285 running Fedora Core 3.
I moved from 4.14 to 4.27 so from beta to beta it is the same.

Particularly as I have another AMD Opteron 285 also running Linux but Fedora Core 6, and it gets 33,000 to 36,000 all the time on app 4.14.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109384656135
RAC: 35939664

RE: Can't say I have

Message 77438 in response to message 77437

Quote:
Can't say I have noticed much in the way of a speed increase.

That's because you are not allowing for the cyclic nature of crunch times and haven't noticed the particular significance of the sequence numbers you have been assigned

Quote:
My first did go under 30,000 seconds for the first time

Because it has a sequence number right near the trough in the cycle.

Quote:
but the next two are back the same as before over 41,000 seconds.

Actually the next three if you check again. They have sequence numbers of 224, 225, 234. Your frequency is 745.20 which means that the period of the cycle is 114.4. There is a cycle peak therefore at 229 so you can see that by pure bad luck your three 41K results bracket the peak very closely. Each peak of the cycle represents the slowest possible crunch time for your computer.

Quote:
So after 3 results it is no faster on an AMD Opteron 285 running Fedora Core 3.

You can't really say that unless you do a proper analysis. Go find Mike Hewson's marvelous Ready Reckoner if you want an easy way to do the analysis. Try reading this thread for some background and then look for Mike's posts.

Cheers,
Gary.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.