Lunatic Linux Rev 2 apps?

Message boards : Number crunching : Lunatic Linux Rev 2 apps?
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · 6 · Next

AuthorMessage
peristalsis

Send message
Joined: 23 Jul 99
Posts: 154
Credit: 28,610,163
RAC: 51
United States
Message 539977 - Posted: 2 Apr 2007, 13:06:06 UTC
Last modified: 2 Apr 2007, 13:08:12 UTC

I'm pretty close to being a linux newbie (i.e. confused) mainly to my not devoting enough of my negligible free time to learning it :(
ANyway, I tried Simon and Teams newest Ap. Borked something up. As soon as I fired BOINC up after installing the two files, it started trashing WUs with signal 11 errors. Lost about a dozen as I shot for the close button. Had been running the previous optimized ap. Replaced the new files with the old and back to crunching.
Running a fairly current version of PCLinixOS. Haven't the slightest idea of which kernel. Going through the thread I see mention of things like "lib-c" (going on my faulty memory) but I'm lost. Anyone willing to help an old codger with some hand holding on this ? TIA...j
ID: 539977 · Report as offensive
Profile michael37
Avatar

Send message
Joined: 23 Jul 99
Posts: 311
Credit: 6,955,447
RAC: 0
United States
Message 540048 - Posted: 2 Apr 2007, 17:10:09 UTC - in response to Message 539977.  

I'm pretty close to being a linux newbie (i.e. confused) mainly to my not devoting enough of my negligible free time to learning it :(
ANyway, I tried Simon and Teams newest Ap. Borked something up. As soon as I fired BOINC up after installing the two files, it started trashing WUs with signal 11 errors. Lost about a dozen as I shot for the close button. Had been running the previous optimized ap. Replaced the new files with the old and back to crunching.
Running a fairly current version of PCLinixOS. Haven't the slightest idea of which kernel. Going through the thread I see mention of things like "lib-c" (going on my faulty memory) but I'm lost. Anyone willing to help an old codger with some hand holding on this ? TIA...j


Linux apps are a bit in a state of flux. The initial release just happened, and the development is polishing up the rough corners. It's not as smooth as Windows apps which have been out for a month and have a nice installer.

To learn which kernel you are running, type 'uname -a'
To learn which processors you have, type 'cat /proc/cpuinfo'

You shouldn't really care which glibc you are using.

ID: 540048 · Report as offensive
Profile KWSN - Chicken of Angnor
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 9 Jul 99
Posts: 1199
Credit: 6,615,780
RAC: 0
Austria
Message 540130 - Posted: 2 Apr 2007, 21:11:22 UTC

Hi,

I just uploaded new compiles that should hopefully solve a lot of incompatibility options.

There are now 2 versions of each app included - one for Kernel 2.4.xx, one for Kernel 2.6.xx.

You can find out which Kernel version you're running - simply type this on a shell/xterm:

uname -r

After downloading the new app of your choice, unpack it ("tar xfvj archive.tar.bz2").

If you want, you can check out Instructions.txt inside the folder it creates; it tells you what I posted above, basically.

To install the app, stop BOINC, then copy the two files inside Files-to-install/Kernel-2.X (where X is either 4 or 6, depending on what "uname -r" said) to BOINC/projects/setiathome.berkeley.edu/.

Before you restart BOINC, you should check whether the app works on your system. This is a bit of a bother, but it'll save your WUs.

You can do this in the BOINC/projects/setiathome.berkeley.edu directory itself, but be sure to clean up again afterwards - preferrably, use an empty folder somewhere and copy the science app (KWSN-2.2B-*) there. You'll also need a work unit; there should be some inside the setiathome.berkeley.edu directory also. They are about 350 KB large. Copy one of them and give it the filename work_unit.sah (cp your.wus.name work_unit.sah) and put it in the same folder as the app before.

Then, run this command inside that folder:

./KWSN-2.2B-XXX -bench

where XXX is your specific version. Wait a few seconds (on a P4, less than 10) and it should finish. If it hasn't finished after a minute, ctrl+c it.

Check out stderr.txt. It'll tell you what happened and whether it had an error.

HTH,
Simon.
Donate to SETI@Home via PayPal!

Optimized SETI@Home apps + Information
ID: 540130 · Report as offensive
Johan Jongerius

Send message
Joined: 10 Sep 99
Posts: 3
Credit: 5,568,169
RAC: 0
Netherlands
Message 540145 - Posted: 2 Apr 2007, 21:35:26 UTC

Hi Simon,

Everything works fine now on my p4-sse2.

Thanks for al the good work.
ID: 540145 · Report as offensive
peristalsis

Send message
Joined: 23 Jul 99
Posts: 154
Credit: 28,610,163
RAC: 51
United States
Message 540166 - Posted: 2 Apr 2007, 22:34:34 UTC - in response to Message 540130.  

THanks to responders. I'll give it another shot tomorrow..j
ID: 540166 · Report as offensive
Dominik S.

Send message
Joined: 4 Jun 03
Posts: 15
Credit: 4,346,294
RAC: 0
Poland
Message 545937 - Posted: 14 Apr 2007, 6:52:38 UTC

Hi Simon,
I just downloaded 64-Bit Linux SSE2-generic, (I have AMD Athlon(tm) 64 Processor 3000+), I was trying to benchmark this app in standalone mode and in stderr.txt I receive this:
SETI@home error -108 Unknown error
from boinc_init_diagnostics()
File: main.cpp
Line: 229


I'm using gentoo with glibc ver. 2.5
any suggestions

ID: 545937 · Report as offensive
Profile michael37
Avatar

Send message
Joined: 23 Jul 99
Posts: 311
Credit: 6,955,447
RAC: 0
United States
Message 545940 - Posted: 14 Apr 2007, 7:03:59 UTC - in response to Message 545937.  
Last modified: 14 Apr 2007, 7:18:50 UTC

Hi Simon,
I just downloaded 64-Bit Linux SSE2-generic, (I have AMD Athlon(tm) 64 Processor 3000+), I was trying to benchmark this app in standalone mode and in stderr.txt I receive this:
SETI@home error -108 Unknown error
from boinc_init_diagnostics()
File: main.cpp
Line: 229


I'm using gentoo with glibc ver. 2.5
any suggestions


What's the output of 'ldd KWSN-R2.2B-64bit-SSE2-generic'?


ID: 545940 · Report as offensive
vl.adam

Send message
Joined: 23 Dec 05
Posts: 3
Credit: 394,025
RAC: 0
France
Message 545941 - Posted: 14 Apr 2007, 7:07:10 UTC
Last modified: 14 Apr 2007, 7:08:22 UTC

Hi Simon and Crunch3r,
here are the benchmark results of the SETI 64 bits linux app, in comparison with the default, the 32 bits, and the old Crunch3r's 64 bits. My system is AMD64@2430MHz.

Current WU: testWU-4
---------------------------------------------------
Testing default-512...
Running S@H client default-512....
Time for default-512:
309.09user 1.61system 5:59.81elapsed 86%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+8200minor)pagefaults 0swaps
---------------------------------------------------
---------------------------------------------------
Testing setiathome-5.12.x86_64-unknown-linux-gnu...
Running S@H client setiathome-5.12.x86_64-unknown-linux-gnu....
Time for setiathome-5.12.x86_64-unknown-linux-gnu:
156.04user 2.18system 2:53.29elapsed 91%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+483698minor)pagefaults 0swaps
---------------------------------------------------
---------------------------------------------------
Testing KWSN-R2.2B-SSE2-generic...
Running S@H client KWSN-R2.2B-SSE2-generic....
Time for KWSN-R2.2B-SSE2-generic:
121.38user 0.86system 2:13.89elapsed 91%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+21431minor)pagefaults 0swaps
---------------------------------------------------
---------------------------------------------------
Testing KWSN-R2.2B-64bit-SSE2-generic...
Running S@H client KWSN-R2.2B-64bit-SSE2-generic....
Time for KWSN-R2.2B-64bit-SSE2-generic:
194.64user 1.08system 3:34.56elapsed 91%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+15140minor)pagefaults 0swaps
---------------------------------------------------

=> As Simon writes "Performance is better than the equivalent 32-bit apps on Intel systems, on AMD systems it's not as clear; some assembly may be required". For me, it seems quite clear indeed ! I keep 32 bits for the moment.

Thanks and congratulations for your work.
Lionel.
ID: 545941 · Report as offensive
Dominik S.

Send message
Joined: 4 Jun 03
Posts: 15
Credit: 4,346,294
RAC: 0
Poland
Message 545943 - Posted: 14 Apr 2007, 7:18:17 UTC - in response to Message 545940.  


No idea, works for me. What's the output of 'ldd KWSN-R2.2B-64bit-SSE2-generic'?


ldd KWSN-R2.2B-64bit-SSE2-generic
libpthread.so.0 => /lib/libpthread.so.0 (0x00002ad52cf93000)
libc.so.6 => /lib/libc.so.6 (0x00002ad52d0ad000)
/lib64/ld-linux-x86-64.so.2 (0x00002ad52ce76000)
ID: 545943 · Report as offensive
Profile michael37
Avatar

Send message
Joined: 23 Jul 99
Posts: 311
Credit: 6,955,447
RAC: 0
United States
Message 545944 - Posted: 14 Apr 2007, 7:19:46 UTC - in response to Message 545943.  
Last modified: 14 Apr 2007, 7:28:02 UTC


No idea, works for me. What's the output of 'ldd KWSN-R2.2B-64bit-SSE2-generic'?


ldd KWSN-R2.2B-64bit-SSE2-generic
libpthread.so.0 => /lib/libpthread.so.0 (0x00002ad52cf93000)
libc.so.6 => /lib/libc.so.6 (0x00002ad52d0ad000)
/lib64/ld-linux-x86-64.so.2 (0x00002ad52ce76000)


I am not gentoo expert, but I am a bit concerned why the first two libraries point to /lib, not to /lib64. Again, this may be just how gentoo does things.

All this is exceedingly bizarre. I tried to run the 64-bit app in a standalone mode and got the same error. I put the same file into projects and edited app_info.xml onto it and it runs fine. No idea why. Wanna try?

ID: 545944 · Report as offensive
Dominik S.

Send message
Joined: 4 Jun 03
Posts: 15
Credit: 4,346,294
RAC: 0
Poland
Message 545949 - Posted: 14 Apr 2007, 7:33:59 UTC - in response to Message 545944.  


No idea, works for me. What's the output of 'ldd KWSN-R2.2B-64bit-SSE2-generic'?


ldd KWSN-R2.2B-64bit-SSE2-generic
libpthread.so.0 => /lib/libpthread.so.0 (0x00002ad52cf93000)
libc.so.6 => /lib/libc.so.6 (0x00002ad52d0ad000)
/lib64/ld-linux-x86-64.so.2 (0x00002ad52ce76000)


I am not gentoo expert, but I am a bit concerned why the first two libraries point to /lib, not to /lib64. Again, may be just be just how gentoo does things.

All this is exceedingly bizarre. I tried to run the 64-bit app in a standalone mode and got the same error. I put the same file into projects and edited app_info.xml onto it and it runs fine. No idea why. Wanna try?

I'm using multilib librarys, there is lib32 and lib64 and lib is only a link to lib64
On download page there is info:
You need to run 64-bit Linux 2.6.x and BOINC to use these apps - 2.4.x is not supported!
They also require glibc 2.3.4 or newer, anything less will NOT work.
Maybe this is the case.
ID: 545949 · Report as offensive
Profile michael37
Avatar

Send message
Joined: 23 Jul 99
Posts: 311
Credit: 6,955,447
RAC: 0
United States
Message 546023 - Posted: 14 Apr 2007, 15:14:35 UTC - in response to Message 545949.  

I'm using multilib librarys, there is lib32 and lib64 and lib is only a link to lib64
On download page there is info:
You need to run 64-bit Linux 2.6.x and BOINC to use these apps - 2.4.x is not supported!
They also require glibc 2.3.4 or newer, anything less will NOT work.
Maybe this is the case.


I am debugging this problem as we speak. I am running a 64-bit application under boinc, but at the same time I am unable to run it as a standalone. No resolution yet.

As I said, try to drain your queue and do try to run the 64-bit application under boinc when you have a chance.

ID: 546023 · Report as offensive
Dominik S.

Send message
Joined: 4 Jun 03
Posts: 15
Credit: 4,346,294
RAC: 0
Poland
Message 546190 - Posted: 14 Apr 2007, 20:14:13 UTC - in response to Message 546023.  


I am debugging this problem as we speak. I am running a 64-bit application under boinc, but at the same time I am unable to run it as a standalone. No resolution yet.

As I said, try to drain your queue and do try to run the 64-bit application under boinc when you have a chance.

I'll give a try, by for now I can't connect to my amd64
ID: 546190 · Report as offensive
Dominik S.

Send message
Joined: 4 Jun 03
Posts: 15
Credit: 4,346,294
RAC: 0
Poland
Message 547771 - Posted: 17 Apr 2007, 13:59:05 UTC - in response to Message 546190.  


I am debugging this problem as we speak. I am running a 64-bit application under boinc, but at the same time I am unable to run it as a standalone. No resolution yet.

As I said, try to drain your queue and do try to run the 64-bit application under boinc when you have a chance.

I'll give a try, by for now I can't connect to my amd64

Under control of boinc it works. I don't know for now about speed improvement but I'll see.
ID: 547771 · Report as offensive
Dominik S.

Send message
Joined: 4 Jun 03
Posts: 15
Credit: 4,346,294
RAC: 0
Poland
Message 548455 - Posted: 18 Apr 2007, 12:01:08 UTC - in response to Message 547771.  


I am debugging this problem as we speak. I am running a 64-bit application under boinc, but at the same time I am unable to run it as a standalone. No resolution yet.

As I said, try to drain your queue and do try to run the 64-bit application under boinc when you have a chance.

I'll give a try, by for now I can't connect to my amd64

Under control of boinc it works. I don't know for now about speed improvement but I'll see.

On my AMD64 3000+ (socket 754) KWSN-R2.2B-64bit-SSE2-generic is slower than KWSN-R2.2B-SSE2-generic about 15%.
Normal WU (this with ~62 point) with KWSN-R2.2B-SSE2-generic take about 10000 sec
with KWSN-R2.2B-64bit-SSE2-generic take about 11500 sec
I revert to 32 bit version.
ID: 548455 · Report as offensive
Profile michael37
Avatar

Send message
Joined: 23 Jul 99
Posts: 311
Credit: 6,955,447
RAC: 0
United States
Message 549345 - Posted: 20 Apr 2007, 3:18:00 UTC - in response to Message 548455.  


On my AMD64 3000+ (socket 754) KWSN-R2.2B-64bit-SSE2-generic is slower than KWSN-R2.2B-SSE2-generic about 15%.
Normal WU (this with ~62 point) with KWSN-R2.2B-SSE2-generic take about 10000 sec
with KWSN-R2.2B-64bit-SSE2-generic take about 11500 sec
I revert to 32 bit version.


This has been reported by some users. They also mentioned that using the ICC Patch (google "naughty intel") speeds up the 64-bit versions to be faster than 32-bit versions. The patched executables are, of course, unsupported, so your mileage may vary.



ID: 549345 · Report as offensive
Dominik S.

Send message
Joined: 4 Jun 03
Posts: 15
Credit: 4,346,294
RAC: 0
Poland
Message 549459 - Posted: 20 Apr 2007, 8:05:21 UTC - in response to Message 549345.  


This has been reported by some users. They also mentioned that using the ICC Patch (google "naughty intel") speeds up the 64-bit versions to be faster than 32-bit versions. The patched executables are, of course, unsupported, so your mileage may vary.

Thanks for suggestion. It works.
ID: 549459 · Report as offensive
Chris Kojiro

Send message
Joined: 26 Oct 99
Posts: 2
Credit: 19,911,638
RAC: 27
United States
Message 554636 - Posted: 27 Apr 2007, 5:02:17 UTC
Last modified: 27 Apr 2007, 5:09:40 UTC

Howdy,
I have some comparison results for processing speed with the new Lunatic Linux 2.2B apps versus other versions of SAH apps, as well as numbers for the even newer 64bit Linux 2.2B versions of the app. These numbers, unless noted, all come from P4 3.6GHz boxes. The results are for the workunits in mid April 2007 that gave 62.4 credits, unless noted.

For a baseline, running the SAH 5.15 "vanilla" app a few months back it took 15,237 seconds for 64.4 credits, which with a simple ratio works out for a "62.4 credit" workunit to a processing time of 14,764 seconds.

s.d. = standard deviation
  "App"                            seconds        Linux kernel  BOINC
SAH 5.15                          14,764            RHEL3 2.4   5.4.9
Lunatic R1.3 P4 - SSE2            13,744 s.d. 182   RHEL3 2.4   5.4.9
Lunatic R2.2B P4 - SSE2           10,710 s.d. 37    RHEL4 2.6   5.4.9 & 5.8.16
Lunatic R2.2B P4 - SSE3           10,675 s.d. 78    RHEL4 2.6   5.4.9 & 5.8.16
Lunatic R2.2B 64bit P4 - SSE3      8,804 s.d. 406   RHEL4 2.6   5.8.17
Lunatic R2.2B 64bit generic SSE2   8,709 s.d. 266   RHEL4 2.6   5.8.17

The conclusion seems obvious, these optimized apps help a lot.
The old R1.3 app gave a noticeable bump of 7%.
The new R2.2B apps gave another kick of 22% over the R1.3 app, and 27.7 % overall.
Then these new 64 bit R2.2B apps add another jump of 18.4% over the 32bit R2.2B app, 36.6% over the R1.3 app, and 41% overall.

Some notes, I don't think we are seeing a statistical difference between the SSE2 and SSE3 apps, at least on RHEL3 or RHEL4. The 64 bit BOINC is Augustine's version.

I had already started my little project when this thread "CPU Time on 62.40 Credit WU ?"was first posted. But, my results seem more pertinent to this thread. My boxes aren't as fast as some of the results reported in the "CPU Time on 62.40 Credit WU ?" thread. Whew!, there are some really fast systems out there.

I hope this encourages more users to try out these optimized chicken fried apps while they are hot.

Czesc, Chris
ID: 554636 · Report as offensive
Profile ML1
Volunteer moderator
Volunteer tester

Send message
Joined: 25 Nov 01
Posts: 21916
Credit: 7,508,002
RAC: 20
United Kingdom
Message 554768 - Posted: 27 Apr 2007, 10:59:22 UTC - in response to Message 554636.  

...I hope this encourages more users to try out these optimized chicken fried apps while they are hot.

Thanks for a very useful comparison.

And thanks to the Chicken Coop for all their hard work!


Any of this lot going into the ALFA multibeam standard app?

Happy crunchin',
Martin

(Still stuck in EDF and other systems too critical to touch yet... :-( )

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)
ID: 554768 · Report as offensive
Uftoun - zemedelec

Send message
Joined: 26 Jul 99
Posts: 11
Credit: 100,017,809
RAC: 0
Czech Republic
Message 554772 - Posted: 27 Apr 2007, 11:04:03 UTC

Hello,
I have problem with SSE2 version on Intel Xeon with SSE2.
When WU paused (by switch to another project, or another WU), process got signal 6 and generate error.
WU : http://setiweb.ssl.berkeley.edu/result.php?resultid=525783256
On Xeon with SSE3 and aplication SSE3 working OK.
Can you help me?
Thanks
Libor
ID: 554772 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 · 6 · Next

Message boards : Number crunching : Lunatic Linux Rev 2 apps?


 
©2025 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.