NEW: LINUX TEST APPLICATION FOR EINSTEIN@HOME
29 Jul 2005 12:27:02 UTC
Topic 189636
(moderation:
Bernd has finished building a new test application for Linux. This detects and uses the SSE instruction set if your computer supports it, which can increase the speed by a significant factor. The test app is available here.
Please report success and/or failure with the test app to this message board thread.
Director, Einstein@Home
NEW: LINUX TEST APPLICATION FOR EINSTEIN@HOME
)
Well so far it is working! I'll do some statistics once I get results about speed increases.
such things just should not be writ so please destroy this if you wish to live 'tis better in ignorance to dwell than to go screaming into the abyss worse than hell
Not feeling the love
)
Not feeling the love here:
2005-07-29 23:02:51 [Einstein@Home] Starting result w1_0954.0__0954.1_0.1_T01_S4hA_1 using einstein version 0.06
2005-07-29 23:02:51 [Einstein@Home] Unrecoverable error for result w1_0954.0__0954.1_0.1_T01_S4hA_1 (process got signal 7)
2005-07-29 23:02:51 [Einstein@Home] Unrecoverable error for result w1_0954.0__0954.1_0.1_T01_S4hA_1 (process got signal 7)
2005-07-29 23:02:51 [---] request_reschedule_cpus: process exited
2005-07-29 23:02:51 [Einstein@Home] Computation for result w1_0954.0__0954.1_0.1_T01_S4hA_1 finished
2005-07-29 23:02:51 [---] schedule_cpus: must schedule
2005-07-29 23:02:51 [LHC@home] Pausing result wjun1C_v6s4hvnom_mqx__17__64.3_59.31__6_8__6__15_1_sixvf_boinc22404_3 (left in memory)
2005-07-29 23:02:51 [Einstein@Home] Starting result w1_0954.0__0954.2_0.1_T01_S4hA_1 using einstein version 0.06
2005-07-29 23:02:51 [Einstein@Home] Starting result w1_0954.0__0954.3_0.1_T01_S4hA_1 using einstein version 0.06
2005-07-29 23:02:52 [Einstein@Home] Unrecoverable error for result w1_0954.0__0954.2_0.1_T01_S4hA_1 (process got signal 7)
2005-07-29 23:02:52 [Einstein@Home] Unrecoverable error for result w1_0954.0__0954.2_0.1_T01_S4hA_1 (process got signal 7)
2005-07-29 23:02:52 [---] request_reschedule_cpus: process exited
2005-07-29 23:02:52 [Einstein@Home] Deferring communication with project for 59 seconds
2005-07-29 23:02:52 [Einstein@Home] Deferring communication with project for 59 seconds
2005-07-29 23:02:52 [Einstein@Home] Computation for result w1_0954.0__0954.2_0.1_T01_S4hA_1 finished
2005-07-29 23:02:52 [---] schedule_cpus: must schedule
2005-07-29 23:02:52 [LHC@home] Resuming result wjun1C_v6s4hvnom_mqx__17__64.3_59.31__6_8__6__15_1_sixvf_boinc22404_3 using sixtrack version 4.66
2005-07-29 23:02:53 [Einstein@Home] Unrecoverable error for result w1_0954.0__0954.3_0.1_T01_S4hA_1 (process got signal 7)
2005-07-29 23:02:53 [Einstein@Home] Unrecoverable error for result w1_0954.0__0954.3_0.1_T01_S4hA_1 (process got signal 7)
2005-07-29 23:02:53 [---] request_reschedule_cpus: process exited
2005-07-29 23:02:53 [Einstein@Home] Computation for result w1_0954.0__0954.3_0.1_T01_S4hA_1 finished
2005-07-29 23:02:53 [---] schedule_cpus: must schedule
2005-07-29 23:02:53 [Einstein@Home] Starting result w1_0954.0__0954.4_0.1_T01_S4hA_0 using einstein version 0.06
2005-07-29 23:02:54 [Einstein@Home] Unrecoverable error for result w1_0954.0__0954.4_0.1_T01_S4hA_0 (process got signal 7)
2005-07-29 23:02:54 [Einstein@Home] Unrecoverable error for result w1_0954.0__0954.4_0.1_T01_S4hA_0 (process got signal 7)
2005-07-29 23:02:54 [---] request_reschedule_cpus: process exited
2005-07-29 23:02:54 [Einstein@Home] Deferring communication with project for 59 seconds
2005-07-29 23:02:54 [Einstein@Home] Deferring communication with project for 59 seconds
2005-07-29 23:02:54 [Einstein@Home] Computation for result w1_0954.0__0954.4_0.1_T01_S4hA_0 finished
What processor you have. Did
)
What processor you have. Did you delete the current tasks before starting boinc over? What version of BOINC are you using?
such things just should not be writ so please destroy this if you wish to live 'tis better in ignorance to dwell than to go screaming into the abyss worse than hell
I have 6 machines testing the
)
I have 6 machines testing the new version right now, 3 are AMD XP 2400+ and the other 3 are AMD XP 2000+.
Both showed almost exactly the same performance increase of around 28%.
Still not as fast as running in windows, but getting pretty close.
Times:
XP 2400+
Old Linux New Linux Windows
28,045.32 39,027.03
28,045.32 39,023.33
28,029.64 39,032.79
XP 2000+
Old Linux New Linux Windows
33,848.51 47,072.46
33,866.03 47,053.06
36,278.04 50,586.85
Could someone tell me some example times for Windows runing Einstein@Home on the 2000+ and 2400+
I'll post some more info when I try the new client on some 2800+ machines and a few celeron and pentium4 machines.
such things just should not be writ so please destroy this if you wish to live 'tis better in ignorance to dwell than to go screaming into the abyss worse than hell
With the old version... My
)
With the old version...
My XP2000+ runs at around 8 hours in Win2K
By comparison it runs at 13Hrs when I'm in Linux.
My Celeron 2.4GHz machine runs at 14Hrs in Linux.
I have overclocked it to 3GHz and it still runs around 13Hrs.
I'll try loading the new Linux version and post some data.
RE: What processor you
)
It's a p4 with HT using Linux 2.6.3-7mdksmp, with BOINC 4.43. There were no current einstein tasks, as it had been suspended long ago because of the linux issues.
But I think the problem may be a bad download of the test file package. I downloaded it again today (3 more times) and I'm getting different file sizes on the download. First time last night I got 305K, and today I got 546K once and 819K twice. All extract without giving any errors, but something is clearly not right there.
Maybe UWM can add the checksum to the download page???
Anyway, I deleted the files put in last night and went with the ones extracted from the last download (one of the 819K files I got twice). Once my penalty box time is over (quota reached), I'll give this one a try and see what happens.
RE: Maybe UWM can add the
)
Good point. Will do.
BM
BM
That is very odd my tarball
)
That is very odd my tarball is 820k according to ls -lh
such things just should not be writ so please destroy this if you wish to live 'tis better in ignorance to dwell than to go screaming into the abyss worse than hell
RE: Good point. Will
)
Thank you. Only the last one I downloaded checks ok, so I'll give it a try.
Darren
It's working fine for me - I
)
It's working fine for me - I see a decrease in processing times of roughly 30%, as others have mentioned. However, the WUs don't seem to claim as much credit as the ones processed on Windows, but perhaps that's the same old Boinc client problem?