Posts by old_user30957

1) Message boards : Number crunching : Main climateprediction.net site down (Message 6660) Posted 6 Dec 2004 by old_user30957 Post: I seem to have received a master file fetch error during a trickle but no model data was lost and the trickle seems to have been receieved correctly. I'm waiting for the next one to see if the same error occurs. Good luck getting the servers back up.
2) Message boards : Number crunching : JavaScript Stats (Message 6495) Posted 1 Dec 2004 by old_user30957 Post: I have written a script that allows people on servers that do not support PHP or other server-side language (such as Geocities, Tripod, etc.) to retreive numeric statistics. This is done by including an external Javascript file. You can choose to display the data in tables or make it fancy with background images. The following link describes how to include the stats. It will be useful to know basic JavaScript before using. <a href="http://infinisoft.34sp.com/boinc/stats.htm">JavaScript Stats</a> Portions of the WAP-parser written by Petrus.
3) Message boards : Number crunching : cpdn extremely slow on boinc 4.56 (Message 6451) Posted 28 Nov 2004 by old_user30957 Post: > > I think the slowdown has something to do with how BOINC is structured or > how > > the WU processing for SETI@home changed. I get a similiar slowdown, > classic > > averaged 6 hours and now its 6.5 to 7 hours. > > > Different slowdown than the one asked about by the first poster. > > The one you mention is due to the Fortran compiler used in BOINC CPDN. While > most Intel CPUs run a little, to significantly faster under BOINC CPDN, AMDs > were either a little to significantly slowed down by the choice of compiler. > AMD64s in particular have shown some significant slowdowns under BOINC. I'm > seeing similar to you with my Athlon64 3400+. It was less than 5.5 hrs per > trickle under classic and now is around 6 hrs. On the other hand, my P4 3.4 > GHz improved from around 6.5 hrs under classic to 4.5 hrs under BOINC (running > one model). > I was referring to a SETI workunit, which I think is in C/C++. My CPDN trickle rate is a bit slow, my current model has about 50 hours CPU time but only 3 trickles (ouch). But your right this is a little off-topic from what the original poster was talking about. Which reminds me I really need to think about defragging my old cruncher machine...
4) Message boards : Number crunching : cpdn extremely slow on boinc 4.56 (Message 6438) Posted 27 Nov 2004 by old_user30957 Post: > What i can tell is that the java environment is a veryvery serious problem in > my opinion to the scientific programms (at all) and the way things are done > within it. For instance: Setiathome Work Units took me about 2hours and 30 > minutes to process on the old client (still c++/c) then the whole stuff was > ported to BOINC which is either just java or a mix of java and c++ which I > don't know exactly but there has to be a lot of Java in it as the Work Units > needed 3 hours to process now. >> As far as I know boinc is ansi c (sience apps c/c++ and fortran) >> where did you read that java was used (out of curiosity) >> Florian >> www.domplatz1.de I thought so too? I haven't seen any .java files in the BOINC source so far, only C source and header files, plus the program itself itself is an executable. Porting to Java doesn't make sense for a computationally intense application, CPDN especially. I think the slowdown has something to do with how BOINC is structured or how the WU processing for SETI@home changed. I get a similiar slowdown, classic averaged 6 hours and now its 6.5 to 7 hours.
5) Message boards : Number crunching : Model Error (Message 6410) Posted 26 Nov 2004 by old_user30957 Post: > Just scrap it. It cannot be recovered. > Alright, thanks.
6) Message boards : Number crunching : Model Error (Message 6398) Posted 26 Nov 2004 by old_user30957 Post: A few days ago I began running BOINC and CPDN. For some reason the computer messed up, causing a few SETI units to have computational errors and a CPDN model to end abnormally. The problem dealt with my computer and seems to be fixed. BOINC removed the old model from the Queue and has replaced it with a new model. However, the old model still exists as a directory under projects/climateprediction.net. Am I safe in assuming that this folder can be deleted? Or is it possible to restart the model from scratch?