Posts by koniiiik

1) Message boards : Number crunching : Problems with Minirosetta 1.76 (Message 61844) Posted 19 Jun 2009 by koniiiik Post: i had crashing WUs when i had faulty RAM installed on my PC. After running mem test and replacing faulty RAM everything was back to normal. Also overclocking RAM can cause some random WU crashes I guess you were replying to my post. Well, this is very unlikely – I just completed a build of OpenOffice.org 3.1.0, without any error. I think you would agree that building OOo is a much more difficult test for the RAM than running a few minirosetta tasks. In fact, minirosetta has been the only program dying on SIGILL on my machine ever since I first attached my machine to the project, only that the crashes were not as frequent as they are with 1.76. I reported this error for version 1.54 (http://boinc.bakerlab.org/rosetta/forum_thread.php?id=4691&nowrap=true#59218) and nobody seemed to care at that time. Later I tried to set rosetta to assign the smallest WUs possible and it did indeed make it possible to run most tasks successfully but, well, now it seems minirosetta is crashing shortly after start. Guess I'll have to switch to a different project.
2) Message boards : Number crunching : Problems with Minirosetta 1.76 (Message 61825) Posted 18 Jun 2009 by koniiiik Post: Like the previous few versions did, although not as often as this one does, most of my tasks die on signal 4, which means illegal instruction. For example, see http://boinc.bakerlab.org/rosetta/result.php?resultid=259586800 or any other task assigned to me, there are about 20 of them in a row with the same error. It is probably using some kind of special processor features which it doesn't detect correctly or whatever. I can provide core dumps if it will be of any help.
3) Message boards : Number crunching : Problems with Minirosetta v1.54 (Message 59218) Posted 1 Feb 2009 by koniiiik Post: The 1.54 version seems to be in conflict with the Linux ABI in FreeBSD. One machine I'm running boinc on is a FreeBSD one, boinc downloads and runs the Linux binaries through the Linuxulator. Version 1.47 worked flawlessly, but the 1.54 version crashes randomly on SIGILL. http://boinc.bakerlab.org/rosetta/results.php?hostid=973136 shows only one successful task, which was run with Rosetta beta rather than minirosetta; all of the minirosetta tasks crashed sooner or later.