Posts by Trog Dog

Message boards : Unix/Linux : Has anyone gotten D@H to run on a Linux 2.4.x machine?

It runs fine on Damn Small Linux (kernel 2.4.26), for some reason this distro also doesn't need the ulimit fix either, even though the default stack limit is 8192.

Now that is really puzzling!

I gave up on RHEL3 and have even removed boinc from it, for security reasons, since it IS a server.

I'm really beginning to wonder about these Fortran compilers. From what I can tell by reading "Understanding the Linux Kernel" 2nd Edition, the expansion of the stack depends on the programs stack pointer being at, or a little below, the memory reference causing the page fault. I'm wondering if Fortran does something funny with the stack frame.

IIRC, someone in either the boinc message boards or developer/projects mailing lists said something about another project having corrupted data if interrupts occurred at the right point during Fortran execution. I was trying to find what was happening with the stack problem and found that running Fortran in the GBD debugger will cause it to get corrupted data errors. I'm beginning to think that the Fortran compiler is outputting code that references data BELOW the stack pointer in memory and it gets corrupted when the return address for the interrupt is pushed on the stack.

I'm also wondering if the reason D@H requires an unlimited stack on some Linux systems is because the stack behavior changes and when the page fault occurs, Linux just allocates the page without checking the stack pointer if it's set to "unlimited" and the page fault occurs in the address space that Linux reserves for the stack to grow into.

Message boards : Unix/Linux : Has anyone gotten D@H to run on a Linux 2.4.x machine?

Message boards : Number crunching : AMD64

I have compiled charmm with g77 successfully and it seemed that it worked correctly.

Message boards : Unix/Linux : Where are the linux alpha testers?

That probably was the "ulimit" issue. If you want to go back to Ubuntu, 6.10 seems to run pretty well.

Nope, I had fix implemented. This was not a problem unique to this project. All projects had the problem, across all linux machines. The jobs would just stall, and the activity monitor would show just the normal OS activity. Quitting and restarting BOINC would solve the issue...until it happened again a day or several days later.

I guess I just haven't run enough BOINC under Linux yet to see this issue or else I saw it and didn't recognize the symptoms.

Message boards : Unix/Linux : Howto: Dealing With 0x1 Error

Message boards : Number crunching : Charmm 5.04 (Windows)

Here is David's answer. It seems to be normal behavior of BOINC. We will try to think up a possible solution for this.

Andre

That's the normal behavior; it doesn't work well in this case.
A couple of workarounds:

1) Make a new application for each version
(a bit of a pain because you have to run a separate
validator and assimilator during the transition).

2) Modify your validator so that if two results are
from different versions, it doesn't compare them, just
picks one of them as canonical.

If anyone can think of a good solution
(preferably without database mods) let me know.

-- David

Message boards : Number crunching : Some more PII/PIII w/ Windows? Anyone? :-)

Message boards : Unix/Linux : Ubuntu 6.10

@clownius

Off topic
Ive had issues with 6.10 mainly arts but im actually on 6.10 now. its a bit of a mess install. I have way to many local packages (i cant remove them or it wants to remove 30 other things i need) and cant get the Intel compiler to install properly. Maybe next time im bored ill consider a fresh install but at the moment the computers still waiting on a SATA DVD burner to come in and i have stuff i cant rescue off the HDD until it arrives.
Ill have another go at the seti optimized for Linux Core 2 Duo when i have the hardware right.

I'm really new at the Linux thing, but I tried a dozen packages and found the Ubuntu family to be the easiest for me to install and configure and it seems to have what I need to interface with Windows. I haven't found a GUI interface yet to use on my Windows machines to work with the Linux machines. I have my Linux machines on KVMs so far, but that only helps me "locally" when I'm out in the computer room.

Message boards : Number crunching : Updated Project List For BOINCView

Message boards : Number crunching : To Linux crunchers: lower credits issue

IIRC, someone actually took a Windows XP machine, loaded VMware on it, and found that a WU ran faster in Linux on VMware on WinXP, than it ran natively on just WinXP on the same machine.

My 3.4GHz Pentium 4 with HT took 5.5 hours for a single WU in Linux under VMWare (while also running BOINC in the host O/S). Running Docking in the host O/S (Windows 2003 SBS), work units took on average about 10 hours.

Whether this is related to disk performance, I don't know - it was using the same hard-drive and (presumably) still had to patch it through the host O/S.