Tuesday, July 14, 2009

The mysterious GAMESSQ program


A recent question from Jacob Lerche mentioned a queueing program for GAMESS called GAMESSQ. Never heard of it. A google search only revealed that the latest MacMolPlt version has a "hook to it". Nothing on the GAMESS page. Strange ...

A few days later the another google search listed Jacob's question, that I posted on the blog comments, as one of the top hits, so now Molecular Modeling Basics looked like an authority on the subject and I decided to look into it.

The program is indeed bundled into the latest GAMESS download (which you can initiate here). It came with an brief manual in html format, so I put it on my web server here. The "about GAMESSQ" feature on the Mac tells me that the program is open source and written by one Jason Ekstrand and "commissioned" by MacMolPlt author Brett Bode.

Anyway, it is a very useful program, and I made a screencast about how to use it. What I don't mention in the screencast (which is already quite long) is (1) that the Jobs>open in wxMacMolPlt doesn't seem to work and (2) the program allows you to submit several jobs at once, which will then run one-after-another. Which of course is the whole point of a queueing program.

10 August, 2009. Update: GAMESSQ has now been officially released.

24 comments:

Jan Jensen said...

Update: "Jobs>open in wxMacMolPlt" now works. No idea why it didn't work before. Very handy!

Anonymous said...

I found this post in the net about gamessQ and downloaded (again) the
gamess tarball.. but there is no hint about gamessQ inside it.

Have I missed the point in your post? It would be nice if you could give me a small hint as I'm quite interested in molecular modeling

Thanks in advance

Jan Jensen said...

I downloaded the the pre-compiled MacOSX version of GAMESS and found the GAMESSQ program in the same directory as the GAMESS executable.

So if you don't see it they probably forgot to include it in the package you downloaded.

Some suggestions:
1. look in the gamess/tools directory

2. try to download gamess for another OS and see if you can find it there (I don't know how transferable the GAMESSQ program is).

3. let me know which package you downloaded and I can check with the GAMESS people.

Has anyone else out there had similar problems or know solutions?

patrick said...

Indeed ... in the macox binary package there is a gamessQ executable...

however I'm a linux guy and I'm downloading the source package for x86 linux. there is no gamessq inside (even with a grep -Ri gamessq * i found no hint about gamessQ there).

it would be cool if someone could provide the source to build it also on linux if possible

Jan Jensen said...

I'll look into it. Watch this space...

Jan Jensen said...

Here's the current status.

GAMESSQ has only been released for MacOSX.

There is a bug in the Windows version that needs to be fixed.

The source code might be released soon. I'll let you know when I know more.

For those non-Mac users that can't wait, you may want to look at GTK-GAMESS. This GAMESS queueing program is no longer supported (as far as I know) but it may still work fine.

Of course there is always the old-fashioned way:

1. create a file called, for example, "runjobs" that looks something like this:

#!/bin/csh
rungms job1 >& job1.log
rungms job2 >& job2.log

2. on the command line type "chmod 755 runjobs"

3. on the command line type "runjobs"

This will run two input files (job1.inp and job2.inp) one after another and create the output files job1.log and job2.log

Let's hope GAMESSQ becomes generally available soon!

chemicalscum said...

What is the advantage of using GAMESSQ over writing a BASH script? It is a very easy way to queue your jobs.

When I learned how to do it as a grad student 20 years ago the hard part was that I had to learn vi to do it.

Jan Jensen said...

Billg,

The advantage of GAMESSQ over a BASH script is that you don't have to learn vi to use it :)

Seriously though, features like the Avogadro input menu and GAMESSQ are great ways to get started.

Once one gets comfortable with modeling there are usually more efficient ways of doing things (such as BASH scripts) but they can seem very intimidating to people starting out.

Jan Jensen said...

Update: GAMESSQ is now available for Windows and Linux as well.

Anonymous said...

Very nice blog! I'm running GAMESSQ now on Windows and it's very nice to be able to easily queue GAMESS runs. I'm a synthetic organic chemist by training but I have some experience in molecular modelling, mostly docking and database screening (pharmacophores etc). Just recently I've started doing some ab initio -related stuff.

About GAMESSQ: I've got a Quad processor but GAMESS is always using only 1 processor i.e. 25% of the resources. If I set "processors" number higher than 1 when setting up a run, GAMESS just gives an error complaining something about DDI. It would be nice if GAMESSQ ran more than one job at a time.

Jan Jensen said...

Anonymous,

Glad to hear you like the blog.

Specifying more than one processor works fine on my mac.

What is the exact wording of the error message? If you don't mind, please post it to the GAMESS google group. That way more people can help.

Anonymous said...

The exact wording of the error message is:

DDI: Compile limit MAX_NODES = 1
DDI: Requested number of nodes = 4
DDI: Increase MAX_NODES and recompile DDI.

I noticed that other people have had the same problem. There were some possible solutions to the problem in the google group you suggested but I tried those and unfortunately they didn't help. I don't know if the fact that I'm running Win XP _home_ has something to do with this.

At least I figured out a way to make GAMESSQ to run several gamess jobs at the same time by queuing several jobs, starting the first one, pausing it, waiting until the next one starts and resuming the first one, etc. :) There's probably a faster way that I don't know of...?

Jan Jensen said...

I have checked with the GAMESS people, and this should work.

Are you running a somewhat old version of the windows version?

If so, please get the latest version and try again.

If it still doesn't work let me know.

Antti Siiskonen said...

I'm the anonymous guy above :)

I downloaded Wingamess just a couple of weeks ago, so I think this is the latest version. Unless there are patches somewhere that I don't know of. Gamess.09.exe is dated 29.5.2009.

At least I can run multiple gamess jobs at the same time (each using just 1 core) but of course it would also be nice to be able to run just 1 job with 4 cores...

Jan Jensen said...

Siiskonen, Sarom Sok from the GAMESS group has recompiled ddikick. Please give it a try, by
(1) downloading it
(2) unzip/untar it
(3) rename it to ddikick.exe

Let me know if that works

Antti Siiskonen said...

I did (1), (2), and (3) and tried to start a job in GAMESSQ with either 1, 2, 3 or 4 processors.

It works perfectly now!

Many, many, many thanks for taking the time to help me with this! And probably helping many other people with the same problem, too.

Best regards,
Antti

Jan Jensen said...

Excellent! Glad to hear it. The real credit goes to Sarom.

Antti Siiskonen said...

Now I noticed something strange with GAMESSQ. When I use GAMESSQ for some time; starting runs and sometimes cancelling them, starting new ones etc. Eventually there are a bunch of gamess processes running that are not using CPU but are using a lot of Windows page file. In the end there's something like 12 GB of page file and Windows complains about that. When I kill the gamess processes manually, the page file size drops and everything is fine again.

I don't know if this is just because of some settings are wrong on my computer...

Jan Jensen said...

Brett Bode (who helped write GAMESSQ) is looking into it and promised to get back to me. Stay tuned...

Jan Jensen said...

Update: the Windows distribution of GAMESS now includes a new version of ddikick allowing for parallel computations.

No word from Brett yet about the GAMESSQ problem.

R said...

Thank you for the Intro! :-) Currently I am building a crystal (beta- Si3N4), but suffer a (finding) lack of documentation.

Best,

R

Jan Jensen said...

Building a crystal is something I have no experience with. Are you using Avogadro? If so, you might want to ask for help on the Avogadro mailing list.

Anonymous said...

Using GamessQ I am running VSCF calculations (using all 4 cores of a quad core processor) on WinGamess. I am getting error message like the following:

DDI Process 3: trapped a segmentation fault (SIGSEGV)


Also, after the completion of the VSCF jobs, GamessQ does not save the ".rst" files. But such ".rst" files are very useful to evaluate overtones from the initially evaluated fundamentals. Is it possible to save the ".rst" files?

Regards,
Jeet

Jan Jensen said...

Jeet - best to handle this on the google GAMESS list.

Please include input and output files. Do you have the same problem with other runtyp's?