Minimum system requirements for GEOS-Chem
The hardware recommendations are: enough memory to run GEOS-Chem (at least 2 GB, but more than 4 GB for running at 2x2.5) and disk storage for met fields. The more diagnostics you save out the more memory it will require.
The Linux flavor (RedHat, SuSE, Fedora, Ubuntu, etc.) 64-bit archi is not an issue with GEOS-Chem.
Can GEOS-Chem run on Microsoft Windows (XP, Vista, etc.)
No. You will need to run GEOS-Chem in a Unix or Linux environment.
However, GEOS-Chem should be able to run in Mac OS Leopard, which is 100% compatible with Unix. As of Mar 2007, we are not aware of anyone having done this yet.
Which compilers does GEOS-Chem support?
The software recommendation is to have Intel Fortran compiler.
Also, you should be aware that because GEOS-Chem uses OpenMP parallelization, you can only run on as many nodes as are shared by the memory. For example, if you had 2 PC's, and each PC w/ 4 cores each, then you can only run on 1 PC at a time (i.e. 4 cores). This is because OpenMP has a requirement that all of the processors on the machine must be able to see all of the memory on the machine. In that case, you could run 2 jobs simultaneously on 4 cores, but not a single job on 8 cores. See www.openmp.org for more information.
MPI (Message Passing Interface) is required for passing memory from one physical system to another. The current version of GEOS-Chem does not yet have this capability.
Hi Yuxuan and congrats on your new position!
We also advise to use the more robust Intel fortran compiler instead of the Portland Group (PGI). Here is an email response from Bob to a similar question:
Bob Yantosca wrote:
I would recommend the Intel Compiler (IFORT) 10.1 instead of the Portland group compiler. I think the Intel compiler is just a better all-around compiler than PGI. We have had various minor compatibility problems with PGI (or at least I should say, other users who use it have). Also, I believe the Intel compiler is free to you if you are at an educational institution.
Intel 10.1 is very optimized for multi-core chipsets. Mike Seymour of Dylan's group at U. Toronto has done some testing with that compiler on their multi-core server. Here are the results (this is on the "platforms.pdf" file that is on the "Platforms and Parallelization" page of the website.
U. Toronto v7-04-11 Platform: Dell Latitude D830 laptop ($CAD2000): Core 2 Duo T7500 2.2 GHz, 2 GB RAM OS: Ubuntu 7.04 64-bit Linux kernel 2.6.20-15-generic Compiler: Intel x86-64 ifort 10.023 2 cores 02:13 walltime Contact Mike Seymour (firstname.lastname@example.org)
U. Toronto v7-04-11 Platform: Dell Poweredge 1950 ($CAD3150): 2x4 core Xeon E5345 2.33 GHz, 4 GB RAM OS: Ubuntu 7.04 64-bit Linux kernel 2.6.20-15-generic Compiler: Intel x86-64 ifort 10.023 8 cores 00:43 walltime Contact Mike Seymour(email@example.com)
You can talk to Mike directly for more of the lowdown on the Intel compiler. Intel also publishes C and C++ compilers (ICC) as well.
I have used the PGI compiler in Greece on a single processor Linux PC, but I had to debug GC (v7-04-03, I think) for it to run smoothly. I have no experience w/ PGI on multi processors.
How much disk space will I need for GEOS-Chem?
It will depend on which type of met data (GEOS-4 or GEOS-5) that you will use, and how many years of data you will download. Typical disk space requirements are:
|GEOS-4 4x5 met (uncompressed)||~13 GB/yr|
|GEOS-4 2x25 met (uncompressed) =||~52 GB/yr|
|GEOS-5 4x5 met (uncompressed)||~30 GB/yr|
|GEOS-5 2x25 met (uncompressed) =||~120 GB/yr|