SLAMM Forum

General Category => Using SLAMM => Topic started by: MSub on January 06, 2016, 09:06:21 AM

Title: SLAMM in Linux
Post by: MSub on January 06, 2016, 09:06:21 AM
We are working on a large study area and are thinking of making use of High Performance Computing (HPC) cluster that we have in campus. The operating system in the HPC is Linux. So, we need to compile SLAMM on Linux. I wanted to get some idea about the possibility of using SLAMM in Linux before I start digging into the Delphi compiler in Linux. Has anyone tried SLAMM on Linux? What would be the best approach to get started for it to make it work?

Thank you for your help.
Title: Re: SLAMM in Linux
Post by: Jonathan S. Clough on January 12, 2016, 10:50:19 AM
We have run large jobs using a parallel-processing version that we execute on a 32-processor Windows server.  The software is not currently cross-platform compiled, however.  All compiled versions that I know of require Windows. 

If you don't output maps to Microsoft Word, though, there's not much interface with the OS other than binary and text file reading and writing.  Therefore it should be compilable on an alternative platform without too much difficulty. 

You'd have to do that yourself with the open-source code for now.  To create a Linux-native version would likely require compiling with Lazarus for Free Pascal.   We originally tried to create a 64-bit version on that software but found the compiler to be quite buggy and we lost a lot of time going down that dead end.  However, that was 6 years ago and the open-source compiler has likely improved in the interim.

Oh, I'm remembering now that we did have some researchers run it in Linux using the "wine" software but we haven't done that yet ourselves.  They had success, though and wrote a paper about their work.  (http://ascelibrary.org/doi/abs/10.1061/41114(371)477)
Title: Re: SLAMM in Linux
Post by: MSub on January 12, 2016, 02:10:40 PM
Thanks Jonathan. By "a parallel-processing version", do you mean SLAMM parallel-processing version? If so, do we need to modify code to make it run as a parallel-processing version? Can you please elaborate more about it?

I have also thought about wine, but am not sure if it will be able to take an advantage of parallel processing in HPC. I'll take a look on the paper.