Recently I’ve hit the limit of my simulation capacity on my machine at work. This is a 8-core (2 x quad-core) Xeon machine with 16 gigs of RAM. So it’s quite capable, albeit somewhat pathetic in the grand scheme of people doing multi-phase fluid dynamics (I need to be submitting jobs to a capable supercomputer). As I moved to larger, more detailed systems, at some point they all just started failing. I was astonished to find out that I had indeed hit the limit of my memory addressing in a 32-bit environment. Not that this is all that incredible … I just had never hit this boundary. I have been getting around it lately by decomposing my domains into smaller pieces to be handled by each individual processor, but at some point I have to reconstruct all this data and analyze it effectively. So this morning I finally bit the bullet and reinstalled my Debian machine AMD64/EM64T. Then I recompiled a whole bunch of stuff and got the appropriate 64-bit applications, and viola, I’m in heaven. Now I can pull in and cache 11GB data sets in one analysis application like it’s nothing. I realize this is nothing too spectacular but it’s new to me. Brilliant. That is all.