From: Jeremy S. <j.a...@da...> - 2011-07-09 21:45:18
|
Hi Nathan, I run MESA here on the COSMOS super computer here in Cambridge and found it significantly faster. On 2 cores I can get a 1M from pre ms to WD in 1.5 hours. The technician helped me a lot and we found that three cores is the most efficient (gaining only a small boost with 4). Jeremy On Jul 9 2011, Nathan Thompson wrote: >Hello, > > My name is Nathan Thompson and I am a graduate student at Wichita State > University. I have been using MESA on my personal computer for about a > year to do computations. It is an Intel i7 860 machine (four cores) and > does well with the computations. > >Recently, I was given an account on a university computer in order to use >that to do computations, a 16 core machine of which I am not 100% sure of >the processor type, but I was assured that each of its cores were faster >than my i7. However, when I ran the same model in MESA on both boxes, my PC >did the computations twice as quickly as the university machine (and I had >full use of all 16 cores). Both versions of MESA were compiled using >gfortran. > > I know the MESA is "thread safe" as stated by the website, but > performance is noted to not increase with additional processors. Is there > an option that I need to set in MESA to activate the use of OpenMP for > multiple threads, or is this enabled "out of the box?" Also, is there a > way to limit the number of cores MESA has access to, so that I can test > to see if one of these cores is indeed faster than my i7? > >Thanks, >Nathan Thompson >Wichita State University > |