Question concerning Beuwolf computing & Relativistic N-Body Problems
- To: mathgroup at smc.vnet.net
- Subject: [mg62536] Question concerning Beuwolf computing & Relativistic N-Body Problems
- From: alan heider <mathphysabs at yahoo.com>
- Date: Sun, 27 Nov 2005 02:40:29 -0500 (EST)
- Sender: owner-wri-mathgroup at wolfram.com
Hello All - I'm considering buying a significant number of Mac's (preferably G5's) for Beuwolf computing. I have lots of questions as it seems that Virginia Tech is the largest current group: I. During down time for the original computations (some code has been written, some not), we'd like to hook up with BOINC at Berkeley for sorting through gravitational wave data. Is there another project more pressing (e.g. protein folding?) II. Code has been written for the calculation of all relevant GR quantities (this is my own and is much more efficient than any previously written, as I have tested it on another machine). Current computations for the inspiral of two black holes are currently utilizing a combination of finite element, finite difference, and spectral methods. Ours is capable of a fully relativistic N-Body simulation. The bottle necks include: After simplification using CAS, huge matrices are generated. Is it faster to send these out to a LinPak application versus trying to either invert or iterate them in Mathematica? Any suggestions would be sincerely appreciated (including current BenchMarks on Feyncalc for perturbation theory...... we need about 20th order) Thanks in advance, Geoffrey Alan Cope __________________________________ Yahoo! Mail - PC Magazine Editors' Choice 2005 http://mail.yahoo.com