[Beowulf] mem consumption strategy for HPC apps?
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Robert G. Brown rgb at phy.duke.eduMon Apr 18 09:41:12 PDT 2005
- Previous message: [Beowulf] mem consumption strategy for HPC apps?
- Next message: [Beowulf] Cluster computing software survey
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Mon, 18 Apr 2005, Robert G. Brown wrote: > The COST of this (in time) on (say) 100 BT is at least 100 usec per face > to start talking to a node plust 0.5/10 MB/sec = 0.05 seconds, bandwidth > dominated. To do all six faces requires 0.3 seconds. If you organize > your code carefully you can probably find a pattern that does all of the > bidirectional communications in parallel for all nodes in the SAME 0.3 > seconds. ... or rather half that (duh) as they're bidirectional. > ... Unless I'm making egregious arithmetic errors again (very > possible as longtime list humans know:-). Obviously. rgb (again) -- Robert G. Brown http://www.phy.duke.edu/~rgb/ Duke University Dept. of Physics, Box 90305 Durham, N.C. 27708-0305 Phone: 1-919-660-2567 Fax: 919-660-2525 email:rgb at phy.duke.edu
- Previous message: [Beowulf] mem consumption strategy for HPC apps?
- Next message: [Beowulf] Cluster computing software survey
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
