[GRASSLIST:434] Cygwin Performance vs. Linux

Has anyone benchmarked WinGrass against a Linux version on the same or similar hardware?

Background...

I have several GIS workstations that I use in my research but the top end machines (> 2Ghz) typically run Windows as most of the really expensive software I use is Windows specific.

I have two GNU/Linux machines that I use for file serving and experimenting as well as running Unix programs that have no good windows port like MB System for multi-beam sonar processing.

I recently tried Libranet on my fastest Linux machine (500 Mhz) and really like the distribution overall, but it has managed to seriously break several development packages required for building grass (like the tk headers of all things!? ) While I was working on this problem, I experimented with WinGrass. It seems to be a little buggy right now and probably lags behind in developer attention relative to Linux, but overall quite usable. It feels a bit slugish to me though.

My question is this: am I better off running the cygwin version of grass on the hot Windows machines, or running it on the slower Linux machine (and loosing that cool Libranet desktop in the process)?

--
David Finlayson
School of Oceanography
Box 357940
University of Washington
Seattle, WA 98195-7940
USA

Office: Marine Sciences Building, Room 112
Phone: (206) 616-9407
Web: http://students.washington.edu/dfinlays

David Finlayson wrote:

My question is this: am I better off running the cygwin version of grass
on the hot Windows machines, or running it on the slower Linux machine
(and loosing that cool Libranet desktop in the process)?

There is a performance penalty for using Cygwin, but I doubt that it
would typically amount to a 4:1 ratio (i.e. making a 2GHz Cygwin
system slower than a 500MHz Linux system).

By far the biggest performance hit with Cygwin is spawning commands
(i.e. fork/exec; this is easily 20x slower on Cygwin than on a "real"
Unix such as Linux or *BSD). File I/O shouldn't be much slower than on
Linux, and there shouldn't be any difference for actual computation.

--
Glynn Clements <glynn.clements@virgin.net>