"O" for Linux Wanted!!

David Konerding dek at socrates.ucsf.edu
Mon Oct 13 00:34:37 EST 1997

In article <61s7cd$2ak$1 at nntp3.u.washington.edu>, Ethan A Merritt wrote:
>In article <slrn4642jgv.lue.dek at socrates.ucsf.edu>,
>David Konerding <dek at socrates.ucsf.edu> wrote:
>>The O page indicates that O has not been released for Linux, even though
>>it seems to compile and work- slowly.  Since O is very graphics intensive
>>and support for hardware graphics support on Linux is minimal, it may not
>>be worth using it.  However, I say, release the Linux version anyway, so
>>those who can't afford SGIs can use it anyway.
>Or just use XtalView instead. It works like a charm under linux.
>I was truly amazed to find that the on-the-fly map calculations are quite
>passably fast on a Pentium Pro200 + Redhat.

A Pentium Pro 200 is about 1/3rd the speed of an 195MHz R10000 for floating point
and about the same as an R5000 (200MHz?).  I agree, that's passably fast.
I have noticed some problems with precision on linux, IE when I do iterative
work of multiply crds by rotation matrices, feeding the result back in,
the loss in precision is significant compared to SGI/DEC/HP.  The problem may
be limited to the math done by Mesa, but I have seen problems with AMBER as
well.  not sure exactly how important this loss in precision is for molecular
modelling- it doesn't seem to lead to large overall differences in structural
parameters at the end of the simulation.

However, I think the big thing now for Linux would be to increase its support
for hardware accelerated graphics.  If that gets done, it'll be really competitive
for molecular modelling.  I'm certainly happy having a dual PPro 200 running
Linux on my desk, but would be quite happy to have a really fast SGI with
hardware accelerated texture graphics :-)

Email: dek at cgl.ucsf.edu    David Konerding     WWW: http://picasso.ucsf.edu/~dek
Snail: Graduate Group in Biophysics
Medical Sciences 926, Box 0446
University of California
San Francisco, CA 94143

More information about the Xtal-log mailing list