Given that the vast majority of PyMOL downloads
logged are not for Linux (nearly 10 to 1 against), given the recently
splintering of the Linux desktop market, and given the
importance of maintaining backward compatibility with older Linux distributions,
we must take a simple lowest-common denominator approach toward our
precompiled Linux binaries.
Thus, our 32-bit Linux builds are currently
prepared in a glibc-2.3 environment using GCC3. These builds deliver
reasonable performance and compatibility on a wide set of distributions.
Obviously, we cannot prepare optimized builds for each member of the diverse
combinatorial population of systems (DISTRO x VERSION x RUNTIME x GPU x DRIVER)
where it needs to run. Based on PyMOL usage alone, nearly 90% of our
platform-specific effort should be directed at Windows and Mac, not Linux.
Supporting one Linux binary out of three total is already an over-allotment
Practically speaking, the only way to achieve top
PyMOL performance on Linux is to build from source code using libraries and
compilers optimal for your specific hardware and distribution. That is one
of the reasons why the PyMOL open-source code is targeted at Linux
Thanks for all the input!
It seems like people
do not notice any difference in speed between the different linux
distributions. Also not between KDE or Gnome. That is good to know.
Apparently I have some other problem (unrelated to pymol, probably
openGL/glx related) which causes the dramatic difference in speed between our
Redhat (5 fully patched) and Suse (10.2 fully patched). We are using the
latest nvidia drivers from the nvidia website.
Accelrys/Insight/Discovery Studio. We got Insight/Material Studio (we do not
have the license for DS1.6 or 1.7) to run under Fedora7. Pymol also runs fine
so probably I will switch both machines to Fedora7. Anyway, thanks
Laboratory of Organic
Swiss Federal Institute of Technology
[mailto:firstname.lastname@example.org] On Behalf
Of Mathias W.
Sent: 13 July 2007 12:48
Subject: Re: [PyMOL] best linux distribution to run pymol
Joris Beld schrieb:
(unfortunately i cannot switch to Suse since i cannot get
Insight/Discovery Studio to run under Suse).
This due to the ignorance of the accelrys developers. Even
the newest version of DS (1.7) does not run on modern
Distributions using glibc-2.4 or higher. They say that they
only support RHEL4 which is real old (and uses glibc-2.3).
I guess it would have been no big deal to test DS on a
glibc-2.4 system and to find the bug which is preventing it
from running (I don't think they are still using
linuxthreads). As of the time DS 1.7 was released glibc-2.4
was already widely spread and so I call this ignorance. I got
a test version of DS 1.7 and I don't want to use an old linux
just because of this modelling program. As you pointed out
you end up with the situation that you cannot run any other
newer program (or only with great effort)...
granted that I am not getting money from accelrys ;) I think we should
differentiate between support to different distributions and
compatibility with different distribution.
I have an "unsupported" DS 1.7 running on RHEL 5 (with very nice native
support for nvidia graphics) and a week ago it was running "unsupported"
on a Kubuntu 7.04 (feisty fawn).
At the moment I can not say how pymol is performing on RHEL 5.
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
PyMOL-users mailing list