From:	<gran(-(at)-)classic.chem.msu.su>
Date:	Fri, 10 Apr 98 13:05:38 +0400
Subject:	Re: Summary: Linux versus win NT

Dear netters,
   Taking into account the recent CCL discussion, this letter, which contains
some comments on the motivation and goals of our PC GAMESS project, probably
may be interesting for the computational chemists community. Please, note,  this
letter is intended solely to clarify our own position. It is not intended  to
initiate useless discussions on "what is the best operating system?" or,  more
globally, "what is the essense of life?" topics.



>      I have been using the windows version of gamess for almost 6
>     months
>      now and i  have had no problems. I still see people trying to
>     install
>      gamess under linux.
>      Is there a good reason why people are still using gamess under
>      linux on PC's ? Are there any serious problems or bugs in the
>     windows
>      version ?
>I expected to get a word from Dr Alex. A. Granovsky who ported gamess to
>win95/nt or from the gamess group , but i didn't . It seems that they are as
>anxious
>as i am to hear the comments of the users before they say anything.


   The main goal of the PC GAMESS project was (and currently is) to use high
level Intel-based workstations as promising production systems for serious and
intensive quantum chemical calculations. There is an old tradition to use UNIX
on the workstation-class computers designated mainly (or exclusively) for
various scientific calculations. On the other hand, the world of PC is a
relatively new and still partially unknown area (although, for example, our
group has used them since at least 1992 in the production  work) for the most of
the computation chemists. This is exactly the case when  the old good receipts
are not necessarily the best ones. Here, a very careful  choice of the operating
system to be used and the corresponding software  available (e.g. compilers) is
critical, if somebody is really interested in  getting the good production
performance. Unfortunately, this simple and  important idea is rarely considered
as seriously as probably should be.  The correct choice plays the decisive role
as really allowing the Intel-based  systems to be competitive with other modern
hardware platforms
(DEC, SGI, IBM, etc..).

   To our opinion, a thorough analysis of the current situation with the PC
operating systems and compilers available  shows that the usage of free UNIX
clones (like Linux, FreeBSD, etc...) on the production level systems still has
serious drawbacks. The freely available gcc/f2c and g77 compilers (and their
clones) are certainly not capable to generate an optimized floating point code
of the required quality. Though there exist several commercial Linux-based
Fortran and C compilers of higher quality, there are still other serious
problems with them, such as the code stability, overoptimization, bad
optimization, and so on. In general, the situation is not surprising, because
Linux is still a relatively new area in the (scientific-oriented) compiler
market.

   Today, the situation with the non-UNIX software is much better.
There is a lot of high quality compilers available, mainly for Win32 (Win NT and
95). Windows NT itself is stable, fast, and powerful enough to run efficiently
the most demanding scientific programs, including various quantum chemistry
packages. It has the built-in support of SMP. If needed, there is a variety of
useful (both free and commercial) software on the market (e.g., telnet daemons,
full POSIX support, UNIX compatibility tools, etc...), which simplifies the
process of transition from the UNIX to the NT  environment.

   This was (and still is) the motivation for our decision to use Windows NT as
the base system for the PC GAMESS project and for our production computers
(though the PC GAMESS itself is available also for other PC operating systems,
including Windows 95, OS/2, and even DOS). According to our estimates,  there
are already about 200-400 persons in the world, who use the PC GAMESS in  their
production work. We think this is large enough.

Just to sum up:
   On the PC, there are multiple reasons to work under Linux. There are also
multiple reasons to use NT instead. Both systems have serious advantages (and
disadvantages as well) as compared to each other. Both of them will definitely
survive in the forthcoming millennium. Both are stable,  powerful, and have the
prospects for the further growth.
For the intensive scientific calculations, NT still seems to be a better choice.
Possibly, the situation will change in future. Possibly, it won't. In any case,
this process may take up to several years. But we are interested to get the
results of our calculations today.


        Yours sincerely,
        Alex. A. Granovsky,
        Laboratory of Chemical
        Cybernetics,
        Moscow State University


All information concerning the current status of the PC GAMESS project can be
found on the  http://classic.chem.msu.su/gran/gamess/index.html




Similar Messages
08/01/1996:  Re: CCL:M:Heat of formation calculation using MOPAC.
06/08/1993:  undergrad computational chem
08/03/1995:  ACS Chicago - CINF Abstracts    - 29 pages document -
04/23/1992:   Huckel MO Theory software
03/11/1996:  Law of conservation of difficulty: violations.
12/15/1997:  SUMMARY of Beowulf for chemistry
02/07/1994:   High Cost of Comp. Chem.
08/01/1995:  Spin contamination, effect on energy and structure.
05/03/1995:  summary (II): teaching material for computational chemistry
11/22/1997:  EXTENDED HUECKEL--MORE INFO AND A FINAL SUMMARY


Raw Message Text