( please pardon my crosspost on this. It won't happen again)
For those involved in high-performance computing (HPC) and computational
science and engineering (CSE)*, Gentoo, is a powerful tool, both on the
desktop and on servers or clusters. With it, users can build environments
that do not waste time offering unwanted network services and system
daemons, and the binaries for the OS can be highly optimized to take full
advantage of the system architecture. In particular, this can improve the
efficiency of numerical simulations that depend on floating point
performance. Shaving 1 or 2 cycles from an operation may not seem like
much, but when the routine is executed 5 million times per calculation,
the savings rapidly add up. Also, Gentoo doesn't have a rigid release
cycle with EOL dates forced on users who have working systems.
Administrators and users have the ultimate control over when upgrades are
made and to which packages. Finally, a group of enthusiastic volunteers
have written ebuilds for over 100 different scientific and engineering
applications and libraries, making them readily accessible to the user
* Computational Science and Engineering is related to but distinct from
Science. It is not usually concerned with research into computing for its
ends, but as a tool for making advances in other fields. It covers areas as
broad as Geographic Information Systems, Bioinformatics, Computational Fluid
Dynamics, and Chemical Reactor Design, with many other fields as well.
With the potential advantages of Gentoo-based HPC and CSE environments, we
feel that the Gentoo community ought to do more to reach out to potential
users and developers in these fields. Gentoo doesn't come without a
learning curve, and even when that curve is conquered the need for a
specialized peer group is not obviated. Furthermore, Gentoo lacks name
recognition in these fields, and even when a developer or administrator
finds Gentoo and discovers the benefits of its philosophy, there is often
a sense of isolation that discourages him or her from moving forward with
Gentoo-based solutions. To address these issues, we propose the
formation of a Gentoo-science project, encompassing both desktop and
server issues, with the following list of potential project goals.
1) To act as a clearinghouse for information about the use of Gentoo
Linux in computational science and engineering, including the
installation of major third-party applications, such as Matlab and
Mathematica, which are often less than completely Gentoo-friendly.
2) To be a presence for Gentoo at technical meetings and trade shows,
both in informal settings, such as BOFs and badge button campaigns,
and in more formal settings, such as organized symposia or vendor
3) To coordinate with the sci herd and other appropriate Gentoo herds
for bug-testing, code validation, and potential tree restructuring
as necessary and desirable.
4) To provide input to project teams about the specific needs of
computational science and engineering users, in particular the
gentoo-server, gentoo-cluster, and gentoo-desktop teams.
5) To advocate with developers and independent software vendors for
increased distribution neutrality in specifying software
pre-requisites for applications, and to encourage use of packaging
systems that are ebuild-friendly.
At present, there is no official irc channel or gentoo forum or mailing
list for this proposed project. Once the level of interest is determined,
a mailing list will probably be set up to make communications easier. The
proponents of this project can often been found in #gentoo-server or
#gentoo-cluster on irc.freenode.net. Please feel free to post your
comments as followups to this message. E-mailed comments or requests to
be added to the possible mailing list should be sent to afant@...
(having "gentoo-science" in the subject would be greatly appreciated).
Thank you to everyone for your attention and thoughts about this exciting
direction for Gentoo.
JFMuggs on freenode.net
firstname.lastname@example.org mailing list