S2PLOT forthcoming release

[David Barnes]

Hi all,

This is an advance notice to members of the S2PLOT Google Group that a
new release of S2PLOT is coming soon.

The next release of S2PLOT will be version 2.4. This is an
intermediate release offering a range of new features, as well as some
important system changes.

New features include:

* a new open source S2PLOT application - s2anim - for displaying time-
series of point-like data
* functions to draw wireframe and transparent solid cubes
* functions to draw textured spheres with control over the orientation
of the texture
* optional reading of control commands coming in on a socket

The architectural changes are probably more important in the
forthcoming release than the new features. Specifically:

* the cygwin architecture has been removed
* gcc compilers versions >= 4.3 are required as S2PLOT now uses OpenMP
internally
* some devices will be dynamically loaded modules
* a few minor bug fixes

Most of the architectural changes are motivated by enabling an Open
Source release of S2PLOT. This is something that has been requested
repeatedly, so we hope you are glad it is on its way!

The approximate timeline at this stage is:

* version 2.4 binary release (linux and darwin) September 12th 2008.
* version 2.4 source release October 10th 2008.

The source release will be delayed in case significant bug(s) are
discovered in the binary release.

At this stage, I am keen to hear from anyone who would like to help
test the source release and build system.

Regards,

David Barnes.

[Rick Wagner]

Hi David,

> * gcc compilers versions >= 4.3 are required as S2PLOT now uses OpenMP
> internally

Two things:

* Is an OpenMP capable compiler actually a requirement, or just an
option for getting better performance?
* Are you sure it's gcc >= 4.3? I believe gcc has supported OpenMP
from 4.2 onward.

> Most of the architectural changes are motivated by enabling an Open
> Source release of S2PLOT. This is something that has been requested
> repeatedly, so we hope you are glad it is on its way!

Yes, this is still eagerly awaited by many of us.

> At this stage, I am keen to hear from anyone who would like to help
> test the source release and build system.

Count me in. I have access to a variety of systems, and would be glad
to help.

--Rick

------------------------------------------------------------------------
-
Rick Wagner, Graduate Student Researcher
UCSD Physics
9500 Gilman Drive
La Jolla, CA 92093-0424
Email: rwa...@physics.ucsd.edu
WWW: http://lca.ucsd.edu/projects/rpwagner
(858) 822-4784 Phone
------------------------------------------------------------------------
-
Creative algorithms are cool,
creative variable names are not.
--Rick Wagner
------------------------------------------------------------------------
-

[David Barnes]

Hi Rick,

thanks for the interest. I'll add you to my list of testers and will
get in touch once the binary release is done and ready for testing of
the source release.

>   * Is an OpenMP capable compiler actually a requirement, or just an  
> option for getting better performance?

It will be required for the binary release because that will be
compiled with -fopenmp, and so any S2PLOT programs will be linked
against libgomp*.so. I decided to force this rather than double up on
all the architectures. Source releases will of course be able to
disable this requirement.

>   * Are you sure it's gcc >= 4.3? I believe gcc has supported OpenMP  
> from 4.2 onward.

You're right, and it may have even been in 4.1 in some form. However
since I do the binary build with 4.3, this defines the minimal
libgomp*.so version I suspect. Again, with the source release this
can be adjusted by the user to suit.

- David Barnes.

[Rick Wagner]

Hi David,

>> * Is an OpenMP capable compiler actually a requirement, or just an
>> option for getting better performance?
>
> It will be required for the binary release because that will be
> compiled with -fopenmp, and so any S2PLOT programs will be linked
> against libgomp*.so. I decided to force this rather than double up on
> all the architectures. Source releases will of course be able to
> disable this requirement.
>
>> * Are you sure it's gcc >= 4.3? I believe gcc has supported OpenMP
>> from 4.2 onward.
>
> You're right, and it may have even been in 4.1 in some form. However
> since I do the binary build with 4.3, this defines the minimal
> libgomp*.so version I suspect. Again, with the source release this
> can be adjusted by the user to suit.

I'm really torn about what to suggest regarding using OpenMP. On the
one hand, I think it's great that it's being adopted. But, OS X 10.4
and 10.5 are both only on GCC 4.0.1., Ubuntu 8.04 is at GCC 4.2.3,
etc. My concern is that you'll be asking users to install a new
version of GCC, when the best part about binary distributions is not
needing to install other things.

However, I'm not sure of what will break, so I'll wait until I try to
run things.

--Rick

[David Barnes]

Hi Rick,

you raise a fair point. On Mac, it is easy enough to simply download
and install the 4.4 compilers (currently experimental) from
hpc.sourceforge.net. On Linux it is generally less easy for users I
agree.

Does anyone else have a view on this?

- David.