Projects in Scientific Computing

• I'm interested in scientific programming languages and methodologies. Most physicists use Fortran-77 as their main language, but for many tasks I believe that this language is unsuitable; I don't believe Fortran-77 is "scale-invariant". Most physicist who do not use Fortran-77 have moved to either C or C++ as their main implement. I think these languages may be fine for those among us who are more or less expert programmers, but I have met very few of these physicists. Something else is needed. So one thrust of my research is into the suitability of various languages for scientific programming by the "lay" programmer.

  Some of the languages I've considered so far:

  • Fortran 90/95. I am involved with a project to convert the ESPACE code at TJNAF's Hall A from VAX Fortran to F95. The theory is that it will not take long since most of the vanilla Fortran stuff will not need to change. We'll see.
  • Python. Python is not really in the same class as Fortran or C, but is a competitor to Tcl or Visual Basic. It is an object-based language, which I find to be very useful. I use this for most of my programs which are not CPU intensive. It's an interpreted language, so you can just type stuff in and run it, or you can put your program in a file and run it. I have a four-vector class I use all the time.
  • Eiffel. I wrote a database manager program in Eiffel, and then threw it away. Eiffel is a very nice language in theory, but I found it lacking in "expressiveness". If you want to do things "the Eiffel Way", you quickly end up with lots of classes flying around. If you want to code in Eiffel in a Fortran-ish fashion, it is difficult because there just aren't many built-in features in Eiffel. Sather is a spinoff of Eiffel which is designed to be more efficient and more aimed at the scientific user, but I have not tried it. Development appears to have stagnated; the last release of Sather was in September 1996.
  • Oberon. Oberon seems to be a really nice language with lots of interesting and useful features. It loses out on the implementation; there are two competing versions of Oberon out there (Linz and ETH), and neither have made it far out of the stone age as far as Unix is concerned. Also, Oberon the language is intimately tied into Oberon the System, and this system is bizarre enough (compared to X11) that it discourages casual users. It's a shame, because it seems like a nice language.
  • Ada 95. This is the new version of the infamous Ada programming language which was the lanugage you were required to use by the U.S. Department of Defense, if you were doing military-related programming work. I thought it was silly then, but now I'm not so sure. They have done a nice job with the language, and Ada 95 includes object orientation. I have not done any substantial programming in this language, but it's at the top of my list.
  • C++. I haven't included a link, one should not have much trouble locating such a link. I have done little programming in C++, but have recently (10/97) read a book (Practical C++ Programming, Steve Oualline, O'Reilly) on the language. My initial reaction: I have never seen an introductory book spend so much time telling you how to avoid many features of the language so that you don't hang yourself. This has always been my impression about C++. I don't consider it a good language for non-experts. You might enjoy an article by M. M. Steiner et al in the September/October issue of Computers in Physics. Nobel Laureate K. G. Wilson is a co-author on this article. They conclude that C++ is "good", but while doing so they indicate that it takes about twice as long to learn the grammar of C++ as it does that of Fortran, and it takes many times this long to learn how to program in an Object-Oriented fashion. The article is peppered with comments that reinforce the above conclusions: C++ is not for pedestrian programmers, unless they are predominantly using libraries of code generated by experts, writing little real code themselves.

• Information about the Linux operating system. I don't use anything else anymore unless I have to. The information here includes some general Linux pointers as well as some information specifically tuned to doing physics on a Linux workstation.

  If you're curious what the Linux working environment looks like, click here to see a screen dump from my workstation.

• The Linux Fortran Information Page which attempts to be an exhaustive reference on Fortran compilers and tools available for PC/Linux platforms.
• I maintain the Linux version of the program ESPACE, which is the data-analysis code for Hall A at Jefferson Lab. This project involved porting CERNLIB to Linux using the Absoft F77 compiler. See the ESPACE/Linux Page for more information.
• More comments on programming languages for scientific work. These comments are my own opinions and should not be construed to be anyone else's.