Declarative Systems & Software Engineering Newsletter
Declarative Systems & Software Engineering NewsletterI have had little in the way of additions to the delivery list, as asked for in the first issue. Remember that your project students might appreciate the feeling of more contact. And we can send it to collaborators elsewhere as well.
If you already have a cv.html with words on it, then you will want to save the words to put back later.
PVM is a software package which allows the user to treat distributed processor resources as a single parallel computer. These processor resources are distributed on a TCP/IP (usually, but not exclusively) network, and they need not all have the same CPU architecture. The PVM daemon does all the low level message passing, leaving the application to construct message buffers and call functions to send/receive messages.
The course started with an overview of what goes on in the HPCC, and was then followed by a general introduction to PVM, with a brief examination of what it gives you and how it achieves it. Example code fragments were shown in Fortran77. This was followed by an introduction to MPI and then a comparison of the two systems. In the afternoon, there was a practical session during which we could try a few exercises. Unfortunately these were all done in Fortran, although we were told that future courses would have the option of C in them.
All the slides and other material is available from the HPCC via the WWW at: http://hpcc.soton.ac.uk/homepage.html (maybe not at the moment, but RSN!) I have bound copies of the slides used which I am willing to lend to people if you just want to have a browse through it.
I felt that reading the Introduction section of the PVM manual + the function definitions in the same manual was sufficient for me to get up and running with PVM itself, although the material was interesting and it was worth going to on that basis.
The HPCC is also running courses on:
For more information about any of the courses mentioned, e-mail support@par.soton.ac.uk
The conference was jolly good, and I have the proceedings if anyone is interested. I was forced to attend all the papers in the formal methods section (rather than nipping out early to C&A next door) by virtue of the fact that I was made session chairman. Glad I stayed. There were quite a few talks on teaching Z (particularly interesting was David Lightfoot's) and much discussion on whether there was any point if you didn't teach proof. Some people were doing very brave things with students who only came in with a couple of Ds and no A level maths. One of the lecturers showed how he tested quite a lot of Z knowledge using multiple choice tests, which are available for all.
I was interested in the use people made of tools. A lecturer in the electrical eng. dept. at Bradford uses LDRA Testbed to teach his students testing. The good students find out for themselves when they have tested all poss. paths. This reminded me of some of the principles behind the Montessori teaching methods for kids. I.e. devise a learning experience where it will be obvious when they have performed correctly, rather than using an external reward system.
On similar lines, there was an excellent speaker from Australia. The aim of the course he was talking about was to get students to accept the concept of quality as a personal ideal, rather than an imposed standard. So he set a piece of code to write, and it had to compile and run first time. But he gave no marks for them achieving this. They were marked on their adherence to software quality techniques. So they had to keep a log of time spent on what, and records of walkthroughs, both for their own work and those in which they participated. They also had to write a page reflecting on the experience. I was very impressed by this particular paper.
All in all, glad I could attend (and I did buy a new coat from C&A!)
Hugh Glaser
Declarative Systems & Software Engineering Group
Department of Electronics and Computer Science
University of Southampton