urbans@admin.kth.se (Urban Sundstr|m) (02/20/91)
This is a matter of much concern to our organisation. There are several critical aspects. The first one is performance. I got a bit disappointed when I read about the Sather language in Rick Jones entry in this newsgroup, "Sather, a news object oriented language", 90-12-12. I quote: >Preliminary benchmarks show a performance improvement over Eiffel >of between a factor of 4 and 50 on basic dispatching and functions >calls. On Stanford shelf benchmarks ...., Sather is slightly faster >than C++ etc. From this I conclude that Eiffel can be 4 to 50 times slower than C++ ! Is that true ? In such case it is a problem at least for us. Another problem is that you need a lot of numerical functions in scientific calculations like trigonometric and other elementary functions. For us are also matrix operations for linear algebra calculations mandatory. E g we are calculating the inverse of matrices with several hundreds of columns in double precision arithmetic. The calculation times are acceptable with Fortran but can very easely become very high, because the number of operations are growing proportionally to the third power of the number of columns. In Sather there seems to be classes for all these problems but we dont know anything about the performance of matrix classes. We would like to have all these features in Eiffel. If anybody has some experience of using Eiffel for this type of calculations we would very much appriciate to hear about it. Kjell Degerstedt