RFC: naming for FrontTransversal and Transversal ranges

[bearophile]

Bill Baxter:
>Strange though, aren't you the guy always telling us how being able to express your algorithm clearly is often more important than raw performance?<

Nope. I have said many things :-) One of the things I have said is that most of the code of most programs isn't performance-critical, so for the non-performance-critical code it's better to have idioms that allow to avoid bugs and write in a short and expressive way. Forcing the programmer to write all the program in a low level or all of it in a high level way is bad for programmer productivity or overall program efficiency.

Also, regarding 'express your algorithm clearly' many of the things inside the Chapel language look clear and not too much difficult to learn and use. I have compiled the Chapel compiler and I have tried to write some small programs, such experiments show me that I am able to write them with not many problems. I can learn 20-30% of such language in a matter of hours. So it shows ways to have fast data-parallel code that is simple enough to write. For example I am trying to implement tiled-based 2D matrix processing in D. It's not an easy thing to do in D, if you want to do it efficiently. In Chapel such (basic, for efficiency) thing is quite easy, and partially done by the compiler already.

Bye,
bearophile