All, Thank you very much for the provided references.
Robby, John, Jerzy: thanks for the pointer to Jeff Siskind. His works on automatic differentiation are very interesting. I should look at his Stalingrad software. I did not think about automatic vs symbolic differentiation before; now I am convinced that I need the former. (Maybe the latter, too, in specific simple cases, with racket-cas, thanks Jens Axel) I agree that nothing that already exists can fullfill my needs completely (and did not expect that); however, some projects I can use, either directly or as examples. For DSL, Honu seems underdone; Magnolisp is richer, but not complete too (for instance, it does not have vectors). Lush is nice, but I wanted something statically typed. Terra is better as it has types. I see no reason not to look outside of the Racket world for DSL-s to be implemented in Racket. For code generation: I am yet to understand whether Fulmar will actually help (it seems to me now that the Fulmar code itself is supposed to be generated). Nanopass seems great, thanks again Jens Axel. Konrad, Neil: I think completely the same way as you do. It is just that I am not quite advanced in doing new and fancy Racket stuff, so I am going to be slow. First I have to focus on one project, and after that, on general use. RAI seems to be the closest to what I need to do. It has a DSL with arrays and matrices, it generates C code, and it even has automatic differentiation, according to the docs. It is designed for DSP, but probably can be extended to non-DSP programming. I should look at it closer. Regards, Dmitry -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.