Marcin 'Qrczak' Kowalczyk escribió: > 2009/9/5 Antonio Vieiro <[email protected]>: > >> Scheme is highly parallelizable, so I don't see a reason >> to program multithreaded code. The evaluator should do that for us. > > It's harder than it seems. I claim that an evaluator of a Scheme-like > language (where mutations to objects shared between parts of the > program can happen anywhere) will never be able to automatically > parallelize non-trivial computations. >
I'd be happy with a parallel evaluation of arguments, for instance, that takes advantage of multicore systems. As in Schemik [1], for instance. A "small Scheme" R7RS that clarifies the semantics of parallel evaluation (where appropriate) would be perfect. Cheers, Antonio [1] http://schemik.sourceforge.net/ "Schemik is a high-level lexically-scoped implicitly-parallel dialect of Scheme and Common LISP, this means the parallel execution of programs is done independently of the programmer and each program written in Schemik always produces the same results no matter which parts of the program are executed simultaneously" _______________________________________________ r6rs-discuss mailing list [email protected] http://lists.r6rs.org/cgi-bin/mailman/listinfo/r6rs-discuss
