[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Vectorizing Dylan?

At 10:23 AM 10/26/92 -0500, Scott_Fahlman@SEF-PMAX.SLISP.CS.CMU.EDU wrote:
>     Since you are working on a language for the 1990ies, where I expect
>     to see more vector processors than today, I wonder whether it would 
>     not be appropriate to include vectorizing constructs into Dylan from
>     the beginning.
> This sounds like a good job for a library.  The nice thing about the way
> Dylan is being organized is that there's no sharp division between what's
> "in" the language and what is available as a well-known and widely used
> library.  The trick will be to get this package out early and in solid
> shape.
> The only tricky thing will be to make sure that the rules for sealing and
> inlining are such that a compiler can take advantage of any primitive
> high-speed vector operations that a machine might support.  But this same
> requirement comes up all over the place (for example, in a machine whose
> ALU supports SIN or MULTIPLY/ACCUMULATE or operations on primitive ASCII
> strings), and I'm sure it's being worked on by the Dylan designers.

Yes, it is.  The purpose of sealing is to allow this kind of efficient