[plt-scheme] R6RS Interoperability

As a newbie, I wonder if explicit typing alone would allow for a large
enough performance gain. I have the vague impression that Stalin for
example, touts "whole program optimization", whatever that is, as the
main reason for its performance.

What order of performance gains can you expect from explicitly typing
vs allowing compiler to infer type, given a whole program and its
typical inputs? (Just curious.)

Or more generally, what would be the first 3 performance optimizations
PLT people would implement, given infinite resources?

--Yavuz

On Fri, Jun 27, 2008 at 09:35, Mark Engelberg <mark.engelberg at gmail.com> wrote:
> On Thu, Jun 26, 2008 at 11:14 PM, Noel Welsh <noelwelsh at gmail.com> wrote:
>> I'd like more speed as much as the next programmer, but I expect
>> getting funding for compiler development is very difficult as the
>> basics have been done many times over and therefore have very little
>> research value.
>>
>> N.
>>
>
> Are there any mainstream languages out there which allow you to code
> your program quickly using a dynamically-typed language, and then
> gradually add type annotations to improve performance?  As I
> understand it, "gradual typing" is an active area of research, and
> Typed Scheme brings new techniques to the table, in terms of providing
> soundness in interaction between modules and allowing fairly
> straightforward annotation of idiomatic Scheme code.  Since
> performance is a key goal of gradual typing, it seems like a
> reasonable litmus test to apply to Typed Scheme -- does Typed Scheme's
> approach offer the same sorts of optimization opportunities as a
> statically-typed language, or are there complications caused by its
> interoperability with dynamically-typed modules?  Seems like a
> genuine, useful research topic to me...
>
> --Mark
> _________________________________________________
>  For list-related administrative tasks:
>  http://list.cs.brown.edu/mailman/listinfo/plt-scheme
>