Am Mi., 23. März 2022 um 15:19 Uhr schrieb Taylan Kammer
<xxxxxx@gmail.com>:

> Macros are inherently compile-time.  Use of continuations can be turned
> into a simple jump instruction after rudimentary static analysis, tricks
> for storing small numbers ("fixnums") on the stack are well-known, and
> the language doesn't force dynamic dispatch on data structures since you
> can use type-specific accessors like vector-ref.  We even have a bunch
> of type-specific equality procedures, plus eq?, eqv?, and equal?.
>
> I'd say Scheme is "high-level in all the right places." :-)

What Scheme does not yet have, though, is an efficient and usable
idiom to write generic algorithms à la those in the C++ template
library.

Some Schemes like Guile experiment with generic methods, but these
actually have a runtime overhead that shouldn't be imposed on the
program (and would be a counter-example to your general statement).
Then we have SRFI 225 using something like runtime type classes. This
will be faster than generics, but there's still some indirection
overhead compared to C++ and I am far from being sure that SRFI 225 is
specified in a way that makes inlining achievable by optimizing
compilers.

I think the right abstraction hasn't been found for Scheme yet. For
zero-time overhead, it will probably have to be based on syntactic
abstraction.