> (similar to the
>native C arrays numpy exposes to C that allow getting the >speed of C).

For this in particular, there is pointer->bytevector, bytevector->pointer and bytevector-TYPE-ref, bytevector-TYPE-set! (it’s not limited to only bytes, also floats are available IIRC).

Don’t do bytevector->pointer + pointer->bytevector + … in long sequences though, or else this loop will use memory proportional to the length to this sequence. (This is intentionally undocumented (*).)

Best regards,
Maxime Devos

(*) For its application bytevector-slice, see https://lists.gnu.org/archive/html/guile-devel/2023-01/msg00042.html, which was ignored.