Re: [Matlisp-users] Matlisp subset in pure CL

[Raymond T. <to...@rt...>]

>>>>> "Nicolas" == Nicolas Neuss <Nic...@iw...> writes:

    Nicolas> Maybe it is also possible with the help of CL vendors and developers to
    Nicolas> reduce the FFI call overhead.

Speculation here, but the FFI call overhead itself is probably quite
low.  The cost currently is turning off GC, extracting the addresses,
putting an unwind-protect around the FFI call, and setting up the FP
modes to what Fortran wants.  If we had immovable foreign vectors and
just set the FP modes to what Fortran wants, then a lot of the
overhead would go away.

    Nicolas> For the more distant future, I dream of a seamless transition between
    Nicolas> Matlisp and CL-Matlisp, e.g. along the following lines:

    Nicolas>   1. Use the CL-implemented version, if there is any, and if the matrices
    Nicolas>      are small.

And the nice thing is we can even get rid of the call overhead if the
function could be inlined.  That's a huge win for small
matrices/vectors. 

Ray