For the heavy lifting and fancy math, gobs of brainpower and R&D dollars
have gone into things like CUDA, BLAS, Torch, TensorFlow, R and
Matlab/Octave. There is no conceivable way we can approach that level of
features and performance from scratch. "If you can't beat them, join
them" would be the winning move here :)
+1. Unfortunately, making such things portable falls down on the FFI problem. For people with really fast compilers and unboxed flonums, modifying the backend of the Fortran-to-CL compiler at <
http://quickdocs.org/f2cl/> to generate Scheme instead of CL might be worth doing.
For prior art on multi-backend matrix stuff that integrates into Lisp,
look into Clojure's core.matrix and Incanter.
Clojure doesn't have portability concerns, as it is only compatible with itself (and not so much).
John Cowan
http://vrici.lojban.org/~cowan xxxxxx@ccil.orgThat you can cover for the plentiful and often gaping errors, misconstruals
and disinformation in your posts through sheer volume -- that is another
misconception. --Mike to Peter\