Inline C for Common Lisp

Introduction

Inline C is an idea taken from Lush - namely, the ability to just switch over to C when anything is slow or awkward in plain LISP. This is particularly useful when using outside libraries, because no LISP bindings are necessary.

Example

Here's an example of the usage (taken from test-cinline.lisp):

(defun symm-eigenvals (matrix)
  "Use GSL and BLAS to find the eigenvalues of a symmetric matrix

Usage: (symm-eigenvals
         (make-array 4 :initial-contents '(1d0 2d0 2d0 1d0)
                       :element-type 'double-float))
==>    #(3.0000000000000004d0 -1.0000000000000002d0)
"
  (declare (type (simple-array double-float (*)) matrix))
  (let* ((n^2 (length matrix))
         (n-r (multiple-value-list (floor (sqrt n^2))))
         (n   (if (zerop (cadr n-r)) (car n-r) (error "matrix isn't square")))
         (eivals (make-array n :element-type 'double-float)))
    (cin:cinline ((double* matrix) (uint n) (double* eivals))
                 """
#include <gsl/gsl_math.h>
#include <gsl/gsl_eigen.h>
                 """
                 """
gsl_matrix_view m = gsl_matrix_view_array ($matrix, $n, $n);
gsl_vector_view o = gsl_vector_view_array ($eivals, $n);
gsl_eigen_symm_workspace *w = gsl_eigen_symm_alloc ($n);

gsl_eigen_symm (&m.matrix, &o.vector, w);
gsl_eigen_symm_free (w);
                 """
                 ("gsl" "blas"))
    eivals))

Status

Right now, Inline C is only implemented for CMUCL and SBCL. UFFI was my first choice, but it doesn't give you access to the data pointer in a vector.

Future Direction

Look at CFFI for portability Provide the ability to place different pieces of C code in one compilation unit