;;; As a simple example of SIMD optimizations using SBCL ...
;;; ... compare disassembly and speed of the following

(defun deep-copy-1 (source destination)
  (declare (type (simple-array s32) source destination))
  ;; "Naive" solution
  (dotimes (index (array-total-size source))
    (setf (aref destination index) 
          (aref source index))))

(defun deep-copy-2 (source destination &aux (length (array-total-size source)))
  ;; This should enforce a bounds check; otherwise it is not functionally
  ;; equivalent to DEEP-COPY-1. You may have to use an ASSERT instead.
  (declare (type (simple-array s32 (length)) source destination))
  ;; Using SB-SIMD
  (do-vectorized (index 0 length)
    (setf (s32-row-major-aref destination index)
	      (s32-row-major-aref source index))))