fast.clj

;; 5 microsecs
                 (loop [a-row (int 0)
                        a-idx a-offset
                        b-idx b-offset
                        c-idx c-offset]
                   (when (< a-row a-rows)
                     (loop [a-col (int 0)
                            a-idx a-idx
                            b-idx b-idx
                            c-idx c-idx]
                       (when (< a-col a-cols)
                         (aset c-data c-idx
                               (* (aget a-data a-idx)
                                  (aget b-data b-idx)))
                         (recur (inc a-col)
                                (+ a-idx (aget c-strides 1))
                                (+ b-idx (aget c-strides 1))
                                (+ c-idx (aget c-strides 1)))))
                     (recur (inc a-row)
                            (+ a-idx (aget a-strides 0))
                            (+ b-idx (aget b-strides 0))
                            (+ c-idx (aget c-strides 0)))))

slow.clj

;; 20 microsecs
                 (loop [a-row (int 0)
                        a-idx a-offset
                        b-idx b-offset
                        c-idx c-offset
                        acc nil]
                   (if (< a-row a-rows)
                     (let [new-acc
                           (loop [a-col (int 0)
                                  a-idx a-idx
                                  b-idx b-idx
                                  c-idx c-idx
                                  acc acc]
                             (if (< a-col a-cols)
                               (let [new-acc
                                     (aset c-data c-idx
                                           (* (aget a-data a-idx)
                                              (aget b-data b-idx)))]
                                 (recur (inc a-col)
                                        (+ a-idx (aget c-strides 1))
                                        (+ b-idx (aget c-strides 1))
                                        (+ c-idx (aget c-strides 1))
                                        new-acc))
                               acc))]
                       (recur (inc a-row)
                              (+ a-idx (aget a-strides 0))
                              (+ b-idx (aget b-strides 0))
                              (+ c-idx (aget c-strides 0))
                              new-acc))
                     acc))