gistfile1.clj

(defn zpmap [f coll]
  (apply concat
         (pmap #(doall (map f %))
              (partition-all (Math/ceil (/ (count coll)
                                           (float (+ 2 (.availableProcessors
                                                        (Runtime/getRuntime))))))
                             coll))))

; This is intended for a fast f and large coll, where pmap performs poorly.
; It provides a nearly 2x speedup over map on dual-core machines I've tried it on;
; a map operation that took 5.5 seconds takes 3 with zpmap. It's not providing any
; further improvement on a quad-core machine. Changing the above code to partition
; in to only two pieces has little to no effect on the time required to run on a
; quad-core machine, though it does reduce CPU usage from 390% to 200%.