Combining Ken and David's tips, this version processes a byte-array of
sizd 1920000 in about 13 milliseconds on my machine:
(def buffer-size 1920000)
(def array (byte-array buffer-size))
(defn java-like [^bytes cpuArray]
(loop [i (int 0)]
(if (< i buffer-size)
(let [b (aget cpuArray i)
g (aget cpuArray (unchecked-add i (int 1)))
r (aget cpuArray (unchecked-add i (int 2)))
a (aget cpuArray (unchecked-add i (int 3)))]
(aset cpuArray i a)
(aset cpuArray (unchecked-add i (int 1)) b)
(aset cpuArray (unchecked-add i (int 2)) g)
(aset cpuArray (unchecked-add i (int 3)) r)
(recur (unchecked-add i (int 4)))))))
user=> (time (java-like array))
"Elapsed time: 13.648662 msecs"
Found something interesting when I tried to emulate how Robert's
version aliased 'unchecked-add' as '+':
(defn java-like [^bytes cpuArray]
(loop [i (int 0)]
(if (< i buffer-size)
(let [+ unchecked-add
b (aget cpuArray i)
g (aget cpuArray (+ i (int 1)))
r (aget cpuArray (+ i (int 2)))
a (aget cpuArray (+ i (int 3)))]
(aset cpuArray i a)
(aset cpuArray (+ i (int 1)) b)
(aset cpuArray (+ i (int 2)) g)
(aset cpuArray (+ i (int 3)) r)
(recur (+ i (int 4)))))))
When I try to compile this, Clojure complains that I'm trying to
rebind i, a primitive local, with a value of the wront type in
'recur'.
It seems that 'unchecked-add' returns a primitive (note that in the
first version, 'recur' happily accepts the return from 'unchecked-add'
without coercion), but when 'unchecked-add' is bound to a new name,
the return gets boxed.
Is this the correct interpretation, or am I missing something?
On Jan 28, 9:06 am, David Nolen <dnolen.li...@gmail.com> wrote:
> On Fri, Jan 28, 2011 at 10:48 AM, Robert McIntyre <r...@mit.edu> wrote:
> > I tried to convert this java code line for line to clojure to compare
> > the speed differences, and boy was I surprised!
>
> > public static void ConvertToAWT(byte[] cpuArray){
> > // Given an array of bytes representing a c-style bgra
> > image,
> > // converts to a java style abgr image
> > int len = java.lang.reflect.Array.getLength(cpuArray);
> > for (int i = 0; i < len; i+=4){
> > byte b = cpuArray[i+0];
> > byte g = cpuArray[i+1];
> > byte r = cpuArray[i+2];
> > byte a = cpuArray[i+3];
> > cpuArray[i+0] = a;
> > cpuArray[i+1] = b;
> > cpuArray[i+2] = g;
> > cpuArray[i+3] = r; }}
>
> > (defn java-like []
> > (loop [i (int 0)] (if (< i buffer-size)
> > (let [ + clojure.core/unchecked-add
> > b (aget cpuArray i)
> > g (aget cpuArray (+ 1 i))
> > r (aget cpuArray (+ 2 i))
> > a (aget cpuArray (+ 3 i))]
> > (aset-byte cpuArray i a)
> > (aset-byte cpuArray (+ 1 i) b)
> > (aset-byte cpuArray (+ 2 i) g)
> > (aset-byte cpuArray (+ 3 i) r)
> > (recur (int (+ i 4)))))))
>
> > (defn clojure-like []
> > (doall (flatten (map (fn [[b g r a]] [a b g r]) (partition 4 4
> > cpuArray)))))
>
> I'm assuming you're using 1.2.0. cpuArray needs to be hinted. All the
> literals also need to be hinted. Don't use aset-byte, use aset.
>
> In 1.3.0 you no longer have to hint the literals and you can use the
> *unchecked-math* compiler flag instead of redefining unchecked-add as a
> local.
>
> David
--
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
