Re: [SISC-devel] Possible bug in MemoizedSymbol

On Fri, Mar 14, 2003 at 09:25:56AM -0600, Scott Miller wrote:
> > BTW, thanks for your fast answers ;-)
> 
> You're absolutely right.  Its a slow leak, but its a leak.  I've checked 
> in a fix into CVS, which does the Java APIs suggestion of wrapping the 
> value in a WeakReference.   Thanks for the help!

As an aside, if you are repeatedly calling into SISC from Java on the 
same piece of code, its best to defining that code as a function and 
then call into SISC using the eval(Procedure, Value[]) method.  This way 
you don't need to compile and optimize the Scheme code every time, just 
execute it.  You can do this with the following style code:

  Procedure p=(Procedure)r.eval("(lambda (a) (* 3 3))");
   

  ...
  Quantity a=Quantity.valueOf(3);
  int sqa=((Quantity)r.eval(p, new Value[] {a}).intValue();


I ran a small test to see the performance difference between the above
and r.eval("(* 3 3)");.  Here are the results, running each 2000 times 
in a loop:

(* 3 3)               8.6 seconds
(p, new Value[] {a})) 0.3 seconds

	Scott

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+cfoXr9IW4v3mHtQRAgNSAJ4l6jIQC5Eu/UQMk/iMCjelMu7ObwCgilTI
y11Tik6mvUXCgIi7MTvqSuA=
=m1GB
-----END PGP SIGNATURE-----