optimization

Re: optimization

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

nos wrote:
[color=blue]
> Does anyone know if the java compiler 1.4.2
> from SUN optimizes subexpressions? I.e.
> in the following two lines
>
> int tcount =
> Thread.currentT hread().getThre adGroup().enume rate(myThreadLi st);
> String tcurrent = Thread.currentT hread().getName ();
>
> will it evaluate Thread.currentT hread();
> only once?
>
> just curious[/color]

Hi,
I'm no expert, but I doubt the *compiler* could do that, since it
doesn't know that Thread.currentT hread() doesn't have any
side-effects. It would have to make sure it gets called twice. After
all, it's theoretically (practically?) possible that the Thread
implementation could be replaced later with something that does have
side-effects. One possible place for such optimization would be a
highly-intelligent JIT, which "knows" about its own attached runtime
(i.e. it knows that Thread.currentT hread() has no side-effects). I
don't think anything like this exists, though.
- --
Chris
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)

iD8DBQE/rC3YwxczzJRavJY RApZmAJ48PgCVef/bVmx4AUeIuo9Usw bn9wCgwPPj
yqrDWpeTkDQFzLo i19FliRY=
=xmxk
-----END PGP SIGNATURE-----

Re: optimization

very good point!
on some earlier project we discovered that when
we had two calls to the random number generator
(not in java) that the compiler optimized it to only
one call -- so we had to put a line number on
the second call (FORTRAN has line numbers)

"Chris" <chris2k01@hotm ail.com> wrote in message
news:o71rb.3050 $jy.544@clgrps1 3...[color=blue]
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> nos wrote:
>[color=green]
> > Does anyone know if the java compiler 1.4.2
> > from SUN optimizes subexpressions? I.e.
> > in the following two lines
> >
> > int tcount =
> > Thread.currentT hread().getThre adGroup().enume rate(myThreadLi st);
> > String tcurrent = Thread.currentT hread().getName ();
> >
> > will it evaluate Thread.currentT hread();
> > only once?
> >
> > just curious[/color]
>
> Hi,
> I'm no expert, but I doubt the *compiler* could do that, since it
> doesn't know that Thread.currentT hread() doesn't have any
> side-effects. It would have to make sure it gets called twice. After
> all, it's theoretically (practically?) possible that the Thread
> implementation could be replaced later with something that does have
> side-effects. One possible place for such optimization would be a
> highly-intelligent JIT, which "knows" about its own attached runtime
> (i.e. it knows that Thread.currentT hread() has no side-effects). I
> don't think anything like this exists, though.
> - --
> Chris
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.2.2 (GNU/Linux)
>
> iD8DBQE/rC3YwxczzJRavJY RApZmAJ48PgCVef/bVmx4AUeIuo9Usw bn9wCgwPPj
> yqrDWpeTkDQFzLo i19FliRY=
> =xmxk
> -----END PGP SIGNATURE-----[/color]