From: Eric Hodel <drbrain@...7.net>
Date: 2003-02-27T12:14:16+09:00
Subject: Re: Ruby Compile-time optimization


--Kj7319i9nmIyA2yE
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

Daniel Carrera (dcarrera@math.umd.edu) wrote:

> Now, as many of you know, Perl is not exactly an interpreted language. =
=20
> Perl first "compiles" the program into some internal bytecode, which is=
=20
> optimized before being run.  For instance, if you have this:
>=20
> use constant  $MYCONST =3D> 3
> $val =3D 2 + $MYCONST
>=20
> Perl turns it into:
>=20
> $val =3D 5
>=20
> I ran the ruby code again, but this time I replaced 'val =3D 2 + 2' by 'v=
al=20
> =3D 4' to simulate the effect of optimization.  Again, the average of 3=
=20
> tries:
>=20
>                  Total     Minus-the-loop
> ruby             2.660     1.637     100.0%
> ruby -optimized  1.793     0.770      47.0%
> perl             1.770     0.630      38.5%
>=20
> Now ther Perl's operation is only about 18% faster, which in this case=20
> particular case is mitigated by Ruby's loop being 11% faster.  Overall,=
=20
> the Perl code is only 1% faster (which is less than experimental error).
>=20
> I realize that the fact that Ruby is a more dynamic language makes=20
> optimization more difficult, but I really think that we should look into=
=20
> it.

If you poke around in the mailing lists, you'll find it has been brought
up before.  In short, Fixnum#+ may be redefined at any time, so
constant folding won't work.

--=20
Eric Hodel - drbrain@segment7.net - http://segment7.net
All messages signed with fingerprint:
FEC2 57F1 D465 EB15 5D6E  7C11 332A 551C 796C 9F04


--Kj7319i9nmIyA2yE
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (FreeBSD)

iD8DBQE+XYKFMypVHHlsnwQRAueIAJ9RUNd3pcfQ0IM4vYpZYSiUsycOtgCgxHTX
eGkM/X6HPrIMVgWeR0bPsFs=
=Qivw
-----END PGP SIGNATURE-----

--Kj7319i9nmIyA2yE--