Performance drops with 1.1.2?!

[Thorsten Kramp]

N8TM@aol.com writes:

> If you're looking for speed, why not '-Os -malign-double -funroll-loops' ?

Yes, with these compile-time options the gap closes.  The performance
of both versions is now similar again with a derivation of 1% at most
(for some tests in favour of the code generated by egcs 1.1.2 and for
others in favour of egcs 1.1b).  Thanks to all who responded!

Best, Thorsten

-- 
Thorsten Kramp                      phone  +49 631 205 3295
University of Kaiserslautern        fax    +49 631 205 3558 (shared)
Department of Computer Science      email  kramp@acm.org
Distributed Systems Group           pgp    available on request