This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

gcc300 vs gcc295.3 benchmarking?


I compiled & installed gcc3.00 yesterday, on a 450mhz PIII system running
Mandrake linux 6.2. Whereas certain problems (like the floating point stack
alignment problem) appear solved in this new release, it creates executables
that, to my astonishment, run significantly slower than those created by v.
2.95.3!
For example, a simple floating point dhrystones benchmark (8byte doubles)
clocks between 780 and 795 dhry/sec on my system, using a set of
optimisation flags that I determined as optimal, when using gcc295.3. With
the same flags, gcc3.0 "only" reaches between 700 and 720 dhry/sec. Using
just the -O2 flag, these figures are approx. 720 and 650 respectively.
Needless to say that this difference is still visible in my own programmes
(notably in realtime animation programmes).
Is this a known issue, why is this so, will this change in the next release,
etc? I for myself won't upgrade to a compiler that gives me slower code!

RJV Bertin
_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]