This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: ANSI-C++ compliance?
- To: "bothner at cygnus dot com" <bothner at cygnus dot com>, "jbuck at synopsys dot com" <jbuck at synopsys dot com>, "Mike Stump" <mrs at wrs dot com>
- Subject: Re: ANSI-C++ compliance?
- From: "Geert Bosch" <bosch at gnat dot com>
- Date: Wed, 16 Sep 1998 12:05:51 +0200
- Cc: "egcs at cygnus dot com" <egcs at cygnus dot com>, "jab3 at hotmail dot com" <jab3 at hotmail dot com>
- Reply-To: "Geert Bosch" <bosch at gnat dot com>
On Tue, 15 Sep 1998 14:13:08 -0700, Mike Stump wrote:
Run egcs 1.1 sparc against the 2.7.2.1 compiler using Bench++ to
compare EH performace on the sparc, and then say this again. I
suspect you'll find that egcs 1.1 is worse.
Are you talking about performance of actual exceptions, or installing/removing
exception handlers? Table based approaches have virtually no overhead when
no exceptions occur, but are slower when actual exceptions occur. As
the occurrence of exceptions is exceptional (or at least should be) it is
much more important to minimize the overhead when no exceptions occur.
Could you please make clear what you mean by "EH" performance?
Artificial benchmarking tests can be very misleading in predicting
performance of actual properly designed applications.
Regards,
Geert