This is the mail archive of the
java@gcc.gnu.org
mailing list for the Java project.
Re: Cross-class inlining
- From: "Robin Garner" <robin dot garner at iname dot com>
- To: "Andrew Haley" <aph at redhat dot com>
- Cc: "java" <java at gcc dot gnu dot org>
- Date: Thu, 23 Oct 2003 00:15:05 -0500
- Subject: Re: Cross-class inlining
> > Still, the "all in 1" approach gets a 49% speedup over just -O3 on
> > individual modules, and 17% better than I could do by hand.
>
> Cool.
>
Actually that was a bit premature - must have been tired. What I was seeing was a 17% improvement over the hand-inlined source. Compiling the unoptimized source all on the one command line actually only gets a 12% improvement.
So: next question - is there any way to manually direct the inlining ? I've tried playing with various values of -finline-limit, but can't get better than this.
JikesRVM has an "exception pragma" idiom to direct the inliner, ie if a method throws VM_PragmaInline, then it gets inlined by the optimizing compiler. It would seem that in this case at least I can make a better judgement than the compiler as to what should be inlined.
tia,
Robin
--
__________________________________________________________
Sign-up for your own personalized E-mail at Mail.com
http://www.mail.com/?sr=signup
CareerBuilder.com has over 400,000 jobs. Be smarter about your job search
http://corp.mail.com/careers