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]
Other format: [Raw text]

Re: gcc compile-time performance


On Sat, May 18, 2002 at 07:36:02PM +0200, Marc Espie wrote:
> In article <78200000.1021742887@gandalf.codesourcery.com> you write:
> >Agreed.  You need to look for order-of-magnitude improvements -- of which
> >there are definitely some to be had in G++.  Precompiled headers, already
> >being worked on, is one excellent example; for many programs this technique
> >has been shown to make a huge difference.
> 
> Forgive me for trying to get things back on another track, but g++ speed
> isn't the only issue.
> 
> The compilation speed problems I've reported about the OpenBSD kernel deal
> EXCLUSIVELY with C code, and not really modern C code at that.
> 
> 20% for the bare-bones compiler is no joke, C++ problems notwithstanding.

... and there are definitely algorithmic improvements to be had in the
C compiler and the generic optimizers.  The old CSE pass has already
been mentioned.  There's at least one nasty quadratic algorithm in
tree->RTL lowering, which both I and Richard Henderson failed to
eliminate.  The garbage collector, with its 2x slowdown on low-memory
machines, has multiple angles of attack -- most significantly, we
ought to be able to get generational GC out of Geoff's PCH work, which
will be O(garbage) and not the current O(total allocated memory).

I'm sure there are others.

zw


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