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 Release Status (2003-08-22)


> On Fri, 29 Aug 2003, Mark Mitchell wrote:
> 
> > I don't think we'll have any trouble getting to a release on the C++
> > front.
> >
> > Compile times are another major story.  I'm not sure what the full
> > situation is at this point, and I'm not sure we even know what
> > apples-to-apples means.
> 
> I'm afraid there are some memory consumption regression presented in
> gcc3.4, i.e. on my machine (.5GB RAM + .5GB swap) it compiles some MICO
> files and cosumes about 300MB memory where gcc3.2.3/3.3.1 consumes only
> about 50-70MB. I'm talking here about C++ compiled with -O0. On the other
> hand, there is one file which 3.4 is not able to compile with -O1/-O2,
> since it is not able to find free fp register (IIRC).
> 
> Yes, I know, I should report both issues, but I guess few MBs of
> preprocessed files are useless for you and on the other hand I've not been
> able yet to find smaller examples for issues duplication. :-(

What -funit-at-a-time switch does to your compile times?
In case it does not help, can you send me the preprocessed files
privately?

Honza
> 
> > For example, should we assume people will be
> > using PCH?
> 
> I'm not sure, since to use gcc's pch, the user will need to do:
> 
> 1) some source code changes to gain maximal speed
> 2) some build framework changes to add support for manual pch processing
> 
> Speaking again for MICO project, I have already done 1. point, since I'm
> using Comeau C++ for MICO development (reason: speed with PCH on). On the
> other hand I'm not sure if it's worth the effort to do 2. point just for
> gcc3.4, especially when gcc compile server seems to be on the good way to
> support some kind of automatic processing... but well, maybe I will do it
> at the end, since Comeau doesn't provide mt-safe run-time...
> 
> BTW: During experiments with gcc's 3.4 PCH, I've been able to compile
> MICO's ORB core directory in about 4 minutes - gcc2.95.4 compiles in
> about 6 minutes and gcc3.2.3 compiles the same in about 12 minutes. Comeau
> with PCH on is about 4-5 minutes.
> 
> Cheers,
> 
> Karel
> --
> Karel Gardas                  kgardas@objectsecurity.com
> ObjectSecurity Ltd.           http://www.objectsecurity.com


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