Re: Size of C++ objects
From: Albert Chin-A-Young (china_at_foo.com)
Date: 12/07/04
- Previous message: David Anderson: "Re: Size of C++ objects"
- In reply to: David Anderson: "Re: Size of C++ objects"
- Next in thread: Michael Murphy: "Re: Size of C++ objects"
- Reply: Michael Murphy: "Re: Size of C++ objects"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Tue, 07 Dec 2004 16:15:42 -0600
David Anderson <davea@quasar.engr.sgi.com> wrote:
> In article <rdydnSzehYpySSjcRVn-hg@speakeasy.net>,
> Albert Chin-A-Young <china@oof.com> wrote:
>>We're building KDE 3.3.1 on IRIX 6.5 with the SGI C++ compiler. The size
>>of C++ libraries/binaries seems to be quite large for this platform.
>>Having built only the arts, kdelibs, kdebase, and kdegraphics packages:
>> $ du -ks /opt/TWWfsw/kde33
>> 609868 /opt/TWWfsw/kde33
>>
>>Comparison with other platforms:
>> (Tru64 UNIX 5.1)
>> $ du -ks /opt/TWWfsw/kde33
>> 472602 /opt/TWWfsw/kde33
>> (RHEL 3.0/x86)
>> $ du -ks /opt/TWWfsw/kde33
>> 409512 /opt/TWWfsw/kde33
>
>>albert chin (china @at@ thewrittenword .dot. com)
>
> Since you don't mention what compiler options are in use
> nor what compiler release is in use (nor which IRIX release)
> it is impossible to make a particularly meaningful comment.
> But I'll try anyway :-)
CFLAGS="-mips4 -r5000 -O1 -OPT:Olimit=0"
CXXFLAGS="-mips4 -r5000 -O1 -OPT:Olimit=0"
$ uname -R
6.5 6.5.23m
$ CC -v
MIPSpro Compilers: Version 7.4.2m
> For example, with C++ code,
> -gslim
> versus
> -g
> versus no -g option at all
> get radically different debug-information sizes on IRIX?
> (Ignoring -g3 here, which is like -g in resulting size.)
I read about -gslim but because we don't use -g, it won't be of any use.
> Unlike gcc, with IRIX CC the operative assumption is that for
> production builds no -g option at all is used but instead -O3
> and/or IPA is used. (-O2 is good, it just depends on how
> crucial performance is for the code. If it matters,
> experiment). Given that disk space is so much cheaper today,
> and that the debug_information space on disk has NO impact on
> peformance or in-memory size (on IRIX), one would think that
> the disk space was not an issue. (Presuming one wanted it to
> make some semblance of debugging possible). Also note that
> with optimization, debugging becomes problematic as (even with
> -g3) IRIX compilers first-and-foremost go for compiled code
> speed and do not try very hard to maintain correspondence between
> the source code and the debug_info.
We use -O1 because -O2 has caused problems for us before (dunno if this
was a 7.4 based compiler though).
> The Compiler Guide on techpubs.sgi.com has more information
> on these issues.
Ok, will look. Thanks.
-- albert chin (china @at@ thewrittenword .dot. com)
- Previous message: David Anderson: "Re: Size of C++ objects"
- In reply to: David Anderson: "Re: Size of C++ objects"
- Next in thread: Michael Murphy: "Re: Size of C++ objects"
- Reply: Michael Murphy: "Re: Size of C++ objects"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
|
