Re: [maemo-developers] gcc options for smallest size, best speed, etc
Hi Koen, On 4/27/06, Koen Kooi [EMAIL PROTECTED] wrote: -BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Øyvind Kolås wrote: On 4/27/06, Philippe Laporte [EMAIL PROTECTED] wrote: Hi, I would assume that this list be a top source for the following question: what are the compile options for ARM for gcc for the smallest possible size, etc. For both smallest size, and best speed, seperately of course. Or where should I look? One of the changes that have mattered for me is adding -march=armv5te to the CFLAGS. - -mtune=armv5te should yield the same improvement while retaining compatibilty with armv4. Just in case someone tries to build maemo for a strongarm platform :) Actually with arm targets -mtune only accepts values valid for -mcpu instead of -march. Unlike x86 where --march implies --mcpu. So armv5te isn't an option but something like arm920t is. Does anybody know which of the options supported by -mcpu in gcc 3.3 included with the maemo 1.1 sdk is closest to the ARM926EJ in the Nokia 770? I realise gcc 3.4 (used for the kernel build) has support for -mcpu/-mtune=arm926ej-s, so at least the future is bright :) regards, Koen Regards, David Connolly ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] gcc options for smallest size, best speed, etc
Hi, I'd expect -g -Os to do all appropriate code size optimisations regardless of the architecture. Well -g is a bit stupid in this scenario, since this means you tell gcc to optimize for size but include debugging informations too. I would recommend -O2 as the best option to use. It does a fine job deciding wether a optimization is really worth the extra bloat, which is especially important when it comes down to the small i-cache the N770's processor has. lg Clemens ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] gcc options for smallest size, best speed, etc
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Clemens Eisserer wrote: Hi, I'd expect -g -Os to do all appropriate code size optimisations regardless of the architecture. Well -g is a bit stupid in this scenario, since this means you tell gcc to optimize for size but include debugging informations too. I would recommend -O2 as the best option to use. It does a fine job deciding wether a optimization is really worth the extra bloat, which is especially important when it comes down to the small i-cache the N770's processor has. - -Os uses the same optimizations as -O2 but takes size into account, so that should be even better for small caches. From my tests -Os is sligtly faster as -O2 since you need to load less from jffs2 (less zlib stuff). regards, Koen -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.1 (Darwin) iD8DBQFEUfAsMkyGM64RGpERAg4QAJ9foVR72/3hV/YGPOXHO24kEO/sFQCdFIBx nm4+tMzQSGRs/dNCVShAFF0= =DQOO -END PGP SIGNATURE- ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
[maemo-developers] gcc options for smallest size, best speed, etc
Hi, I would assume that this list be a top source for the following question: what are the compile options for ARM for gcc for the smallest possible size, etc. For both smallest size, and best speed, seperately of course. Or where should I look? Thanks, -- Philippe Laporte Software Gatespace Telematics Första Långgatan 18 41328 Göteborg Sweden Phone: +46 702 04 35 11 Fax: +46 31 24 16 50 Email: [EMAIL PROTECTED] ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] gcc options for smallest size, best speed, etc
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Philippe Laporte wrote: Hi, I would assume that this list be a top source for the following question: what are the compile options for ARM for gcc for the smallest possible size, etc. For both smallest size, and best speed, seperately of course. Or where should I look? Have a look at http://free-electrons.com/articles/optimizations/ , that explains the most commonly used optimizations. regards, Koen -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.1 (Darwin) iD8DBQFEUIyAMkyGM64RGpERAvvoAJ9rVeBogUkxZM+BuB1LmkEcb1nw/QCgtuXs xhjNNDCWmQWlfRXLI8MRpX4= =Y58a -END PGP SIGNATURE- ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] gcc options for smallest size, best speed, etc
On Thu, 2006-04-27 at 11:03 +0200, Philippe Laporte wrote: Hi, I would assume that this list be a top source for the following question: what are the compile options for ARM for gcc for the smallest possible size, etc. For both smallest size, and best speed, seperately of course. I'd expect -g -Os to do all appropriate code size optimisations regardless of the architecture. I've noticed that, for C++, you can give -g -Os a lot of extra help if you can use -fno-exceptions too. -- Murray Cumming [EMAIL PROTECTED] www.murrayc.com www.openismus.com ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] gcc options for smallest size, best speed, etc
On 4/27/06, Philippe Laporte [EMAIL PROTECTED] wrote: Hi, I would assume that this list be a top source for the following question: what are the compile options for ARM for gcc for the smallest possible size, etc. For both smallest size, and best speed, seperately of course. Or where should I look? One of the changes that have mattered for me is adding -march=armv5te to the CFLAGS. /Øyvind Kolås -- «The future is already here. It's just not very evenly distributed» -- William Gibson http://pippin.gimp.org/http://ffii.org/ ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers