On 4/24/05, Ivor Hewitt <[EMAIL PROTECTED]> wrote: > Cheers, pretty interesting.... although it does show that -O2 is best > on an Nemehia, and theres not much between -O2, -O3 and -Os on an > Ezra. Although the inclusion of "kernel size" as a benchmark is > clearly going to favour the -Os! :) > > I think its still worth benchmarking a multimedia app, like myth or > mplayer, for a real world conclusion. > > However, in summary, you're not going to go far wrong by simply > choosing "-march=i586 -O2" for any EPIA. > > Regards, > > > On 4/23/05, MagicITX <[EMAIL PROTECTED]> wrote: > > On 4/23/05, Ivor Hewitt <[EMAIL PROTECTED]> wrote: > > > The gcc compiler hasn't got any scheduling logic for the c3 since via > > > hasn't released that information or contributed to the gcc > > > development..... and since the c3 design is based off the old cyrix > > > dies I suspect that the 586 optimisations might be a reasonable match. > > > > > > -Os is all very well but that just tries to pack everything into the > > > smallest space and so disables all word alignment and doesn't try to > > > optimise the code for speed. -O2 seems like a reasonable compromise > > > for the c3. It doesn't unroll loops but does do some optimisations. > > > > > > Personally I always found "586 and -O2" gave decent performance... > > > with the added benefit I can use the same binaries on P3, C3 and C3-2 > > > which helps testing and debugging (for me). > > > > > > If I have time I might run some benchmarks with different options. > > > > > > > > > On 4/22/05, Devan Lippman <[EMAIL PROTECTED]> wrote: > > > > I'd have to agree with Ivor's flag settings, I've found on my TC that > > > > i586 works better than C3 > > > > > > > > On 4/22/05, Ivor Hewitt <[EMAIL PROTECTED]> wrote: > > > > > On 4/22/05, Neale Swinnerton <[EMAIL PROTECTED]> wrote: > > > > > > Hi, > > > > > > > > > > > > I've got a Epia ME6000 working well as a Myth frontend. I've got > > the > > > > > > XvMC playback so that watching Live TV or recordings is great (I've > > got > > > > > > a backend with 2 x Hauppauge DVB-T cards). > > > > > > > > > > > 1. I'd have thought an m6000 would struggle. Out of interest what cpu > > > > > load are you seeing in mplayer? > > > > > 2. Why are you transcoding dvb-t recordings anyway? do you really > > need > > > > > to save disk space that badly. Wouldn't you rather watch the original > > > > > quality? > > > > > > > > > > > -march=c3 -Os -fomit-frame-pointer > > > > > > > > > > > What about just -march=i586 -O2 > > > > > _______________________________________________ > > > > > mythtv-users mailing list > > > > > [email protected] > > > > > http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users > > > > > > > > > > > > > > > > > -- > > > > Thanks, > > > > Devan Lippman <[EMAIL PROTECTED]> > > > > > > > _______________________________________________ > > > mythtv-users mailing list > > > [email protected] > > > http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users > > > > > > > I did a comparison of compiler optimization settings for the ME6000 > > and M10000. The results are mixed. You can find the articles on > > www.magicitx.com under the VIA Epia menu -> Optimization if you're > > interested. > > > > -- > > Tim > > www.magicitx.com > > >
I agree. I you happen to test compiler optimizations on multimedia apps and want somewhere to publish it. Let me know. I'm certainly interested. -- Tim www.magicitx.com _______________________________________________ mythtv-users mailing list [email protected] http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
