Fisrt thing is I still don't catch why you can overclock... I understand that someone is afraid to (I am too, so I don't until my system is well behind new market products so the increase of perfomance is desired using many programs and I break somthing is cheap to replace or can upgrade to something better ;-) Anyway, there are programs that do the overclock directly into mainboard/graphic bios so you can do them from DOS/Windows and make them permanent (dangerous, as well).
I think option -O4 maybe specific to a certain compiler (egcs, for example). In the past I used djgpp (amazing gcc port to DOS) and also used a now dead fork of it for pentium specific and had up to -O6 !!! I just googled a bit and have seen it exists in egcs :-)) and there are also some proposal of -O4 optimizaion standard. I don't know about voodoo ;-)) (have nvidia based) but maybe you can have a look in wonderful /proc tree: cat /proc/dma --> You should have dma on cat /proc/devices --> My nvidia device is listed here, so should Vodoo cat /proc/pci --> A list with all devices attached to pci bus (not very useful in this case IMHO) In my case (nvidia) I have (the most informational): cat /proc/driver/nvidia/version --> Driver version an gcc version used to compile it. cat /proc/driver/nvidia/cards/0 --> IRQ and graphic card bios cat /proc/driver/nvidia/agp/card --> Card facilities used/available (1x, 2x,... fast writes,...) cat /proc/driver/nvidia/agp/host-bridge --> Same as before but for motherboard bridge Humm, I didn't know all the info it provided! ;-)) If you want to go on with gcc optimizations, you can try adding: -fexpensive-optimizations to Makefile and playing with: -funroll-loops -funroll-all-loops -malign* options (these will speed up things but may and probably should overfill cpu cache so effect can be positive or negative depending on each program and each cpu power/cache size ratio). Good luck, again! El sáb, 28-12-2002 a las 00:36, Olaf Marzocchi escribió: > You are right!! unfortunately, I already use an mplayer I compiled for i686 > with the option -O4 (it was automatically set by the ./configure command, I > don't know what -O4 is: I always heard of -O3 as best optimization level). > I asked about xfree because I know the kernel shouldn't improve the speed much. > I'll try -O3 even if I think that the O4 switch, not reported as error at > compile time, already optimize the code well. > Unfortunately, I cannot oveclock the system: read the specs below. > So, are there around xfree srpms I could use? If not, is following the > instructions provided in the xfree site trouble free (with mandrake, I mean)? > > Maybe there is another choice... linux didn't recognize my vodoo banshee > pci well: mplayer cannot use it even if I enabled the appropriate options > at the compile time. It seems that the 3dfx (or similar, I don't remember) > device is not present. > If someone tells me (in addition to the previous question) how to be sure > to have the video card recognized, I can try to use that as vodeo output (I > use xv now, I tried everything but it's the fastest with my system). > > Thank you very much > Olaf > > At 21.15 27/12/2002, you wrote: > >Don't know how hard it will be, but why don't you try to recompile divx > >player/driver yourself? Probably you will get better improvement with > >this than playing with kernel if you are not short of memory or running > >many services at the same time. I have never played with Xfree, but it > >seems to me that you could get more improvement than playing with kernel > >in this precise case. > >If you want to have a look on recompiling divx stuff, you'll probably > >find in source files an archive called something like 'Makefile' in wich > >you can change the optimization levels for gcc. Most programs (including > >kernel) are optimized in level 2 (-O2) and compatible with any 386 or > >(as in Mandrake) compatible with classic Pentium. Maybe you can change > >this to your specific processor and change some optimizations: > >-O3 --> Mainly will tell gcc that it can re-order machine level > >instructions to minimize queues of instructions waiting to be procesed > >-march=athlon --> Object code will be athlon processor specific. > >Depending on your processor it will be i686, i586, pentium, ... > >See 'man gcc' for more information are there are more options, but they > >will probably give you less improvement. > > > >I have had up to 12% improvement on a program I did myself just by using > >-O3 and processor specific. > > > >Another thing is trying to overclock a bit the graphic card and/or main > >board/processor. > > > >Good luck!!! (and tell us about!!! ;-)) > > > > ********************************** > Configuration: Celeron 400 chipset sis620, 224 MB ram cas2, 40 GB HD IBM > GXP 60, integrated audio CMI 8738, integrated ethernet davicom901A, PCI > Vodoo Banshee 16MB (plus the sis 620 integrated in the chipset, actually > disabled), LG 8080B CD-RW. > Mandrake 9 download edition. > > > ---- > > Want to buy your Pack or Services from MandrakeSoft? > Go to http://www.mandrakestore.com -- Diego Dominguez __/\__ | | Andalucia / \ Spain \ / |__ __| \/ ___________________________________________________ Yahoo! Postales Felicita las Navidades con las postales más divertidas desde http://postales.yahoo.es
Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com