You are not logged in.
NEW:
There are now three databases to make things easier. Add any or all of these to pacman.conf
[pentium-m]
Server = http://www.naderehvandi.net/arch
[lifebookp7010]
Server = http://www.naderehvandi.net/arch
[nightfrost]
Server = http://www.naderehvandi.net/arch
[Pentium-m]
(contains packages optimized for pentium-m processors)
glibc 2.3.5
cairo 1.0.2 (compiled with support for glitz)
glitz 0.4.4
qt 3.3.5
gtk2 2.8.7
XorgRC3
e17: (groups: efl, e17, e17-extra)
The e17 packages are based on shadowhand's old pkgbuilds (thanks a lot for that!). The troubles I'm currently having with e17 involve entrance and the additional modules. Entrance is a bit tricky to configure nicely, and some of the modules are just causing me various troubles. (I hope this isn't gonna cause problems with founix's e17 repo; I'm offering these here as an alternative, not a general replacement, the same way I'm offering qt and gtk2.)
[nightfrost]
(contains packages optimized for i686 and above)
anacron
aspell-sv
audio-convert
aurbild
bin2iso
bmp-crossfade
bmp-cueinfo
bmp-cvs
bmp-mac
bmp-musepack
cairo-gtk-engine-cvs
ecamegapedal
fceu
kdebluetooth
kdmtheme
ksplash-engine-moodin
mac
metabar
powernowd
qpkg
udftools
vbetool
xvkbd
yakuakecd
subfonts (this installs font-arial-iso-8859-1 to /usr/share/mplayer)
[lifebookp7010]
(contains packages specifically for the fujitsu-siemens lifebook p7010)
kernel26archck5-2_lifebookp7010-2.6.14archck5 (configured for the fujitsu-siemens lifebook p7010)
ieee80211 (needed for ipw2200)
ipw2200
rt2x00
slmodemd (don't know if this one works)
-------------------------------------
I have a bunch of packages that I use for myself compiled with the following flags for some extra tweaky performance:
export CFLAGS="-march=pentium-m -O2 -pipe -fomit-frame-pointer"
export CXXFLAGS="-march=pentium-m -O2 -pipe -fomit-frame-pointer"
export LDFLAGS="-Wl,-O1 -Wl,--as-needed -Wl,--sort-common -s"
I'm not sure how nicely I can maintain the repo, and I can't guarantee that it's gonna work out for you. If it does, good for you. If it doesn't post your troubles here and I'll see what i can do.
Those packages that are also part of the official repos have been compiled using the official pkgbuilds, with the exception of cairo to which I added --enable-glitz.
Offline
(Note: if anyone's planning to use the kernel, please tell me first. Right now there's neither a conflict= array nor a provides= array. Is it possible to add "conflicts=('kernel26archck=2.6.14.archck5')"? If not, how do I make it conflict with one particular version of a package?)
I will not use the repo since I don't have a pentium-m, but just a note on the kernel:
If you've changed the name, which you basically did, it doesn't conflict with anything, and it doesn't have to provide anything, so you are good to go
Offline
Thanks for the input. However, it installs modules to /lib/modules/2.6.14-archck5, as does dibble's archck package from community...
Offline
Change the _patchver variable in the PKGBUILD to something else and you're set.
Offline
Thanks. I'll try that at my next kernel compile... right now; I'm gonna give my old laptop a rest
Offline
Uhhh... Why optimization? Beyond -O2 and -march=i686, the impact starts dropping pretty fast on x86 machines.
(Ahh well, have it your way, but it seems to me that optimizing for specific processors is a waste of time.)
Offline
Uhhh... Why optimization? Beyond -O2 and -march=i686, the impact starts dropping pretty fast on x86 machines.
(Ahh well, have it your way, but it seems to me that optimizing for specific processors is a waste of time.)
but it matters to them, just as using kde might be a waste of memeory this might be a waste of cpu-ti,me, but then again maybe kde does the job for one guy, and this optimization might get more speed out of the computer´...it is all up to each one to decide
I will give the repo a try before I say if it is good or bad
http://www.linuxportalen.com -> Linux Help portal for Linux and ArchLinux (in swedish)
Dell Inspiron 8500
Kernel 2.6.14-archck1 (selfcompiled)
Enlightenment 17
Offline
Uhhh... Why optimization? Beyond -O2 and -march=i686, the impact starts dropping pretty fast on x86 machines.
(Ahh well, have it your way, but it seems to me that optimizing for specific processors is a waste of time.)
Yep; you're probably right on the loss of impact. You might see this more as an aesthetic thing. Packages compiled for my processors feel nice to me aesthetically. It turned out someone else was interested, so I opened it up.
@Cybertron: Hope you'll enjoy it
Offline
I'm using this repo on my Celeron 2.6 machine and I feel little (but noticable) difference (yeah, I know - this is probably placebo effect but whatever ).
You can see earlier discusion and few tests here: http://bbs.archlinux.org/viewtopic.php?t=16655.
Offline
also, I'm a philosophical phenomenologist by day so I just say "placebo schplacebo!"
Offline
take two placebos and call me in the morning.
"Be conservative in what you send; be liberal in what you accept." -- Postel's Law
"tacos" -- Cactus' Law
"t̥͍͎̪̪͗a̴̻̩͈͚ͨc̠o̩̙͈ͫͅs͙͎̙͊ ͔͇̫̜t͎̳̀a̜̞̗ͩc̗͍͚o̲̯̿s̖̣̤̙͌ ̖̜̈ț̰̫͓ạ̪͖̳c̲͎͕̰̯̃̈o͉ͅs̪ͪ ̜̻̖̜͕" -- -̖͚̫̙̓-̺̠͇ͤ̃ ̜̪̜ͯZ͔̗̭̞ͪA̝͈̙͖̩L͉̠̺͓G̙̞̦͖O̳̗͍
Offline
I think the only few packages that are really useful in the whole repository are:
- glibc optimized for pentium-m
- cairo compiled with glitz (depends on your hardware if it's faster)
- Xorg 6.9RC2
- kernel
Kernel can have quite some speed differences when optimized, same for the basic glibc system library. Glitz can speed up cairo in some cases and Xorg 6.9 is completely different from 6.8 when it comes to the performance of RENDER operations, which are used extensively by cairo.
Offline
thanks for the input, JGC. You don't think qt & gtk2 add anything noticeble to speed then? I'm curious to know why.
(In the end though, I'm happy with the aesthetics-argument ;-))
Offline
There's not much stuff you can optimize in GTK and QT. These toolkits are pretty optimized with i686 optimizations. The only benefit you will see from extensive processor optimizations is that pipe lengths, cache optimization strategies, etc are done for that specific CPU. For a graphical toolkit however, the bottleneck isn't the CPU, but the X11 system with poor protocol implementations in some drivers. Xorg can gain some extra speed by compiling for SSE, 3DNow or MMX, but optimizing the toolkits is useless: they are waiting for Xorg to finish rendering more than they are running through their internal loops.
Offline
Hm... that's very interesting. So what you're saying is that Xorg is being the bottleneck here, but that that's merely contingent. Does this imply that, if things were different (i.e. if X was better implemented), the gui toolkits could benefit from optimizations as well?
Offline
<EDIT: Post removed. i should know better than to do this. :oops: Sorry all.>
~Peter~
Offline
...and they say Gentoo users are ricers. *sigh*
You know that compiling for your specific CPU type rather than i686 will give you almost no noticable performance improvement, right?
Dude; I've stated my reasons why I've opened up the repo and why I have a few apps compiled for a specific CPU. I don't necessarily give a horse's arse about speed; and I'm *not* recompiling the whole friggin system with über-optimizations, just a few apps: most of which aren't in the official repo anyway, so don't accuse me for being a ricer and that handy derogatory manner of yours. I just like the fact that I have some pkgs compiled for a specific processor, because it feels nice to me to know that I have. Is that so hard to understandä?
Here's a comparison for you. Whenever I eat dinner, I like to have my drink to the right of my plate; not because it's more efficient or because of some other rational reason for it. I have my glass there beacuse it feels nice to me.
Now, on top of that; according to JGC, there might be some speed gain, so perhaps it's not all that clear on that point either.
If you wonder why I do this, that's fine (although I've already answered). If you want to state that you would never do that yourself, go ahead. But if you just want to nag because I'm doing something that one doesn't *need* to do, and something that's slightly different, then screw you, man.
(OK; I just woke, had a bad night's sleep, and looking to go through a shitty day; normally I would just not reply to this).
Offline
(...)
(... No harm here...)
There's not much stuff you can optimize in GTK and QT. These toolkits are pretty optimized with i686 optimizations. The only benefit you will see from extensive processor optimizations is that pipe lengths, cache optimization strategies, etc are done for that specific CPU.
Yeah. Your probably right and fact that I use pentium-m on my Celeron 2.6Ghz machine can even make it worse - optimization strategy may be using facts that are true on -M but not on pentium4 core. The most speedup expected is from using SSE and SSE2 instructions that as you can see (in erlier thread about this idea) on simple math loop are giving about 20%. Of course in real app. - where math operations aren't dominating this will result in "<1%" but ... if someone can have 1% speedup without big costs (sharing binaries) then whole idea is still sane and profitable.
(In the end though, I'm happy with the aesthetics-argument )
I'm happy with it too. Hope you even add some more binaries (kdebase & kdelibs are those on which I'm counting on specialy).
Offline
<EDIT: Post removed. i should know better than to do this. :oops: Sorry all.>
Obviously I over-reacted a tad. So, I'm sorry too...
Offline
added a few packages from the AUR that I use regularly. More info in the first post of this thread.
Offline
updated e17...
Offline
I'm thinking is there any benefit from having packages with no binaries (like subfonts) in pentium-m repo. Maybe what you'd like to do is spliting repo into:
[pentium-m] - these you compile for speed
[nightfrost] - your personal (but still shared) repo.
Offline
You know what; that's probably a good idea. It's not gonna be what I was planning from the beginning thouugh. But it's probably the best for the community.. I'll post here when that's done.
Offline
As soon as iphitus releases archck6, I will split my repo into three:
[lifebookp7010]
customized kernel for the lifebook p7010 + relevant modules
[pentium-m]
glibc
cairo (with glitz support)
glitz
gtk2
qt
xorg6.9
e17
[nightfrost]
various packages from the AUR
More info will be posted later.
Offline
I've updated stuff according to my last post (except for archck6, which is not yet released). Check out the first post of this thread for more info.
Offline