You are not logged in.
sHyLoCk wrote:I remember Allan mentioning somewhere about USE flags. Not sure which post.
I'm fairly sure I did not.... but you can use srcpac which will aplly a sed line to a PKGBUILD before building it to update.
This is the first time I heard of srcpac. I did a quick search and it looked interesting. I also found customizepkg. How are these two tools different?
Offline
sHyLoCk wrote:I remember Allan mentioning somewhere about USE flags. Not sure which post.
I'm fairly sure I did not.... but you can use srcpac which will aplly a sed line to a PKGBUILD before building it to update.
Ah right, sorry about that. I was confused. Here is the post. It is srcpac that I was referring to.
Offline
I don't mean to hijack the thread, but I admit I've been wondering whether an Intel Atom-optimized rebuild would make a significant difference. It seems to be a truly unique processor compared to most other mainstream CPU's and I'm surprised I haven't found any similar endeavours on google--then again maybe there's a good reason for that =]
thayer williams ~ thayerwilliams.ca
Offline
http://www.linux-mag.com/id/7574/3/
Gentoo Optimizations Benchmarked
Gentoo is a source based distribution which lets the user decide how to optimize their system in many ways. Linux Magazine benchmarks three of the most common GCC optimizations; -Os, -O2 and -O3, and throws in Ubuntu for good measure.
Offline
I don't mean to hijack the thread, but I admit I've been wondering whether an Intel Atom-optimized rebuild would make a significant difference. It seems to be a truly unique processor compared to most other mainstream CPU's and I'm surprised I haven't found any similar endeavours on google--then again maybe there's a good reason for that =]
You don't want to be compiling on netbooks. My 6 year old desktop compiles faster than my netbook.
You won't see much speed increase either.
Offline
http://www.linux-mag.com/id/7574/3/
Gentoo Optimizations Benchmarked
Gentoo is a source based distribution which lets the user decide how to optimize their system in many ways. Linux Magazine benchmarks three of the most common GCC optimizations; -Os, -O2 and -O3, and throws in Ubuntu for good measure.
sigh...
It's almost guaranteed that ubuntu isn't using the same versions, starting with the kernel.
Phoronix also does benchmarks like these... arch vs ubuntu is just as subjective.
Offline
If you have an idea of what you might want tested, I have a Gentoo-like install and an Arch install on my main rig.
Offline
thayer wrote:I don't mean to hijack the thread, but I admit I've been wondering whether an Intel Atom-optimized rebuild would make a significant difference. It seems to be a truly unique processor compared to most other mainstream CPU's and I'm surprised I haven't found any similar endeavours on google--then again maybe there's a good reason for that =]
You don't want to be compiling on netbooks. My 6 year old desktop compiles faster than my netbook.
You won't see much speed increase either.
On the contrary I compile on my netbook routinely...just never rebuilt the entire system from source. With respect to the speed increase you may be right, but it's interesting that I can't find mention of anyone even trying it. You would think that, with the quirks of the Atom (in-order instructions, etc.) an optimized system just might yield tangible improvement.
Edit: Hah! It's Thayer btw, forgot I was logged into the wife's account...
Last edited by libby (2009-10-30 02:23:35)
Offline
Is there are rule about impersonating one's wife around here?
A bit of info here:
http://ivoras.sharanet.org/blog/tree/20 … -atom.html
--EDIT--
Interestingly, with -march=native in makepkg.conf, most of the optimizations that are listed in that page are automagically detected by GCC. I guess the GCC people are staying on top of it.
Offline
It's almost guaranteed that ubuntu isn't using the same versions, starting with the kernel.
Phoronix also does benchmarks like these... arch vs ubuntu is just as subjective.
It says ubuntu was "thrown in for good measure"...
The point of the article and this thread is to figure out if its worth to optimize packages.
Judging from that article:
1. theres not a single best optimization for all packages
2. optimization does make a difference with some packages
So id say its useless to recompile every package but might be worth it on some cpu intensive packages.
And you'd have to benchmark to find the right optimizations.
Last edited by Mikko777 (2009-10-30 09:45:08)
Offline
thayer wrote:I don't mean to hijack the thread, but I admit I've been wondering whether an Intel Atom-optimized rebuild would make a significant difference. It seems to be a truly unique processor compared to most other mainstream CPU's and I'm surprised I haven't found any similar endeavours on google--then again maybe there's a good reason for that =]
You don't want to be compiling on netbooks. My 6 year old desktop compiles faster than my netbook.
You won't see much speed increase either.
I suppose you could build your packages in a chroot with your atom-specific gcc hacks on a more powerful machine and then set up a repo on your local network to download your optimised packages from there. Probably not worth the hassle though.
Offline
In my personal experience, I found that KDE and Firefox were a bit faster on Gentoo than Arch.
By faster, I mean that KDE started about 10 seconds sooner on Gentoo than Arch. It's really not a huge difference.
Also, I think you can gain the most speed by compiling a highly configured kernel. Remove support for everything you don't need and compile what you do need into the kernel and not as modules. You should see a noticeable decrease in boot times.
Offline
Yaourt have a build function
yaourt -Sb
but as i know it downloads all dependencies as binarys ![]()
Offline
I have a Gentoo install alongside my Arch install on my main box. I used to use Gentoo regularly. I actually thought that I was going to toss Arch on here, play with it a bit, then switch to Gentoo as my main OS after I got it built alongside this in my spare time, since I knew Arch would install and configure more quickly. I used to do the same thing with Ubuntu for a bit when I wanted to do some heavy reconfiguring of Gentoo. I haven't timed alot of operations, but I can't tell any difference at all between simple things like Flux and Firefox load times and running Huludesktop and bringing up Thunderbird. Firefox load for example (with exact same setup of add-on's, bookmarks, options, etc) is down to the second exact. (maybe there is a difference in the 'below a second threshhold', but not a noticeable one. I have my fluxbox autostart script set to start firefox, sleep for several seconds, then toss up some terminal windows so that they will get auto-placed properly beside the firefox window, and there is no noticeable difference in the gap between either box; i was expecting to shave like maybe a second off or something noticeable after all the work I put into the Gentoo install too). Open Office is quite fast in either environment. Nothing noticeable in difference, although I suppose I could pull out the stopwatch and compare them or something. Gentoo is reasonable fun to play with though, but performance doesn't really seem a solid reason to switch to it, at least for the primary things that I do on a day to day basis. I like the fbcondecor (which I could theoretically get working in Arch, but I don't think a larry the cow or 'emerge world' background would fit in). I do like having pictures of Larry around better than the Ronald McDonald wallpapers that I have on my Arch box. Just kidding about Ronald - he came to mind when I first wondered if Arch had a mascot though (since the whole Golden Arches thing). I should look around the forums and see if Arch has a mascot I suppose.
But yes, I've been using Arch a couple of weeks now without looking at my Gentoo install. If you're really curious, I could try to to run something on each and time it. There might be a difference, but I can't say that I have noticed one in anything I did, and I switched over to the Gentoo for a bit playing around with the normal stuff I do. My machine might not be the same as yours though [i7; 3.8 Ghz; 12 GB RAM; x86_64 on Gentoo & Arch], but it could give you more info to play with (although I think the simple answer of "I don't notice it; these other people don't seem to; you probably wouldn't either" answers your question well enough, and you will definitely notice the compile times of things, but I don't recommend against trying Gentoo for the fun of it if you are so inclined [but I do think Arch is better for me personally].
To understand recursion, you must understand recursion.
Offline
Thank you very much for your detailed and well thought-out response; I'm convinced that Gentoo isn't worth my time now:) Although I can only dream about 12G of RAM.
Offline
All that ram isn't all as great as you would think. I just put it in since i was building a new comp and wanted to max out what the MB could handle. It is alot easier to get the wife to agree to a large upfront expenditure than a medium sized one now and another one down the road, so I wanted to make sure my bases were covered. Its neat if I want to do something silly like see how many VM's I can run all at one time or something.
To understand recursion, you must understand recursion.
Offline
I used FreeBSD for a while (5 & 6), and compiling EVERYTHING from source with compiler flags adjusted (to -O2 plus arch specific - had an amd back then) made no difference whatsoever. I knew this once I tried PcBSD, with comes with vanilla packages, and there wasn't any difference in speed between my "optimized" FreeBSD install and my PcBSD system. In fact, PcBSD came with a "fat" kernel, they needed to support a variety of hardware, mine was streamlined to support only what I have - and that didn't make any difference either... There might be some small difference for a few applications (mostly mencoder and friends), but otherwise you'll gain nothing for the time spent rebuilding your whole system.
Offline
Can I be lazy and add a feature request. Drop to console on fail.. most times one must send a few stop signals even if --no-confirm is off. NIce script ![]()
We may only attempt to reverse engineer the future.. for now.
walls/screens|.conkyrc
Offline