Slashdot Mirror
Measuring The Benefits Of The Gentoo Approach
An anonymous reader writes "We're constantly hearing how the source based nature of the Gentoo distro makes better use of your hardware, but no-one seems to have really tested it. What kind of gains are involved over distros which use binary packaging? The article is here."
Portage can be used to install binary (precompiled tbz2 packages of ebuilds).
.deb packages or slackware .tgz packages.
From emerge --help:
--usepkg (-k short option)
Tell emerge to use binary packages (from $PKGDIR) if they are available, thus possibly avoiding some time-consuming compiles.This option is useful for CD installs; you can export PKGDIR=/mnt/cdrom/packages and then use this option to have emerge "pull" binary packages from the CD in order to satisfy dependencies.
--usepkgonly (-K short option)
Like --usepkg above, except this only allows the use of binary packages, and it will abort the emerge if the package is not available at the time of dependency calculation.
You can also, of course, emerge rpm and install any RPM packages. I'm not sure about debian
Gentoo is also accept pre-orders for it's upcoming 1.4 release. Information can be found here, at the Gentoo Store.
They even have precompiled packages optimizaed for Athlon-XP's - drool!
Looks like someone failed to set their USE flag properly. If you have it set right you will get support for all you want. Or if you do "emerge -vp packagname" before doing an actual emerge you can see what optional flags aren't getting used. People that use Gentoo but don't read the portage/emerge/use documents are asking for this. Gentoo isn't for all, it is only for the willing.
Please go here and reas as much as possible for installing Gentoo so you don't do something stupid.
Unstable Apps: Our Android Apps Don't Suck
One of my favorite uses for Gentoo is optimizing for size rather than execution speed. As you say, the CPU is rarely the bottleneck these days, but loading files from the disk can be a factor in a program starting up. I've done the benchmarks, and some rather large programs see significantly reduced load times when optimized with -Os.
You obviously don't use GIMP or analyse OS mapping data much. I have the RAM to get the info into memory, IO is not an issue for much of my work.
Having said that, portage is the main reason I've converted all my machines to Gentoo; it's just not a serious option to go back to RPM based systems after using it for a week or so.
TWW
"Encyclopedia" is to "Wikipedia" what "Library" is to "Some people at a bus stop"
Changing the CPU multiplier will not give you what you are looking for as it will likely change your FSB speed and L1/L2 cache access rates. The most common bottlenecks on systems are heirarchical IO bandwidth related. For example, having scads of RAM for buffer/cache will help you with disk IO woes.
Compiling binaries with optimization for a particular processor help with i-cache and d-cache utilization. The fewer instructions fetched (or the order in which they are fetched) makes a big difference in performance.
Boosting CPU cache size (up to practical cache limits), increasing FSB speed and avoiding disk IO are much more significant than CPU M/GHz.
Compilation, especially optimized (-0X) compilation is _VERY_ CPU intense. If you have enough RAM to avoid the disk thrashing caused by writing numerous intermediate files you will peg your CPU.
Most user activities (aside from games) on computers are not bottlenecked by CPU but by various heirarchical IO constraints and hence the previous poster was correct that the CPU is not a significant bottleneck on modern systems.
Your "Is -mmmmx and such worth it?" guide is a little unfair. Thing is, you notice that only -O3 really made much of a difference. Well, that's because each of your tests is just one big loop and -O3 does heavy loop unrolling. You basically chose the absolute optimal case for -O3 to win (besides possibly having a small function call inside that loop so that -O3 could inline it). And you didn't do any floating point multiplies or divides which is where -mmmx+sse and -m3dnow would help you.
If anything you were also using a relatively small dataset. If you get a large enough data set (or code size) -O3 might actually hurt you (loop unrolling and function inlining will bloat both code and data size and make it much more likely to have a cache miss).
Anyways, synthetic benchmarks are one thing but your is so synthetic as to be rediculous.
Agreed. Back when I was a hardware tech in the early 90's, I recall building pools of identical machines for customer orders. For burn-in, I would loop benchmarks for 24 hours before shipping them out. There was typically 1-2% difference in identical systems.
People tend to forget the complexity of a PC and the inevitable, microscopic differences each part made. Thus differences in resistance, heat generated, and performance.