With that I fully agree!

And just like you say, my point was actually, that a selection of a platform is usually being made based on more than one criterion. Decisions are made based also on load tests (if they are not benchmarks), vulnerability analyses and stability under conditions resembling later usage scenarios in a laboratory environment to ensure comparability of different offerings.

So benchmarks can be useful, but only if they are done properly and consistently for all candidates to be assessed, and they are just one of several pieces of the big picture.

And that's why Slackware is my distro of choice, and BTW, it feels pretty fast to me, in my particular use cases.

gargamel

You are missing the point. This is exactly why people do benchmarks: distros are different and hence perform differently. Why on earth would you want to remove all the variables?

Hi,

Benchmarks are made using specifics, so to use that benchmark as a means to indicate universal status is moot. The benchmark data is for the environment and system variables at the time performed. Really cannot use this to compare between equipment that is not equivalent or better than the test bench equipment.

If Gnu/Linux(a) is compared/bench marked then to do same tests on Gnu/Linux(b) on the same equipment then a somewhat fair test. Most users do not interpret the benchmark properly, most users then make assumptions on the test as compared to their known environment, which is comparing apples & oranges.

just like kiss

Yet they even haven't included Gentoo in the test. Slackware is naturally slow since it's meant to be generic to run with most hardware and not optimized to native ones e.g. it doesn't include machine-specific instructions when building the packages. I'm a "Slacker" (it's my first distro, and it's my alternative distro to Gentoo), but really, there's a noticeable difference when it comes to performance as I compared it to Gentoo.

P.S. Distro doesn't necessarily mean binary distros right? So Gentoo should have been included.

P.P.S. Slackware could still be made faster if it's recompiled from sources with machine-specific optimization flags.

Yeah, your optimized distro is about 1 second faster every 10 years.
Please read all the thread

Clearly a stubborn non-technical based reply. Why do I have to read them for when I used both distros (for a long time) myself? Anyway, you are wrong.

Yeah, a stubborn non-technical reply based on facts:
http://www.linuxquestions.org/questi...ml#post4792625

1 members found this post helpful.

Where's the test about Gentoo? And that's not even a general test lol.

You can do it and show us the result.
But really man you have to study how processors works and then you will understand why it's impossible to get big result from that kind of optimization. And you may even believe that but you can't believe that you will get back the processor time that you wasted while you compile and recomplie that package every time it gets updated.

Gentoo is hard to test, I mean it is compiled from source, so it depends entirely on CFLAGS.

I will test the live CD if you want.

What do you mean by general test ? If you want me to test something else, tell me. I was gonna test mencoder, but as I said, too many dependencies in the one included with the Arch live CD.

Studying how processor works? Well I'm halfway there - at least seeing how using special instructions would help big time (e.g. having 1 CPU cycles instead of 4; using special string operations, etc.), and making proper use of caches. And sorry but I don't have the time for tests. You should have at least tried optimizing your system yourself, and/or study some concepts of GCC before making an aggressive reply to my post.

Yes it does, and sometimes more hardcore is that trying to build a more native system from earlier stages than stage3 (whereas you would change CHOST).
Not necessary for me. I don't really mind proving it. But Gentoo is not really about live packages.I mean based from system-wide usage, so it's more like in most environment most people use like the effect on the desktop manager, X, etc. But if you would really make a test, it would be best to test not-really data dependent tasks, and softwares that has their own instruction patterns since some explicitly use their own arrangement of instructions most notably those that handle media files.

Wrong conclusion.

Most softwares (like Microsoft Windows) don't include machine-specific instructions, CPU vendors know this and optimize their hardware for it (and not for Gentoo). A modern x86 CPU has an instruction decoder to feed RISC-like execution units with micro-instructions for the optimal execution speed. Of course exotic instructions get a penalty and will be emulated in micro-code.

So as a compiler you are aiming for using only the most generic instructions, because the others are slow.

Performance-wise the best thing you can do on modern hardware is to implement multi-threading and make use of specialised units like GPGPU and AESNI. Both you can't do by compiler flags, you have to change the algorithm. Case closed.

http://www.funroll-loops.info/ ;-)

In modern hyper-threaded superscalar pipelined out-of-order-execution CPUs there is no concept of "cycles". How much cycles an instruction takes depends on the instructions around it, instructions of other threads, which will be executed on the same core or module, the results of the branch prediction and so on...

Using "special" CISC instructions from early x86 assembly is a good way to slow your program down.