DragonFlyBSD, CentOS, Ubuntu, Solaris Benchmarks

[LightBit]

binary incompatibility is a huge issue

Especially for proprietary software.

[Cthulhux]

Incompatible package formats are a very minor problem while binary incompatibility is a huge issue and one that BSD has and linux has not.

Oh, yes, binary compatibility.

BSD:
* runs BSD binaries
* runs ELF binaries

Linux:
* runs ELF binaries

Oh well.

FAIL.

[LightBit]

No, BSD BullShitDistro is trying to inform you BSD zealots is that the BSD init system really is inferior and that system V and lately systemd (to a much more greater degree) is just superior. I agree with him in full. Here's the reasons:

BSD init:
-slow initialisation
-big shell overhead
-prone to freezing the entire system
-Dependency based script startup
-limited to serializing script startup
-cannot be taylored to any graphical configuration tool
-complex interface

System V:
-faster initialisation
-smaller shell overhead
-more robust
-capable of parallel startups
-Simpler interface
-Many graphical config tools are taylored to it

Why not add Systemd to your outstanding comparison?

[Cthulhux]

Your nickname alone makes me wonder that you actually know the word "behaviour".
And while we're at fragmenting: How many Linux distributions with incompatible package formats are out? Hmm.

[curaga]

GCC version alone is not enough, as it can be compiled without OpenMP. (libgomp needs separate porting to platforms)

[XorEaxEax]

The GraphicsMagick tests shouldn't be legal. It uses OpenMP to scale and the DragonFly compiler in 3.2 doesn't have support for OpenMP and is therefore much slower than Linux in this test. The current development version of DragonFly has GCC 4.7 and OpenMP is enabled on that. In every other test, it is competitive to Linux.

You sure about this? The Phoronix test data sheet states that the DragonFlyBSD 3.21, 3.03 tests where done with GCC 4.47, which as far as my memory goes certainly has openmp support, the test results do indicate that you are right though. I fully agree though that for a test lke this to be anything close to fair the same compiler toolchain/version needs to be used across all systems, as it stands it's pretty worthless.

You do realize that CentOS uses a kernel that is 3 years older than the kernel Ubuntu is using and it performs better in most of those tests right?

Well to cut Ubuntu some slack, I imagine this is vanilla Ubuntu which is running dash and lots of other stuff together with compositing etc which can affect benchmarks.

Also I'd say that in many (most?) of these tests done here the kernel version does not have a big impact, compiler toolchain, filesystem used and the settings applied, number of applications/services running by default etc seem more likely to affect the outcome unless the benchmark targets an area which has had great improvements between kernel versions. Overall I think these tests just highlight the general sloppiness which have become synonymous with 'Phoronix benchmarking'.

[Cthulhux]

Your behaviour is insulting

Please re-read your nickname and your previous posting, then consider this statement again.

[crazycheese]

You guys should just STFU and get back to coding under whichever license you prefer. Otherwise there will be less both BSD and GPL code - which means there would be more proprietary code.

Or you can continue trollwars, with hope someday to create trollcompiler so that each of your claims here actually work. Which will never happen. Because you will all be banned and your time wasted.

[Cthulhux]

The superior GNU technology fails to finish its own kernel for about 30 years now.

[froznen]

This is how DragonflyBSD performed in previous benchmarks so these actually no improvement which is classic BSD.

You do realize that CentOS uses a kernel that is 3 years older than the kernel Ubuntu is using and it performs better in most of those tests right?