Phoronix FreeBSD 8 vs Ubunto 9.10

I just saw a new article on Phoronix testing the performance of FreeBSD 8.0 vs Ubunto 9.10.
I think a lot of the tests are influenced by the difference of gcc compilator versions that are used - 4.2.1 for FreeBSD (pretty old already) and 4.4.1 for Ubuntu. But I don't think that memory and threads tests have anything to do with the gcc version difference thou :)
 
IIRC 8.0 (and 7.2 too) still have quite a lot of debugging options turned on by default. I wonder how the figures would be when they're turned off.
 
I was wondering about that too. If they have WITNESS and other stuff on, the comparison is totally false.
 
DEBUG=-g, KTRACE and AUDIT are in the 7.2 kernel, and on by default. Then again, I have no idea about the kind of stuff that's in the linux kernel and on by default.
 
Maybe I am missing something, but what is the point of comparing FreeBSD with Ubuntu? The two projects have completely different goals and are usually used for very different tasks.

I suppose some people have to much free time and need the ad revenue.
 
I don't see any use of such benchmarks. There is some gzip, some lame, an aged byte benchmark etc., Povray. I would like to see the state of the art render engine Yafray, h264 and aac encoding instead of mp3, bzip2 instead of gzip etc. pp. And there you see the problem: a benchmark is a very personal point of view.
 
There is another thing: on an dual core ubuntu laptop if I have a load bigger than 2, the sistem responds very slow. With FreeBSD I have immediate response when I launch an app, there is a very big difference.

I have FreeBSD 8.0 RC1, I've read that debug is not compiled but from time to time I have detailed messages on console, and it seems like kernel debug messages.

Code:
lock order reversal:
 1st 0xc6b81df4 pseudofs (pseudofs) @ /usr/src/sys/kern/vfs_syscalls.c:4096
 2nd 0xc0e0be7c allproc (allproc) @
/usr/src/sys/fs/pseudofs/pseudofs_vnops.c:757
KDB: stack backtrace:
db_trace_self_wrapper(c0cb94c3,e710aaf0,c09029f5,c08f377b,c0cbc396,...) at 
db_trace_self_wrapper+0x26
kdb_backtrace(c08f377b,c0cbc396,c45305a8,c45280d0,e710ab4c,...) at 
kdb_backtrace+0x29
_witness_debugger(c0cbc396,c0e0be7c,c0cb549d,c45280d0,c0cac20d,...) at _witness_debugger+0x25
witness_checkorder(c0e0be7c,1,c0cac20d,2f5,0,...) at witness_checkorder+0x839
_sx_slock(c0e0be7c,0,c0cac20d,2f5,c4ae7500,...) at _sx_slock+0x85
pfs_readdir(e710ac20,0,c6b81d9c,0,e710ac58,...) at pfs_readdir+0x131
VOP_READDIR_APV(c0d99080,e710ac20,c0cc4670,1000,1000,...) at VOP_READDIR_APV+0xa5
kern_getdirentries(c8c78900,4,2839f000,1000,e710ac74,...) at kern_getdirentries+0x214
getdirentries(c8c78900,e710acf8,10,c0cbd05e,c0d9cd50,...) at getdirentries+0x31
syscall(e710ad38) at syscall+0x2a3
Xint0x80_syscall() at Xint0x80_syscall+0x20

Code:
laptop# uname -a
FreeBSD laptop 8.0-RC1 FreeBSD 8.0-RC1 #0: Mon Sep 21 17:41:07 EEST 2009
ov@laptop:/usr/obj/usr/src/sys/LAPTOP  i386

Mystery solved: I have 8.0 RC1 but config file was from 8.0Beta.

You're right guys, only DEBUG=-g in 8.0RC1.

Can I remove DEBUG=-g line?
 
Yep, not a perfect test really, and I posted my 2c in the article's discussion thread (on phoronix, not slashdot). :)
 
overmind said:
There is another thing: on an dual core ubuntu laptop if I have a load bigger than 2, the sistem responds very slow. With FreeBSD I have immediate response when I launch an app, there is a very big difference.

I have FreeBSD 8.0 RC1, I've read that debug is not compiled but from time to time I have detailed messages on console, and it seems like kernel debug messages.

Code:
lock order reversal:
 1st 0xc6b81df4 pseudofs (pseudofs) @ /usr/src/sys/kern/vfs_syscalls.c:4096
 2nd 0xc0e0be7c allproc (allproc) @
/usr/src/sys/fs/pseudofs/pseudofs_vnops.c:757
KDB: stack backtrace:
db_trace_self_wrapper(c0cb94c3,e710aaf0,c09029f5,c08f377b,c0cbc396,...) at 
db_trace_self_wrapper+0x26
kdb_backtrace(c08f377b,c0cbc396,c45305a8,c45280d0,e710ab4c,...) at 
kdb_backtrace+0x29
_witness_debugger(c0cbc396,c0e0be7c,c0cb549d,c45280d0,c0cac20d,...) at _witness_debugger+0x25
witness_checkorder(c0e0be7c,1,c0cac20d,2f5,0,...) at witness_checkorder+0x839
_sx_slock(c0e0be7c,0,c0cac20d,2f5,c4ae7500,...) at _sx_slock+0x85
pfs_readdir(e710ac20,0,c6b81d9c,0,e710ac58,...) at pfs_readdir+0x131
VOP_READDIR_APV(c0d99080,e710ac20,c0cc4670,1000,1000,...) at VOP_READDIR_APV+0xa5
kern_getdirentries(c8c78900,4,2839f000,1000,e710ac74,...) at kern_getdirentries+0x214
getdirentries(c8c78900,e710acf8,10,c0cbd05e,c0d9cd50,...) at getdirentries+0x31
syscall(e710ad38) at syscall+0x2a3
Xint0x80_syscall() at Xint0x80_syscall+0x20

Code:
laptop# uname -a
FreeBSD laptop 8.0-RC1 FreeBSD 8.0-RC1 #0: Mon Sep 21 17:41:07 EEST 2009
ov@laptop:/usr/obj/usr/src/sys/LAPTOP  i386

Mystery solved: I have 8.0 RC1 but config file was from 8.0Beta.

You're right guys, only DEBUG=-g in 8.0RC1.

Can I remove DEBUG=-g line?

You have WITNESS enabled (because of lor message)
 
linux is a respectable operating system but that article is quite the contrary. as we all know, they are comparing apples to oranges. nuff said.
 
Those tests are subjective as hell and it is clear as day to see which ones are relying on clean code rather than debug code. It is blatantly obvious that the performance "issues" are caused by debug code running; lots of bounds checking, etc. especially with the memory testing. Why not run a proper test with both operating systems at max performance? Or better yet, skip the testing altogether because in reality, no one really gives a rat about these benchmarks except (1) uber-geeks and (2) people trying to sell a product.
 
Personally, i find ubuntu to be terrible under load. I'm not a programmer so i have no idea why, but similar applications, especially network based ones, really seem to run better, and at lower load on FreeBSD.

But what's more, even when the system DOES get loaded, you can still enter commands...This is just not true for me on ubuntu. The only thing i use ubuntu for anymore is my xbmc htpc's, and as soon as someone gets xbmc with vdpau working well on freebsd, I'm sure i'll end up switching those boxes as well.
 
DutchDaemon said:
DEBUG=-g, KTRACE and AUDIT are in the 7.2 kernel, and on by default. Then again, I have no idea about the kind of stuff that's in the linux kernel and on by default.

Isn't it kind of a hindrance to end users to enable DEBUG -g in a final version release?
 
irkkaaja said:
Isn't it kind of a hindrance to end users to enable DEBUG -g in a final version release?

Most users probably won't notice and it'll help finding where a bug occurs. More 'advanced' users compile their own kernel anyway.
 
irkkaaja said:
Isn't it kind of a hindrance to end users to enable DEBUG -g in a final version release?
-g doesn't hinder performance. Debug info is kept in separate file and only loaded when debugging.
 
mtippett said:
Hi all,

Please look at

http://phoronix.com/forums/showthread.php?p=94222#post94222

In particular the debug code status. ie: RC-1 was used, and the debugging code was turned off.

I am truly interested in seeing any further views about the testing on the thread.

Regards,

Matthew

Why should someone care about it? Because e.g. gzip is faster in Ubuntu? What I do know: FreeBSD _is_ stable, FreeBSD doesn't need a plethora of updates week after week just to fix some of the primary bugs etc. For example, do you care about Povray? It's nice but lightyears behind present render-technologies. They're using multi-gpu technologies according to their changelog, is there maybe some impact on the performance while using a ATI-card (with fglrx driver) in Linux and some lousy free driver in FreeBSD? What are the compile-time options of the used applications and libraries, this is important: FreeBSD != Linux. There are lots of questions, but almost no answers just some numbers without any significance.
 
oliverh said:
There are lots of questions, but almost no answers just some numbers without any significance.
And the numbers are pretty hard to read. As someone on Slashdot noted, the graphs are gray on gray. For some lower is beter, others higher. It's bloody hard to read in any case. The guy that created them must be seriously colorblind.

Not that I really care about those numbers anyway. I'm not even considering trading in my beautiful FreeBSD desktop and servers for that horrible piece of #@%^#$@# called Ubuntu x(
 
Eponasoft said:
No way...

What? Why not?

FreeBSD has been using an outdated version of gcc by default ever since the switch to GPLv3. Both gcc 4.4 and llvm produce significantly faster code than gcc 4.2, which we're currently using, and that's not to mention that the GPLv2 variant of gcc is essentally no longer maintained. Look at this:

http://multimedia.cx/eggs/icc-vs-gcc-smackdown-round-3/

gcc 4.4 is way faster than gcc 4.2. We won't see any improvements like that until we start using a compiler that's actively developed, be it clang or open64.
 
Back
Top