Ruby Benchmarks

[Zafer Aydoğan]

Hello Folks,

here are a few charts benchmarking Ruby on FreeBSD, NetBSD, OpenBSD
and Linux done by Jaime Fournier (ober@), hosted by me:

http://www.aydogan.net/ruby/mauthesis.com/ruby_bench/

Please have a look. NetBSD is doing very well.

Cheers, Zafer.

[Aleksej Saushev]

Hi!

Post by Zafer Aydoğan
here are a few charts benchmarking Ruby on FreeBSD, NetBSD, OpenBSD
http://www.aydogan.net/ruby/mauthesis.com/ruby_bench/
Please have a look. NetBSD is doing very well.

What do they do in bm_gc* tests? Such big difference looks suspicious.
Do they assume/rely on threaded Boehm GC? If yes, can you rebuild with
threads enabled in devel/boehm-gc and ruby itself and rerun benchmarks?

Can anyone with pbulk setup run partial bulk build with threaded boehm-gc?
PKG_OPTIONS.boehm-gc=threads

[Zafer Aydoğan]

Post by Zafer Aydoğan
Hello Folks,
here are a few charts benchmarking Ruby on FreeBSD, NetBSD, OpenBSD
http://www.aydogan.net/ruby/mauthesis.com/ruby_bench/
Please have a look. NetBSD is doing very well.
Cheers, Zafer.

I have updated the Benchmarks. They include now OpenSolaris and DragonflyBSD.

http://www.aydogan.net/ruby/mauthesis.com/ruby_bench/

The old benchmarks are available here:

http://www.aydogan.net/ruby/mauthesis.com/ruby_bench_old/

Thanks go to Jaime Fournier for doing the benchmarks.

Cheers, Zafer.

[Zafer Aydoğan]

Post by Zafer Aydoğan

Post by Zafer Aydoğan
Hello Folks,
here are a few charts benchmarking Ruby on FreeBSD, NetBSD, OpenBSD
http://www.aydogan.net/ruby/mauthesis.com/ruby_bench/
Please have a look. NetBSD is doing very well.
Cheers, Zafer.

I have updated the Benchmarks. They include now OpenSolaris and DragonflyBSD.
http://www.aydogan.net/ruby/mauthesis.com/ruby_bench/
http://www.aydogan.net/ruby/mauthesis.com/ruby_bench_old/
Thanks go to  Jaime Fournier for doing the benchmarks.
Cheers, Zafer.
From the 56 different tests

Scores best:

Linux 29 times
NetBSD 23 times
DragonflyBSD 3 times
FreeBSD 1 time

Scores Last:
OpenBSD 37 times
FreeBSD 8 times
DragonflyBSD 7 times
OpenSolaris 3 times
NetBSD 1 time

Additionally NetBSD scores 18 times second best.


Zafer.

[Pawel Jakub Dawidek]

Post by Zafer Aydoğan

Post by Zafer Aydoğan

Post by Zafer Aydoğan
Hello Folks,
here are a few charts benchmarking Ruby on FreeBSD, NetBSD, OpenBSD
http://www.aydogan.net/ruby/mauthesis.com/ruby_bench/
Please have a look. NetBSD is doing very well.
Cheers, Zafer.

I have updated the Benchmarks. They include now OpenSolaris and DragonflyBSD.
http://www.aydogan.net/ruby/mauthesis.com/ruby_bench/
http://www.aydogan.net/ruby/mauthesis.com/ruby_bench_old/
Thanks go to  Jaime Fournier for doing the benchmarks.
Cheers, Zafer.
From the 56 different tests

Linux 29 times
NetBSD 23 times
DragonflyBSD 3 times
FreeBSD 1 time

That's a bit disappointing, I must admit:) What FreeBSD version did you
guys test? If this was 8.0-BETA, you must remember that all kind of
debuging is on and will be turned off just before the release.
I read there is no WITNESS and INVARIANTS in the kernel, but for example
malloc will fill each allocated byte with 0xa5 and also every byte that
is freed. This has huge performance impact. To turn off this particular
behaviour one has to do the following:

# ln -s jz /etc/malloc.conf

I'm not sure if that all, but it's definiately a good start.

Post by Zafer Aydoğan
OpenBSD 37 times
FreeBSD 8 times
DragonflyBSD 7 times
OpenSolaris 3 times
NetBSD 1 time
Additionally NetBSD scores 18 times second best.

[Aleksej Saushev]

Hello!

Post by Pawel Jakub Dawidek

Post by Zafer Aydoğan
Linux 29 times
NetBSD 23 times
DragonflyBSD 3 times
FreeBSD 1 time

That's a bit disappointing, I must admit:) What FreeBSD version did you
guys test? If this was 8.0-BETA, you must remember that all kind of
debuging is on and will be turned off just before the release.
I read there is no WITNESS and INVARIANTS in the kernel, but for example
malloc will fill each allocated byte with 0xa5 and also every byte that
is freed. This has huge performance impact. To turn off this particular
# ln -s jz /etc/malloc.conf
I'm not sure if that all, but it's definiately a good start.

Just because of this, there should be simple way to turn all those
safety checks off in one place. So that there would be less
misunderstanding.

[Marcin M. Jessa]

Post by Zafer Aydoğan

Post by Zafer Aydoğan

Post by Zafer Aydoğan
Hello Folks,
here are a few charts benchmarking Ruby on FreeBSD, NetBSD, OpenBSD
http://www.aydogan.net/ruby/mauthesis.com/ruby_bench/
Please have a look. NetBSD is doing very well.
Cheers, Zafer.

I have updated the Benchmarks. They include now OpenSolaris and DragonflyBSD.
http://www.aydogan.net/ruby/mauthesis.com/ruby_bench/
http://www.aydogan.net/ruby/mauthesis.com/ruby_bench_old/
Thanks go to Jaime Fournier for doing the benchmarks.
Cheers, Zafer.
From the 56 different tests

Linux 29 times
NetBSD 23 times
DragonflyBSD 3 times
FreeBSD 1 time
OpenBSD 37 times
FreeBSD 8 times
DragonflyBSD 7 times
OpenSolaris 3 times
NetBSD 1 time
Additionally NetBSD scores 18 times second best.

That's an awesome achievement compared to the numbers of developers each
of the O.Ss have.
Please don't forget to thank for such great job and donate to the NetBSD
Fundation ! :)


Marcin

[o***@linbsd.org]

Hello

Post by Pawel Jakub Dawidek
# ln -s jz /etc/malloc.conf

Thanks for the advice.
Will make sure I set that on the next run and compare.
Check the benchmarks on Ruby 1.9.x where native threads kick in.
The story is a little different.
http://mauthesis.com/RubyBenchMarksTwo.html

Thanks.
***@netbsd.org