Hello everyone,

I'm facing a performance issue on linux.

I have two system on the same hardware: one ubuntu and one slitaz. Slitaz is a small distro. I use it because I need a light system.

My problem is that slitaz is much slower than ubuntu.

I joined the results of a benchmark done using geekbench.

If you have any idea of where the problem might be please tell me.

Thank you very much.

[file name=geekbench.txt size=6073]http://www.linux.com/media/kunena/attachments/legacy/files/geekbench.txt[/file]

I've done another bench with 0 load and it tells that the slitaz system has a memory problem.

Ubuntu :

Memory
Read Sequential
single-threaded scalar 1830 |||||||
Write Sequential
single-threaded scalar 1582 ||||||
Stdlib Allocate
single-threaded scalar 882 |||
Stdlib Write
single-threaded scalar 1267 |||||
Stdlib Copy
single-threaded scalar 2546 ||||||||||

Slitaz :
Memory
Read Sequential
single-threaded scalar 1770 |||||||
Write Sequential
single-threaded scalar 1525 ||||||
Stdlib Allocate
single-threaded scalar 809 |||
Stdlib Write
single-threaded scalar 505 ||
Stdlib Copy
single-threaded scalar 1011 ||||

As you can see, the Stdlib write and copy are more than twice better on ubuntu than on slitaz.
I guess this is because of my custom kernel on slitaz, but I don't know what options could improve memory management.

The whole slitaz bench (ubuntu one has not changed) : [file name=geekbench_slitaz.txt size=3035]http://www.linux.com/media/kunena/attachments/legacy/files/geekbench_slitaz.txt[/file]

It seems that it's a cache issue.

Someone knows which kernel option can custom the cache size (not the hardware one of course) and/or the number of slab objects avaibles ?

I just downloaded Geebench for linux and runned ./geekbench_x86_32.

It outputed the results I posted.

I'm not a bench expert but in my case theses benchmarks highlighted a memory problem on slitaz. And further investigation (using slabtop) revealed that it's a cache problem.