cp2k speedup on multicore machines

[cavallo]

Dear All,

I just started to work with cp2k, and I would like to know the speedup
I can expect with cp2k on dual-core em64t and opteron machines.

I tested a simple job on an em64t. Running on 2 cpus the speedup is
1.07 instead of 2. Going up to 4 cpus the speedup is 1.28 instead of
4.  Teo Laino was kind enough to test my input on a Cray, and he got a
speedup of 1.50 on 2 cpus, and 2.79 on 4 cpus, which is definitely
better.

Since this is my first cp2k installation, I would like to know if the
speedup I got is in line with other people experience on em64t and
opterons multicore machines (I hope not), or I have to work to compile
a better executable (in the case any tip is very wellcome). There's no
problem of memory or hd access, since two serial jobs running
simultaneously are not slowed down.

So far, I compiled cp2k on a fedora 7, intel 10.0.025, intel-mkl
libraries, and gcc compiled fftw3 and mpich2.

Thanks,
Luigi