|
|
Brian Dobbins wrote:
On Thu, Oct 9, 2008 at 10:13 AM, Jeff
Squyres <jsquyres@cisco.com>
wrote:
On Oct 9, 2008, at 8:06 AM, Sangamesh B wrote:
OpenMPI
: 120m 6s
MPICH2 : 67m 44s
That seems to indicate that something else is going on -- with -np 1,
there should be no MPI communication, right?
Wow. Yes. Ditto.
I'd be more inclined to double-check how the Gromacs app is
being compiled in the first place
E.g.,
mpicc -show
Someone correct me if I'm wrong, but if MPICH2 embeds
optimization flags in the 'mpicc' command and OpenMPI does not, then if
he's not specifying any optimization flags in the compilation of
Gromacs, MPICH2 will pass its embedded ones on to the Gromacs compile
and be faster.
Yes, I have one established example of this. I built MPICH2 with
CFLAGS=-O2. I compiled a non-MPI code with "mpicc" (no flags) and got
optimized performance with MPICH2 but non-optimized performance with
OMPI. About 3x difference in performance for my particular test case.
Not a single bit of MPI in the test code.
I'm rusty on my GCC, too, though - does it default to an O2
level, or does it default to no optimizations?
When I tried it, default gcc seemed to be no optimization. In my
MPICH2 "mpicc" (with optimization built in) I had to specify "mpicc
-O0" explicitly to turn optimization back off again.
|
|
|
![[Flag]](../../../../../community/mirrors/flag-us.gif){kind=small}