Ocean Modeling Discussion

well it looks as if you're having trouble creating your history file - does it work fine for less processor cores? Are you using parallel NetCDF output? In any case, your benchmark numbers so far show that you're spending (even for this number of timesteps) an inordinate amount of time for ...

Furthermore, to my experience, the i7-family CPUs machines (by at least a factor of 2) outperform all previous generations of CPUs, including Core 2 and 5400-series Xeons (that is ROMS running 8 theads on a Core i7 920 is faster than on 8 threads dual-quad Xeon 5420s). This is a statement that ...

I have not examined the NetCDF 4.1.1 library. As far as I know, UNIDATA NetCDF does not support neither distribute the pnetcdf library. NetCDF 4.1.1 allows pnetcdf to be used when using the classic v3 mode - the configure script has a --with-pnetcdf= option - UNIDATA does not distribute the pnetcdf ...

More generally speaking this error is due to a mixing of PIC and non-PIC objects when the target executable is expected to have more than 2GB of address space extent. It is a "feature" of AMD64/EM64T that AMD introduced by employing 4 different 64bit addressing models (kernel, small (default ...

(2) Find its maximum and then its corresponding U-, V- and W- components, (3) From (2) find the (I,J,K) grid locations as to where this maximum CFL occurs and, [snip] I do not know how to code up (1)-(4) in MPI code because of steps (2) and (3) where we need to do some sorting to find the ...

My boss is looking to get such a test system to play with. He says they are good at single precision floating point, but double precision is not so fast. I think the people who would get the easiest benefit are using libraries such as lapack where you just change your function calls or the library ...

It is strange but true that on Intel x86-family CPUs a serial run will normally go faster if there are multiple tiles. Actually one would expect this type of behaviour on all cache-based processors (possibly including cache-based vector processors like the later Cray ones - though things could be ...

I can't imagine that UPWELLING is big enough for you to need a larger stacksize. It must be something else - I will try it on an SGI. Edit: I tried it on: IRIX64 video2 6.5 10070055 IP27 and could not reproduce the problem. I did run into linker troubles from mp_numthreads and another similar ...

A very basic question: Have you tried resetting the stack limit to unlimited?

For the Bourne shell family:
ulimit -s unlimited
For the C-shell family:
unlimit stacksize

In many cases this solves the problem.

Constantinos

I would not expect anywhere near a factor of 7 (or even of 2) performance difference (for a program spending 50% of its time communicating exclusively very large messages - a rather unlikely configuration - 7 times higher bandwidth would at best buy you 43% less wallclock time). ROMS' (as well as ...