Hi,

I have an important problem with management of memory under unix os (kernel : 2.4.18-6mdkenterprise). I would like to submit a job using 2 Go of ram. We have a quadriprocessor with 4 Go of ram. But the os seems to divide the total memory ressources by the number of processors (impossible to run a job exceeding 1 Go of ram). Can you give me some advice concerning modifications of our system os in order to allow large computation (without pro-hpc license).

Cordially yours.

Either you switch to a 64bit platform or you run the case in parallel.

There is no other way to allocate the total amount of memory to a single processor?

My guess is that this is set by kernel parameters, for which a given amount of mem is set per processor. Can you check that?

High memory is activated, but submission of large simulation (executable of 2 Go) is still impossible (segmentation fault). Have you please some advice about kernel parameter in order to allow my computation ?

BTW, which version of the code are you using and with which kind of operating system? Is it by chance Linux and v3.150A?

I used linux with release v3.150A. I hope that you have the solution or informations about configuration of pro hpc on multi proc linux server.

Thancks by advance

Then the reason is why you cannot run a simulation requiring more than 1 GB of memory comes from the dynamic linking of libraries. Solution: You need to compile a static executable with the additional compile option -Y, hence from the command line: star -Y -snf star-star.exe instead of the default options: star -svf star-star.exe

Note that star-star.exe is the default executable name. For getting all the starlink options => starlink -h

Enjoy... ;-)

Always trouble with compilation. Y option do not appears in my compilator options (Absoft fortran 77 compiler 5.0). In your previous message you recommand to use : star -Y -snf star.exe star is not a valid command (you may be think to starlink ??).

Actually i'm looking to used prohpc, but i have some trouble with mpich environment setup but i keep faith

Help is welcome

Iy you want to use between 1GB and 2GB with your version you have to add the -Y switch when doing the starlink.

> starlink -Y .....

First, I apologize. You should have read starlink instead of star (that happens when you write it at late hours...). Anyway, for parallel runs, please install the mpich libraries the STAR-CD support office should have provided you with (latest version is 1.2.4 to my best knowledge). If they are already installed, check the environment variables MPI_ROOT, MPI_DEVICE and MPI_ARCH. Adding $MPI_ROOT/$MPI_ARCH/$MPI_DEVICE/bin to you $PATH should allow you to find mpirun. That's something to start with... For additional details, please read the installation guide...

Hope this helps you further.

respected sir i m looking for hte detail of the topic according to the operating system point of view your positive action in this regard will be highly appriciated

i need the complete documentation and code if provided please consider my request thanking you

please send me the notes

i want details of this topic

mesg from mem.mang on unix

plz send me the the a breif detail about memory management in unix system. i will be greatfull to you. girish gaurav

sir, can u please send me this article u have written on unix memory mangement as i need it for my project work..

thanking you

sarvani k

send me notes