2. System configuration

---

Setting DAMASK options

Adjust the settings in DAMASK/CONFIG using any text editor, the default values should work on most systems.

---

Setting shell environment

Depending on your shell, execute one of the following commands to set the DAMASK environment.

• in bash (default on most Linux systems):

   > source DAMASK/env/DAMASK.sh 

• in the C-Shell:

   > source DAMASK/env/DAMASK.csh 

• in the Z Shell:

   > source DAMASK/env/DAMASK.zsh 

Sourcing the file is necessary before compiling or running DAMASK, hence it is recommended to do it automatically when logging in. This can be done by adding the source command to shell initialization files such as .bashrc or .profile located in your home directory.

Result

The script reports system environment settings in part derived from DAMASK/CONFIG. Additionally, it

• exports the variable $DAMASK_NUM_THREADS according to the settings in DAMASK/CONFIG
• adds DAMASK/python to $PYTHONPATH such that the damask module can be properly imported in Python
• adds $DAMASK_BIN to $PATH if necessary and possible
• sets the limits for stack, heap, virtual memory, and physical memory via ulimit

---

A few words on memory limits

Please note: If you don't change the stack size, you might experience a stack overflow for large models, especially when using openMP. Should your simulation terminate with a segmentation or memory fault at an early stage this is most likely the reason.

Increase the stack size by using the ulimit command in the parent shell.

• The easiest way would be to run

  > ulimit -s unlimited

  or

  > ulimit -s hard

  which sets the stack size to unlimited (Linux) or 64 MiB (OS X). If you want a different stack limit, you basically have to try what size is necessary for your model.

• Another option (for the spectral solver and Intel Fortran only) is to append

  -Wl,-stack_size,0x40000000

  to the COMPILE_OPTIONS in DAMASK/code/makefile with 0x40000000 being an exemplary stack size of 1 GiB.

Determining the optimal stack size is difficult. A good start to read about it is http://software.intel.com/en-us/forums/topic/501500.