How to set dimensions in ORAC : The config.h file

Being written mostly in fortran77 language, the ORAC program does not dynamically allocate the required memory. Memory allocation is done statically and dimensions throughout the code are given in a single file named config.h. To adapt the size of the program to other problems the config.h file need to be changed and the program recompiled. In the current distribution an ancillary awk script that builds the config.h file has been provided. This script is called configure and can be found in the tests directory. configure parses a general input file for ORAC and produces, to the standard output, the corresponding config.h file.

A certain number of ORAC routines contains INCLUDE statements. The corresponding include files, which may contain PARAMETER, COMMON and general dimension statements ( REAL, INTEGER etc.), have by convention a .h suffix and are generated by the standard preprocessor (/lib/cpp) from .inc files and the config.h file. The .inc files are templates of include files. where constants are initialized to character symbols (some are listed below). When making the executable, these character symbols are replaced by the standard preprocessor with their numeric values assigned in the config.h file.

The meaning of most of the character symbols contained in the config.h is explained in the file itself. Here, it is worth mentioning a few:

• PRESSURE
  The statement #define PRESSURE is found in the distribution config.h file. It implies that the single time step non-bonded force routines will be generated including the pressure computation section. Since force routines not including the pressure calculation are faster of about 10-20 %, it might be useful in simulation at constant volume to replace the statement with:
  #undef PRESSURE
  With the current version of ORAC , after this change all the *.CPP.f files must be removed by hand and the program recompiled.
• _SIT_SOLU_
  This is the maximum number of atoms in the system (it includes the solvent and solute atoms. The highly misleading name is due to historical reasons.
• _TYP_SOLU_
  This is the maximum number of possible different units type as coded in the topology database.
• _NRES_
  This is the maximum number of possible units in the solute (i.e. the number of entries in the JOIN structured command).
• _TGROUP_
  This is the maximum number of groups in the system.

• _LMAX_   _MMAX_   _NMAX_.
  These parameters controls the dimension of the sine/cosine work in the standard Ewald Method. In the config.h provided in the ORAC distribution archive, written for SPME simulations, these parameters are all defined to be 1
• _NAT_WW_   _NAT_WP_   _NAT_PP_.
  These parameters control the neighbor list dimensions. E.g. the three neighbor lists for the solvent (since a maximum of three shell for r-RESPA are allowed) are integer arrays of dimensions _NAT_WW_$\times$_MOL_SOLV_.
• _FFT1_   _FFT2_   _FFT2_   MORD
  These parameters control the dimensions of the $Q$ charge array and of the $M$ polynomials for PME computation (see section 4.1)

The total size of the code depends on the number of particles in the system and on the kind of calculation to be carried out. To give an idea, an 8000 atoms system, running with PME, linked cell and computing e.g. the VACF, requires about 25 Mb of memory. The equilibration of the solvated reaction center (33000 atoms) requires around 85 Mb.