PLEASE NOTE THIS DOES NOT WORK CURRENTLY AND HAS NOT BEEN WORKED ON FOR SOME TIME. PLEASE CONSIDER IT ABANDONED AS OF MAR 12 2015.
This is the calculation engine for dynamics runs.
This program takes an input file generated by a GUI and runs the appropriate calculation, ideally faster than it would run in the GUI.
Shared Memory Runs with OpenMP
To use OpenMP capabilities, edit the namelist OMP.nml. Here you can set the max number of threads, and turn on/off the dynamic allocation and nested parallelism features of OpenMP. In addition, if you would like to use the environment variables
$OMP_NUM_THREADS, $OMP_DYNAMIC & $OMP_NESTED
rather than the namelist, then you can set the variable 'useNameList' to .FALSE., in which case the compiler defaults will be used, unless the above environment variables are set, in which case they take precedence. Any variables omitted from the namelist will go to their program defaults:
numThreads = OMP_get_num_threads() dynamicAllocation = .FALSE. nestedParallelism = .FALSE. useNamelist = .TRUE.
The OMPTools module is designed to allow the program to setup the OMP environment, print the status of its variables and print the properties of a parallel region, where that latter two of these are intended for debugging and testing the parallelisation of the application. It consists of the following routines:
OMP_setup(exitTest)
This subroutine attempts setup of the OMP environment, returning .TRUE. if no pathological errors are found. A log file is created called "OpenMP-Data.txt" which is used to store all output from the module.
OMP_check_capabilities()
This subroutine tests whether dynamic allocation and nested parallelism are available to the user.
OMP_teardown()
This subroutine cleans up the module, closing the output file.
OMP_read_namelist()
This subroutine sets up the program defaults for the OMP variables that will be used for any variables not specified by the user. It then attempts to read the namelist OMP.nml. If succesful, the values in that file are honored and program defaults are used for the variables. If the file is not present or not properly formatted, the compiler defaults will be used.
OMP_get_parallel_info(message)
This subroutine tests whether the current region is parallel, and if so returns the information made available by the OMP API about the region. The output is labelled with the message passed to the routine.
OMP_get_general_info(message)
This subroutine prints general information about the OMP environment that is available to the program via the OMP API. The output is labelled with the message passed to the routine.
Timing
Three intrinsic routines are used to time the execution. CPU_TIME gives the CPU time, SYSTEM_CLOCK gives the wall time and DATE_AND_TIME gives the date and time. These are wrapped in a module named TimeTools.f90, which allows the starting and measurement of the two clocks, and the printing of the differences in time. A logical flag tells the measurement routines whether to print or not. In comparing CPU and wall times, it is recommended that you measure the two times first, then print them to avoid delay.