Doxygen Generated Documentation From Interface Source Code
IO_writeCheckpoint.F90 File Reference

Go to the source code of this file.


subroutine IO_writeCheckpoint ()

Function/Subroutine Documentation

◆ IO_writeCheckpoint()

subroutine IO_writeCheckpoint

This is a generic call to write the important simulation data to a checkpoint file. A checkpoint file writes a few different types of data to a file, first the physical data like, temperature, pressure, density etc. which are stored in each cell on the grid. Second, in order to recreate the simulation from a checkpoint file a number of other single quantities are needed as well. We call these scalar values which include simTime, dt, nstep, globalNumBlocks etc. We also store descriptive strings that describe the simulation run.

The same IO_writeCheckpoint routine is called regardless of the type of file being written, (such as hdf5 parallel, hdf5 serial or pnetcdf) IO_writeCheckpoint prepares the Grid_ioData (like getting the globalNumBlocks) and collects the scalars wanting to be checkpointed from each unit. IO_writeCheckpoint then calls four methods, io_initFile, io_writeData, IO_writeParticles and io_closeFile. Each of these routines _is_ specific to the type of io library used and have their own implementation. In addition, io_writeData has its own implementation for io library and type of grid (UG, Paramesh, or other)

In FLASH IO_writeCheckpoint is called from IO_output (or IO_outputInitial or IO_outputFinal) IO_output checks to see if enough wall clock time, simTim, or nsteps has passed to checkpoint.

We have put IO_writeCheckpoint in the API because a user may want to write a checkpoint at another time or for another reason without having to go through IO_output. For most flash users IO_writeCheckpoint will only ever be called through IO_output


For those familiar with FLASH2, breaking up the checkpoint routine into these four different methods is a change. Because FLASH3 now supports different grid packages and we are committed to supporting both hdf5 and parallel netCDF having each grid and io library writing its own checkpoint file proved to be a lot of code duplication. We believe that while dividing up the checkpoint routines created more files it will in the end be easier to maintain.

SEE ALSO IO_output

Definition at line 69 of file IO_writeCheckpoint.F90.