HDF5 reader/writer for performance metrics about openmoc

[wbinventor]

This might be most relevant for you to implement Tim, since it will help you keep a permanent copy of performance metrics from your experiments once your PAPI instrumentation is implemented.

[wbinventor]

One additional feature to this would be to write the data using Python's "pickle" function - ie, openmoc.process would performance data from both hdf5 and pickling. This would allow the data to be easily read back into Python, and would be more platform-independent than HDF5 which will require h5py to be installed. For machines like BGQ without numpy and h5py, all performance data (timers, papi counters, etc) can be pickled and the file copied to another machine with numpy, matplotlib, etc. for data processing and plotting.

[wbinventor]

Functionality is implemented as the process.storeSimulationState(...) routine. This method stores the simulation date to either HDF5 or as a Pickled Python dictionary based on the "use_hdf5" parameter (false by default). The "fluxes" boolean parameter indicates whether to store the FSR scalar fluxes (false by default).

commit (master): 1f89772