2011-11-25 17:22:27 |
Lars Butler |
description |
Tasks for the UHS computation should be parallelized over 1) the logic tree samples and 2) the sites. |
Tasks for the UHS computation should be parallelized over 1) the logic tree samples and 2) the sites.
The result of each task will be a list of 1d arrays (1 array per PoE). Each array represents the y-axis (SA values) of a Uniform Hazard Spectra curve. Note: The x values of all result curves are defined in the job config as 'UHS_PERIODS'.
Once the results are obtained from the Java calculator, these intermediate results should be written to a temporary hdf5 somewhere on the NFS. During the course of the computation, these results are collected and reduced into a few final result files (1 file per PoE). See bug # 888171 for more details on the calculation flow. |
|