[Back] [Up] [Next]

Time database correlation

A time index database correlation, like any other database correlation, involves multiple input databases where each database potentially has a different number of time states. The number of time states in a time database correlation is not directly related to the number of time states in each input database. The number of time states in the database correlation are instead determined by counting the number of unique time values for every time state in every input database. The times from each input database are arranged on a number line and each unique time value is counted as one time state. Time values from different input databases that happen to have the same time value are counted as a single time state. Once the time values have been arranged on the number line and counted, VisIt calculates a list of time state indices for each database that identify the right time state to use for each database with respect to the time database correlation's time state. The first time state for each database is always the first time state index stored for a database. The first time state is used until the time exceeds the first time on the number line, and so on.

Time database correlations are useful in many of the same situations as stretched index database correlations since they are both used to align different databases in time. Unlike a stretched index database correlation, the time database correlation does a better job of aligning unrelated databases in actual simulation time rather than just spreading out the time states until each input database has an equal number. Use a time database correlation when you are correlating two or more databases that were generated with different dump frequencies or databases that were generated by totally different simulation codes. figure shows the behavior of databases A and B when using a time database correlation.


Figure 9