# Parallel Agnostic Random Number Generator Improvements

MOOSE's random number generator capabilities were improved during the month of August. While MOOSE has been able to generate parallel-decomposition-independent random numbers in most field-based calculations for some time, support for the initial condition system was notably absent. Developers may now access random numbers from within an initial condition that will remain consistent as the number of processors or threads is changed.

# Improved API Behavior for Periodic Distance-Based Methods

MOOSE has several methods that can return distances between points while taking into account periodicity. Previously, these methods only worked with lines, rectangles, and cubes. These methods can now be used on any domain shape and they will degrade properly (e.g. will return physical distances) when used on irregular domains.

# More GrainTracker Performance Improvements

Performance of the Polycrystal initial condition object has been significantly improved. The complexity of the old algorithm has been reduced from to , where represents the number of partial grain pieces and represents the number of order parameters used in the simulation. This improvement drastically reduces the setup time of very large simulations.

# New External Problem Interface for MOOSE-wrapped Apps

A new ExternalProblem object has been merged into MOOSE. This new object improves the interface used for wrapping and coupling to external modeling and simulation software.