1Lawrence Berkeley National Laboratory, Berkeley, California, U. S. A.
2Lawrence Livermore National Laboratory, Livermore, California, U. S. A.
Because of the intensive computing requirements for such simulations, we modified an implementation of our Adaptive Mesh Refinement (AMR) method for compressible gas-dynamic for distributed memory parallel computers. We use a hybrid C++/Fortran programming model, where Fortran provides the numerical core of the calculation, and C++ provides the control-flow, data-management framework. To minimize the impact on existing applications that rely on our foundation library, the message passing and synchronization constructs required for data that is shared among processors were placed in the library. Other codes using older versions of our library require only minimal changes to become fully parallel. Our initial application was for a explosion confined in a small chamber.