In highly adaptive irregular problems such as many Particle-In-Cell (PIC) codes and Direct Simulation Monte Carlo (DSMC) codes, data access patterns may vary from time step to time step.
To efficiently parallelize such adaptive irregular problems on distributed memory parallel computers, effective methods for domain partitioning must be addressed.
A simple one-dimensional domain partitioning method is implemented and compared with unstructured mesh partitioners such as recursive coordinate bisection and recursive inertial bisection.
Detailed multi-dimensional numerical simulations provide an ideal vehicle to study fundamental properties of hydrocarbon flames. The physical processes in hydrocarbon flames are highly complex and interact in a strongly non-linear fashion. Numerical experimentation is an excellent way to isolate physical processes, study their interactions, or predict important properties such as flammability limits. The study of flammability limits is of great practical importance to fire safety. Near the flammability limit, several physical processes, especially chemistry, become very important and the interaction among them becomes crucial in determining the flammability limit. Only highly detailed models which include detailed chemistry and diffusive processes can obtain the correct flammability limits.
The extinction of hydrocarbon flames is a multidimensional, transient process. To date, sufficiently detailed calculations for hydrocarbon flames has only been carried out for steady-state flames. Some preliminary calculations of transient methane flames with moderately detailed flames have been carried out at the Naval Research Laboratory. These calculations clearly show the need for including detailed chemistry in hydrocarbon flame modeling. The reaction scheme used in these calculations used 15 species and 35 reactions. A complete reaction mechanism for methane would involve 50 species and 200 reactions. For higher hydrocarbons, which are of more interest to the Navy, the number of species and reactions is far greater. These calculations are currently beyond the capabilities of current supercomputers.
Parallel computers are only means currently available to perform these calculations. Thus it is imperative that the current detailed flame code be ported to a parallel machine. The existing sequential code has been extensively tested and verified, so to ensure reliability, the code should be ported with only minimum modification. The PARTI software will enable us to port the code to a wide variety of parallel computes with limited disruption to the existing sequential code.
The flame code is not restricted to hydrocarbon fuels and can be generalized to other energetic fuels. A parallel reactive flow code will provide the Navy with a very powerful tool to study these fuels.
Questions about the system or webserver: webmaster@cs.umd.edu
Problems with publications homepage: wes@cs.umd.edu
