On dynamic parallel execution of large scientific codes

Due to the character of the original source materials and the nature of batch digitization, quality control issues may be present in this document. Please report any quality issues you encounter to digital@library.tamu.edu, referencing the URI of the item.Includes bibliographical references.Detection and utilization of the natural parallelism of a program is the key to exploiting the computing power of a massively parallel processing (MPP) computing system. In this thesis some large scientific codes written in the Fortran language are analyzed to study the looping behavior of these codes to determine loops that merit parallelization. By inserting special tags in the code the parallelism analyzer can communicate to the compiler to insert special instructions for the dynamic allocation of processors at run-time of an MPP system. The target MPP architecture which features the Dynamic Processor Allocator coprocessor can allocate free processors to multiple tasks at their requests through a dynamic processor allocation protocol. The simulation results show that the proposed scheme utilizes the system resources only when necessary, and the parallelism of application codes can be effectively exploited