Real time algorithms for efficient dynamic memory allocation preemptive scheduler and searching using openmp

Abstract

This thesis entitled REAL TIME ALGORITHMS FOR EFFICIENT DYNAMIC MEMORY ALLOCATION PREEMPTIVE SCHEDULER AND SEARCHING USING OPENMP is concerned with Algorithms for Real time systems that demand less latency and processes that can be executed in parallel Serial programs are less suited for realtime systems since the different tasks or process are executed serially To suit realtime applications example Automotive systems biological process variation study etc concurrent program structure combined with effective scheduler and context switching modules using semaphores mutex need to coexist Additionally the use of multicores with dynamic affinity switching among the different cores depending on the load on an individual core will accelerate the speedup and reduce latency The focus of this research is to integrate the scheduler concepts for multiqueue handling multicore units with hybrid operating system handling capability priority based preemptive scheduling semaphore implementation and dynamic memory allocation and deallocation To achieve deterministic execution time for tasks The requirement in earlier reported works to reserve static set of locations apriori is eliminated in this thesis In addition to simulation the study is done on an actual hardware and practical implementation constraints are also handled effectively The algorithms which were earlier in single threaded mode are transformed to multithreaded and along with partition schemes achieve reduced searching time To perform parallel programming openMP collection of compiler directives and library functions based application program development is done The demonstrated applications include searching and sorting of text and strings in large database example genome database Medical database etc The average speedup achieved with one added core compared to reported single core are fourteen points seventy nine percentage for search and two points thirty six percentage for sorting respectively Higher speedups can be achieved with more cores run in parallel