THE METRICS ABOUT PERFORMANCE TESTS

In order to determine the efficiency of the pcylitx platform, some performance tests have been performed. In these tests, the performance of the C++ standard threading library ( std::thread ) and the performance of the libraries that are built by "pcynlitx" has been compared. On the tests, a dedicated workstation computer is used. The specifications of the hardware which is used in the tests are given in below.

The hardware which is used on the tests

On the tests, well known, standardized benchmark algorithms have been used. The algorithms are selected from PARSEC benchmark suite and RODINIA benchmark suite. PARSEC is an abbreviation for ”the Princeton Application Repository for Shared-Memory Computers”. The main contributors of the PARSEC project are Intel Corp. and Princeton University. RODINIA benchmark has been developed by the University of Virginia, the Dept. of Computer Science for parallel computing applications. You can reach the PARSEC and RODINIA benchmarks web sites from the links given in below.

PARSEC Benchmark Suite

RODINIA Benchmark Suite

THE METHOD THAT IS USED ON THE TESTS

In each test, the aim is to obtain the time spending ( CPU user execution time ) between the creation and joining the threads. Therefore, at the end of each test, the time which are spent on the execution of the threads are determined. Then, the average values of the execution times are compared. The ratio of excution times gives the performance losses caused by the autonomuos thread management system.