Accelerating the Resolution of an Inverse Thermal Problem Via Posix Thread Api and Optmization Flags

Abstract

A major difficulty for the resolution of an inverse thermal problem is its high computational cost. Therefore, thermal tomography has found a limited application to industrial processes where fast execution is required, e.g., mapping imperfections in the process of manufacturing castings. In an attempt to improve the computational performance of the resolution of an inverse thermal problem developed, the computational power of multi-core CPUs was used by programming tool parallel architectures shared memory POSIX Threads (Pthreads). Tests were performed on processors with different microarchitectures (Intel® Core 2 Duo, Intel® Core 2 Quad and Intel® Core i7) and different amounts of threads. The parallel implementation of our algorithm exploiting the processing power of multi-core CPUs with the aid of optimization flags arrived to get a performance of up to four times faster than the serial version of the code on the same CPU representing a significant gain for simulations involving the resolution of inverse thermal problems.