In this projet we try to develop a machine Learning model capable of predicting(suitability) of OpenMP GPU offloading using a set of miro-architecture independant features. These features will reflect the nature of the program. further we try to predict suitability of GPU offloading by features extracted from the CPU execution.