General-purpose computing on graphics processing units is increasingly being used to solve problems. In this paper, we compare the OpenMP interface with the already established OpenCL framework in terms of the difficulty of setting up the environment, the complexity of programming and the execution time.
We tested the tool on various algorithms that cover many areas of computer science. We used different compilers for testing. The algorithms were tested on an integrated and discrete graphics processing unit.
The end results in OpenMP are worse than in OpenCL, except with the Clang compiler. With the Clang compiler on algorithms for computing the scalar product and sorting, the OpenMP tool achieved better results than the implementation in OpenCL.
|
