I try to wrote a program which provide a kind of "map" skeleton that wraps OneAPI calls hiding hardware targeting issues through some parameter specifying the kind of target (CPU or GPU/Accelerator). In the former case (CPU targeting) the map can be implemented using TBB parallel for and the alike. In the latter (GPU or accelerator) the map should be wrapped on top of SyCL kernels. The Idea is to execute a complex function over each elements of a vector or a set, to understand with increasing of vector size and more complex function it is better to execute my code on GPU or on CPU. I thought also about mandelbrot, Julia set, and Sierpinski Carpets. could you introduce me a famous function for example like Mandelbrot?
Aucun commentaire:
Enregistrer un commentaire