Abstract
In order to deal with the inefficiency of trational serial tool path algorithms and incompatibility issues on the heterogeneous hardware platforms, this paper suggests a tool path planning method based on CPU-GPU(Central Processing Unit-Graphic Processing Unit) heterogeneous parallel computing. The method contra poses NURBS(Non-Uniform Rational B-Splines) surface which is abstracted as a matrix multiplication on the principle of isoparametric line tool path planning method. Then a parallel algorithm in accordance with Open CL(Open Computing Language) specification is proposed. Adopting data parallel programming model, the method executes multiple work-items of the GPU on the core under control of the CPU logic, and reconstructs the isoparametric line method as parallel execution instead of traditional serial execution. Simulation results show that this algorithm takes less time to generate tool paths on the CPU- GPU heterogeneous platforms, reduced by 1.5 to 15.9 times compared with traditional serial algorithm and it is of great significance to the tool path planning’s real-time or quasi realtime generation.