A Unified Programming Model For Heterogeneous Computing With CPU And Accelerator Technologies: An Axiomatic Approach
Yuqing Xiong · 2024
A new theory for computer science based on an axiomatic approach is created in this paper. We take it as a postulate that an accelerator and a program executting on it can be regarded as a general printer. From it, we naturally and directly build a unified programming model for heterogeneous computing with CPU and accelerator (like GPU, FPGA, Google TPU, Atos QPU, and more) technologies. Furthermore, We get some interesting results from it: it allows OpenMPlPthread-based code to be separated from the entire program code, resulting in higher modularity; makes any comper system a single-core CPU computer system, which makes programming of any computer systems simpler; and provides another new approach to coupled applications computing. This indicates that the theory should be able to provide more solutions for other problems (like hybrid cloud computing, multicloud computing, and so on).