Porting C++containers based on domestic heterogeneous platforms

With the development of the times, the demand for high-performance computing power is also increasing, and heterogeneous computing collaborative parallel computing has become an important development direction of computing technology. In recent years, domestic heterogeneous platforms have flourished, and it is of great significance to transplant and develop suitable algorithms and software for the architecture of domestic heterogeneous platforms. The C++Standard Template Library (STL) is an important component of the C++programming language, providing a series of universal and reusable template classes and functions that can help developers better develop software. In terms of STL heterogeneous porting, the progress of domestic heterogeneous platforms lags far behind NVIDIA, which hinders the development of heterogeneous computing in China. Therefore, this article migrates commonly used containers in STL to broaden the adaptation range of domestic heterogeneous software. By porting several iconic containers, a solution for container heterogeneous porting is proposed, providing a reference method for STL heterogeneous porting. The final experiment showed that compared to non heterogeneous operations, this scheme significantly improves the performance of heterogeneous porting of containers in large data volumes. At the same time, it also has a huge improvement in the performance of internal operation functions of containers compared to the original STL parallel operation method.