Method and device for solving triangular linear equation set of multiprocessor system

Abstract

The invention provides a method and a device for solving a triangular linear equation set of a multiprocessor system. The multiprocessor system comprises at least one core processor and a plurality ofaccelerators. The method comprises the following steps: dividing matrixes which need to solve the triangular linear equation set into a plurality of sub-matrixes according to the sizes of the scheduled sub-matrixes; calculating and updating the sub-matrixes iteratively; and leading the accelerators to calculate and update the sub-matrixes in parallel, wherein each iteration comprises the following step: distributing the sub-matrixes to the accelerators from the sub-matrixes which are not calculated and updated in the sub-matrixes according to a row direction and the predefined sequence of theaccelerators. The loads of the accelerators can be balanced by distributing matrixes to be solved one by one through the sub-matrixes, and consequently, the calculation ability of the accelerators issufficiently utilized to calculate and update peak value performance.

Description

technical field [0001] The invention relates to the field of data processing, in particular to a method and a device for solving triangular linear equations of a multiprocessor system. Background technique [0002] LAPACK (Linear Algebra PACKage, linear algebra package) is a very effective, powerful and widely used linear algebra function library jointly developed by Oak Ridge National Laboratory, University of California, Davis, and University of Illinois. It is used in different high-performance computing environments. Efficiently solve numerical linear algebra problems, which has been effectively working for HPC (High Performance Computing, high-performance computing) and computing science organizations for more than 20 years. For details about LAPACK, see http: / / netlib.amss.ac.cn / lapack / index.html. [0003] As a professional linear algebra library, LAPACK provides a variety of linear algebra subroutines, including routines for solving trigonometric linear equations acco...

AI Technical Summary

Problems solved by technology

[0017] Thus, in the above-mentioned parallelism implementation scheme, since the sub-matrices in a given matrix are allocated by column, there is a problem that the load allocated to multiple processors is unbalanced due to the mismatch between the number of columns and the number of processors. possibility

Therefore, due to the unbalanced load of multiple processors, when the processor with high load is working, the processor with low load must be idle for a period of time, so that the computing power of these multiple processors cannot be obtained. while fully utilizing to achieve peak performance

Images