SW26010-Pro processor-oriented high-performance implementation method for 1-level and 2-level BLAS function library

Abstract

The invention discloses an SW26010-Pro processor-oriented high-performance implementation method for a 1-level and 2-level BLAS function library. The method comprises the following steps: performing task division on a problem to generate a plurality of sub-problems, the structure of the problem comprising a vector, a common matrix, a symmetric matrix or a triangular matrix; if the matrix is a vector, a common matrix or a symmetric matrix, allocating the operation of each sub-problem to a corresponding thread; if the sub-problem is a triangular matrix, distributing the operation of the diagonal part of the sub-problem to the 0 # thread, and distributing the operation of the non-diagonal part of the sub-problem to other corresponding threads; and splicing the operation results of the threads to obtain a solution of the problem. According to the method, the parallelization of the BLAS 1-level function and the BLAS 2-level function is realized, the problem of data dependence among threads is solved, and the performance of the functions is further improved through a self-adaptive tuning mechanism.

Description

technical field [0001] The invention relates to the field of realization of basic linear algebra library BLAS (Basic Linear Algebra Subprograms), in particular to a high-performance method for implementing 1st and 2nd-level BLAS function libraries for SW26010-Pro processors. Background technique [0002] BLAS is a basic linear algebra subroutine library, which mainly includes the basic operations of vectors and matrices. It is one of the most basic and important mathematical libraries and is widely used in scientific computing, weather forecasting, astrophysics and other fields. The BLAS library is the core of many professional software. Among them, BLAS 1 and 2 functions will be called repeatedly many times by almost all applications related to matrix operations and dense linear algebra algorithm software packages (such as LAPACK, ScaLAPACK). Practices in numerical matrix analysis and deep learning have shown that BLAS level 1 and level 2 functions are of great significance...

AI Technical Summary

Problems solved by technology

[0006] The present invention provides a high-performance implementation method of 1st and 2nd-level BLAS function libraries for SW26010-Pro processors, so as to meet the needs of BLAS 1st-level and 2nd-level functions on SW26010-Pro many-core processors, and solve the existing problems The problem of low performance of open source math library

Images