Function Description

The Scalable Linear Algebra PACKage (ScaLAPACK) is a high-performance, distributed-parallel linear algebra library tailored for large-scale parallel and distributed computing. It comprises three key components: ScaLAPACK, PBLAS, and BLACS, serving as an extension of the LAPACK project for parallel environments.

The Basic Linear Algebra Communication Subprograms (BLACS) is a message-passing library that acts as the communication component for ScaLAPACK. Its functions include data synchronization and reception. Unlike MPI, its communication targets are primarily submatrix blocks.

The Parallel Basic Linear Algebra Subprograms (PBLAS) is a distributed foundational math library. Featuring interfaces similar to BLAS, it is built on top of BLAS and BLACS to provide the distributed-memory standard BLAS routines required by ScaLAPACK.

The relationship between ScaLAPACK, the underlying communication library, and the foundational math library is illustrated below:

Currently, Netlib ScaLAPACK defines specifications for more than 130 ScaLAPACK interfaces, including interface names, parameter lists, and functions. Kunpeng HPCKit has optimized some of these interfaces. For details, see Table 1. Other interfaces will be packaged into a complete ScaLAPACK function library and released in later versions. Currently, the community documentation contains only the description of optimized interfaces. For details about other interfaces, see the Netlib ScaLAPACK documentation.