The performance issues with the MATLAB version of the code continued as, in the mid-1970s, vector architectures became available for scientific computations. Vector architectures exploit pipeline processing by running mathematical operations on arrays of data in a simultaneous or pipelined fashion. Most algorithms in linear algebra can be easily vectorized. Therefore, in the late 1970s there was an effort to standardize vector operations for use in scientific computations. The idea was to define some simple, frequently used operations and implement them on various systems to achieve portability and efficiency. This package came to be known as the Level-1 Basic Linear Algebra Subprograms (BLAS) or Level-1 BLAS.
The term Level-1 denotes vector-vector operations. As we will see, Level-2 (matrix-vector operations), and Level-3 (matrix-matrix operations) play important roles as well.
In the 1970s, the algorithms of dense linear algebra were implemented in a systematic way by the LINPACK project. LINPACK is a collection of Fortran subroutines that analyze and solve linear equations and linear least-squares problems. The package solves linear systems whose matrices are general, banded, symmetric indefinite, symmetric positive definite, triangular, and tridiagonal square. In addition, the package computes the QR and singular value decompositions of rectangular matrices and applies them to least-squares problems.
LINPACK uses column-oriented algorithms, which increase ...