Project 50: Modules for linear algebra
This project was closed in 1989. Compare with project 55 (BLAS).
------------------------------------------------------------------
Pasadena 1984:
==============
(Rice): The objectives for a set of modules for linear algebra are
1) machine efficiency,
2) software parts (higher level means of describing computations),
3) standardization,
4) to serve as a model for Ada packages, Fortran 8x modules, etc..
Several existing efforts are discussed. The need for rapid
coordination of these efforts is expressed. A subgroup of Aird, Ford,
Hanson, Lawson, Reid, Rice and Smith is formed to meet on 28 June to
initiate this coordination effort.
(see § 5.2. review of technical projects).
Como 1987:
==========
Rice reported that the intention was to create an abstract framework
for linear algebra computations with the following properties:
a) it naturally includes the basic mathematical operations of linear algebra
(solve Ax-b, least squares, matrix multiplication, eigenvalues,...),
b) it does not reflect any programming language or machine characteristics,
c) it reflects certain computational or algorithmic aspects such as special
matrix types (band, symmetric, square, etc.)
The objective is to provide consistency of concepts as a guide to both
"lower level" implementers (chip builder, package builders) and
"higher level" implementers (e.g., PSE's for partial differential
equations or statistics).
During the discussion it was noted that the proposed modules would be
something like a very extended language independent BLAS. Delves
suggested that an effort should be made at standardizing portable
specifications for linear algebra operations. The first order of
business would be to determine what operations are to be part of this
standard, and then how to bind them to programming languages. He plans
to bring a paper to the next meeting that would outline a suitable set
of operations. Ford said that one could divide the operations into
level 1 (low level), level 2 BLAS (higher level), level 3 BLAS (carry
the thinking further to parallel machines), and perhaps go even
further and think of language independence. Ford and Stetter expressed
their support for Delves' proposal.
Stanford 1988:
==============
J. Rice reported that the project was still active.
Jerusalem 1990:
===============
On Rice's suggestion this project was closed.