A Reference Implementation for Extended and Mixed Precision BLAS

This paper describes a C implementation of the proposed new BLAS Standard. Permitting mixtures of input/output types and precisions, as well as higher internal precision, the new BLAS standard contains many more subroutines than the existing standard. We have developed an automated process of generating and systematically testing these large numbers of routines. We believe our methodology could be applicable to the other languages besides C. In particular, our algorithms used in the testing code would be very valuable to all the other BLAS implementors. 1 Introduction This library of routines is part of a reference implementation for the Dense and Banded BLAS routines, along with their Extended and Mixed Precision versions, as documented in Chapters 2 and 4 of the new BLAS Standard. This is obtainable from the followin URL, to which we refer the reader for details: http://www.netlib.org/cgi-bin/checkout/blast/blast.pl EXTENDED PRECISION is only used internally; the input and o..