By using a combination of 32-bit and 64-bit floating point arithmetic, the
performance of many dense and sparse linear algebra algorithms can be significantly
enhanced while maintaining the 64-bit accuracy of the resulting solution. The
approach presented here can apply not only to conventional processors but also to
exotic technologies such as Field Programmable Gate Arrays (FPGA), Graphical
Processing Units (GPU), and the Cell BE processor. Results on modern processor
architectures and the Cell BE are presented.
Buttari, A.
,
Dongarra, J.
,
Kurzak, J.
,
Luszczek, P.
, and
Tomov, S.
(n.d.).
Using mixed precision for sparse matrix computations to enhance the
performance while achieving 64-bit accuracy
, ACM Transactions on Mathematical Software, 1, to
appear December 2008.
2.
Demmel, J.W.
(1997). Applied Numerical Linear Algebra,
SIAM, Philadelphia, PA.
3.
Higham, N.J.
(1996). Accuracy and Stability of Numerical Algorithms
,
SIAM, Philadelphia, PA.
4.
Kurzak, J.
and
Dongarra, J.J.
(n.d.a). Implementation of mixed precision in solving systems of linear
equations on the CELL processor, Concurrency Computat. Pract. Exper, to appear.
5.
Kurzak, J.
and
Dongarra, J.J.
(n.d.b). Mixed precision dense linear system solver based on cholesky
factorization for the CELL processor, Concurrency Computat. Pract. Exper,
in preparation.
6.
Moler, C.B.
(1967).
Iterative refinement in floating point
, J. ACM, 14(2):
316—321
.
7.
Stewart, G.W.
(1973). Introduction to Matrix Computations},
Academic Press, New York, NY
.
8.
Stewart, G.W.
(2001). Matrix algorithms,
Philadelphia, PA: Society for Industrial and
Applied Mathematics
.
