TechRxiv
article.pdf (397.2 kB)

Multiple-Precision BLAS Library for Graphics Processing Units

Download (397.2 kB)
preprint
posted on 30.06.2020, 06:55 by Konstantin Isupov, Vladimir Knyazkov
The binary32 and binary64 floating-point formats provide good performance on current hardware, but also introduce a rounding error in almost every arithmetic operation. Consequently, the accumulation of rounding errors in large computations can cause accuracy issues. One way to prevent these issues is to use multiple-precision floating-point arithmetic. This preprint, submitted to Russian Supercomputing Days 2020, presents a new library of basic linear algebra operations with multiple precision for graphics processing units. The library is written in CUDA C/C++ and uses the residue number system to represent multiple-precision significands of floating-point numbers. The supported data types, memory layout, and main features of the library are considered. Experimental results are presented showing the performance of the library.

Funding

This work was supported by the Russian Science Foundation (grant number 18-71-00063)

History

Email Address of Submitting Author

ks_isupov@vyatsu.ru

ORCID of Submitting Author

0000-0003-0239-0404

Submitting Author's Institution

Vyatka State University

Submitting Author's Country

Russian Federation

Licence

Exports

Licence

Exports