Block algorithm for Householder transformations for hybrid architecture computers

O.V. Popov, O.V. Rudich

Abstract


A parallel block cyclic algorithm for two-sided Householder transformations for hybrid architecture computers with multi-core and graphic processors is dealt with in the paper. Results of testing are presented together with investigation of algorithm’s characteristics (scalability, acceleration) depending both on its parameters and parameters of matrix being transformed.

Prombles in programming 2014; 2-3: 99-106


References


Уилкинсон Дж.Х., Райнш К. Справочник алгоритмов на языке Алгол. Линейная алгебра. – М.: Машиностроение, 1976. – 389 с.

http://www.top500.org

Химич А.Н., Молчанов И.Н., Попов А.В., Чистякова Т.В., Яковлев М.Ф. Параллельные алгоритмы решения задач вычислительной математики. – К.: Наук. думка, 2008. – 248 с.

http://www.netlib.org/ lapack/

http://developer.download.nvidia.com/CUBLAS.pdf [Електронний ресурс] CUBLAS

http://developer.nvidia.com/cuda-toolkit-4.0 [Електронний ресурс]

CUDA TOOLKIT 4.0


Refbacks

  • There are currently no refbacks.