54
reads

2012-10-19
09:30:00 - 12:00:00

103 , Mathematics Research Center Building (ori. New Math. Bldg.)

We will study the paper "Accelerating scientific computations with mixed precision algorithms" by Baboulina, Buttarib, Dongarrac, Kurzakc, Langouc, Langouf, Luszczekg, and Tomov in this student seminar. The paper has been published in Computer Physics Communications, 2009. It shows how we can use combination of 32-bit and 64-bit floating point arithmetic to significantly enhanced the performance of linear algebra algorithms, while maintaining the 64-bit accuracy of the resulting solutions. The approach can be applied to conventional processors and latest technologies such as GPU, FPGA, and Cell processors.