A Group Block Distribution Strategy for a Heterogeneous Machine

J. Barbosa, J. Tavares, and A. Padilha (Portugal)


Heterogeneous computing, LU factorization, static distri bution, load balancing


This paper discusses the data distribution problem for in herently sequential algorithms, such as the LU factoriza tion in linear algebra, when computed on heterogeneous machines. These algorithms present additional difficulties to optimize the processing time due to the fact that the com putational load for data matrix columns increases with their index, requiring a fine tuned load assignment and distribu tion. A new static load distribution strategy is presented that achieves better results than other algorithms developed for the same purpose as shown in the results section. The strategy is called static because it distinguishes the com putation environment implementation from the algorithm implementation, as explained below.

Important Links:

Go Back