Code Modernization: Case Study Using Matrix Multiplication

• Nickolas R. Machado UFRRJ
• Juliana M.N.S. Zamith UFRRJ

Abstract

Arithmetic operations involving matrix structures are processes that consume significant execution times depending on the size of the instance to be resolved. This work proposes to optimize and modernize a code that solves the matrix multiplication problem in order to improve its performance. For this, we used parallelization techniques such as vector instructions (AVX), OpenMP and the use of graphics cards. Results were obtained that improve up to 735 times the computational time of the modernized codes when compared to the non-modernized ones.

Keywords: Code Modernization, Matrix Multiplication, Parallel Programming

References

[1] Jason Sanders & Edward Kandrot, (2010) “CUDA by example : an introduction to
general-purpose GPU programming”.
[2] Tim Mattson, (2012) “A “Hands-on” Introduction to OpenMP”.
[3] Intel Corporation, “Intrinsics Guide”.