Ensino de Organizações de Memória em Arquiteturas Paralelas usando Placas Gráficas Aceleradoras

Abstract

This work proposes simple assignments by using Graphics Processing Units (GPU) to teach parallel architectures. NVIDIA GPU has become very popular in less than one decade, since the CUDA framework has appeared in 2007. A GPU is an interesting didactic resource as it is a parallel and programmable architecture, and it has several memory organizations to be explored as: main memory, shared memory in distributed banks, switch on/off and resize cache L1, specialized memories (textures and constant).