Parallel game programming

Abstract

On this TFG I wanted to find out if parallel programing is and will be used to develop and optimize games. We have this impression because of the fact that nowadays processors improve its power not raising the processing speed but raising the processing units. At the end of the TFG through the development of a windows platform layer we found out that modern low-level APIs give a lot of options to use parallel programming in the form of multithreading of rendering tasks, using SIMD instructions and having the compute pipeline to perform GPUGP tasks. We also found out why the GPU outperforms the CPU at parallel algorithm processing and how we can optimize GPU work focusing mainly on coalesced memory access and memory throughput but also measuring the occupancy and concurrency.