Marcelo Zamith

Automatic Dynamic Task Distribution between CPU and GPU for Real-Time Systems

The increase of computational power of programmable GPU (graphics processing unit) brings new concepts for using these devices for generic processing. Hence, with the use of the CPU and the GPU for data processing come new ideas that deals with distribution of tasks among CPU and GPU, such as automatic distribution. The importance of the automatic distribution of tasks between CPU and GPU lies in three facts. First, automatic task distribution enables the applications to use the best of both processors. Second, the developer does not have to decide which processor will do the work, allowing the automatic task distribution system to choose the best option for the moment. And third, sometimes, the application can be slowed down by other processes if the CPU or GPU is already overloaded. Based on these facts, this paper presents new schemes for efficient automatic task distribution between CPU and GPU. This paper also includes tests and results of implementing those schemes with a test case and with a real-time system.

An adaptative game loop architecture with automatic distribution of tasks between CPU and GPU

This article presents a new architecture to implement all game loop models for games and real-time applications that use the GPU as a mathematics and physics coprocessor, working in parallel processing mode with the CPU. The presented model applies automatic task distribution concepts. The architecture can apply a set of heuristics defined in Lua scripts in order to get acquainted with the best processor for handling a given task. The model applies the GPGPU (general-purpose computation on GPUs) paradigm. In this article we propose an architecture that acquires knowledge about the hardware by running tasks in each processor and, by studying their performance over time, finding the best processor for a group of tasks.

A Neighborhood Grid Data Structure for Massive 3D Crowd Simulation on GPU

Simulation and visualization of emergent crowd in real-time is a computationally intensive task. This intensity mostly comes from the

An architecture for game interaction using mobile

O(n^2)

A game loop architecture for the GPU used as a math coprocessor in real-time applications

complexity of the traversal algorithm, necessary for the proximity queries of all pair of entities in order to compute the relevant mutual interactions. Previous works reduced this complexity by considerably factors, using adequate data structures for spatial subdivision and parallel computing on modern graphic hardware, achieving interactive frame rates in real-time simulations. However, the performance of existent proposals are heavily affected by the maximum density of the spatial subdivision cells, which is usually high, yet leading to algorithms that are not optimal. In this paper we extend previous neighborhood data structure, which is called neighborhood grid, and a simulation architecture that provides for extremely low parallel complexity. Also, we implement a representative flocking boids case-study from which we run benchmarks with simulation and rendering of up to 1 million boids at interactive frame-rates. We remark that this work can achive a minimum spee up of 2.94 when compared to traditional spatial subdivision methods with a similar visual experience and with lesser use of memory.

A bidimensional data structure and spatial optimization for supermassive crowd simulation on GPU

One of the most important factors in digital games is the immersion of the players to enhance their experience in the virtual world created by games. This immersion can be achieved in different ways, such as graphics effects, realistic physics, artificial intelligence and even from user inputs. The Nintendo Wii controller, Sony Playstation Move controller and Microsoft Xbox Kinect, to name a few, present new forms of user interaction using the players movements, which has led to new forms of gameplay, as well as great forms of feedback. In addition, mobile devices are providing new features, like accelerometers, touch screens, cameras and GPS, which allow new forms of interaction in games played on these devices. In response to this evolution, this paper presents a multi-platform architecture that supports a desktop game being controlled by mobile devices, using their built-in features such as vibration feedback, sound feedback, image feedback and unique inputs like location, voice recognition, camera and gestures in order to enhance player immersion during game play. Additionally, the architecture supports different types of user data input, such as gesture and motion, that can be utilized according to user preferences and desires. Moreover, with this architecture, the mobile screen device can be used to give players feedback, like visual effects, scores and statistics. These available schemes have proven to be very useful and attractive for players, giving them choices that best fit their preferences and abilities. This architecture can also be used in collaborative games, where each player uses his device to interact with games running both on mobile devices and desktop computers. In order to demonstrate our architectures effectiveness, we have developed a desktop game and done a case study with a pilot group of users, where some usability factors were evaluated. Finally, using a mobile device for game data input and player feedback provides people who are resistant to complex game controllers the opportunity to enjoy playing games using the same device they are accustomed to using daily, thereby minimizing the time needed to learn what is usually necessary to start enjoying a game. This can, in many situations, lower the barrier encountered by experienced smartphone users who are novice game players, as demonstrated in our pilot usability tests, which found high fun factors in such players.

Performance Evaluation of Optimized Implementations of Finite Difference Method for Wave Propagation Problems on GPU Architecture

This article concerns the use of a graphics processor unit (GPU) as a math co-processor in real-time applications in special games and physics simulations. To validate this approach, we present a new game loop architecture that employs GPUs for general-purpose computations (GPGPUs). A critical issue here is the process distribution between the CPU and the GPU. The architecture consists of a model for distribution, and our implementation offers many advantages in comparison to other approaches without the GPGPU stage. This architecture can be used either by a general-purpose language such as the Compute Unified Device Architecture (CUDA), or shader languages such as the High-Level Shader Language (HLSL) and the OpenGL Shading Language (GLSL). Although the architecture proposed here aims at supporting mathematics and physics on the GPU, it is possible to adapt any kind of generic computation. This article discusses the model implementation in an open-source game engine and presents the results of using this platform.

An Architecture with Automatic Load Balancing and Distribution for Digital Games

Computing and presenting emergent crowd simulations in real time is a computationally intensive task. This intensity is mostly due to the complexity of the traversal algorithm needed for the interactions of all elements against each other on the basis of a proximity query. By using special data structures such as grids, and due to the parallel nature of graphics hardware, relevant previou work reduced this complexity considerably, making it possible to achieve interactive frame rates. However, existing proposals tend to be heavily bound by the maximum density of such grids, which is usually high, leading to arguably inefficient algorithms. In this article we propose the use of a fine- grained grid and accompanying data manipulation, to lead to scalable algorithmic complexity. We also implement a representative flocking boids case study, from which we ran benchmarks with more than one million simulated and rendered boids at nearly 30fps. We remark that related previous work achieved no more than 15,000 boids with interactive frame rates.

Techniques for designing GPGPU games

The scattering of acoustic waves in non-homogeneous media has been of practical interest for the petroleum industry, mainly in the determination of new oil deposits. A family of computational models that represent this phenomenon is based on finite difference methods. The simulation of these phenomena demands a high computational cost. In this work we employ GPU for the development of solvers for a 2D wave propagation problem with finite difference methods. Although there are many related works that use the same implementation presented in this paper, we propose a detailed and novel performance and memory bottleneck analysis for this hardware architecture.

A Distributed Architecture for Mobile Digital Games Based on Cloud Computing

Distributed computing is being used in several fields to solve many computation intensive problems. In digital games, it is used mainly in multi-player games, where the majority of the game logic is processed in a mainframe or cluster. Single player games could also use distributed computing to process the game logic, devoting host processing to renderization, which is usually the task that digital games spend most of its processing time. By using distributed computing, games could need softer system requirements, since the game loop would be distributed. This paper presents a game loop that can be applied in both multi-player and single-player games, using automatic load balancing and distributing game logic computation among several computers.