We describe and evaluate a software-only implementation of a novel mechanism for accessing and streaming GPU-rendered content from the cloud to low-end user devices. The unique properties of our implementation enable the trivial cloud-deployment of graphics-intensive applications, even ones that were not originally intended to run in the cloud. We achieve this goal by creating virtual GPU nodes that appear to the application like hardware devices, but that do not incur the overhead of virtualization. The low-level access to the frame buffer maximizes the number of applications that work out-of-the-box without the system imposing any specific display manager or windowing system.

We present a compression algorithm and a streaming protocol designed for streaming of computer-desktop graphics. The encoder has low memory requirements and can be broken into a large number of independent contexts with a high degree of data locality. The encoder also uses only simple arithmetic, which makes it amenable to hardware or highly parallel software implementation. The decoder is trivial and requires no memory, which makes it suitable for use on devices with limited computing capabilities. The streaming protocol runs over UDP and has its own unique error recovery mechanism specifically designed for interactive applications.

A widely useful technique for implementing complex programs is to use multiple threads of execution. Unfortunately, when we multithread a program we introduce the potential for bugs caused by unprotected critical sections. A widely useful technique for reducing the number of critical sections, and hence the number of bugs, is that of active objects. This paper presents a design for active objects that is easy to understand, implement and use in any sufficiently powerful object-oriented programming language. We show the implementation and use of our design in both the JavaTM and C++ programming languages. © 1998 John Wiley & Sons, Ltd.

We present an algorithm for updating data flow information derived from a program, in response to program edits. Our algorithm, applicable to intraprocedural or interprocedural data flow problems, is more general than previous methods because it can update any monotone data flow problem defined on a reducible flow graph and can handle arbitrary program edits.Rather than design yet another special-purpose data flow update algorithm, we show how to reduce the class of data flow problems to another class of problems which already has a fast update algorithm. More specifically, we reduce a monotone data flow problem formulated for solution using Graham-Wegman elimination [12] to the problem of constructing and decorating an attributed tree structurally isomorphic to the dominator tree of the program flow graph.To update this dominator tree in response to program changes, we first show how to express domination in terms of two local properties of nodes and edges in the flow graph — niceness and deepness. Domination is then updated in response to an edge addition or deletion by repeatedly performing two local operations on the dominator tree, each operation restoring the local properties violated by the edge change. Second, we extend Reps' attributed tree update techniques [19,21,20] and his complexity analysis, to update attribute values in response to changes in this structure.