Programmability with increased performance? New strategies to attain this goal include two approaches to data flow architecture: data flow multiprocessors and the cell block architecture.

Data Flow Supercomputers

Novel data-driven and demand-driven computer architectures are under development in a large number of laboratories in the United States, Japan, and Europe. These computers are not based on the tradlUonal von Neumann organization; instead, they are attempts to identify the next generation of computer. Basmally, m data-driven (e.g., data-flow) computers the availability of operands triggers the execution of the operation to be performed on them, whereas in demand-driven (e.g, reduction) computers the reqmrement for a result triggers the operation that will generate it. Although there are these two distinct areas of research, each laboratory has developed its own mdlvxdual model of computation, stored program representation, and machine organization. Across this spectrum of designs there m, however, a significant sharing of concepts. The aim of this palaer m to identify the concepts and relationships that exist both within and between the two areas of research. It does thin by examlmng data-driven and demand-driven architecture at three levels, computation organizatmn, (stored) program organization, and machine organLzation. Finally, a survey of various novel computer architectures under development is given.

Data-Driven and Demand-Driven Computer Architecture

Dataflow machines are programmable computers of which the hardware is optimized for fine-grain data-driven parallel computation. The principles and complications of data-driven execution are explained, as well as the advantages and costs of fine-grain parallelism. A general model for a dataflow machine is presented and the major design options are discussed.Most dataflow machines described in the literature are surveyed on the basis of this model and its associated technology. For general-purpose computing the most promising dataflow machines are those that employ packet-switching communication and support general recursion. Such a recursion mechanism requires an extremely fast mechanism to map a sparsely occupied virtual space to a physical space of realistic size. No solution has yet proved fully satisfactory.A working prototype of one processing element is described in detail. On the basis of experience with this prototype, some of the objections raised against the dataflow approach are discussed. It appears that the overhead due to fine-grain parallelism can be made acceptable by sophisticated compiling and employing special hardware for the storage of data structures. Many computing-intensive programs show sufficient parallelism. In fact, a major problem is to restrain parallelism when machine resources tend to get overloaded. Another issue that requires further investigation is the distribution of computation and data structures over the processing elements.

http://web.cecs.pdx.edu/~alaa/ece588/papers/veen_acmcs_1986.pdf

Dataflow machine architecture

This paper describes the development of a wavefront-based language and architecture for a programmable special-purpose multiprocessor array. Based on the notion of computational wavefront, the hardware of the processor array is designed to provide a computing medium that preserves the key properties of the wavefront. In conjunction, a wavefront language (MDFL) is introduced that drastically reduces the complexity of the description of parallel algorithms and simulates the wavefront propagation across the computing network. Together, the hardware and the language lead to a programmable wavefront array processor (WAP). The WAP blends the advantages of the dedicated systolic array and the general-purpose data-flow machine, and provides a powerful tool for the high-speed execution of a large class of matrix operations and related algorithms which have widespread applications.

Wavefront Array Processor: Language, Architecture, and Applications

Data-flow languages and visual programming environments are reviewed. The iconic data flow (IDF) system, a visual programming environment initially oriented toward the development of image processing and computer vision algorithms, is described. It is intended to exploit the intuitive nature of flow charts and the human capability for comprehending pictures, while providing a fully featured, easy to use programming language, including a full set of data types and control flow operators. Experimental performance results on the size of the programs that can be described and how fast they run are reported. Areas for future work are indicated. >

IDF: A graphical data flow programming language for image processing and computer vision

In an attempt to increase the performance of computing machines, there appears to be two main approaches—(1) to use faster components in existing architectures and (2) to design new architectures which are capable of exploiting some form of concurrency. The first approach is inherently limited in that the effects of reduced integrated circuit geometry, new process technology, and new logic families can be expected to increase overall system performance by only a couple of orders of magnitude. While this is initially impressive, it does not allow the desired machine performance projected to be necessary to solve large physics problems, or needed for accurate weather prediction. 16 The second approach, while being a considerably more difficult organizational problem, is inherently unlimited in nature. There are numerous levels at which concurrency can be exploited in digital computers, i.e. multiple data paths, more concurrent realization of low-level circuit functions, overlap and pipeline processing within a single processing element, multiple processors, etc. In developing any new "fast as possible" machine, it is important to attempt to implement all of the above suggestions. However the work reported here will mainly be concerned with solving the problem of how to utilize and organize systems containing large numbers of independent processors.

A. L. Davis

Papers

A data flow evaluation system based on the concept of recursive locality