Architectures and Compilation Techniques for Fine and Medium Grain Parallelism: Proceedings of the IFIP WG10.3 Working Conference, Orlando, FL, USA, 20-22 January 1993(v. A-23 IFIP Transactions A: Computer Science and Technology)

Medium and especially fine grain parallelism has been a focus of the data-research area since its inception in the 1970s. Much experience has been gained but the interest both in the academia and the industry continues to flourish. The development of multiple ALU superscalars/superpipelined machines/VLIWs (small resources Von Neumann machines) has meant the related software and hardware topics for finding higher degrees of fine and medium grain parallelism on such machines has become increasingly important. This volume presents new parallelization ideas being discovered in this relatively unchartered research area, including some which are likely to have immediate practical impact. With invited papers from prominent specialists, it is hoped it also offers a critical review of the accomplishments of the data-flow research area to date.

Table of Contents:
Part 1 Compilation for parallelism: decomposed software pipelining - a new approach to exploit instruction level parallelism for loop programs, J. Wang and C. Eisenbeis; software pipelining - petri net pacemaker, V.H. Allan et al; efficient execution of doacross loops on distributed memory systems, A. Zaafrani and M.R. Ito; balancing fine- and medium-grained parallelism in scheduling loops for the XIMD architecture, C.J. Newburn et al; a new loop partition method-clustering, S.Y. Tseng et al; flow-sensitive inter-procedural analysis method for parallelization, T. Iitsuka; code generation for multi-threaded architectures from dataflow graphs, S. Murer and P. Farber; the initial performance of a bottom-up clustering algorithm for dataflow graphs, W.A. Najjar et al. Part 2 Architectures: synchronization and parallelism control in the BARDE dataflow processor, H. Seebauer and J. Siemers; the instruction set architecture of the inference processor UNIRED II, K. Shimada et al; a functional data-flow architecture dedicated to real-time image processing, J. Serot et al; DART - a data-driven processor architecture for real-time computing, W.G. Farquhar and P. Evripidou; two fundamental limits on dataflow multiprocessing, D.E. Culler et al; the realities of parallel processing and dataflow's role in it - a NASA HPCC perspective, T.L. Sterling and M.J. MacDonald. Part 3 Communication issues: integrated network barriers for d-dimensional meshes, J.A. Solworth and J. Stamatopoulos; parallel dimension permutations on star graph, S. Latifi; on the effectiveness of interleaved memories for binary trees, R.V. Boppana. Part 4 VLIW processors: contribution of compilation techniques to the synthesis of dedicated VLIW architectures, G. Menez et al; selective scheduling framework for speculative operations in VLIW and superscalar processors, S.M. Moon et al; URSA - a Unified ReSource Allocator for registers and functional units in VLIW architectures, D.A. Berson et al. Part 5 Experiences in parallel computing: data flow computing and the conjugate gradient method, N. Rubin; results of parallel implementations of the selection problem using SISAL, M. Daumas and P. Evripidou; a parallel implementation of the travelling salesman problem on a sequent symmetry multiprocessor, A. Sohn. Part 6 Language issues: exploring the stream data type in SISAL and other languages, A.L. Wendelborn and H. Garsden; multidimensional streams rooted in dataflow, E.A. Lee. Part 7 Compiling practice: high level compiling for low level machines, C. O'Donnell; modelling instruction-level parallelism for software pipelining, A. Adl-Tabatabai et al.