Today's Workstations Research Articles

Rapid system prototyping for high performance reconfigurable computing

Proposes a soft-reconfigurable coarse-grain platform, based on a very few types of computational cells, and shows how a range of advanced applications can be mapped for achieving high performance. Two very flexible cells are proposed that account for the versatility of the approach. These are the MA PLUS, which is an enhanced multiply-add cell, and the UNL (Universal NonLinear) cell. These two cells together provide an unprecedented capability for 'coarse-grain reconfigurable computing'. Although the applications range from FIR filtering of images to large-scale inverse problems, this paper demonstrates the mapping of two advanced problems, namely (1) reconstruction of computerized tomography images from fan-beam projections, and (2) color conversion of video from the RGB to HSI domain. A speedup by factors of 20 or more over today's workstations is estimated.

Multilevel hypergraph partitioning: applications in VLSI domain

In this paper, we present a new hypergraph-partitioning algorithm that is based on the multilevel paradigm. In the multilevel paradigm, a sequence of successively coarser hypergraphs is constructed. A bisection of the smallest hypergraph is computed and it is used to obtain a bisection of the original hypergraph by successively projecting and refining the bisection to the next level finer hypergraph. We have developed new hypergraph coarsening strategies within the multilevel framework. We evaluate their performance both in terms of the size of the hyperedge cut on the bisection, as well as on the run time for a number of very large scale integration circuits. Our experiments show that our multilevel hypergraph-partitioning algorithm produces high-quality partitioning in a relatively small amount of time. The quality of the partitionings produced by our scheme are on the average 6%-23% better than those produced by other state-of-the-art schemes. Furthermore, our partitioning algorithm is significantly faster, often requiring 4-10 times less time than that required by the other schemes. Our multilevel hypergraph-partitioning algorithm scales very well for large hypergraphs. Hypergraphs with over 100 000 vertices can be bisected in a few minutes on today's workstations. Also, on the large hypergraphs, our scheme outperforms other schemes (in hyperedge cut) quite consistently with larger margins (9%-30%).

Interactive real-time motion blur

Motion blurring of fast-moving objects is highly desirable for virtual environments and 3D user interfaces. However, all currently known algorithms for generating motion blur are too slow for inclusion in interactive 3D applications. We introduce a new motion-blur algorithm that works in three dimensions on a per object basis. The algorithm operates in real time even for complex objects consisting of several thousand polygons. While it only approximates true motion blur, the generated results are smooth and visually consistent. We achieve this performance break-through by taking advantage of hardware-assisted rendering of semitransparent polygons, a feature commonly available in today's workstations.

The uc berkeley system for interactive visualization of large architectural models

Realistic-looking architectural models with furniture may consist of millions of polygons and require gigabytes of data-far than today's workstations can render at interactive frame rates or store ...

The UC Berkeley System for Interactive Visualization of Large Architectural Models

Realistic-looking architectural models with furniture may consist of millions of polygons and require gigabytes of data—far than today's workstations can render at interactive frame rates or store in physical memory. We have developed data structures and algorithms for identifying a small portion of a large model to load into memory and render during each frame of an interactive walkthrough. Our algorithms rely upon an efficient display database that represents a building model as a set of objects, each of which can be described at multiple levels of detail, and contains an index of spatial cells with precomputed cell-to-cell and cell-to-object visibility information. As the observer moves through the model interactively, a real-time visibility algorithm traces sightline beams through transparent cell boundaries to determine a small set of objects potentially visible to the observer. An optimization algorithm dynamically selects a level of detail and rendering algorithm with which to display each potentially visible object to meet a userspecified target frame time. Throughout, memory management algorithms predict observer motion and prefetch objects from disk that may become visible during imminent frames. This paper describes an interactive building walkthrough system that uses these data structures and algorithms to maintain interactive frame rates during visualization of very large models. So far, the implementation supports models whose major occluding surfaces are axis-aligned rectangles (e.g., typical buildings). This system is able to maintain over twenty frames per second with little noticeable detail elision during interactive walkthroughs of a building model containing over one million polygons.

Proposal for a less cumbersome anesthesia workstation.

With adaptation and implementation of these concepts, current problems with today's anesthesia workstations can be overcome. The problems of information management and workstation layout will be of less significance.

Automated determination of complex surface structures by LEED

Conventional surface crystallography by low-energy electron diffraction (LEED) employs a trial-and-error search controlled at each step by human effort. This trial-and-error approach becomes very cumbersome and unreliable to solve complex surfaces with a large number of unknown structural parameters. We discuss automatic optimization procedures for LEED, which combine numerical search algorithms with efficient methods of determining the diffracted intensities for varying structures. Such approaches can reduce the computer time required for an entire structure determination by many orders of magnitude, while fitting many times more unknown structural parameters. Thereby, relatively complex structures, with typically 10 adjustable atoms (or 30 adjustable coordinates), can be readily determined on today's workstations. These include non-symmetrically relaxed structures, surface reconstructions and adsorbate-induced substrate distortions. We also address the theoretical and experimental requirements for an accurate structural determination.

System support for time-critical applications

A number of interesting applications have time constraints associated with them — most notably those in the area of multimedia. Today's workstations do not provide adequate support for applications whose definition of correctness is a function of time. This paper presents a set of system-level mechanisms that permit the incorporation of multiple streams of sustained, high-bandwidth, time-critical information into the distributed workstation computing environment. The goal of this work is to enable programmers to write applications that manipulate (i.e. acquire, transport, process, store and present) and synchronize time-critical information streams, and do so with an ease comparable to that of manipulating text or graphics on a workstation today. This paper provides a description of the different types of temporal relationships required by time-critical applications, and provides a means of expressing these relationships to the underlying system in a minimally constraining fashion (i.e. through a technique called ‘loose synchronization’). The central concept behind the mechanisms described here is the notion of time-regulated transport of time-critical information being provided by means of mechanisms that are software analogues to Phase-Locked Loops (PLL).

Fitting dozens of coordinates by LEED: automated determination of complex surface structures

We present the first comprehensive discussion of automatic optimization procedures for surface-structure determination by LEED. These procedures combine numerical search algorithms with efficient methods of determining the diffracted intensities for varying structures. Such approaches can reduce the computer time required for an entire structure determination by many orders of magnitude, while fitting many times more unknown structural parameters. Thereby, relatively complex structures, with typically 10 adjustable atoms (or 30 adjustable coordinates), can be readily determined on today's workstations. We list over two dozen structures so determined, many as yet unpublished.

Estimating effective population size from samples of sequences: a bootstrap Monte Carlo integration method.

We would like to use maximum likelihood to estimate parameters such as the effective population size N(e) or, if we do not know mutation rates, the product 4N(e) mu of mutation rate per site and effective population size. To compute the likelihood for a sample of unrecombined nucleotide sequences taken from a random-mating population it is necessary to sum over all genealogies that could have led to the sequences, computing for each one the probability that it would have yielded the sequences, and weighting each one by its prior probability. The genealogies vary in tree topology and in branch lengths. Although the likelihood and the prior are straightforward to compute, the summation over all genealogies seems at first sight hopelessly difficult. This paper reports that it is possible to carry out a Monte Carlo integration to evaluate the likelihoods approximately. The method uses bootstrap sampling of sites to create data sets for each of which a maximum likelihood tree is estimated. The resulting trees are assumed to be sampled from a distribution whose height is proportional to the likelihood surface for the full data. That it will be so is dependent on a theorem which is not proven, but seems likely to be true if the sequences are not short. One can use the resulting estimated likelihood curve to make a maximum likelihood estimate of the parameter of interest, N(e) or of 4N(e) mu. The method requires at least 100 times the computational effort required for estimation of a phylogeny by maximum likelihood, but is practical on today's work stations. The method does not at present have any way of dealing with recombination.

Multicast support for group communications

Abstract This paper describes a multicast model which can be integrated into existing unicast communication systems to provide better support for group communications. Multicast services are becoming more important, as more and more of today's network workstation environments are used to provide group communications for the exchange of multimedia information [29]. These environments allow users to exchange information in the form of ‘documents’ containing text, graphics and voice; some systems support both store-and-forward (e.g., mail) and real-time (e.g., conferencing) material. In this paper, various multicast design issues are addressed and solutions are proposed to provide an effective multicast extension for a wide range of existing unicast protocols.

The enabling states approach: designing usable telecommunications services

Unless it is possible to use a broadband communications terminal for directly engaging in the tasks and activities which are important to the user (without the necessity of learning complex skills like those required to operate today's workstations) uptake will be restricted and usage will be inefficient. The enabling states approach, which distinguishes between the users' goal tasks, the states of the communication system which enable the performance of those tasks, and the enabling tasks required to create such enabling states, is described. The intention is to improve usability (and hence both uptake and usage) by designing interfaces that automatically identify and create the appropriate enabling states, allowing the user to concentrate on goal tasks rather than the enabling tasks. The approach is illustrated and discussed in the context of integrated broadband communication systems. >

Intel Darpa to Develop Prototype of 2000‐Processor Computer System Using New Intel Superchip

INTEL Scientific Computers announced that it had signed an agreement with the Defence Advanced Research Projects Agency to develop a prototype of a system that will ultimately contain as many as 2000 processors, each with the power of a Cray‐1. A significant portion of the funds will go toward the development of software to make prototypes of the envisioned systems as easy to use as today's technical workstations. Prototypes of the system are to be demonstrated by the end of 1991.

The precursors of personal computing

I am compiling an anthology of historical papers which laid the groundwork for today's workstations and personal computers. The following list is preliminary, based mostly on personal recollection and a little library research and suggestions from others.

Report: Speech Processing for AT&T Workstations

Speech processing can exploit the growing overlap of information processing and information transportation typical in many of today's advanced workstations that combine computing and telephony. With speech input and output capabilities added, the ubiquitous telephone becomes a remote terminal for the personal workstation, allowing access to its features from anywhere in the world. For example, features such as voice mail, remote access to text mail using text-to-speech synthesis, and access control using speech recognition can reduce time lost to the unproductive game of telephone tag.