High-performance Graphics Research Articles

Cooperative Visualization of Computational Fluid Dynamics

AbstractTempus Fugit/Interview is a computational fluid dynamics visualization application for which processing is distributed between high performance graphics workstations and supercomputers. Facilities are provided in the application for more than one user to view shared images creating a cooperative visualization environment. The way in which the computation is partitioned between the super computer and the workstations is critical to the capability of the application to present simultaneous, identical, animated images of fluid dynamics to more than one user.

PIX: An object-oriented network graphics environment

This paper describes PIX, a system to enhance X with PHIGS' three dimensional graphics capability. Our goal is to provide the user with the conveniences and features of PHIGS—which is an international standard for 3D computer graphics programming—in a network-oriented, hardware-independent environment. To ensure portability of our system, we have designed and implemented PIX under the X Window System, a network-transparent windowing system developed at MIT. Some of PIX's important and unique features are: an object-oriented data structure design, a network-oriented system architecture, separable server processes to take advantage of a dedicated 3D graphics hardware. In particular, simulations with the 3DP, a three Dimensional Processor developed at Computer Motion Inc., have indicated promising potential for PIX to evolve into a high performance graphics system.

Use of transputers for real time dose calculation and presentation for three-dimensional radiation treatment planning

Purpose : Real-time 3-dimensional dose calculation will allow display of isodose contours and other metrics for a planner to assess plan effectiveness during plan development, facilitating optimization. Methods and Materials : Parallel processing provides an effective means to calculate 3-dimensional dose distribution in real-time while plan parameters are being chosen and adjusted. An array of 20 transputers and a high performance graphics workstation have demonstrated the feasibility of real-time 3-dimensional beam parameter specification, dose calculation, and dose-distribution presentation for evaluation. A mesh connected set of processors using surface processors to generate and terminate rays, and ray processors to calculate ray attenuation and dose distribution has been developed to efficiently utilize large numbers of processors and provide good load sharing, even for small beams that intersect only a small part of the volume. Results : Our feasibility study has calculated dose distribution by the Effective Path Length method in about one second per beam for a treatment volume of 56,400 voxels. We expect to reduce the total time for computation, communication, and display, with even larger volumes, to less than one second. The number of processors can easily be increased for larger treatment volumes or more accurate and computation-intensive dose-calculation algorithms. Transputers provide an elegant and economical method for harnessing up to hundreds of powerful general-purpose processors for computational tasks including dose calculation and isodose contour generation. The same distributed-memory parallel-processing configuration is also suitable for calculation of isodose contours and dose-volume histograms for plan evaluation, automatic calculation of apertures and filters as beam parameters are manipulated, and more accurate dose calculation algorithms that incorporate the effects of scatter. Conclusion : Parallel processors can efficiently provide real-time calculation of the information necessary to evaluate treatment plans as they are developed allowing the planner to optimize the plan based on dose distribution and its effects on tumor control and complications.

Breaking the frame-buffer bottleneck with logic-enhanced memories

Logic-enhanced memory chips that can remove the rasterizer/frame buffer bottleneck which limits the performance of current image-generation architectures are discussed. Putting pixel memory on-chip with rasterizing processors provides the two to three orders of magnitude improvement in access rates needed to support realistic shading models and aliasing in interactive systems. Current high-performance graphics systems and logic-enhanced memory architectural issues are reviewed. The design of the PixelFlow Enhanced Memory Chip (EMC), which exploits advances in semiconductor technology and circuit techniques to build compact, high-performance rasterizers, is described. >

Computer‐generated three‐dimensional reconstruction of the mediastinm correlated with sectional and radiological anatomy

AbstractWe describe a three‐dimensional computer‐generated model of the human mediastinum. Contours were manually traced from photographs of a serially sectioned, unfixed frozen cadaver block, Using software developed in‐house, the contours were digitized, automatically assembled and placed in registration, and displayed on a high performance graphics workstation as surfaced images. The model thus produced has advantages over traditional representations of anatomical structures; it is realistic, accurate, efficient, and flexible. Each anatomical component of the mediastinum is stored as a separate file in the computer and can be selectively accessed, displayed, and manipulated, allowing the model to be disassembled, reassembled, dissected, and viewed in any orientation. The model can be divided into a series of transversely cut “slabs,” the surfaces of which correlated with the anatomy displayed in the original cryosections and with radiological scans, thus permitting 2‐D sectional anatomy to be viewed in the context of the 3‐D model. Animation of the model can be accomplished by recording pre‐planned scripts onto videodiscs. The videodiscs serve as image archives that can be viewed in an interactive, self‐paced manner or accessed from a computer‐based multimedia system that is being developed by the Digital Anatomist Program. Anatomical reconstructions based on datasets of x‐, y‐, z‐coordinates have potential clinical applications as models that can be displayed to match the sectional planes of radiological images. The mediastinum model we describe is part of a large image database being developed by the Digital Anatomist Program that will form the spatial component of a knowledge base in structural biology. © 1992 Wiley‐Liss, Inc.

High-speed networks, visualization, and massive parallelism in the Advanced Computing Laboratory

Very high performance massively parallel machines are now available to solve significant problems for the nation that have been heretofore inaccessible. However, these machines, with their large memories are of little use for these problems without existing in an environment in which the scientist can perceive what is happening in the calculation and analyze the results effectively. This requires being able to transmit enormous amounts of data from these machines through very high performance networks in a way which is easily comprehended and manipulated by the scientist. In order to provide such an environment for these Grand Challenge applications, researchers in the Advanced Computing Laboratory have been developing a very high-speed network (100 Mbytes/s) based on the standard HIPPI protocol developed at Los Alamos. This is used to send data over a multiple cross bar network between supercomputers, high performance graphics machines, and high-speed framebuffers. Examples of scientific problems being explored in this way will be given, including approaches to high-speed archival data storage and interactive data exploration over these networks.

Myofiber orientation in the weanling mouse heart

This study provides a quantitative description at the cellular level of myofiber orientation throughout the ventricles of the mouse heart. We employed computer-based methods of three-dimensional reconstruction from 3 microns plastic-embedded serial sections. Registration marks were introduced by drilling minute holes into each plastic block. Subfields of selected sections were photographed at 20 x magnification, using a computer-controlled microscope. The 35-mm film frames were projected onto a digitizer tablet and the epi- and endocardial boundaries were digitized manually. The "heads" and "tails" of linear segments of a representative myofiber sample present in each projected image were digitized in point mode. The many x-, y-, z-coordinate tables generated by digitization were reassembled automatically, giving a numerical description of the myofiber pattern. This pattern was studied interactively on a high-performance graphics workstation. We find that the heart wall is, to a first approximation, a "sandwich," in which the myofibers in the middle layer run mainly circumferentially, whereas those in the inner and outer layers run parallel or oblique to the apical-basal axis, a variant of the classical model of the myofiber pattern. We observed a "sleeve" in the interventricular septum, formed by longitudinal and oblique myofibers, a feature which apparently has not been described previously. Myofibers not running parallel to the transverse or longitudinal planes were not resolved in this study. We conclude that three-dimensional reconstruction of the cardiac myofiber pattern at the light-microscopic level, while laborious, is technically feasible and scientifically worthwhile.

Biological applications of scanning tunnelling microscopy: novel software algorithms for the display, manipulation and interpretation of STM data

The development of the scanning tunnelling microscope, with its inherent and unique problems associated with image interpretation, has led to the need for specially tailored protocols for post-acquisition data manipulation. In an attempt to address this need, the authors have developed an integrated visualization system for the concurrent display of complementary data types, both experimental and theoretical, which may themselves be subject to appropriate data processing. The requirement for real-time image manipulation has led to the implementation of the software on high performance computer graphics workstations, with minimal dependence on specific hardware being achieved through adoption of appropriate standards including UNIX, ANSI 'C', PHIGS, and OSF/Motif. The authors demonstrate the potential of this visualization system by its application to physical and biological data sets.

Scientific visualization in high-speed network environments

In several cases, new visualization techniques have vastly increased the researchers' ability to analyze and comprehend data. Similarly, the role of networks in providing an efficient supercomputing environment has become more critical and continues to grow and at a faster rate than the increase in the processing capabilities of supercomputers. A close relationship between scientific visualization and high-speed networks in providing an important link to support efficient supercomputing is identified. The two technologies are driven by the increasing complexities and volume of supercomputer data. The interactions of scientific visualization and high-speed networks in a Computational Fluid Dynamics simulation/visualization environment are given. Current capabilities supported by high-speed networks, supercomputers, and high-performance graphics workstations at the Numerical Aerodynamic Simulation Facility (NAS), NASA Ames Research Center are described. Applied research in providing a supercomputer visualization environments to support future computational requirements are summarized.

A SIMD parallel trapezoid rasterization algorithm

An algorithm for the rasterization of trapezoids on the Connection Machine TM (CM) is described. The input consists of an array of trapezoids, with two horizontal sides, arranged with one trapezoid per processor. (Unless otherwise indicated, “processor” should be taken to mean virtual processor.) Each trapezoid is converted to an edge record and the edge records are then distributed to enough processors so that each processor is responsible for one scanline of one trapezoid. Each processor computes a scan record for its scanline, and the scan records are then distributed to enough processors so that one processor is responsible for a single pixel. Final interpolation of position, and possibly shading information, is performed in parallel for all pixels thus created, and the pixels are then broadcast into a frame buffer array, with depth comparisons being performed at the receiving end to ensure that the nearest pixel appears in the array. The Connection Machine-specific features used by the algorithm are logarithmic time cumulative summing, general communication with comparisons for collision arbitration, and virtual processor sets. Performance is similar to that of good graphics workstations. The intended application is to display data already resident in the machine as the result of some previous computation, when a high-performance graphics workstation is not available.

Efficient use of workstations for passive monitoring of local area networks

Effective management of a local area network (LAN) requires not only a protocol to manage the active entities, but also a means to monitor the LAN channel. This is especially true in shared-channel LANs, such as Ethernet, where the behavior of the LAN as a whole may be impractical to deduce from the states of the individual hosts. Passive monitoring can be done using either a dedicated system or a general-purpose system. Dedicated monitors have been favored for several reasons, but recent workstations, when carefully programmed, are sufficiently powerful to serve this function. Using a workstation offers high-performance graphics and a more flexible environment for collecting and presenting LAN behavior.

Meaningful graphics workstation performance measurements

We present how graphics workstation performance is currently advertised by workstation manufactures. We suggest that the manufacturers' performance numbers are not the best indicators of how well the system will perform. Four levels of graphics system performance measurement have been defined by the NCGA Graphics Performance Characterization group[1]. These levels are low-level primitives (points, lines, and polygons per second), pictures, systems (input and action response times), and applications. We suggest that the best indicator of a graphics workstation's performance capabilities is that of an application running on the system. For applications level performance measurements, there are many potential applications from which to choose. We suggest that three-dimensional visual simulators are one of the better applications to study as that type of system has a realistic mix of processor and graphics computational requirements. We describe one such system, the Moving Platform Simulator (MPS), developed at the Naval Postgraduate School with the goal of fully listing its capabilities and utilized graphics techniques. We then present how that system performs on a currently available, high-performance graphics workstation.

Conversion of contoured topography to digital-terrain data

High-performance graphics systems coupled with video digitizers provide the capability to obtain digital topography from contour maps. A technique is presented for conversion of contoured data to gridded data. This technique provides a cost-efficient option for obtaining digital topography where such data are not available, for obtaining average elevation data in support of gravity surveys, or for gridding hand-contoured data. The method is interactive, makes use of standard image processing and computer graphics techniques, and requires a user-operator who can read topographic maps. The resultant digital data retain the smooth trends of the contour data and are well within the error limits reflected in the placement of the original contour lines.

Hardware acceleration for Window systems

Graphics pipelines are quickly evolving to support multitasking workstations. The driving force behind this evolution is the window system, which must provide high performance graphics within multiple windows, while maintaining interactivity. The virtual graphics system presented by [7] provides a clean solution to the problem of context switching graphics hardware between processes, but does not solve all the problems associated with sharing graphics pipelines.The primary difficulty in context switching a graphics accelerator is the pipeline latency encountered during a pipeline flush. This latency removes the responsiveness and interactivity of the graphics system. As primitives become more complex and pipelines become longer, pipeline latency grows. Hardware solutions are described which further accelerate the window system by eliminating the need for pipeline flushing and resynchronization. An overview of the entire system is presented, highlighting the hardware mechanisms which contribute to window acceleration.

A display system for the Stellar graphics supercomputer model GS1000

This paper describes a high performance display system that has been incorporated into the overall architecture of the Stellar Graphics Supercomputer Model GS1000. The display system is tightly coupled to the CPU, memory system and vector processing unit of this supercomputer, and is capable of rendering 150,000 shaded triangles/sec, and 600,000 short vectors/sec. The goal of the architecture is to share hardware resources between the CPU and display system and achieve a high bandwidth connection between them. This coupling of the display system and the processor, the architecture of the rendering processor, and the two ASICs that are used to implement the rendering processor are described.In addition, the display system architecture is contrasted to other approaches to high performance graphics, and design trade-offs and possible extensions are described. The implementation of popular display algorithms on the architecture is discussed, and their performance specified. The reader is advised that Stellar Computer Inc. is seeking patent protection for work described in this paper.

Subanosecond pixel rendering with million transistor chips

The desire for higher performance and higher resolution continuously increases the pixel update rates needed in high performance graphics systems. The increasing density of memory chips on the other hand reduces the pixel update rate that can be provided by the frame buffer. We present the design of a VLSI chip and a graphics system that can sustain sub-nanosecond pixel rendering rates for three-dimensional polygons and can be used to render about a million Z-Buffered and Gourard shaded polygons per second. The chip has been designed at the IBM Research Division's Thomas J. Watson Research Center.

Computer-aided research into a natural neural network

While artificial neural networks are being applied to solving problems in signal processing and other domains, much remains to be discovered about how even the simpler biological neural networks function. Studying simpler examples of natural systems promises to advance our understanding of principles of organization of neural tissue wherever it occurs. It also may prove useful in the development of new computer architectures. Accordingly, the authors have begun the study of the macular linear bioaccelerometers, or balance organs of mammals, using the rat as the model for the class. This effort includes creating a computer-based workbench that a scientist can use to generate geometric reconstructions of neural tissue from electron microscope serial sections, to create a functional model of information flow within the neural tissue, and ultimately to generate computer animations to visualize how the network functions. Their work to data is based upon the study of long series of sections in a transmission electron microscope. The sections are photographed and the photographs are assembled into montages. Selected nerves and receptor units synapsing with them (their receptive fields) are next traced onto transparencies from the montages. The tracings, which are cross-sectional contours, are digitized with a tablet and stored inmore » data files on a personal computer. The files are transferred to a high performance graphics workstation, where software has been developed to reconstruct these sets of contours as polygonal objects, display them in wireframe or solid form, and create sequence files that can be used to produce a computer animation on videotape.« less

Numerical flow field visualization

The application of computer-generated imagery to analyze numerical solutions of fluid flow is presented. High-performance graphics workstations, their current capabilities, and their use in flow field analysis are discussed. Interactive flow field computations about complex three-dimensional bodies applying computer-generated imagery are projected for the future.

Coordinated three-dimensional reconstruction from serial sections at macroscopic and microscopic levels of resolution: the human heart.

This report describes procedures we have developed for obtaining correlated quantitative structural information at two very different levels of resolution. Accurate reconstructions of entire organs and samples of tissue within organs were produced in correct scalar and topographical relationship using computer-assisted techniques. A specially designed sectioning apparatus, a macrovibratome, was used to section serially the ventricles of the human heart macroscopically. Photographs were taken of every slice. A tissue block excised from a slice at a specified locus in the left ventricular wall was embedded in plastic; serial-3 micron sections were cut in each of two orthogonal orientations. Photomicrographs were taken by semi-automated microscopy. Images of both macroscopic and microscopic sections were projected onto a bitpad and manually digitized. The resulting tables of x-, y-, and z-coordinates were reassembled on a VAX 11/750 computer, then transferred to a high-performance graphics workstation and displayed as three-dimensional images. Microscopic images were shown in the correct reference frame with respect to the macroscopic (parent) structure.

Graphics Controllers: An Overview

Computer graphics is a topic of rapidly growing importance in the field of computer science. It has always been one of the most spectacular branches of computer technology. Since vision provides one of the richest method of communication, it is therefore of extreme importance to improve the quality of visual communication. Also for applications like three dimensional graphics, animation etc., a high performance graphics system is inevitable.Due to all these reasons, special purpose hardware is used along with general purpose microprocessors to design graphics workstations. With the advance in VLSI technology, this special purpose hardware has been designed using a variety of chips. Depending on the extent to which these chips can unload the host processor of graphics related activities, we have classified the chips into four generations. A survey of these generations is presented in this paper.