scientific article; zbMATH DE number 1131498
The goal of the book is to describe the building of computing devices specialized to solve particular kind of problems. Such computing devices are called ``special-purpose computers, opposed to the general-purpose computers. A general-purpose one can, in principle, solve any numerical problem in a certain time having the corresponding amount of memory. The main idea with special-purpose computers is to solve the problems faster at the same price. The book contains 7 chapters. Chapter 1 is the introduction. Chapter 2 considers the evolution of general-purpose computers, especially the factors determining the computing speed: the clock frequency, the number of floating point units, the average number of clock cycles for a floating point operation and the efficiency of the computer on a given problem. Chapter 3 explains the necessity of special-purpose computers. The silicon semiconductor technology resulted in a drastic increase of transistors, one can pack tens of multipliers into a chip, but we do not know how to construct a programmable machine which can use more than a few arithmetical units. One way to solve this problem is to give up programmability and design a machine dedicated to some limited class of problems. A brief overview presents several successful attempts to develop special-purpose machines: Ising processors, Caltech Hypercube and other QCD machines, molecular dynamics machines. Chapter 4 deals with the GRAPE systems. They are specialized to the time integration of gravitational \(N\)-body systems. The most straightforward way to handle them is to calculate directly the force on each particle as the sum of the forces from the other \(N-1\) particles. GRAPE is designed essentially for this purpose. A GRAPE system consists of two components, the host processor and the GRAPE processor. The host processor sends the positions and mass of the particles to the GRAPE processor, which calculates the forces on the particles. In 1988 Chikada raised the idea of building a special-purpose computing device for astrophysical simulations. Before Chikada's proposal the authors were thinking of developing something along these lines, Chikada's proposal stated that such hardware design would not be too difficult. This work led to a series GRAPE systems. GRAPE-1 was intended a design/development exercise to determine the gravitational acceleration and force, GRAPE-2 was motivated by the necessity of higher accuracy. WINE-1 performed the Ewald method, which is a standard technique to evaluate gravitational or Coulomb interaction in periodic boundary, GRAPE-3/3A are used for simulation of clusters of galaxies, merging of galaxies with and without black holes, evolution of small groups of galaxies, the formation of galaxies, etc. HARP-1 is an extension of the GRAPE architecture to give hardware support for the Hermite integration scheme. The goal of the GRAPE-4 project was to develop a machine with peak-speed 1 Tflops for the simulation of collisional systems with high relative accuracy of the force and offering a reasonable performance for the individual time-step algorithm. Chapter 5 describes the software for GRAPE systems, Chapter 6 considers the possible fields of their application, Chapter 7 tries to predict the future of GRAPE and other special-purpose computers. The book is the description of a successful special-purpose computer design project by its participants concentrating their attention on the practical aspects, it can be recommended both for specialists dealing with the design of similar equipments in different areas and for scientists specializing in this topic of physics. It will be an excellent guide for practitioners how to realize similar projects.
- Do-it-yourself computational astronomy. Hardwares, algorithms, softwares, and sciences
- Direct \(N\)-body simulations
- GRAPE project
- Acceleration of moment method using CUDA
- An improved DNA-sticker addition algorithm and its application to logarithmic arithmetic
- Solvers for \(\mathcal{O}(N)\) electronic structure in the strong scaling limit
- scientific article; zbMATH DE number 194005 (Why is no real title available?)
- A sparse octree gravitational N-body code that runs entirely on the GPU processor
- The Earth Simulator
- The Polder Computing Environment: a system for interactive distributed simulation
- scientific article; zbMATH DE number 833688 (Why is no real title available?)
- Extending DNA-sticker arithmetic to arbitrary size using staples
- Parallelization, Special Hardware and Post-Newtonian Dynamics in Direct N - Body Simulations
- Educational \(N\)-body websites
- Relativistic binaries in globular clusters
- Relativistic binaries in globular clusters
- SUE: A special purpose computer for spin models
- Performance prediction of N-body simulations on a hybrid architecture
- GRAPE project: special-purpose computers for many-body simulations.
- On the use of a heterogeneous MIMD-SIMD platform to simulate the dynamics of globular clusters with a central massive object
- Relativistic binaries in globular clusters
- Fast vortex method calculation using a special-purpose computer
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4382676)
