SBIR Phase II: Use of Transputers for Wavelet Calculations

9321351德鲁基尔拟议研究的目标是为科学和应用工程应用开发具有极高计算能力的低成本台式和便携式系统。在第一阶段，我们开发了一个基于Transputer的硬件平台，加上一个非常强大的分析工具--小波变换。我们证明，通过适当的编码，小波算法可以实现几乎完全的并行化，即计算速度实际上与所使用的Transputer的数量成正比。第一阶段的结果表明，我们开发的硬件/软件系统在速度和内存上都可以扩展两个数量级。第二阶段的工作将包括二维小波变换的软件开发及其在下一代HTRAM Transputer上的实施。2D算法将针对高效率(25：1或更高)实时图像压缩以及许多其他应用进行优化。这些课程包括复杂机械结构中破坏性波传播的分析，非线性动力系统的特性，以及随机网和渗流簇的特性。这些分析在湍流扩散、等离子体物理中的输运性质、流体力学和磁场产生方面有重要的应用。***

9321351 Drukier The goal of the proposed research is to develop low cost desktop and portable systems with very high computational capabilities for both scientific and applied engineering applications. In Phase I we developed a transputer-based hardware platform coupled with a very powerful analytical tool, the wavelet transform. We demonstrated that with appropriate coding the wavelet algorithm allows almost perfect parallelization, i.e. the computational speed is practically proportional to the number of transputers used. The Phase I results demonstrate that the hardware/software system we developed is scalable over two orders of magnitude in speed and memory. The Phase II effort will consist of software development for two-dimensional wavelet transforms and its implementation on next-generation HTRAM transputers. The 2D algorithms will be optimized for high efficiency (25:1 or better) real time image compression as well as for a number of other applications. These include analysis of destructive wave propagation in complex mechanical structures, properties of nonlinear dynamical systems, and the properties of stochastic webs and percolation clusters. These analyses have important applications in turbulent diffusion, transport properties in plasma physics, hydrodynamics, and magnetic field generation. ***