Em Photonics, Inc.

Fernando E Ortiz

James P Durbano

Dennis W Prather

John R Humphrey

Eric Kelmelis

Peterson F Curt

An accelerator for the speckle atmospheric compensation algorithm may enable real-time speckle processing of video feeds that may enable the speckle algorithm to be applied in numerous real-time applications. The accelerator may be implemented in various forms, including hardware, software, and/or machine-readable media.

Method and system for enabling real-time speckle processing using hardware platforms

A hardware-based acceleration platform for computational electromagnetic algorithms, specifically the finite-difference time-domain ('FDTD') method, comprises reformulating the FDTD algorithm in order to make it more hardware friendly. This reformulation makes use of split fields at every node in the mesh, and combines total- and scattered-field formulations into a single, hybrid formulation. By precomputing coefficients for the nodes in the mesh, it is possible for a single set of equations to support plane waves, point sources, PML ABCs, and PEC walls. In the method sources are determined by means of a lookup table, rather than through direct hardware computations. All of these modifications enable the hardware designer to much more easily develop an FDTD accelerator.

Reformulation of the finite-difference time-domain algorithm for hardware-based accelerators

Disclosed herein is an organization of cache memory for hardware acceleration of the FDTD method. The organization of cache memory for hardware acceleration of the FDTD method provides a substantial speedup to the finite-difference time-domain (FDTD) algorithm when implemented in a piece of digital hardware. The organization of cache memory for hardware acceleration of the FDTD method utilizes a very high bandwidth dual-port on-chip memory in a particular way. By creating many small banks of internal memory and arranging them carefully, all data dependencies can be statically wired. This allows for a many-fold speedup over SRAM-based solutions and removes the burden of data dependence calculation that streaming SDRAM-based solutions must perform.

Growing community of inventors

Newark, DE, United States of America

Fernando E Ortiz