Functional Memory Custom Computing Machine (FMCCM)
Description The FMCCM implements and combines ideas from Field Programmable Gate Array (FPGA) Custom Computing Machines (FCCMs), reduced instruction set computing (RISC) and spreadsheet (tabular) programming. The FMCCM approach differs from others in that it follows the conventional processor/memory architecture except there is no computational circuitry inside the processor where it is normally implemented. Instead of expressions taking several steps to compute in a conventional microprocessor, they are computed in one step in FPGA chips connected in parallel with the memory. Functional memory is analogous to a spreadsheet where most locations simply store data while others contain formulas and display expression results when read. Furthermore, functional memory is reprogrammable and can be loaded with different formulas for different user programs. Therefore, functional memory can best enhance performance in time critical software containing large expressions.
Applications The technique is general purpose and can be applied to all programs (e.g. image processing, image enhancing and editing, digital video processing, pattern recognition in character and image identification systems as well as computer vision). Main Advantages Large expressions, for example, convolution functions for image enhancement, will be performed much faster because of the parallel processing in functional memory. The FMCCM will be fully compatible with existing Peripheral Component Interconnect (PCI) bus-based image processing systems as well as other non-image processing applications.
Type of Offer:
« More Computer Science Patents