Mega Parallel Processor (MPP)

MPP is a hardware and software parallel processor engine specialized for solving problems involving intensive manipulation of multiple data steams in the heterogeneous hierarchical computational environments.

Main target areas of application for MPP are the following:

video compression, encoding, decoding, image processing, computer graphics
video and audio streaming over inter and intra nets
computer vision, image recognition, artificial intelligence
medicine, security, professional sports.

Architecture of MPP is essentially heterogeneous, distributed, hierarchical and hardware platform and OS independent. As a result it is highly customizable and scalable. Concrete hardware realizations of MPP can be built with standard or special custom hardware components.

Computationally MPP is especially effective when application algorithms are modified for mega parallel implementation, and code presenting application problem is directly integrated into the software core of mega parallel machine.

Custom implementations of MPP were used to solve a number of scientific and industrial problems. It includes 2D and 3D modeling of fields, kinetics, dynamics and thermodynamics in complex heterogeneous systems, image and video processing, computer vision, artificial intelligence, etc.

Currently one most effective realization of MPP is build as an expandable grid of standard x86 multiple core processor units physically placed on the single or/and multiple processor motherboards, connected via gigabit Ethernet lines. Depending on the complexity of applied problem and intensity of data steams exchanged between individual mega parallel segments, MPP may include from 1 to 64 and more processor cores.