Contact Information
Parneu
PARNEU project targets are parallel hardware
implementation, high level software development environment and
parallel application development. Hardware research is consentrated in
modular and scalable hardware topologies and run-time reconfigurable
systems. The main emphasis in software design is in high level system
development environments and parallel compiler for both hardware
configuration and executable code.
PARNEU project started from soft computing and
digital signal processing algorithm analyses. The hardware design was
based on experiences on TUTNC. Currently the research work is focused
on software development.
PARNEU system is implemented on modular printed
circuit boards, but the experiences and results will be also used in
system-on-a-chip implementations.
PARNEU applications are low bit-rate video encoding,
motion estimation and various soft computing algorithms, like SOM, MLP
and SDM.
The project is supported by the Academy of Finland,
TEKES and several companies.
Parneu Design
PARNEU system is designed to speed up computationally
demanding signal processing applications including several different
soft computing algorithms like Multilayer Perceptron, Self-Organized
Map and Radial-Basis function networks. In addition to the raw
performance, the system requirements for PARNEU were expandability,
flexibility and reliability in various industrial environments and
embedded systems. Expandability is achieved by placing functional units
into cards, which can be connected together and increased in number
without limits. Flexibility and reliability is reached with general
purpose DSPs, which are basic computing units.
External communication is done via PCI bus to PC or workstation, but
the configurable host interface allows the system to be connected to
other buses, too. In addition, each DSP has a direct high-speed
connection for external devices like sensors.
In general, PARNEU is a general-purpose co-processor system which
can be hosted by a PC or workstation. The host initializes program
execution, monitors operation and collects the results while functional
units inside the co-processor system perform the parallel computation.
Extranet login
