New modalities, new architectures are available to build computers and intelligent systems. Now that the physical limits of the traditional way of increasing the power of the single processor has been touched on, Europe, the United States and Japan are competing to define the standards of the new architectures. Europe has decided to launch a program to coordinate research projects and Italy, with its National Institute for Nuclear Physics, has a role of leadership.
Europe has decided to coordinate efforts to win the techno-industrial battle against the United States and Japan on the architectures of consumer electronics of the future. Mobile phones, avionics, robots and environmental intelligence providing services to people as well as numerical super-calculus applications for medicine, physics and engineering require new logical architectures able to overcome the limits imposed on today's configurations.
Along this frontier, a very strong global competition is raging, which sees the Old Continent potentially gaining an advantage over the American and Asian rivals thanks to its traditional competencies on “embedded” electronics (mobile phones, aeronautics, automobiles…), in fact, intelligent systems.
In this endeavour of scientific and technological endeavour, the European Commission decided to fund the SHAPES international project, coordinated by Italy, in order to aggregate select research institutes and European industrial partners around the competencies of the INFN and of its industrial spin-offs.
This decision was also taken thanks to the urging on the part of research in this sector from the National Institute for Nuclear Physics (INFN) and its industrial spin-offs. In fact, the INFN has realized – with the series of APE super-calculators – the only European example of a platform for calculus built by a research institution, while ATMEL Roma built on this experience to develop the DIOPSIS, an advanced MPSoc (Multi processor System on Chip). With these activities, the Roman researchers have also become leaders in tracing methodologies and guidelines in the development of future computing architectures foreseen in the 7° Framework Program.
Within the SHAPES project, the roman researchers have also been charged with the task of organizing the first international Workshop on parallel numerical architectures for embedded and scientific applications: CASTNESS (Computer Architectures and Software Tools for Numerical Scalable Embedded System) to which 90 academic and industrial researchers active in this field participated.
IN DEPTH: TWO SCENARIOS
“In these last years two distinct scenarios have come to life in computing – explains Pier Stanislao Paolucci, part-time researcher at INFN and CTO at ATMEL Roma, coordinator of the SHAPES European project – on one hand there exists a European specialisation in so-called embedded systems, meaning computers inserted in avionics, in cell phones, in automobiles and in other machines dedicated to interacting with man and the environment. Such systems must produce a small quantity of heat, use little energy, be transportable. In this field, Europe is the leader on the international market”.
On the other hand, there is computing for super-calculus, which increases its power every year by finding a way of inserting ever more transistors inside a single processor. In this field, Americans are leaders. “But the criterion that has led development in this sector has now touched its physical limits – explains Piero Vicini, from the INFN in Rome and responsible for the APE project –. In the last twenty years, technology has allowed the number of transistors available to build a processor to double every eighteen months. Therefore, traditional processors have gradually become faster and larger. However, today it is no longer possible to increase the power of the single processor since the electrical signals, because of the intrinsic physical limitations, are not able to reach the entire processor within a single cycle of the clock.
Furthermore, if speed increases greatly, the energy requirements grow too much and the system heats up excessively, thus rendering it difficult, if not impossible, to assemble”.
“A possible solution to the problem does exist” –adds Paolucci – “in not further increasing the complexity and the speed of the single processor, but rather in using a <> instead, that in years to come will become smaller and smaller thanks to technological advances and in integrating many of these Tiles within a single chip, thus building a system for calculus . The INFN is one of the best centres for centres in Europe for the design of parallel calculus systems, which have been used for scientific applications until now, but now have become essential to maintain the European leadership in electronics for embedded systems”.
“With APE, the INFN has created the first bridge between Europe and America's vocations – adds Vicini – And now that the EU has decidedly chosen to go the way of new architectures, Italy finds itself in an advanced position, and is able to play an important role in this sector of scientific and technological innovation”.
INFN - Communication Office
P.zza dei Caprettari 70
Tel: +39 06 6868162
Fax: +39 06 68307944