INFN APE Parallel Computers (since 1984...)

The APE Group of INFN designed several generations of Massive Parallel Computers, based on Custom VLIW Processors and Custom Programming Languages, dedicated to the simulation of Physical Systems.

The APE Project has been created and initially led by two great masters: Nicola Cabibbo and Giorgio Parisi.

Since its foundation (1983), Pier Stanislao PAOLUCCI participated to the research and development activities of the APE Massive Parallel Design group of INFN.

The APE project created four generations of computers: APE (1984-1988), APE100 (1989-1994), APEmille (1995-2000) and APENext (2001-2006). 

Here below, a photo of the custom VLIW processor, named MAD (Multiply and Accumulate Device), of the APE 100 rack, and of the APE100 Board (1994).

The know-how of APE has been essential for the development of the mAgic VLIW Floating-point processor used in the DIOPSIS 740 and DIOPSIS 940 MPSoCs, and for the launch of the two European projects SHAPES and EURETILE, which has been coordinated by Pier Stanislao Paolucci.

Among the applications of APE Parallel computers: Lattice Quantum Chromo Dynamics, Climatology and Meteorology (Cubed Sphere), Lattice Boltzmann Equations, Synthetic Aperture Radar  (SAR) Processing, Seismic Migration for Oil Search.

Specific Custom VLIW Processors have been designed for the APE machines. Paolucci partecipated to the definition of the processors, to the design of the compilation chain and to the development of efficient application kernels.

Pier Stanislao Paolucci, together with Simone Cabasino, and Gianmarco Todesco invented the TAO Evolving Programming language, and designed and developed its compiler.

Together with Gianmarco Todesco and Simone Cabasino, Paolucci invented also the concept of Evolving Grammars and Dynamic Parser (Zz) and used Zz to develop the TAO Compiler.

Paolucci invented the "Cubed Sphere" algorithm, for an efficient implementation of Climatological and Meteorological simulations on 3D toroidal meshes of processors. Together with Corrado Ronchi and Giuseppe Jacono, a first complete implementation of the Cubed Sphere program has been ported on the APE100 machines.

Together with Nicola Cabibbo, Paolucci invented an efficient SIMD algorithm for Matrix Transposition and FFT on 3D toroidal mesh machines.

Paolucci designed also an efficient algorithm for simualtion of fully connected Neural Networks and Gravitational Systems on APE machines.