We describe APEmille, the latest generation of the APE parallel processors. This machine, an evolution of the APE100 concept, is very efficient for LGT simulations as well as for a broader class of applications requiring massive floating point computations. Several new features characterise this evolution. In particular local addressing capabilities are added to all computing nodes. APEmille also exhibits a higher degree of integration with a network of workstations acting as a global host system. An APEmille system in the Teraflops range will be completed in three-four years. The architecture proposed in this paper is being currently simulated and evaluated. ((orig.)).
Bartoloni, A;  Battista, C;  Bellacci, M;  Cabasino, S;  Cabibbo, N;  Cosimi, M;  Dagostino, U;  Marzano, F;  Panizzi, E;  Paolucci, P S;  Rapuano, F;  Rinaldi, W;  Sarno, R;  Todesco, G M;  Torelli, M;  Vicini, P;  Laico, F;  Tripiccione, R 
- Universita di Roma `La Sapienza` (Italy). Dipartimento di Fisica and INFN
- INFN Sezione di Pisa and Dipartimento di Fisica, Universita di Pisa, Via Livornese, 56010 Pisa (Italy)