University of Pavia, Italy Faculty of Engineering Microcomputer lab - PowerPoint PPT Presentation

University of Pavia, Italy Faculty of Engineering Microcomputer lab Application specific processors for Application specific processors for intensive computing applications intensive computing applications Gianni Danese – Full Professor Francesco Leporati – Associate Professor Mauro Giachero – Ph.D. Student Nelson Nazzicari – Ph.D. Student

Overview of our activities  FPGA for high performance computing  Montecarlo simulation  Fingerprint matching  Real time video compression of Formula 1 sensor data  Motor control for wind tunnel equipment  Intelligent cameras

FPGA for high performance computing (1/2)  Dipole cubic lattice model – Energy calculation  Montecarlo simulations need a lot of computation

FPGA for high performance computing (1/2)  a lot of repeated, regular calculation high-performance VLIW style architecture 

FPGA for high performance computing (1/2)  Our current architecture features a speedup of 2 compared to a 3GHz Pentium 4 processor  The 1S40 FPGA can host 2 accelerator, bringing the speedup factor to 4  Using the more recent Stratix II or Stratix III the achievable speedup can be even greater  This way is possible to build up an accelerator cluster whose cost/performance ratio is much smaller than that obtainable with a cluster of PCs

FPGA for high performance computing (2/2)  Fingerprint matching need a lot of computation  one new image vs. a big archive

FPGA for high performance computing (2/2)  (band limited) phase only correlation algorithm correlation phase-only correlation

A little spin-off: high performance DFT on FPGA  commercial solutions are not satisfying (too slow, big or inaccurate)  butterfly algorithm (created for parallel implementations)  different attempts (fixed vs floating point, data size, etc.)  better than ALTERA FFT Core and other commercial products (using our application- specific metric)

Video Compression on F1 vehicle  thermic camera to control tyre temperature  normal b/w camera to control tyre deformation  low band high noise channel between car and boxes

Video Compression on F1 vehicle  need for real time on board video compression to decrease the amount of data to be transmitted  some on board data elaboration

Video Compression on F1 vehicle Image and video acquisition/compression Temperature evolution of four tyres platform (with extra debug/develop as seen by thermic cameras features)

Motor control for wind tunnel equipment  actuator control for 1:2 model placement  position and speed sensors  treadmill rolling control  closed loop control to maintain belt alignment

VISoc: Intelligent vision system  intelligent camera with embedded neural network  object recognition  shape count (cars, persons...)

Partnerships  Magneti Marelli  Ferrari  Neuricam  Assomac  ST Microelectronics  National Institute for the Physics of Matter  National Institute of Nuclear Physics  Polytechnic of Milan