Reconfigurable Architecture and Dataflow for Memory Traffic Minimization of CNNs Computation

Computation of convolutional neural network (CNN) requires a significant amount of memory access, which leads to lots of energy consumption. As the increase of neural network scale, this phenomenon is further obvious, the energy consumption of memory access and data migration between on-chip buffer and off-chip DRAM is even much more than the computation energy on processing element array (PE array). In order to reduce the energy consumption of memory access, a better dataflow to maximize data reuse and minimize data migration between on-chip buffer and external DRAM is important. Especially, the dimension of input feature map (ifmap) and filter weight are much different for each layer of the neural network. Hardware resources may not be effectively utilized if the array architecture and dataflow cannot be reconfigured layer by layer according to their ifmap dimension and filter dimension, and result in a large quantity of data migration on certain layers. However, a thorough exploration of all possible configurations is time consuming and meaningless. In this paper, we propose a quick and efficient methodology to adapt the configuration of PE array architecture, buffer assignment, dataflow and reuse methodology layer by layer with the given CNN architecture and hardware resource. In addition, we make an exploration on the different combinations of configuration issues to investigate their effectiveness and can be used as a guide to speed up the thorough exploration process.

Component reuse methodology for multi-clock Data-Flow parallel embedded Systems

International audience The growing complexity of new chips and the time-to-market constraints require fundamental changes in the way systems are designed. Systems on Chip (SoC) based on reused components have become an absolute necessity to embedded systems companies that want to remain competitive. However, the design of a SoC is extremely complex because it encompasses a range of difficult problems in hardware and software design. This paper focuses on the design of parallel and multi-frequency applications using flexible components. Flexible parallel components are assembled using a scheduling method which combines the synchronous data-flow principle of balance equations and the polyhedral scheduling technique. Our approach allows a flexible component to be modelled and a full system to be assembled and synthesized with automatically generated wrappers. The work presented here is an extension of previous work. We illustrate our method on a simplified WCDMA system. We discuss the relationship of this approach with multi-clock architecture, latency-insensitive design, multidimensional data-flow systems and stream programming