Reconfigurable Stream-based Tensor Unit with Variable-Precision Posit Arithmetic
Nuno Neves, Pedro Tomas and Nuno Roma
Instituto Superior Tecnico Universidade de Lisboa, Portugal
The increased adoption of DNN applications drove the emergence of dedicated tensor computing units to accelerate multi-dimensional matrix multiplication operations. Although they deploy highly efficient computing architectures, they often lack support for more general-purpose application domains. Such a limitation occurs both due to their consolidated computation scheme (restricted to matrix multiplication) and due to their frequent adoption of low-precision/custom floating-point formats (unsuited for general application domains). In contrast, this paper proposes a new Reconfigurable Tensor Unit (RTU) which deploys an array of variable-precision Vector Multiply Accumulate (VMA) units. Furthermore, each VMA unit leverages the new Posit floating-point format and supports the full range of standardized posit precisions in a single SIMD unit, with variable vector-element width. Moreover, the proposed RTU explores the Posit format features for fused operations, together with spatial and time-multiplexing reconfiguration mechanisms to fuse and combine multiple VMAs to map high-level and complex operations. The RTU is also supported by an automatic data streaming infrastructure and a pipelined data movement scheme, allowing it to accelerate the computation of most data-parallel patterns commonly present in vectorizable applications. The proposed RTU showed to outperform state-of-the-art tensor and SIMD units, present in off-the-shelf platforms, in turn resulting in significant energy-efficiency improvements.