TY - CONF T1 - MASR: A Modular Accelerator for Sparse RNNs T2 - International Conference on Parallel Architectures and Compilation Techniques Y1 - 2019 A1 - Udit Gupta A1 - Brandon Reagen A1 - Lillian Pentecost A1 - Marco Donato A1 - Tambe, Thierry A1 - Rush, Alexander A1 - Gu Wei A1 - David Brooks KW - accelerators KW - compression KW - hardware software co-design KW - hardware support for machine learning KW - neural network accelerators AB - Recurrent neural networks (RNNs) are becoming the de facto solution for speech recognition. RNNs exploit long-term temporal relationships in data by applying repeated, learned transformations. Unlike fully-connected (FC) layers with single vector matrix operations, RNN layers consist of hundreds of such operations chained over time. This poses challenges unique to RNNs that are not found in convolutional neural networks (CNNs) or FC models, namely large dynamic activation. In this paper we present MASR, a principled and modular architecture that accelerates bidirectional RNNs for on-chip ASR. MASR is designed to exploit sparsity in both dynamic activations and static weights. The architecture is enhanced by a series of dynamic activation optimizations that enable compact storage, ensure no energy is wasted computing null operations, and maintain high MAC utilization for highly parallel accelerator designs. In comparison to current state-of-the-art sparse neural network accelerators (e.g., EIE), MASR provides 2x area 3x energy, and 1.6x performance benefits. The modular nature of MASR enables designs that efficiently scale from resource-constrained low-power IoT applications to large-scale, highly parallel datacenter deployments. JF - International Conference on Parallel Architectures and Compilation Techniques UR - https://doi.org/10.48550/arXiv.1908.08976 ER -