7 papers
Rethinking Neural Nonlinearity as Gating
Muhammad Sabih, Frank Hannig, Jürgen Teich
Activation functions are considered an essential primitive for neural nonlinearity, i.e., they enable neural networks to serve as universal approximators. In this paper, we show th…
Co-Design of CNN Accelerators for TinyML using Approximate Matrix Decomposition
José Juan Hernández Morales, Georgios Mentzos, Frank Hannig +4
The paradigm shift towards local and on-device inference under stringent resource constraints is represented by the tiny machine learning (TinyML) domain. The primary goal of TinyM…
Symbolic Polyhedral-Based Energy Analysis for Nested Loop Programs
Avinash Mahesh Nirmala, Dominik Walter, Frank Hannig +1
This work presents a symbolic approach for estimating the energy consumption for nested loop programs when mapped and scheduled on parallel processor array accelerator architecture…
Loop Control Management in Tightly Coupled Processor Arrays (TCPAs)
Dominik Walter, Frank Hannig, Jürgen Teich
Multidimensional loop kernels often suffer from control overhead that can dominate execution time on parallel loop accelerators. Tightly Coupled Processor Arrays (TCPAs) offload lo…
Hardware/Software Co-Design of RISC-V Extensions for Accelerating Sparse DNNs on FPGAs
Muhammad Sabih, Abrarul Karim, Jakob Wittmann +2
The customizability of RISC-V makes it an attractive choice for accelerating deep neural networks (DNNs). It can be achieved through instruction set extensions and corresponding cu…
Evaluation of CGRA Toolchains
Dominik Walter, Marita Halm, Daniel Seidel +4
Increasing demands for computing power also propel the need for energy-efficient SoC accelerator architectures. One class for such accelerators are so-called processor arrays, whic…