Approximated Canonical Signed Digit for Error Resilient Intelligent Computation

Lowering the energy consumption in applications operating on large datasets is one of the main challenges in modern computing. In this context, it is especially important to lower the energy required to transfer data from/to the memory. Usually, this is obtained by applying smart encoding techniques to the data. In this work, we show how to reduce the switching activity in buses and floating-point units by an approximated canonical signed-digit encoder. The precision of the encoding is programmable and can be chosen depending on the application’s required accuracy.

[1]  Alberto Nannarelli.  Fused Multiply-Add for Variable Precision Floating-Point , 2019, 2019 32nd IEEE International System-on-Chip Conference (SOCC).

[2]  R. Hashemian.  A new method for conversion of a 2's complement to canonic signed digit number system and its representation , 1996, Conference Record of The Thirtieth Asilomar Conference on Signals, Systems and Computers.

[3]  Mark Horowitz,et al.  1.1 Computing's energy problem (and what we can do about it) , 2014, 2014 IEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC).

[4]  Sandeep Saini.  Low Power Interconnect Design , 2015 .

[5]  Milos D. Ercegovac,et al.  Digital Arithmetic , 2003, Wiley Encyclopedia of Computer Science and Engineering.

[6]  Song Han,et al.  Learning both Weights and Connections for Efficient Neural Network , 2015, NIPS.

[7]  Massoud Pedram,et al.  Deploying Customized Data Representation and Approximate Computing in Machine Learning Applications , 2018, ISLPED.

[8]  Majid Ahmadi,et al.  Application of neural networks with CSD coefficients for human face recognition , 2013, 2013 IEEE International Symposium on Circuits and Systems (ISCAS2013).

[10]  Qian Wang,et al.  Energy efficient parallel neuromorphic architectures with approximate arithmetic on FPGA , 2017, Neurocomputing.

[11]  Alberto Nannarelli.  Tunable Floating-Point for Energy Efficient Accelerators , 2018, 2018 IEEE 25th Symposium on Computer Arithmetic (ARITH).

[12]  Alberto Nannarelli.  Tunable Floating-Point Adder , 2019, IEEE Transactions on Computers.