Schuyler Eldridge
Research Interests
Schuyler Eldridge has a total of 35 co-authors. Here are the top co-authors by number of shared publications:
Ajay Joshi
2 shared publicationsAlper Buyuktosunoglu
IBM T. J. Watson Research Center
2 shared publications- Mircea R. Stan
Department of Electrical and Computer Engineering, University of Virginia, Charlottesville, VA 22904, USA
2 shared publications Pradip Bose
IBM T. J. Watson Research Center
2 shared publicationsSwagath Venkataramani
IBM T. J. Watson Research Center
2 shared publications- Ajay Joshi
Boston University
1 shared publication
Latest Publications
- Open Access
- 18 December 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Computer Architecture Letters
- Vol. 17 (1) , 92-95
Nile: A Programmable Monitoring Coprocessor
Researchers widely employ hardware performance counters (HPCs) as well as debugging and profiling tools in processors for monitoring different events such as cache hits, cache misses, and branch prediction statistics during the execution of programs. The collected information can be used for power, performance, and thermal management of the system as well as detecting anomalies or malicious behavior in the software. However, monitoring new or complex events using HPCs and existing tools is a challenging task because HPCs only provide a fixed pool of raw events to monitor. To address this challenge, we propose the implementation of a programmable hardware monitor in a complete system framework including the hardware monitor architecture and its interface with an in-order single-issue RISC-V processor as well as an operating system. As a proof of concept, we demonstrate how to programmatically implement a shadow stack using our hardware monitor and how the programmed shadow stack detects stack buffer overflow attacks. Our hardware monitor design incurs a 26% power overhead and a 15% area overhead over an unmodified RISC-V processor. Our programmed shadow stack has less than 3% performance overhead in the worst case.
- 18 December 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Computer Architecture Letters
- Vol. 17 (1) , 92-95
Open AccessNile: A Programmable Monitoring Coprocessor
Researchers widely employ hardware performance counters (HPCs) as well as debugging and profiling tools in processors for monitoring different events such as cache hits, cache misses, and branch prediction statistics during the execution of programs. The collected information can be used for power, performance, and thermal management of the system as well as detecting anomalies or malicious behavior in the software. However, monitoring new or complex events using HPCs and existing tools is a challenging task because HPCs only provide a fixed pool of raw events to monitor. To address this challenge, we propose the implementation of a programmable hardware monitor in a complete system framework including the hardware monitor architecture and its interface with an in-order single-issue RISC-V processor as well as an operating system. As a proof of concept, we demonstrate how to programmatically implement a shadow stack using our hardware monitor and how the programmed shadow stack detects stack buffer overflow attacks. Our hardware monitor design incurs a 26% power overhead and a 15% area overhead over an unmodified RISC-V processor. Our programmed shadow stack has less than 3% performance overhead in the worst case.
- 1 November 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 601-604
Very Low Voltage (VLV) Design
This paper is a tutorial-style introduction to a special session on: Effective Voltage Scaling in the Late CMOS Era. It covers the fundamental challenges and associated solution strategies in pursuing very low voltage (VLV) designs. We discuss the performance and system reliability constraints that are key impediments to VLV. The associated trade-offs across power, performance and reliability are helpful in inferring the optimal operational voltage-frequency point. This work was performed under the auspices of an ongoing DARPA program (named PERFECT) that is focused on maximizing system-level energy efficiency.
- 1 November 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 601-604
Very Low Voltage (VLV) Design
This paper is a tutorial-style introduction to a special session on: Effective Voltage Scaling in the Late CMOS Era. It covers the fundamental challenges and associated solution strategies in pursuing very low voltage (VLV) designs. We discuss the performance and system reliability constraints that are key impediments to VLV. The associated trade-offs across power, performance and reliability are helpful in inferring the optimal operational voltage-frequency point. This work was performed under the auspices of an ongoing DARPA program (named PERFECT) that is focused on maximizing system-level energy efficiency.
- 1 October 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 99-112
Towards General-Purpose Neural Network Computing
Machine learning is becoming pervasive, decades of research in neural network computation is now being leveraged to learn patterns in data and perform computations that are difficult to express using standard programming approaches. Recent work has demonstrated that custom hardware accelerators for neural network processing can outperform software implementations in both performance and power consumption. However, there is neither an agreed-upon interface to neural network accelerators nor a consensus on neural network hardware implementations. We present a generic set of software/hardware extensions, X-FILES, that allow for the general-purpose integration of feedforward and feedback neural network computation in applications. The interface is independent of the network type, configuration, and implementation. Using these proposed extensions, we demonstrate and evaluate an example dynamically allocated, multi-context neural network accelerator architecture, DANA. We show that the combination of X-FILES and our hardware prototype, DANA, enables generic support and increased throughput for neural-network-based computation in multi-threaded scenarios.
- 1 October 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 99-112
Towards General-Purpose Neural Network Computing
Machine learning is becoming pervasive, decades of research in neural network computation is now being leveraged to learn patterns in data and perform computations that are difficult to express using standard programming approaches. Recent work has demonstrated that custom hardware accelerators for neural network processing can outperform software implementations in both performance and power consumption. However, there is neither an agreed-upon interface to neural network accelerators nor a consensus on neural network hardware implementations. We present a generic set of software/hardware extensions, X-FILES, that allow for the general-purpose integration of feedforward and feedback neural network computation in applications. The interface is independent of the network type, configuration, and implementation. Using these proposed extensions, we demonstrate and evaluate an example dynamically allocated, multi-context neural network accelerator architecture, DANA. We show that the combination of X-FILES and our hardware prototype, DANA, enables generic support and increased throughput for neural-network-based computation in multi-threaded scenarios.
- Open Access
- 20 May 2014
- Published by Association for Computing Machinery (ACM)
- p. 169-174
Neural network-based accelerators for transcendental function approximation
- 20 May 2014
- Published by Association for Computing Machinery (ACM)
- p. 169-174
Open AccessNeural network-based accelerators for transcendental function approximation