Research Areas

HPLP members involved: Vaibhav Verma and Nick P. Moon

AI_RISC is a hardware/software codesign solution for future edge devices which incorporates AI functionality into the CPU by adding the AI hardware units to the processor core, new AI extensions to the ISA and provide software support to make it easy for programmers to target this scalable processor. AI_RISC presents an end-to-end optimized system-level solution for enabling the next generation of AI edge devices. 

Collaborators:

VELVET is a very-low voltage system-on-chip consisting of an open-source RISC-V processor and a deep neural network accelerator designed with cross-layer reliability techniques to prevent low-voltage processor failures. This heterogeneous system employs circuit, system and application level resiliency techniques to enable reliable low-voltage operation.

HPLP members involved: Robert E. West III and Junhan Han

The thermal performance of Processing in Memory(PiM) 3D IC's  are of critical importance. An emphasis is given on developing a robust thermal modelling network that investigates into novel cooling methods, such as dynamic flow rate microchannels based on the location of hotspots. 3D IC thermal management have been addressed by dynamic thermal management method which controls the direction and magnitude of microfluidic flows in the thermal chamber by multiplexing inlets and outlets.

Collaborators:

The run time required for various emerging technologies can be accurately emulated by employing a flexible FPGA spatial fabric. FPGA’s are also introduced to create prototypes for near and in-memory processing.

Collaborators:  Alif Ahmed, Ersin Cukurtas, Joshua Fixelle, Yimin Gao, Xinfei Guo, Preslav Ivanov, Samira Khan, Sergiu Mosanu, Mohammad Nazmus Sakib,  Kevin Skadron, II Tommy Tracy, Vaibhav Verma, Mircea Stan.

Our goal is to increase chips’ endurance. Old/used electronics can regain their functionality with either passive or active recovery. Developing an efficient method on recovery is crucial for practicality. Chips need to do exercise.

Collaborators:

HPLP members involved: Elisa Pantoja

Unwanted access to private intellectual property and circuit designs during manufacturing and distribution has led to a serious need to establish protective measures against malicious attacks. We make use of electronic printing, obfuscation techniques, and design for security to build defense features that protect hardware against malicious modification, reverse engineering, and vulnerability exploitation.

HPLP members involved: Patricia Gonzalez- Guerrero, Rahul Sreekumar

The idea of an end to end asynchronous system is of great interest for ultra low power applications where there is a clear trade-off between power and accuracy. Stochastic computing when implemented asynchronously proves to be an innovative solution towards lesser hardware and silicon footprint.

HPLP members involved: Mohammad Nazmus Sakib

The compact size, high stability and ultra-low depinning current density make skyrmion a potential candidate for future memory technology. Skyrmion can be used to design conservative logic system, stochastic computing. The goal of this project is to design energy efficient memory and logic circuits and systems to replace CMOS based technologies by exploiting the non-volatility of highly stable magnetic skyrmion.