Engineering topics

Asifuzzaman, Kazi

Publications and source records attributed to Asifuzzaman, Kazi.

3 records

DOE OSTI2024

Neuro-Spark: A Submicrosecond Spiking Neural Networks Architecture for In-Sensor Filtering

Neuro-Spark, which is a new neuromorphic architecture with a field-programmable gate array (FPGA) implementation for ultrafast spiking neural network (SNN) inference at the edge, facilitates smart-pixel in-sensor filtering for high-energy physics experiments at the Large Hadron Collider (LHC). Utilizing the evolutionary optimization for neuromorphic systems (EONS) training method, we generate compact SNN models with 91% signal efficiency, akin to convolutional neural networks but with half the parameters. However, deploying near the detector poses a challenge because the SNN must handle a sustained input data rate exceeding 1013 GB/s. To overcome this, we propose a novel hardware architecture that uses high-level synthesis to construct a tuned architecture for the EONS-trained SNN. In addition to the analysis and validation with an AMD Xilinx Artix-A7 FPGA, our solution consumes only ç24% of FPGA LUT and flipflops. We also introduce an innovative quantization method that reduces FPGA resource utilization by ç15% without compromising accuracy. Our FPGA implementation achieves computing latency of ç10 ns for smart-pixel application inference on an edge FPGA.

Miniskar, Narasinga Rao↗

DOE OSTI2024

CMOS-Based Single-Cycle in-Memory XOR/XNOR

Big data applications are on the rise, and so is the number of data centers. The ever-increasing massive data pool needs to be periodically backed up in a secure environment. Moreover, a massive amount of securely backed-up data is required for training binary convolutional neural networks for image classification. XOR and XNOR operations are essential for large-scale data copy verification, encryption, and classification algorithms. The disproportionate speed of existing compute and memory units makes the von Neumann architecture inefficient to perform these Boolean operations. Compute-in-memory (CiM) has proved to be an optimum approach for such bulk computations. The existing CiM-based XOR/XNOR techniques either require multiple cycles for computing or add to the complexity of the fabrication process. Here, we propose a CMOS-based hardware topology for single-cycle in-memory XOR/XNOR operations. Our design provides at least 2× improvement in the latency compared with other existing CMOS-compatible solutions. We verify the proposed system through circuit/system-level simulations and evaluate its robustness using a 5000-point Monte Carlo variation analysis. This all-CMOS design paves the way for practical implementation of CiM XOR/XNOR at scaled technology nodes.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

A review of crosstalk polymorphic circuits and their scalability

Using a control variable, the functionality of Polymorphic circuits can be modified, making them adaptable and useful for reconfiguring circuit behavior — all the way from gate level to system level. State-of-the art polymorphic circuits are based on custom non-linear circuit design or emerging devices such as ambipolar FET, configurable magnetic devices etc. While some of these approaches are inefficient in performance, others involve exotic devices. The Crosstalk computing based polymorphic circuits offer a fresh perspective. In Crosstalk, the interconnect interference between nanoscale metal lines is intentionally engineered to exhibit the programmable Boolean logic behavior. This approach relies on the coupling between metal lines and not on the transistors for computing, resulting in better scalability, security by obscurity, and fault tolerance by reconfiguration. Our novel approach is backed by the mathematical formulation that conveys the rationale to generalize and achieve a wide variety of polymorphic circuits. Our experiments, including design, simulation, and Power Performance Area (PPA) characterization results indicate that crosstalk circuits provide significant improvement in transistor count (about 3x), switching energy (2x), and speed (1.5x) for polymorphic logic circuits. In the best-case scenario, the transistor count reduction is 5x. This paper presents Crosstalk computing’s fundamentals, polymorphism and the scalability aspects to compete/co-exist with CMOS for digital logic implementations below 10 nm. Our scalability study uses Open Source 7 nm PDK, considers all process variation aspects and accommodates worst-case scenarios. The study results for various benchmark circuits show that the Crosstalk technology is a viable alternative to CMOS for digital logic implementations below 10 nm, having 48% density, 57% power, and 10% performance gains over equivalent CMOS counterparts. Finally, we compare Crosstalk Polymorphic Circuit design technique with similar approaches described in related works and discuss its features and constraints.

97 MATHEMATICS AND COMPUTING↗