Search Results (5)

In this paper, a mathematical model for obtaining energy consumption of IMC architectures is constructed. This model provides energy estimation based on the distribution of a specific dataset. In addition, the estimation reduces the required simulation time to create an energy consumption model of SRAM-based IMC architectures. To validate our model with realistic data, the energy consumption of IMC is compared by using NeuroSim V3.0 for the CIFAR-10 and MNIST-like datasets. Furthermore, an application is created with our model to select highest performing quantization mapping based upon the parameters of energy consumption and accuracy. Full article

Hybrid magnetic tunnel junction/complementary metal oxide semiconductor (MTJ/CMOS) circuits based on in-memory-computation (IMC) architecture is considered as the next-generation candidate for the digital integrated circuits. However, the energy consumption during the MTJ write process is a matter of concern in these hybrid circuits. In this regard, we have developed a novel write circuit for the contemporary three-terminal perpendicular-MTJs that works on the voltage-gated spin orbit torque (VG+SOT) switching mechanism to store the information in hybrid circuits for IMC architecture. Investigation of the novel write circuit reveals a remarkable reduction in the total energy consumption (and energy delay product) of 92.59% (95.81) and 92.28% (42.03%) than the conventional spin transfer torque (STT) and spin-Hall effect assisted STT (SHE+STT) write circuits, respectively. Further, we have developed all the hybrid logic gates followed by nonvolatile full adders (NV-FAs) using VG+SOT, STT, and SHE+STT MTJs. Simulation results show that with the VG+SOT NOR-OR, NAND-AND, XNOR-XOR, and NV-FA circuits, the reduction in the total power dissipation is 5.35% (4.27%), 5.62% (3.2%), 3.51% (2.02%), and 4.46% (2.93%) compared to STT (SHE+STT) MTJs respectively. Full article

Artificial intelligence (AI) has revolutionized present-day life through automation and independent decision-making capabilities. For AI hardware implementations, the 6T-SRAM cell is a suitable candidate due to its performance edge over its counterparts. However, modern AI hardware such as neural networks (NNs) access off-chip data quite often, degrading the overall system performance. Compute-in-memory (CIM) reduces off-chip data access transactions. One CIM approach is based on the mixed-signal domain, but it suffers from limited bit precision and signal margin issues. An alternate emerging approach uses the all-digital signal domain that provides better signal margins and bit precision; however, it will be at the expense of hardware overhead. We have analyzed digital signal domain CIM silicon-verified 6T-SRAM CIM solutions, after classifying them as SRAM-based accelerators, i.e., near-memory computing (NMC), and custom SRAM-based CIM, i.e., in-memory-computing (IMC). We have focused on multiply and accumulate (MAC) as the most frequent operation in convolution neural networks (CNNs) and compared state-of-the-art implementations. Neural networks with low weight precision, i.e., <12b, show lower accuracy but higher power efficiency. An input precision of 8b achieves implementation requirements. The maximum performance reported is 7.49 TOPS at 330 MHz, while custom SRAM-based performance has shown a maximum of 5.6 GOPS at 100 MHz. The second part of this article analyzes the FinFET 6T-SRAM as one of the critical components in determining overall performance of an AI computing system. We have investigated the FinFET 6T-SRAM cell performance and limitations as dictated by the FinFET technology-specific parameters, such as sizing, threshold voltage (V_th), supply voltage (V_DD), and process and environmental variations. The HD FinFET 6T-SRAM cell shows 32% lower read access time and 1.09 times better leakage power as compared with the HC cell configuration. The minimum achievable supply voltage is 600 mV without utilization of any read- or write-assist scheme for all cell configurations, while temperature variations show noise margin deviation of up to 22% of the nominal values. Full article

State-of-the-art IoT technologies request novel design solutions in edge computing, resulting in even more portable and energy-efficient hardware for in-the-field processing tasks. Vision sensors, processors, and hardware accelerators are among the most demanding IoT applications. Resistance switching (RS) two-terminal devices are suitable for resistive RAMs (RRAM), a promising technology to realize storage class memories. Furthermore, due to their memristive nature, RRAMs are appropriate candidates for in-memory computing architectures. Recently, we demonstrated a CMOS compatible silicon nitride (SiN_x) MIS RS device with memristive properties. In this paper, a report on a new photodiode-based vision sensor architecture with in-memory computing capability, relying on memristive device, is disclosed. In this context, the resistance switching dynamics of our memristive device were measured and a data-fitted behavioral model was extracted. SPICE simulations were made highlighting the in-memory computing capabilities of the proposed photodiode-one memristor pixel vision sensor. Finally, an integration and manufacturing perspective was discussed. Full article

(1) Background: DNA sequence alignment process is an essential step in genome analysis. BWA-MEM has been a prevalent single-node tool in genome alignment because of its high speed and accuracy. The exponentially generated genome data requiring a multi-node solution to handle large volumes of data currently remains a challenge. Spark is a ubiquitous big data platform that has been exploited to assist genome alignment in handling this challenge. Nonetheless, existing works that utilize Spark to optimize BWA-MEM suffer from higher overhead. (2) Methods: In this paper, we presented PipeMEM, a framework to accelerate BWA-MEM with lower overhead with the help of the pipe operation in Spark. We additionally proposed to use a pipeline structure and in-memory-computation to accelerate PipeMEM. (3) Results: Our experiments showed that, on paired-end alignment tasks, our framework had low overhead. In a multi-node environment, our framework, on average, was 2.27× faster compared with BWASpark (an alignment tool in Genome Analysis Toolkit (GATK)), and 2.33× faster compared with SparkBWA. (4) Conclusions: PipeMEM could accelerate BWA-MEM in the Spark environment with high performance and low overhead. Full article