Pimball

Resch, Salonik; Khatamifard, S. Karen; Chowdhury, Zamshed I.; Zhao, Zhengyang; Wang, Jianping; Sapatnekar, Sachin S.; Karpuzcu, Ulya R.

doi:10.1145/3357250

Cited by 27 publications

(6 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Resch et al [74] present PIMBALL, an adaptive BNN accelerator rooted in Processing In Memory (PIM) technology. PIMBALL uses computational RAM (CRAM) like spintronic PIM [75] but adds a transistor for efficient BNN operations.…”

Section: ) Binary Neural Network (Bnn)mentioning

confidence: 99%

“…This approach aimed to maximize data locality and minimize data movement. On the other hand, Resch et al [74] employed a data duplication technique in their PIMBALL design to enable efficient computation of BNN.…”

Section: A Design Guidelines 1) Memory Mappingmentioning

confidence: 99%

“…Proper data organization techniques, such as data compression, data encoding, and data grouping, should be employed to reduce memory footprints, improve data access efficiency, and exploit the parallelism offered by STT-MRAM devices. For instance, Kim et al [81] and Resch et al [74] proposed a memory array design that incorporates the Multiply-Accumulate (MAC) operation. They achieved this by partitioning the memory array into multiple sub-computing arrays, which are utilized for various operations, such as adding partial products and accumulating results.…”

Section: ) Data Organizationmentioning

confidence: 99%

“…Pipelining techniques can also be employed to overlap computation and memory access stages, improving overall throughput. Resch et al [74] implemented pipelining in their CNN computations by utilizing a 9-stage pipeline. They employed multiple memory arrays to perform computations at each layer simultaneously for different input images.…”

Section: ) Parallelism and Pipeliningmentioning

confidence: 99%

See 3 more Smart Citations

Domain-Specific STT-MRAM-Based In-Memory Computing: A Survey

Yusuf,

Adegbija,

Gajaria

2024

IEEE Access

View full text Add to dashboard Cite

In recent years, the rapid growth of big data and the increasing demand for high-performance computing have fueled the development of novel computing architectures. Among these, in-memory computing architectures that leverage the high-density and low-latency nature of modern memory technologies have emerged as promising solutions for domain-specific computing applications. STT-MRAM (Spin Transfer Torque Magnetic Random Access Memory) is one such in-memory computing technology that holds great potential for this field due to its non-volatility, high endurance, and low power consumption. In this survey paper, we aim to provide a comprehensive overview of the state-of-the-art in STT-MRAM-based domainspecific in-memory computing (DS-IMC) architectures. We examine the challenges, opportunities, and trade-offs associated with these architectures from the perspective of various application domains, like machine learning, image and signal processing, and data encryption. We explore different experimental research tools used in studying these architectures, guidelines for efficiently designing them, and gaps in the state-of-the-art that necessitate future research and development.INDEX TERMS domain-specific architectures, in-memory computing, spin-transfer torque RAM.

show abstract

Section: ) Binary Neural Network (Bnn)mentioning

confidence: 99%

Section: A Design Guidelines 1) Memory Mappingmentioning

confidence: 99%

Section: ) Data Organizationmentioning

confidence: 99%

Section: ) Parallelism and Pipeliningmentioning

confidence: 99%

See 2 more Smart Citations

Domain-Specific STT-MRAM-Based In-Memory Computing: A Survey

Yusuf,

Adegbija,

Gajaria

2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Neuro-inspired signal processing presently relates to the most intensively studied topics of technical sciences in a wide range of completely different realizations. It can be implemented from an electronic perspective, using, for example, field-programmable gate array (FPGA) circuits [1][2][3][4] or employing the recent achievements of spintronics [5][6][7], phase change materials [8,9] or other components.…”

Section: Introductionmentioning

confidence: 99%

Neuro-Inspired Signal Processing in Ferromagnetic Nanofibers

et al. 2021

View full text Add to dashboard Cite

Computers nowadays have different components for data storage and data processing, making data transfer between these units a bottleneck for computing speed. Therefore, so-called cognitive (or neuromorphic) computing approaches try combining both these tasks, as is done in the human brain, to make computing faster and less energy-consuming. One possible method to prepare new hardware solutions for neuromorphic computing is given by nanofiber networks as they can be prepared by diverse methods, from lithography to electrospinning. Here, we show results of micromagnetic simulations of three coupled semicircle fibers in which domain walls are excited by rotating magnetic fields (inputs), leading to different output signals that can be used for stochastic data processing, mimicking biological synaptic activity and thus being suitable as artificial synapses in artificial neural networks.

show abstract