{"id":{"repo_id":"nus","oai_identifier":"oai:scholarbank.nus.edu.sg:10635/207958"},"canonical_url":"https://search.dev.ndltd.org/etd/nus/oai:scholarbank.nus.edu.sg:10635/207958","repository":{"repo_id":"nus","name":"National University of Singapore","base_url":"https://scholarbank.nus.edu.sg/oai/request"},"display":{"title":"NON-VOLATILE IN-MEMORY COMPUTING WITH SKYRMIONS AND PHASE CHANGE MEMORIES","abstract":"The non-volatile in-memory compute engine (NVIMCE), which saves on the latency and energy associated with data movement between memory and processing elements in the conventional von Neumann architectures, is a crucial design technique for enabling ultra-low power intelligent edge devices. Due to power/area versus throughput trade-offs (such as to increase the complexity of periphery circuitry to enable higher parallelism or result precision), the memory subsystem in NVIMCE needs to be carefully designed to avoid incurring large power and area overheads. Furthermore, the emerging non-volatile memory technologies that are suitable for application in NVIMCE suffer from issues such as read/write errors, low endurance, and cycle-to-cycle variations. Hence, this dissertation addresses key design issues using a device-circuit-architecture co-design framework and demonstrates promising design strategies for NVIMCEs using skyrmions and phase-change memories. The proposed design strategies achieved ultra-low power consumption and low latencies needed for future edge devices in Internet of Smart Things.","abstract_html":"The non-volatile in-memory compute engine (NVIMCE), which saves on the latency and energy associated with data movement between memory and processing elements in the conventional von Neumann architectures, is a crucial design technique for enabling ultra-low power intelligent edge devices. Due to power/area versus throughput trade-offs (such as to increase the complexity of periphery circuitry to enable higher parallelism or result precision), the memory subsystem in NVIMCE needs to be carefully designed to avoid incurring large power and area overheads. Furthermore, the emerging non-volatile memory technologies that are suitable for application in NVIMCE suffer from issues such as read/write errors, low endurance, and cycle-to-cycle variations. Hence, this dissertation addresses key design issues using a device-circuit-architecture co-design framework and demonstrates promising design strategies for NVIMCEs using skyrmions and phase-change memories. The proposed design strategies achieved ultra-low power consumption and low latencies needed for future edge devices in Internet of Smart Things.","abstract_has_math":false,"creators":["MIRIYALA VENKATA PAVAN KUMAR"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2020,"date_issued":"2020-08-21","date_published":"2020-08-21","updated_at":"2026-07-24T03:31:00Z","subjects":["Edge AI, Spintronics, Phase-change memories, Skyrmions, Binary neural networks, Neuromorphic computing"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":null,"outbound_label":null,"outbound_source":null},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["MIRIYALA VENKATA PAVAN KUMAR"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.issued","label":"Date","values":["2020-08-21"]},{"key":"dc:relation.isreferencedby","label":"Dc Relation Isreferencedby","values":["https://scholarbank.nus.edu.sg/handle/10635/207958"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Edge AI, Spintronics, Phase-change memories, Skyrmions, Binary neural networks, Neuromorphic computing"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://scholarbank.nus.edu.sg/bitstreams/d34249d9-d8d9-4f3c-b0e2-b308b7a267b3/download"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["The non-volatile in-memory compute engine (NVIMCE), which saves on the latency and energy associated with data movement between memory and processing elements in the conventional von Neumann architectures, is a crucial design technique for enabling ultra-low power intelligent edge devices. Due to power/area versus throughput trade-offs (such as to increase the complexity of periphery circuitry to enable higher parallelism or result precision), the memory subsystem in NVIMCE needs to be carefully designed to avoid incurring large power and area overheads. Furthermore, the emerging non-volatile memory technologies that are suitable for application in NVIMCE suffer from issues such as read/write errors, low endurance, and cycle-to-cycle variations. Hence, this dissertation addresses key design issues using a device-circuit-architecture co-design framework and demonstrates promising design strategies for NVIMCEs using skyrmions and phase-change memories. The proposed design strategies achieved ultra-low power consumption and low latencies needed for future edge devices in Internet of Smart Things."]},{"key":"dc:format.checksum.md5","label":"Dc Format Checksum Md5","values":["db0b2bd2341d1d46244e57f6f210c60b","4cfcfaf0e72fcf33c371229d8ee04de5"]},{"key":"dc:title","label":"Title","values":["NON-VOLATILE IN-MEMORY COMPUTING WITH SKYRMIONS AND PHASE CHANGE MEMORIES"]}]}],"canonical_facts":{"dc:creator":["MIRIYALA VENKATA PAVAN KUMAR"],"dc:date.issued":["2020-08-21"],"dc:description.abstract":["The non-volatile in-memory compute engine (NVIMCE), which saves on the latency and energy associated with data movement between memory and processing elements in the conventional von Neumann architectures, is a crucial design technique for enabling ultra-low power intelligent edge devices. Due to power/area versus throughput trade-offs (such as to increase the complexity of periphery circuitry to enable higher parallelism or result precision), the memory subsystem in NVIMCE needs to be carefully designed to avoid incurring large power and area overheads. Furthermore, the emerging non-volatile memory technologies that are suitable for application in NVIMCE suffer from issues such as read/write errors, low endurance, and cycle-to-cycle variations. Hence, this dissertation addresses key design issues using a device-circuit-architecture co-design framework and demonstrates promising design strategies for NVIMCEs using skyrmions and phase-change memories. The proposed design strategies achieved ultra-low power consumption and low latencies needed for future edge devices in Internet of Smart Things."],"dc:format.checksum.md5":["db0b2bd2341d1d46244e57f6f210c60b","4cfcfaf0e72fcf33c371229d8ee04de5"],"dc:identifier.uri":["https://scholarbank.nus.edu.sg/bitstreams/d34249d9-d8d9-4f3c-b0e2-b308b7a267b3/download"],"dc:relation.isreferencedby":["https://scholarbank.nus.edu.sg/handle/10635/207958"],"dc:subject":["Edge AI, Spintronics, Phase-change memories, Skyrmions, Binary neural networks, Neuromorphic computing"],"dc:title":["NON-VOLATILE IN-MEMORY COMPUTING WITH SKYRMIONS AND PHASE CHANGE MEMORIES"],"dc:type":["Thesis"]},"updated_at":"2026-07-24T03:31:00Z"}