Dragomir Milojevic
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Dragomir Milojevic has a total of 43 co-authors. Here are the top co-authors by number of shared publications:
- Dwaipayan Biswas
IMEC, Leuven, Belgium
6 shared publications Julien Ryckaert
IMEC, Leuven, Belgium
5 shared publications- James Myers
IMEC, Leuven, Belgium
4 shared publications - Mohamed Naeim
IMEC, Leuven, Belgium
4 shared publications - Sudipta Das
IMEC, Leuven, Belgium
4 shared publications - Bhawana Kumari
Department of Electronics and communication Engineering, Indian Institute of Technology (Indian School of Mines), Dhanbad, India
2 shared publications
Latest Publications
- 9 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Embedded Systems Letters
- Vol. PP (19430663) , 1
Enhancing Multi-Core SoC Performance through AuC Integration for Advanced 3D-IC Systems
This paper explores the application of Array-Under-CMOS (AuC) technology for the cache hierarchy in a commercial 4-core mobile System-on-Chip (SoC). This study evaluates the impact of AuC integration on Power, Performance & Area (PPA) metrics for L1, L2 caches at SoC level. We evaluated memory partitioning not only in comparison to the 2D baseline, but also in relation to the 3D Memory-on-Logic (MoL) integration. Our findings indicate that AuC technology can boost the SoC frequency of up to 71% and achieve 56% higher CPU efficiency compared to traditional 2D implementations. Furthermore, the study highlights the advantages of heterogeneous integration, where SRAM bit-cells are in the N3 FinFET (N3-FF) technology and the periphery in the A10 Nanosheet (A10-NS) technology, resulting in 12% power savings and 18% better CPU efficiency compared to baseline 2D. These results demonstrate the potential of AuC technology to optimize the design of SRAM, contributing to the overall efficiency of computing systems by addressing the memory bottleneck.
- 9 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Embedded Systems Letters
- Vol. PP (19430663) , 1
Enhancing Multi-Core SoC Performance through AuC Integration for Advanced 3D-IC Systems
This paper explores the application of Array-Under-CMOS (AuC) technology for the cache hierarchy in a commercial 4-core mobile System-on-Chip (SoC). This study evaluates the impact of AuC integration on Power, Performance & Area (PPA) metrics for L1, L2 caches at SoC level. We evaluated memory partitioning not only in comparison to the 2D baseline, but also in relation to the 3D Memory-on-Logic (MoL) integration. Our findings indicate that AuC technology can boost the SoC frequency of up to 71% and achieve 56% higher CPU efficiency compared to traditional 2D implementations. Furthermore, the study highlights the advantages of heterogeneous integration, where SRAM bit-cells are in the N3 FinFET (N3-FF) technology and the periphery in the A10 Nanosheet (A10-NS) technology, resulting in 12% power savings and 18% better CPU efficiency compared to baseline 2D. These results demonstrate the potential of AuC technology to optimize the design of SRAM, contributing to the overall efficiency of computing systems by addressing the memory bottleneck.
- 22 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Vol. 15 (4) , 648-658
Iterative Layout-Aware Power, Thermal, and IR-Drop Co-Optimization: Ensuring Convergency in 3D-ICs
Continuous scaling of integrated circuits and the adoption of 3D integration have significantly increased power density, creating critical challenges for thermal and power integrity. Rising power densities can trigger thermal runaway, negatively impacting device reliability and performance. This paper presents an electrothermal coupling framework built upon commercial Electronic Design Automation (EDA) tools, designed for precise iterative analysis of power, thermal and IR-drop characteristics in 2D and 3D configurations of a many-core RISCV SoC. The proposed framework automates iterative Power-Temperature (P-T) simulations to evaluate thermal convergence and potential thermal runaway scenarios in Memory-on-Logic (MoL) and Logic-on-Memory (LoM) stacked configurations. It identifies thermal hotspots, tracks their progression and evaluates various cooling strategies. Initial results indicate that the first P-T iteration provides up to 10% power savings in 3D due to a 11% wirelength reduction compared to 2D. However, by the fifth iteration, MoL power saving reduces to 4%, whereas LoM maintains the 10% saving. LoM exhibits a 6°C lower peak temperature compared to MoL under equivalent cooling conditions. Compared to 2D, the range of power densities (110 - 260 W/cm2) results in temperature variations of −1°C to +3°C in LoM. A 10°C rise in temperature increases IR-drop by 11%; however, physical design-aware adjustments, such as a tighter Power Delivery Network (PDN) pitch, effectively reducing IR-drop by 54%, mitigating thermal impacts.
- 22 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Vol. 15 (4) , 648-658
Iterative Layout-Aware Power, Thermal, and IR-Drop Co-Optimization: Ensuring Convergency in 3D-ICs
Continuous scaling of integrated circuits and the adoption of 3D integration have significantly increased power density, creating critical challenges for thermal and power integrity. Rising power densities can trigger thermal runaway, negatively impacting device reliability and performance. This paper presents an electrothermal coupling framework built upon commercial Electronic Design Automation (EDA) tools, designed for precise iterative analysis of power, thermal and IR-drop characteristics in 2D and 3D configurations of a many-core RISCV SoC. The proposed framework automates iterative Power-Temperature (P-T) simulations to evaluate thermal convergence and potential thermal runaway scenarios in Memory-on-Logic (MoL) and Logic-on-Memory (LoM) stacked configurations. It identifies thermal hotspots, tracks their progression and evaluates various cooling strategies. Initial results indicate that the first P-T iteration provides up to 10% power savings in 3D due to a 11% wirelength reduction compared to 2D. However, by the fifth iteration, MoL power saving reduces to 4%, whereas LoM maintains the 10% saving. LoM exhibits a 6°C lower peak temperature compared to MoL under equivalent cooling conditions. Compared to 2D, the range of power densities (110 - 260 W/cm2) results in temperature variations of −1°C to +3°C in LoM. A 10°C rise in temperature increases IR-drop by 11%; however, physical design-aware adjustments, such as a tighter Power Delivery Network (PDN) pitch, effectively reducing IR-drop by 54%, mitigating thermal impacts.
- Open Access
- 11 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Exploratory Solid-State Computational Devices and Circuits
- Vol. 10, 125-134
System-Technology Co-Optimization for Dense Edge Architectures Using 3-D Integration and Nonvolatile Memory
High-performance edge artificial intelligence (Edge-AI) inference applications aim for high energy efficiency, memory density and small form factor, requiring a design space exploration across the whole stack – workloads, architecture, mapping and co-optimization with emerging technology. In this paper, we present an system-technology co-optimization (STCO) framework that interfaces with workload-driven system scaling challenges and physical design-enabled technology offerings. The framework is built on three engines that provide the physical design characterization, dataflow mapping optimizer, and system efficiency predictor. The framework builds on a systolic array accelerator to provide the design-technology characterization points using advanced imec A10 nanosheet CMOS node along with emerging, high-density voltage-gated spin-orbit-torque (VGSOT) MRAM, combined with memory-on-logic fine-pitch 3D wafer-to-wafer hybrid bonding. We observe that 3D system integration of SRAM-based design leads to 9% power savings with 53% footprint reduction at iso-frequency w.r.t. 2D implementation for the same memory capacity. 3D NVM-VGSOT allows 4× memory capacity increase with 30% footprint reduction at iso-power compared to 2D SRAM 1×. Our exploration with two diverse workloads – image resolution enhancement (FSRCNN) and eye tracking (EDSNet) – shows that more resources allow better workload mapping possibilities which are able to compensate peak system energy efficiency degradation on high memory capacity cases. We show that a 25% peak efficiency reduction on a 32× memory capacity can lead to a 7.4× faster execution with 5.7× higher effective TOPS/W than the 1× memory capacity case on the same technology.
- 11 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Exploratory Solid-State Computational Devices and Circuits
- Vol. 10, 125-134
Open AccessSystem-Technology Co-Optimization for Dense Edge Architectures Using 3-D Integration and Nonvolatile Memory
High-performance edge artificial intelligence (Edge-AI) inference applications aim for high energy efficiency, memory density and small form factor, requiring a design space exploration across the whole stack – workloads, architecture, mapping and co-optimization with emerging technology. In this paper, we present an system-technology co-optimization (STCO) framework that interfaces with workload-driven system scaling challenges and physical design-enabled technology offerings. The framework is built on three engines that provide the physical design characterization, dataflow mapping optimizer, and system efficiency predictor. The framework builds on a systolic array accelerator to provide the design-technology characterization points using advanced imec A10 nanosheet CMOS node along with emerging, high-density voltage-gated spin-orbit-torque (VGSOT) MRAM, combined with memory-on-logic fine-pitch 3D wafer-to-wafer hybrid bonding. We observe that 3D system integration of SRAM-based design leads to 9% power savings with 53% footprint reduction at iso-frequency w.r.t. 2D implementation for the same memory capacity. 3D NVM-VGSOT allows 4× memory capacity increase with 30% footprint reduction at iso-power compared to 2D SRAM 1×. Our exploration with two diverse workloads – image resolution enhancement (FSRCNN) and eye tracking (EDSNet) – shows that more resources allow better workload mapping possibilities which are able to compensate peak system energy efficiency degradation on high memory capacity cases. We show that a 25% peak efficiency reduction on a 32× memory capacity can lead to a 7.4× faster execution with 5.7× higher effective TOPS/W than the 1× memory capacity case on the same technology.
- 16 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 33 (2) , 346-357
Bandwidth-Latency-Thermal Co-Optimization of Interconnect-Dominated Many-Core 3D-IC
The ongoing integration of advanced functionalities in contemporary system-on-chips (SoCs) poses significant challenges related to memory bandwidth, capacity, and thermal stability. These challenges are further amplified with the advancement of artificial intelligence (AI), necessitating enhanced memory and interconnect bandwidth and latency. This article presents a comprehensive study encompassing architectural modifications of an interconnect-dominated many-core SoC targeting the significant increase of intermediate, on-chip cache memory bandwidth and access latency tuning. The proposed SoC has been implemented in 3-D using A10 nanosheet technology and early thermal analysis has been performed. Our workload simulations reveal, respectively, up to 12- and 2.5-fold acceleration in the 64-core and 16-core versions of the SoC. Such speed-up comes at 40% increase in die-area and a 60% rise in power dissipation when implemented in 2-D. In contrast, the 3-D counterpart not only minimizes the footprint but also yields 20% power savings, attributable to a 40% reduction in wirelength. The article further highlights the importance of pipeline restructuring to leverage the potential of 3-D technology for achieving lower latency and more efficient memory access. Finally, we discuss the thermal implications of various 3-D partitioning schemes in High Performance Computing (HPC) and mobile applications. Our analysis reveals that, unlike high-power density HPC cases, 3-D mobile case increases$T_{\max }$ only by$2~^{\circ } $ C–$3~^{\circ } $ C compared to 2-D, while the HPC scenario analysis requires multiconstrained efficient partitioning for 3-D implementations.- 16 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 33 (2) , 346-357
Bandwidth-Latency-Thermal Co-Optimization of Interconnect-Dominated Many-Core 3D-IC
The ongoing integration of advanced functionalities in contemporary system-on-chips (SoCs) poses significant challenges related to memory bandwidth, capacity, and thermal stability. These challenges are further amplified with the advancement of artificial intelligence (AI), necessitating enhanced memory and interconnect bandwidth and latency. This article presents a comprehensive study encompassing architectural modifications of an interconnect-dominated many-core SoC targeting the significant increase of intermediate, on-chip cache memory bandwidth and access latency tuning. The proposed SoC has been implemented in 3-D using A10 nanosheet technology and early thermal analysis has been performed. Our workload simulations reveal, respectively, up to 12- and 2.5-fold acceleration in the 64-core and 16-core versions of the SoC. Such speed-up comes at 40% increase in die-area and a 60% rise in power dissipation when implemented in 2-D. In contrast, the 3-D counterpart not only minimizes the footprint but also yields 20% power savings, attributable to a 40% reduction in wirelength. The article further highlights the importance of pipeline restructuring to leverage the potential of 3-D technology for achieving lower latency and more efficient memory access. Finally, we discuss the thermal implications of various 3-D partitioning schemes in High Performance Computing (HPC) and mobile applications. Our analysis reveals that, unlike high-power density HPC cases, 3-D mobile case increases$T_{\max }$ only by$2~^{\circ } $ C–$3~^{\circ } $ C compared to 2-D, while the HPC scenario analysis requires multiconstrained efficient partitioning for 3-D implementations.- Open Access
- 26 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Exploratory Solid-State Computational Devices and Circuits
- Vol. 10, 67-74
Toward Fine-Grained Partitioning of Low-Level SRAM Caches for Emerging 3D-IC Designs
Scaling on-chip memory capacity is one of the primary approaches to mitigate memory wall bottlenecks. Various 2.5/3D integration schemes, leveraging novel partitioning is being actively explored to improve system performance. However, fine-grained functional partitioning of memory macros is not widely reported. As SRAM scaling stagnates with emerging CMOS logic roadmap, we propose a partitioning of low-level (faster access) caches in 3D using an Array Under CMOS (AuC) technology paradigm. Our study focuses on partitioning and optimization of SRAM bit-cells and peripheral circuits, enabling heterogeneous integration, achieving up to 12% higher operating frequency with 50% leakage power reduction in the memory macros. Applied on a 64-bit mobile SoC CPU core, we achieve upto 60% higher energy efficiency compared to 2D baseline and 14% increase in operating frequency compared to standard memory-on-logic 3D partitioning scheme.
- 26 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Exploratory Solid-State Computational Devices and Circuits
- Vol. 10, 67-74
Open AccessToward Fine-Grained Partitioning of Low-Level SRAM Caches for Emerging 3D-IC Designs
Scaling on-chip memory capacity is one of the primary approaches to mitigate memory wall bottlenecks. Various 2.5/3D integration schemes, leveraging novel partitioning is being actively explored to improve system performance. However, fine-grained functional partitioning of memory macros is not widely reported. As SRAM scaling stagnates with emerging CMOS logic roadmap, we propose a partitioning of low-level (faster access) caches in 3D using an Array Under CMOS (AuC) technology paradigm. Our study focuses on partitioning and optimization of SRAM bit-cells and peripheral circuits, enabling heterogeneous integration, achieving up to 12% higher operating frequency with 50% leakage power reduction in the memory macros. Applied on a 64-bit mobile SoC CPU core, we achieve upto 60% higher energy efficiency compared to 2D baseline and 14% increase in operating frequency compared to standard memory-on-logic 3D partitioning scheme.
- 25 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 32 (11) , 2144-2148
Multidie 3-D Stacking of Memory Dominated Neuromorphic Architectures
Event-driven neuromorphic processors for artificial intelligence (AI) inference on edge/IoT devices require largeon-chip memory capacity, for efficient execution of spiking neural networks (NNs). In this work, we evaluate 3-D stacking benefits on SENECA, a digital neuromorphic accelerator core, sweeping itson-chip memory capacity from 2 up to 32 Mb in both legacy planar and advanced nanosheet CMOS logic nodes. In a planar CMOS node (GF-22 nm), two-die memory-on-logic (MoL) partitioning enables $8\times $ moreon-chip memory, and it boosts operating frequency by 7% with 26% less power than the 2-D. Moving to an advanced nanosheet technology (imec A10), multidie (up to 7 dies) MoL stacking enables a performance increase of up to 29% and power savings up to 31%. Furthermore, a core folding (CF) partitioning in A10 shows up to 16% performance improvement with 12% total power savings with respect to the 2-D implementation on the same technology. We also demonstrate no thermal overhead for multidie stacking at advanced nodes for designs exhibiting low power density. These physical design explorations lay the foundation for system technology co-optimization studies for edge devices.
- 25 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 32 (11) , 2144-2148
Multidie 3-D Stacking of Memory Dominated Neuromorphic Architectures
Event-driven neuromorphic processors for artificial intelligence (AI) inference on edge/IoT devices require largeon-chip memory capacity, for efficient execution of spiking neural networks (NNs). In this work, we evaluate 3-D stacking benefits on SENECA, a digital neuromorphic accelerator core, sweeping itson-chip memory capacity from 2 up to 32 Mb in both legacy planar and advanced nanosheet CMOS logic nodes. In a planar CMOS node (GF-22 nm), two-die memory-on-logic (MoL) partitioning enables $8\times $ moreon-chip memory, and it boosts operating frequency by 7% with 26% less power than the 2-D. Moving to an advanced nanosheet technology (imec A10), multidie (up to 7 dies) MoL stacking enables a performance increase of up to 29% and power savings up to 31%. Furthermore, a core folding (CF) partitioning in A10 shows up to 16% performance improvement with 12% total power savings with respect to the 2-D implementation on the same technology. We also demonstrate no thermal overhead for multidie stacking at advanced nodes for designs exhibiting low power density. These physical design explorations lay the foundation for system technology co-optimization studies for edge devices.
- Open Access
- 23 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-6
GNN-assisted Back-side Clock Routing Methodology for Advance Technologies
The back-side metal layers exhibit lower parasitics compared to the front-side layers in advanced technologies, making them suitable for clock-net distribution. In this study, we explore the advantages of using back-side metal layers for clock routing, which is shared with a power delivery network. Our Graph Neural Network (GNN) based framework, effectively distributes the clock-tree between the front and back sides. We address the back-side clock nets' creation by incorporating back-side buffers. Our results demonstrate better clock and full-chip metrics represented by an increase of up to 13% in the effective frequency with equivalent power consumption, using 3 nm technology.
- 23 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 1-6
Open AccessGNN-assisted Back-side Clock Routing Methodology for Advance Technologies
The back-side metal layers exhibit lower parasitics compared to the front-side layers in advanced technologies, making them suitable for clock-net distribution. In this study, we explore the advantages of using back-side metal layers for clock routing, which is shared with a power delivery network. Our Graph Neural Network (GNN) based framework, effectively distributes the clock-tree between the front and back sides. We address the back-side clock nets' creation by incorporating back-side buffers. Our results demonstrate better clock and full-chip metrics represented by an increase of up to 13% in the effective frequency with equivalent power consumption, using 3 nm technology.
- 1 September 2009
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Automated Pathfinding tool chain for 3D-stacked integrated circuits: Practical case study
New technologies for manufacturing 3D Stacked ICs offer numerous opportunities for the design of complex and efficient embedded systems. But these technologies also introduce many design options at system/chip design level, hard to grasp during the complete design cycle. Because of the sequential nature of current design practices, designers are often forced to introduce design margins to meet required specifications, resulting in sub-optimal designs. In this paper we introduce new design methodology and practical tool chain, called PathFinding Flow, that can help designers to easily trade-off between different system level design choices, physical design and/or technology options and understand their impact on typical design parameters such as cost, performance and power. Proposed methodology and the tool chain will be demonstrated on a practical case study, involving fairly complex Multi-Processor System-on-Chip using Network-on-Chip for communication medium. With this example we will show how High-Level Synthesis can be used to quickly move from high-level to RTL models, necessary for accurate physical prototyping for both computation and communication. We will also show how the possibility of design iteration, through the mechanism of feedback based on physical information from physical prototyping, can improve design performance. Finally, we will show how we can move in no time from traditional 2D to 3D design and how we can measure benefits of such design choice.
- 1 September 2009
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Automated Pathfinding tool chain for 3D-stacked integrated circuits: Practical case study
New technologies for manufacturing 3D Stacked ICs offer numerous opportunities for the design of complex and efficient embedded systems. But these technologies also introduce many design options at system/chip design level, hard to grasp during the complete design cycle. Because of the sequential nature of current design practices, designers are often forced to introduce design margins to meet required specifications, resulting in sub-optimal designs. In this paper we introduce new design methodology and practical tool chain, called PathFinding Flow, that can help designers to easily trade-off between different system level design choices, physical design and/or technology options and understand their impact on typical design parameters such as cost, performance and power. Proposed methodology and the tool chain will be demonstrated on a practical case study, involving fairly complex Multi-Processor System-on-Chip using Network-on-Chip for communication medium. With this example we will show how High-Level Synthesis can be used to quickly move from high-level to RTL models, necessary for accurate physical prototyping for both computation and communication. We will also show how the possibility of design iteration, through the mechanism of feedback based on physical information from physical prototyping, can improve design performance. Finally, we will show how we can move in no time from traditional 2D to 3D design and how we can measure benefits of such design choice.