Microelectronics System-LEvel Design
μSLED
The Microelectronics System-Level Design (μSLED) research group develops methodologies, tools, and architectures to design and optimize complex digital systems across abstraction levels, from system modeling and high-level synthesis (HLS) down to RTL design and implementation. Our goal is to improve productivity, performance, energy efficiency, and predictability in modern computing platforms, especially those combining heterogeneous compute, memory, and communication resources.
μSLED’s research spans system-level design, hardware/software co-design, and design space exploration (DSE), enabling rapid evaluation of architectural trade-offs early in the development cycle. We target platforms ranging from embedded systems to FPGA- and ASIC-oriented accelerators, with a strong emphasis on practical design flows and reusable components.
A key focus area is the design of domain-specific architectures for data-intensive workloads, including systolic arrays and other structured accelerator fabrics. In this context, we also investigate next-generation machine learning architectures, including modular and composable solutions, analyzing their implications in terms of scalability, reuse, and adaptability across application domains. We study how algorithmic structure, data movement, and architectural constraints interact, and we develop techniques to map applications efficiently onto regular and semi-regular architectures (e.g., tiled islands interconnected via NoCs). At the implementation level, we work on RTL microarchitectures and optimization strategies that complement higher-level synthesis and modeling.
Another core direction is Embedded AI, where we investigate architectures and tool flows for deploying machine learning inference on resource-constrained platforms. This includes co-optimizing model characteristics (e.g. quantization and compute patterns), memory hierarchy, and accelerator design, as well as the joint definition and optimization of advanced neural architectures, to meet tight constraints on power, latency, and cost while maintaining robust accuracy.
μSLED also addresses the scalability bottlenecks of large digital designs, including the modeling and optimization of communication and memory subsystems (e.g., dataflow designs, NoC-based architectures, buffering strategies), and the integration of architectural and physical constraints into system-level methodologies. We combine analytical modeling, simulation, and data-driven and AI-driven approaches to enable efficient exploration of complex design spaces, including the use of machine learning techniques to guide DSE and design automation.
The group collaborates with academic and industrial partners in EDA, embedded computing, intelligent systems, and ASIC design, contributing both foundational research and practical tool development. μSLED is active in national and European projects and is committed to technology transfer and impact in real-world applications.
μSLED’s research spans system-level design, hardware/software co-design, and design space exploration (DSE), enabling rapid evaluation of architectural trade-offs early in the development cycle. We target platforms ranging from embedded systems to FPGA- and ASIC-oriented accelerators, with a strong emphasis on practical design flows and reusable components.
A key focus area is the design of domain-specific architectures for data-intensive workloads, including systolic arrays and other structured accelerator fabrics. In this context, we also investigate next-generation machine learning architectures, including modular and composable solutions, analyzing their implications in terms of scalability, reuse, and adaptability across application domains. We study how algorithmic structure, data movement, and architectural constraints interact, and we develop techniques to map applications efficiently onto regular and semi-regular architectures (e.g., tiled islands interconnected via NoCs). At the implementation level, we work on RTL microarchitectures and optimization strategies that complement higher-level synthesis and modeling.
Another core direction is Embedded AI, where we investigate architectures and tool flows for deploying machine learning inference on resource-constrained platforms. This includes co-optimizing model characteristics (e.g. quantization and compute patterns), memory hierarchy, and accelerator design, as well as the joint definition and optimization of advanced neural architectures, to meet tight constraints on power, latency, and cost while maintaining robust accuracy.
μSLED also addresses the scalability bottlenecks of large digital designs, including the modeling and optimization of communication and memory subsystems (e.g., dataflow designs, NoC-based architectures, buffering strategies), and the integration of architectural and physical constraints into system-level methodologies. We combine analytical modeling, simulation, and data-driven and AI-driven approaches to enable efficient exploration of complex design spaces, including the use of machine learning techniques to guide DSE and design automation.
The group collaborates with academic and industrial partners in EDA, embedded computing, intelligent systems, and ASIC design, contributing both foundational research and practical tool development. μSLED is active in national and European projects and is committed to technology transfer and impact in real-world applications.
-
Scientific coordinators
-
Research team
Research area
Research topics
- HLS and RTL design methodologies
- Hardware/software co-design and design space exploration
- System-level design
- FPGA- and ASIC-oriented design
- Architectures and accelerators for Embedded AI and advanced neural architectures
Skills
ERC sectors
SDG
Keywords
Pubblicazioni da archivio istituzionale
Publications from the institutional repository
-
Manca, Edward; Urbinati, Luca; Casu, Mario R. (2025)
Accelerating Mixed-Precision QNN Inference on RISC-V MCUs With the STAR-MAC Unit. In: IEEE ACCESS, vol. 13, pp. 208533-208548. ISSN 2169-3536 -
Bosio, Roberto; Brignone, Giovanni; Urso, Teodoro; Lazarescu, Mihai T.; Lavagno, ... (2025)
Low-Power Subgraph Isomorphism at the Edge Using FPGAs. In: IEEE ACCESS, vol. 13, pp. 67127-67135. ISSN 2169-3536 -
Bosio, Roberto; Minnella, Filippo; Urso, Teodoro; Casu, Mario R.; Lavagno, Luciano; ... (2025)
NN2FPGA: Optimizing CNN Inference on FPGAs With Binary Integer Programming. In: IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, vol. 44, pp. 1807-1818. ISSN 0278-0070 -
Brignone, Giovanni; Bosio, Roberto; Ottati, Fabrizio; Sansoe', Claudio; Lavagno, Luciano (2025)
SILVIA: Automated Superword-Level Parallelism Exploitation via HLS-Specific LLVM Passes for Compute-Intensive FPGA Accelerators. In: ACM TRANSACTIONS ON RECONFIGURABLE TECHNOLOGY AND SYSTEMS, vol. 18. ISSN 1936-7406 -
Manca, Edward; Urbinati, Luca; Casu, Mario R. (2024)
Accelerating Quantized DNN Layers on RISC-V with a STAR MAC Unit. In: 54th Annual Meeting of the Italian Electronics Society, Noto (SR), Italia, September 6-8, 2023, pp. 43-53. ISSN 1876-1100. ISBN: 978-3-031-48710-1 -
Bosio, Roberto; Brignone, Giovanni; Minnella, Filippo; Jamal, MUHAMMAD USMAN; Lavagno, ... (2024)
LESS: Low-Power Energy-Efficient Subgraph Isomorphism on FPGA. In: 2024 Design, Automation & Test in Europe Conference & Exhibition (DATE), Valencia (Spain), 25-27 March 2024, pp. 1-2. ISBN: 978-3-9819263-8-5 -
Lagostina, Lorenzo; Minnella, Filippo; Cortadella, Jordi; Casu, Mario R.; Lazarescu, ... (2024)
Mix & Latch: Comparison With State-of-the-Art Retiming On a RISC-V Benchmark. In: IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, vol. 43, pp. 2229-2233. ISSN 0278-0070 -
Manca, Edward; Urbinati, Luca; Casu, Mario R. (2024)
STAR: Sum-Together/Apart Reconfigurable Multipliers for Precision-Scalable ML Workloads. In: Design, Automation & Test in Europe Conference & Exhibition (DATE), Valencia (Spain), 25-27 March 2024, pp. 1-6. ISBN: 978-3-9819263-8-5 -
Minnella, Filippo; Cortadella, Jordi; Casu, Mario R.; Lazarescu, Mihai T.; Lavagno, ... (2023)
Mix & Latch: An Optimization Flow for High-Performance Designs with Single-Clock Mixed-Polarity Latches and Flip-Flops. In: IEEE ACCESS, vol. 11, pp. 1-1. ISSN 2169-3536 -
Shan, Junnan; Lazarescu, MIHAI TEODOR; Cortadella, Jordi; Lavagno, Luciano; Casu, MARIO ... (2022)
Fast Energy-Optimal Multi-Kernel DNN-like Application Allocation on Multi-FPGA Platforms. In: IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, vol. 41, pp. 1186-1190. ISSN 0278-0070 -
Shah, NASIR ALI; Lavagno, Luciano; Lazarescu, Mihai T.; Quasso, Roberto; Scarpina, ... (2022)
FPGA Acceleration of 3GPP Channel Model Emulator for 5G New Radio. In: IEEE ACCESS, vol. 10, pp. 119386-119401. ISSN 2169-3536 -
Shan, Junnan; Lazarescu, Mihai T.; Cortadella, Jordi; Lavagno, Luciano; Casu, Mario R. (2021)
CNN-on-AWS: Efficient Allocation of Multi-Kernel Applications on Multi-FPGA Platforms. In: IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, vol. 40, pp. 301-314. ISSN 0278-0070 -
BIN TARIQ, Osama; Shan, Junnan; Floros, Georgios; Sotiriou, Christos P.; Casu, Mario R.; ... (2021)
High-Level Annotation of Routing Congestion for Xilinx Vivado HLS Designs. In: IEEE ACCESS, vol. 9, pp. 54286-54297. ISSN 2169-3536 -
Bin Tariq, O.; Lazarescu, M. T.; Lavagno, L. (2021)
Neural Networks for Indoor Person Tracking With Infrared Sensors. In: IEEE SENSORS LETTERS, vol. 5, pp. 1-4. ISSN 2475-1472 -
Arif, Arslan; Barrigon, Felipe A.; Gregoretti, Francesco; Iqbal, Javed; Lavagno, ... (2020)
Performance and energy-efficient implementation of a smart city application on FPGAs. In: JOURNAL OF REAL-TIME IMAGE PROCESSING, vol. 17, pp. 729-743. ISSN 1861-8200 -
Shan, Junnan; Lazarescu, Mihai T.; Cortadella, Jordi; Lavagno, Luciano; Casu, Mario R. (2020)
Power-Optimal Mapping of CNN Applications to Cloud-Based Multi-FPGA Platforms. In: IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS. II, EXPRESS BRIEFS, vol. 67, pp. 3073-3077. ISSN 1549-7747 -
Shan, Junnan; Casu, Mario R.; Cortadella, Jordi; Lavagno, Luciano; Lazarescu, Mihai T. (2019)
Exact and Heuristic Allocation of Multi-kernel Applications to Multi-FPGA Platforms. In: Design Automation Conference 2019, Las Vegas, NV (USA), 2-6 giugno 2019, pp. 1-6. ISBN: 9781450367257 -
BIN TARIQ, Osama; Lazarescu, MIHAI TEODOR; Iqbal, Javed; Lavagno, Luciano (2017)
Performance of machine learning classifiers for indoor person localization with capacitive sensors. In: IEEE ACCESS, vol. 5, pp. 12913-12926. ISSN 2169-3536 -
RAMEZANI AKHMAREH, Alireza; Lazarescu, MIHAI TEODOR; Bin Tariq, Osama; Lavagno, Luciano (2016)
A Tagless Indoor Localization System Based on Capacitive Sensing Technology. In: SENSORS, vol. 16. ISSN 1424-8220 -
Christos, Antonopoulos; Katerina, Asimogloy; Sarah, Chiti; Luca, D'Onofrio; Simone, ... (2016)
Integrated Toolset for WSN Application Planning, Development, Commissioning and Maintenance: The WSN-DPCM ARTEMIS-JU Project. In: SENSORS, vol. 16, pp. 804-843. ISSN 1424-8220