25 research outputs found
Classes and Subclasses in Actor-Oriented Design
No full textEdward Lee and Stephen Neuendorffer EECS Department University of California at Berkeley Berkeley, CA 94720, U.S.A
Implementation Issues in Hybrid Embedded Systems
No full textThis paper presents an approach to the implementation of electronic computation systems whose behavior is tightly integrated with the physical world. We call such systems hybrid embedded systems. Such systems are challenging from a design perspective because their behavior is governed by both continuous-state dynamics from the physical world and discrete-state dynamics from the computation. There are several difficulties that appear in such systems. For instance, understanding of the passage of time during computation is critical to understanding how the computation system affects the state of the physical world. Hybrid embedded systems are also inherently concurrent; the computation system operates concurrently with the dynamics of the physical world, in addition to any concurrency that may be designed into the system. In addition, hybrid embedded systems must generally operate within the constraints of traditional embedded systems. They are inevitably constrained computationally, often have a complex computational architecture, and must perform predictably. This paper presents an approach to the design of embedded systems utilizing component-based system models capable of representing concurrency, the passage of time, and both continuous and discrete behaviors. These models allow for automatic generation of system implementations from high-level abstractions as well as the consideration of low-level architectural details where necessary. We show how this technique can be ued to approach difficulties in the design of a complex digital control system
Scalable High-Level Synthesis for AI accelerator design
No full textSubmission published under a 24 month embargo labeled 'Closed Access', the embargo will last until 2027-05-01The student, Hanchen Ye, accepted the attached license on 2025-05-01 at 12:56.The student, Hanchen Ye, submitted this Dissertation for approval on 2025-05-01 at 13:32.This Dissertation was approved for publication on 2025-05-01 at 14:30.DSpace SAF Submission Ingestion Package generated from Vireo submission #22119 on 2025-10-19 at 19:54:49The growing scale and complexity of deep neural networks (DNNs) present significant challenges for efficient hardware acceleration. High-Level Synthesis (HLS) has emerged as a promising methodology to enhance design productivity for FPGA- and ASIC-based accelerators. However, existing HLS workflows struggle to scale effectively due to their inability to address cross-stack co-design challenges spanning architecture design, compiler infrastructure, and Electronic Design Automation (EDA) algorithms. This dissertation proposes a comprehensive and scalable HLS methodology for AI accelerator design by innovating across three synergistic levels. At the design level, we present HybridDNN and DNNExplorer, two frameworks that facilitate the generation and exploration of hardware accelerators through algorithm-aware modeling and fine-grained design space exploration. At the compiler level, we develop ScaleHLS, HIDA, and StreamTensor, which together form a scalable HLS compiler stack that supports multi-level intermediate representations, design space optimizations, and hardware-aware scheduling for both generic and dataflow-based accelerators. At the EDA level, we introduce ISDC, an iterative scheduling algorithm that integrates feedback from downstream tools to significantly enhance resource utilization. Collectively, these contributions constitute a full-stack solution to scalable HLS, advancing the productivity, performance, and adaptability of AI accelerator design
Chapter 2 ACTOR-ORIENTED MODELS FOR CODESIGN Balancing Re-Use and Performance
No full textAbstract: Most current hardware engineering practice is deeply rooted in discrete-event modeling and synchronous design. Most current software engineering is deeply rooted in procedural abstractions. The latter says little about concurrency and temporal properties, whereas the former lacks many of modularity capabilities of modern programming languages. Actor-oriented design emphasizes concurrency and communication between components while maintaining modularity. Components called actors execute and communicate with other actors. In contrast to the interfaces in object-oriented design (methods, principally, which mediate transfer of the locus of control), interfaces in actor-oriented design (which we call ports) mediate communication. But the communication is not assumed to involve a transfer of control. This paper discusses the structure of actor-oriented models and shows how data and behavioral type systems enhance modularity and re-use potential while enabling designs that embrace concurrency and time. This paper shows how components can be designed for re-use through parameterization and behavioral polymorphism, and how component specialization can offset the performance costs of doing so. Keywords: Actor-oriented design, behavioral types, behavioral polymorphism, component specialization, code generation, parameterization, Ptolemy.
