After successful completion of the course, students are able to apply the knowledge about the basics of application-specific computing systems and architectures. They understand different characterization of image/video processing algorithms requiring application-specific architectures and design constraints. They know the interplay of application-specific architectures and run-time management techniques for high power/energy efficiency or performance-per-power efficiency. They are aware of the modern architectural trends and corresponding research themes. Moreover, they are able to design, develop, and apply concepts to real-world applications of camera-based processing systems.

Application-Specific Computing Systems are ubiquitous, ranging from Internet-of-Things, Automotive, Medical Imaging, Security, Consumer, etc. The continuously increasing user demands have resulted in a significant growth of advanced computing architectures and systems, which has been well complemented by the technological growth steered by Moore’s law. Moreover, algorithmic complexities and data rates are also ramping up as the application use cases are becoming more and more sophisticated. An excellent example would be a multi-camera system where, on the one hand video resolutions have increased to Ultra High-Definition and Super-Vision Multi-View Processing which requires massive data rate processing, and on the other hand, these computing architectures are subjected to stringent design constraints in terms of power, energy and area. Therefore, designing such application-specific architectures for high energy efficiency is a significant challenge.

 This lecture aims at providing (1) an overview of application-specific architecture and commercial design tools; and (2) an insight on innovative architectures and run-time systems for highly energy-efficient application-specific computing systems. A key focus will be on hardware-software collaborative techniques and how to identify important application-specific characteristics to optimize and adapt underlying computing architectures (that provide the computation capabilities), and even the executing algorithms (that determine the computation requirements). The techniques and concepts will be explained with the help of real-world applications of embedded image and video processing, which have widely proliferated in our daily life.

 In this lecture, the following topics will be explained along with a perspective to the actual research works and practical applications to real-world application-specific systems:

 The lecture will cover the following topics:

• Introduction of application-specific computing architectures
• Analysis of application-specific properties with respect to algorithms’ performance, power, and memory requirements
• State-of-the-art techniques and commercial design tools for application-specific systems.
• Complexity reduction techniques.
• Advanced trends for application-specific (multi-/many-core) processor architectures and design methods
• Advanced on-chip memory architectures and application-specific optimizations and management.
• Run-time systems for efficient application-driven resource and power management
• Emerging trends for application-driven extreme energy-efficiency, like Approximate Computing and Neuromorphic Architecture.

Lecture slides, research papers and presentation of examples and calculations

!!! The first lecture appointment is on 10.Oct.2019. !!!

An overview of all (master level) courses offered by the Embedded Computing Systems Group (191-02) in the winter semester 2019 will be presented on 01.Oct.2019, 9:00 am, Seminarraum Technische Informatik.

ECTS breakdown:

15h ...   presence in the lecture
35h ...   solution of assignments
25h ...   preparation for exam
---------------
75h ...  equals to 3 ECTS

=================================

Note: The course "191.105 Advanced Computer Architecture" is also relevant in this field of study.

The students will be evaluated based on a final exam, a quiz, and a course project.