Moritz Hädrich: AI specific Functional Units for Application Specific Instruction-set Processors (ASIPs) (Seminar Work)

30.09.2021, 1:00 p.m.

Invitation to the presentation of Moritz Hädrich

Topic: AI specific Functional Units for Application Specific Instruction-set Processors (ASIPs)​

Project: Seminar work

Supervisor: Muhammad Ali

Abstract: This work investigates the latest research in the field of application-specific instruction-set processors (ASIPs) for artificial intelligence (AI) applications. AI is becoming popular and more research in this direction contributes to more computationally intensive and memory-hungry algorithms. For these constraints, ASIPs are considered as a good solution since ASIP provide a bridge between general-purpose processors and application-specific data path. ASIPs have application-specific instructions which are tailor-made for the application in order to improve the computational performance while maintaining the flexibility and programmability of the system. In this work, the student needs to research different publications related to ASIPs for different machine learning applications. A classification needs to be defined either based on hardware optimization or the application test case used in the references. When classifying it should be kept in mind that it should be a suitable comparison between the research work. Tools and design methodologies used should also be mentioned. A comparison is to be drawn between different publications that are investigated during the seminar work. In the end, future trends need to be stated. The student needs to document his/her findings in a report as the final submission.