Making Reusable Hardware Design IPs Usable: an NoC perspective

Subscribers:
344,000
Published on ● Video Link: https://www.youtube.com/watch?v=WBBsR6wmvn0



Duration: 1:21:48
78 views
1


Reusable hardware design IPs have become increasingly parameterizable and even taken on the form of fully on-demand generators. The flexibility of user customization is key to achieving performance and energy efficiency by tailoring the IP instantiations to fit the specifics of the application. At the same time, unfortunately, the complexity and flexibility of IPs have reached a level where the task of understanding and tuning the myriad of parameters can become itself unmanageable for the IP users. This talk argues for a new paradigm for generator-based IPs that encapsulates the IP authors' knowledge to assist the users' interactions with the IP. An IP should not only capture the microarchitectural and structural view of a design, but it should also include (1) characterization meta-data; (2) high-level tuning knobs that are tailored to the specific domain and meaningful to the IP user; (3) domain-aware simulation-time and runtime monitoring and introspection mechanisms; (4) auxiliary supporting material and tools that enhance how the user interacts with the IP. In doing so, the effort that would otherwise be repeated by each IP user is now performed only once by the IP author. To demonstrate the potential of this new IP paradigm we present a case study of extending the CONNECT Network-on-Chip IP generator (http://users.ece.cmu.edu/~mpapamic/connect/) to embody the above principles. This work is the subject of Michael Papamichael's on-going PhD thesis work at Carnegie Mellon University.




Other Videos By Microsoft Research


2016-07-26Intersection of two passions
2016-07-26Analysis of Boolean Functions: advances and challenges
2016-07-26Two basic problems in finite stochastic optimization
2016-07-26Decision Learning: Learning with Strategic Decision Making
2016-07-26The CUBE: examples and research challenges in large scale interaction space
2016-07-26The Promises and Pitfalls of Demographic Inference in Social Media
2016-07-26Probabilistic Elastic Part Model: A Pose-Invariant Representation for Real-world Face Recognition
2016-07-26Checking App Behavior Against App Descriptions
2016-07-26Scalable learning of Bayesian network classifiers
2016-07-26Everything you always wanted to know about web-based device fingerprinting (but were afraid to ask)
2016-07-26Making Reusable Hardware Design IPs Usable: an NoC perspective
2016-07-26Lessons from Megaprojects: The Creators and Destroyers of Capital
2016-07-26Unsupervised Transcription of Historical Documents
2016-07-26Online Learning and Adaptation Over Networks
2016-07-26Easy Generation and Efficient Verification of Unsatisfiability Proofs
2016-07-26Learning to Understand Natural Language in Physically-Grounded Environments
2016-07-26Frontiers of Accessibility: From the Body to the Mind, the Heart, and the Soul
2016-07-26Analyzing neurological disorders using functional and structural brain imaging data
2016-07-26Collaborative, Large-Scale Data Analytics and Visualization with Python
2016-07-26Gap Probabilities for Zeroes of Stationary Gaussian Functions
2016-07-26Random Walks on Groups and the Kaimanovich-Vershik Conjecture for Lamplighter Groups



Tags:
microsoft research