Research Focus

Developing systems theoretic methodologies for designing optimal and reliable controllers for architecture resource management.


  1. Raghavendra Pradyumna Pothukuchi, Amin Ansari, Petros Voulgaris, Josep Torrellas, "Using Multiple Input, Multiple Output Formal Control to Maximize Resource Efficiency in Architectures", International Symposium on Computer Architecture (ISCA), June 2016.
  2. Raghavendra Pradyumna Pothukuchi, Josep Torrellas, "A Guide to Design MIMO Controllers for Architectures", Technical Report, April 2016.
  3. Pradyumna P.R., Tarun C.K.S., Bhanot, S., Remote experimentation of “No-load tests on a transformer” in electrical engineering, IEEE International Conference on Engineering Education: Innovative Practices and Future Trends (AICERA), pp.1-6, 19-21 July 2012.

Research description

  1. Implementing MIMO Control in Real Systems
    • We are working to deploy MIMO controllers for existing processors.
  2. Introducing MIMO Control for Architecture Research Efficiency [ ISCA'16, Technical Report]
    • There are many resources that can be dynamically tuned in modern and upcoming processors to meet several interrelated objectives such as power, performance and temperature.
    • Most current approaches rely heavily on heuristics to manage the processor resources to meet design targets. Such a heavily heuristic approach is difficult to design, tune or verify and does not provide any assurances to the designer about correct operation in scenarios outside the training set.
    • We propose the use of MIMO (Multiple Input Multiple Output) control theory for architecture resource management and developed a methodology to design MIMO controllers for processors. Our work combines concepts from Multiple Input Multiple Output (MIMO) Control theory, System Identification and Robust Control Theory to facilitate and generalize the use of MIMO Control theory in architecture.
    • Our work on providing the architectural insights into the MIMO design methodology is summarized in our paper at ISCA'16. The details of all steps in the proposed MIMO design methodology with MATLAB scripts and descriptions of the design choices are documented in a technical report.