Wenhao Jia

About Me

I am a Systems Engineer at Scaled Inference, a company that aims to empower people through intelligent computing. Previously, I was a researcher at Qualcomm Research Silicon Valley. Prior to that, I graduated from Princeton University in 2014 with a Ph.D. degree in Electrical Engineering.

Research

At Qualcomm, I investigated how to utilize the many heterogeneous processors in a mobile system using an asynchronous parallel software library called MARE.

Previously, during my Ph.D. study, I focused on making massively parallel processors such as graphics processing units (GPUs) easier to program and run faster. Toward that goal, I have worked on a few projects:

  • To help GPU designers and users make tuning decisions, I apply statistical techniques such as linear regression and decision trees to model GPU hardware and software design spaces. The result is two fully automated statistical tools that efficiently sample designs, accurately capture performance/power trends using these samples, and output easy-to-understand models.
  • To transparently improve the memory system performance for GPU users, I study and improve the utility of general-purpose, read-write caches in GPUs. I developed both microarchitectural units and compile-time algorithms to make GPU caches more effective without the need for user intervention.

In general, I'm interested in studying and improving the power and performance of emerging computing platforms such as accelerators, mobile systems, and distributed systems. I want to make these platforms perform well while being easy to use.

Work Experience

  • Scaled Inference Inc., Palo Alto, CA, since October, 2015
    Systems Engineer
    Developing a machine learning online service platform
  • Qualcomm Inc., Santa Clara, CA, October, 2014–October, 2015
    Senior Engineer
    Researched and developed an asynchronous parallel software library
  • Google Inc., Mountain View, CA, June–September, 2011
    Software Engineering Intern
    Studied function-level program code multi-versioning in GCC
  • IBM T.J. Watson Research Center, Hawthorne, NY, June–September, 2010
    Research Intern
    Parallelized and optimized an MPI-based business analytics application

Education

  • Princeton University, Department of Electrical Engineering
    Ph.D., 2014
    M.A., 2011
  • Tsinghua University, Department of Electrical Engineering
    B.S., 2008

Publications

  • GPU Performance and Power Tuning Using Regression Trees
    Wenhao Jia, Elba Garza, Kelly A. Shaw, Margaret Martonosi
    Volume 12, Issue 2, Ariticle 13, ACM Transactions on Architecture and Code Optimization (TACO)
  • Analysis and Optimization Techniques for Massively Parallel Processors
    Wenhao Jia
    Ph.D. Thesis, advised by Prof. Margaret Martonosi and Prof. Kelly A. Shaw
  • MRPB: Memory Request Prioritization for Massively Parallel Processors [pdf | talk]
    Wenhao Jia, Kelly A. Shaw, Margaret Martonosi
    The 20th Int. Symp. on High Performance Computer Architecture (HPCA 2014)
  • Starchart: Hardware and Software Optimization Using Recursive Partitioning Regression Trees [pdf | talk]
    Wenhao Jia, Kelly A. Shaw, Margaret Martonosi
    The 22nd Int. Conf. on Parallel Architectures and Compilation Techniques (PACT 2013)
  • Characterizing and Improving the Use of Demand-Fetched Caches in GPUs [pdf | talk]
    Wenhao Jia, Kelly A. Shaw, Margaret Martonosi
    The 26th Int. Conf. on Supercomputing (ICS 2012)
  • Stargazer: Automated Regression-Based GPU Design Space Exploration [pdf | talk]
    Wenhao Jia, Kelly A. Shaw, Margaret Martonosi
    The 2012 Int. Symp. on Performance Analysis of Systems and Software (ISPASS 2012)
    [Best Paper Award Nomination]

Presentations and Posters

  • Design Space Analysis for Heterogeneous Systems [pdf]
    Wenhao Jia, Tae Jun Ham, Kelly A. Shaw, Margaret Martonosi
    Intel Science &Technology Center for Cloud Computing (ISTC-CC) Annual Retreat 2013
  • Analyzing and Optimizing GPU Communication and Computation [pdf]
    Wenhao Jia, Kelly A. Shaw, Margaret Martonosi
    Intel Science & Technology Center for Cloud Computing (ISTC-CC) Annual Retreat 2012
  • A System-Level ISA and Its Applications to Energy-Performance-Reliability Scheduling and Scratchpad Allocation [pdf]
    Yavuz Yetim, Wenhao Jia, Margaret Martonosi, Sharad Malik, Kelly A. Shaw
    Gigascale Systems Research Center (GSRC) Annual Symposium 2010
  • System-level ISA (SISA) [pdf]
    Wenhao Jia, Margaret Martonosi, Kelly A. Shaw
    Gigascale Systems Research Center (GSRC) Annual Symposium 2009

Software Releases

  • Stargazer: A linear regression-based GPU design space exploration tool.
  • Starchart: A regression tree-based GPU hardware/software optimization tool.
  • HotCRP Beamer Generator: Scripts for generating program committee meeting slides from HotCRP.