Induprakas Kodukula

710, Rhodes Hall
Cornell University, Ithaca, NY 14853
Res: (408) 720-8504
Off: (607) 254-8833
prakas@CS.Cornell.Edu

I'm a graduate student in Computer Science at Cornell University. My advisor is Professor Keshav Pingali. Prior to joining Cornell, I did my undergraduate studies at the Indian Institute of Technology, Madras. I joined Cornell in the fall of 1992.

Research

My work centers on the interplay between applications, compilers runtime systems for obtaining high performance on current architectures. My current work is motivated by the concern that as modern hardware is constantly evolving towards more complexity and better performance, there is not a corresponding gain in application performance. In particular, compiler and runtime system technology has not paid off to expectations because much of this technology has been developed without a proper study of applications. In this context, my thesis is an attempt at deriving applications sensitive technology that reduces the semantic distance from the underlying hardware. Specifically, I've developed restructuring compiler technology suitable for transforming programs from a wide range of domains. I've also developed runtime system technology for application support when applications do not lend themselves to compiler analysis. This work is in the context of the Bernoulli project.

• Induprakas Kodukula, Keshav Pingali, Robert Cox, Dror Maydan. An experimental evaluation of tiling and shackling for memory hierarchy management . ICS '99.
• Nikos Chrisochoides, Induprakas Kodukula, Keshav Pingali. Compiler and run-time support for irregular and adaptive applications . ICS '97.
• Induprakas Kodukula, Nawaaz Ahmed, Keshav Pingali. Data-centric multi-level blocking . PLDI '97.
• Nikos Chrisochoides, Induprakas Kodukula, Keshav Pingali. Compiler support for easing the programmer's burden. Workshop on Structured Adaptive Mesh Refinement Grid Methods, '97.
• Nikos Chrisochoides, Induprakas Kodukula, Keshav Pingali. Data Movement and Control Substrate for network-based parallel scientific computing . LNCS, 1199. Springer-Verlag '97.
• Induprakas Kodukula, Keshav Pingali. Transformations of Imperfectly Nested Loops . Supercomputing '96. (nominated for best student paper award).
• David Bau, Induprakas Kodukula, Vladimir Kotlyar, Keshav Pingali, Paul Stodghill. Solving Alignment using Elementary Linear Algebra . LNCS, 892. Springer-Verlag '95.

• Performance Enhancing Program Transformations(CS712) in Spring '97
  
• Systems Programming and Operating Systems(CS414) in Summer '96
  

• Data-centric Multi-level Blocking . ACM SIGPLAN Conference on Programming Languages Design and Implementation, Las Vegas. Presented in Jun '97. Also presented at Silicon Graphics Inc. (Mar '97), HP Labs Palo Alto (Feb '97).
• Data Movement and Control Substrate for network-based parallel scientific computing . Workshop on Communication, Architecture, and Applications for Network-based Parallel Computing, San Antonio. Presented in Feb '97.
• Data-centric Transformations . Workshop for Automatic Data Layout and Performance Prediction, Barcelona, Spain. Presented in Jan '97.
• Transformations of Imperfectly nested loops . Invited presentation at INRIA Rocquencourt, IRISA Rennes (France). Presented in Jan '97. Also presented at Supercomputing(Nov '96).
• A linear framework for transformations of general nested loops . International Loop Parallelization seminar. Schloss Dagstuhl, Germany. Apr '96.
• Loop transformations in a VLIW compiler. IBM Watson labs. Aug '95.
• The Necessity of Imperfectly Nested loops . HP labs, Chelmsford. Feb '95.

• I am the GNU czar for Cornell CS. See the CS GNU Home for details.
• I co-taught a class called ``Do you see a Computer in this Picture?'' as part of a graduate outreach program sponsored by Cornell University at local high schools.