Albert Hartono 

Dept. of Computer Science and Engineering 
Dreese Labs 
2015 Neil Ave., 395 
Columbus, OH 43210 
(614)370-5269
hartonoa@cse.ohio-state.edu 
http://www.cse.ohio-state.edu/~hartonoa 

RESEARCH INTEREST
Compile-time and run-time optimizations for High Performance Computing (HPC); 
Loop transformations for locality and parallelism; Automation of empirical 
performance tuning of parallel applications; Parallel programming models. 

EDUCATION
2003-2009: Ph.D. Computer Science and Engineering, Ohio State University, Columbus, OH. 
2001-2003: M.S. Computer Science, Indiana University, Bloomington, IN. 
1996-2000: B-Tech. Electrical Engineering, Trisakti University, Jakarta, Indonesia. 

EXPERIENCE:
Su 2003-Au 2009: Graduate Research Associate, Ohio State University, Columbus, OH. 
  Department of Computer Science and Engineering 
  Advisor: Prof. P. Sadayappan 

Su 2007: Research Intern, IBM Almaden Research Center, San Jose, CA. 
  GPFS (General Parallel File System) team 
  Supervisor: John Palmer 

Au 2006-Sp 2008: Graduate Research Co-op, Argonne National Laboratory, Argonne, IL. 
Su 2006: Wallace Givens Research Fellow, Argonne National Laboratory, Argonne, IL. 
  Mathematics and Computer Science Division 
  Supervisor: Dr. Boyana Norris 

Su 2004: Software Engineer Intern, Bell Laboratories, Lucent Technologies, Murray Hill, NJ. 
  Lucent nmake Product Builder, Software Technology Center 
  Supervisor: Gary M. Selzer 

2002-2003: Associate Instructor, Indiana University, Bloomington, IN. 
  Department of Computer Science 

1998-2000: Laboratory Assistant, Trisakti University, Jakarta, Indonesia. 
  Department of Electrical Engineering 

1998-1999: Teaching Assistant, Trisakti University, Jakarta, Indonesia. 
  Department of Electrical Engineering 

JOURNAL PUBLICATIONS
Performance Optimization of Tensor Contraction Expressions for Many Body Methods in 
Quantum Chemistry.
  Albert Hartono, Qingda Lu, Thomas Henretty, Sriram Krishnamoorty, Huaijian Zhang, 
  Gerald Baumgartner, David Bernholdt, Marcel Nooijen, Russell Pitzer, J. Ramanujam, 
  P. Sadayappan. Journal of Physical Chemistry A. (accepted)

BOOK CHAPTER
Annotations for Productivity and Performance Portability.
  Boyana Norris, Albert Hartono, and William Gropp. Petascale
  Computing: Algorithms and Applications.  Computational
  Science. Chapman and Hall / CRC Press, Taylor and Francis Group,
  2007.

REFEREED CONFERENCE PUBLICATIONS
Parametric Tiled Loop Generation for Effective Parallel Execution on Multicore Processors.
  Albert Hartono, Muthu Manikandan Baskaran, J. Ramanujam, and P. Sadayappan. 
  In IEEE International Parallel and Distributed Processing Symposium (IPDPS), 
  April 2010, Atlanta, Georgia. (submitted)

Parameterized Tiling Revisited.
  Muthu Manikandan Baskaran, Albert Hartono, Thomas Henretty, J. Ramanujam, and P. Sadayappan. 
  In IEEE/ACM International Symposium on Code Generation and Optimization (CGO), 
  April 2010, Toronto, Canada. (accepted)

Parametric Multi-Level Tiling for Imperfectly Nested Loop.
  Albert Hartono, Muthu Manikandan Baskaran, Cedric Bastoul, Albert
  Cohen, Sriram Krishnamoorthy, Boyana Norris, J. Ramanujam, and
  P. Sadayappan. In ACM International Conference on Supercomputing
  (ICS), June 2009, Yorktown Heights, New York.

Generating Empirically Optimized Composed Matrix Kernels from MATLAB Prototypes.
  Boyana Norris, Albert Hartono, Elizabeth Jessup, and Jeremy Siek. In
  International Conference on Computational Science (ICCS), May 2009,
  Baton Rouge, Louisiana.

Annotation-Based Empirical Performance Tuning Using Orio.
  Albert Hartono, Boyana Norris, and P. Sadayappan. In IEEE
  International Parallel and Distributed Processing Symposium (IPDPS),
  May 2009, Rome, Italy.

A Polyhedral Framework for Automatic Parallelization and Locality Optimization.
  Uday Bondhugula, Muthu Baskaran, Albert Hartono, Sriram
  Krishnamoorthy, J. Ramanujam, Atanas Rountev, and P. Sadayappan. In
  Workshop on Compilers for Parallel Computing (CPC), January 2009,
  Zurich, Switzerland.

A Practical Automatic Polyhedral Parallelizer and Locality Optimizer.
  Uday Bondhugula, Albert Hartono, J. Ramanujam, and P. Sadayappan. In
  ACM SIGPLAN Programming Language Design and Implementation (PLDI),
  June 2008, Tucson, Arizona.

Towards Effective Automatic Parallelization for Multicore Systems.
  Uday Bondhugula, Muthu Baskaran, Albert Hartono, Sriram
  Krishnamoorthy, J. Ramanujam, Atanas Rountev, and P. Sadayappan. In
  IPDPS Workshop on Next Generation Software (NGS), April 2008, Miami,
  Florida.

Designing High Performance and Scalable MPI Intra-Node Communication Support for Clusters.
  Lei Chai, Albert Hartono, and Dhabaleswar K. Panda. In IEEE
  International Conference on Cluster Computing (CLUSTER), September
  2006, Barcelona, Spain.

Identifying Cost-Effective Common Subexpressions to Reduce Operation Count in Tensor Contraction Evaluations.
  Albert Hartono, Qingda Lu, Xiaoyang Gao, Sriram Krishnamoorthy,
  Marcel Nooijen, Gerald Baumgartner, Venkatesh Chopella, David
  E. Bernholdt, Russell M. Pitzer, J. Ramanujam, Atanas Rountev, and
  P. Sadayappan. In International Conference on Computational Science
  (ICCS), May 2006, Reading, UK.

Automated Operation Minimization of Tensor Contraction Expressions in Electronic Structure Calculations.
  Albert Hartono, Alexander Sibiryakov, Marcel Nooijen, Gerald
  Baumgartner, David E. Bernholdt, So Hirata, Chi-Chung Lam, Russell
  M. Pitzer, J. Ramanujam, and P. Sadayappan. In International
  Conference on Computational Science (ICCS), May 2005, Atlanta,
  Georgia.

RESEARCH REPORTS
PrimeTile: A Parametric Multi-Level Tiler for Imperfect Loop Nests.
  Albert Hartono, Muthu Manikandan Baskaran, Cedric Bastoul, Albert
  Cohen, Sriram Krishnamoorthy, Boyana Norris, J. Ramanujam, and
  P. Sadayappan. Research Report OSU-CISRC-2/09-TR04, Computer Science
  and Engineering Department, Ohio State University, February 2009.

Generating Empirically Optimized Composed Matrix Kernels from MATLAB Prototypes.
  Boyana Norris, Albert Hartono, Elizabeth Jessup, and Jeremy
  Siek. Preprint ANL/MCS-P1581-0209, Mathematics and Computer Science
  Division, Argonne National Laboratory, February 2009.

Annotation-Based Empirical Performance Tuning Using Orio.
  Albert Hartono, Boyana Norris, and P. Sadayappan. Preprint
  ANL/MCS-P1556-1108, Mathematics and Computer Science Division,
  Argonne National Laboratory, November 2008.

Annotations for Productivity and Performance Portability.
  Boyana Norris, Albert Hartono, and William Gropp. Preprint
  ANL/MCS-P1392-0107, Mathematics and Computer Science Division,
  Argonne National Laboratory, January 2007.

Automated Operation Minimization of Tensor Contraction Expressions in Electronic Structure Calculations.
  Albert Hartono, Alexander Sibiryakov, Marcel Nooijen, Gerald
  Baumgartner, David E. Bernholdt, So Hirata, Chi-Chung Lam, Russell
  M. Pitzer, J. Ramanujam, and P. Sadayappan. Research Report
  OSU-CISRC-2/05- TR10, Computer Science and Engineering Department,
  Ohio State University, February 2005.

SOFTWARE DEVELOPMENTS
PrimeTile: A Parametric Multi-Level Tiler for Imperfect Loop Nests. 
  http://primetile.sourceforge.net/ 

Orio: An Annotation-Based Empirical Performance Tuning Framework. 
  http://trac.mcs.anl.gov/projects/performance/wiki/Orio 


SELECTED PROJECTS
Loop transformations for data locality and parallelism.
  Tiling is a crucial loop transformation for generating
  high-performance code on modern architectures, aiming at
  maximizing data locality and exploiting coarse-grained
  parallelism. An automatic source-to-source transformation
  tool, Pluto, has been developed to generate parallel
  OpenMP tiled code from sequential untiled code. Pluto
  generated tiled code has fixed tile sizes. Tiled codes with
  variable tile sizes are important for autotuning systems such
  as ATLAS. Multi-level tiled codes are important for increasing
  data reuse in deep memory hierarchy. A code generator, called
  PrimeTile, has been developed to generate efficient
  parametric multi-level tiled loop nests. Experimental results
  show high speedups for both sequential and parallel executions
  over the best production compilers.

Empirical performance tuning using annotations.
  An extensible annotation-based empirical performance tuning
  tool, called Orio, has been developed to achieve both
  performance and productivity in scientific computations. Users
  annotate the source code with performance hints that will
  trigger a set of low-level optimizations. Based on the given
  values of various performance parameters, Orio generates many
  tuned code versions, selectively and empirically evaluates the
  performance of each version, and finally selects the
  best-performing code. Orio framework has a clean design and
  can be easily extended with external code optimization
  tools. Experiments show that Orio can deliver performance
  improvements when used alone or in conjunction with other
  transformation tools.

Operation minimization of tensor contraction expressions
  Complex tensor expressions arising in many science and
  engineering domains have many algebraically equivalent forms
  that can differ drastically in the number of arithmetic
  operations required for evaluating these expressions. The
  problem of finding an operation-minimal form of tensor
  expressions is, however, NP-hard. We have developed effective
  operation minimization techniques for tensor expressions using
  algebraic transformations and common-subexpression
  elimination. Using our automated operation minimization tool,
  scientists can find new more effective formulations of tensor
  contraction expressions.

High performance cluster interconnects
  We have designed and implemented an optimized MPI intra-node
  communication using user space memory copy scheme, for NUMA
  and multicore systems. Our approach has efficient utilization
  of L2 cache and significantly reduces memory consumption. The
  implementation is faster, more efficient, and scales better,
  and currently is part of the MVAPICH release.

High-performance parallel file systems
  I designed and implemented a light-weight performance
  analysis tool to track and visualize locking contentions
  inside IBM's General Parallel File System (GPFS). Using the
  tool, I was able to identify a set of well-behaved and bad
  mutexes caused by bad coding styles. 

Automatic dependency-based Java build support
  I implemented an extension for Lucent nmake Product Builder 
  to automatically extract implicit build dependencies from 
  Java program source containing new J2SE 1.5 language 
  constructs using an open-source tool, called JavaDeps.

AWARDS AND HONORS
2009: Travel award, International Conference on Supercomputing (ICS)
2009: Travel award, International Parallel and Distributed Processing Symposium (IPDPS)
2000: Ranked 4th out of the ten best graduates of Electrical Engineering Department, Trisakti University, Jakarta, Indonesia.
1999: Undergraduate engineering scholarship, Trisakti University, Jakarta, Indonesia.

REFERENCES
Dr. P. Sadayappan, Professor.
Ohio State University
saday@cse.ohio-state.edu

Dr. Boyana Norris, Computer Scientist.
Argonne National Laboratory
norris@mcs.anl.gov

Dr. J. Ramanujam, Professor.
Louisiana State University
jxr@ece.lsu.edu

Dr. John Palmer, Research Staff Member and Manager (GPFS - General Parallel File Systems).
IBM Almaden Research Center
jpalmer@almaden.ibm.com

Gary M. Selzer, Manager (Lucent nmake Product Builder).
Bell Laboratories, Lucent Technologies (now Alcatel-Lucent)
gmselzer@alcatel-lucent.com