This is the home page for our 16th major research exhibit at the IEEE/ACM Supercomputing conference. The exhibit is again under the title Aggregate.Org / University of Kentucky, the informal research consortium led by our KAOS (Compilers, Hardware Architectures, and Operating Systems) group here at the University of Kentucky's Department of Electrical and Computer Engineering.

A Maze Of Twisty Little Passages

Much like last year, the physically big thing in our exhibit this year is a technical demonstration consisting of a large wooden maze with four balls in it. Each of the colored balls has a different path to take (MIMD), yet it is perfectly feasible to get all the balls to their respective destinations by a series of tilts of the table (SIMD). Yes, you really can execute MIMD code on SIMD hardware with good efficiency... Click on the maze above for a ~50MB video showing the maze in action. Fundamentally, this is what our latest software does for GPUs (Graphics Processing Units). Specifically, it can take shared-memory MIMD code written in C and efficiently execute it on an NVIDIA CUDA GPU.

Why do this? GPUs thus far have not had stable, portable, programming support for general-purpose use, so there is virtually no code base for supercomputing applications. Our technology allows codes written for popular cluster and SMP target models to be used directly. We plan to support both C and Fortran with both shared memory and MPI message passing. The current shared memory model uses "parallel subscripting" in which a[||b] means a in processor b's memory; we initially assumed OpenMP would be the prefered model, but have had requests for Posix Threads. It is surprisingly easy to support dynamic thread creation, although there are performance issues involving memory bank conflicts in sharing the complete memory map.

How well does MIMD code perform? A lot better than it did last year. In fact, even the MIMD interpreter is fast enough to be usable for real applications.

Work Presented

• A Maze Of Twisty Little Passages
  (about the MIMD On GPU -- MOG -- work we are doing)
  The following are the key publications for understanding our MIMD-on-SIMD technology; the first two from the original concept and the last giving details of the current MOG system:
  • H. G. Dietz, "Common Subexpression Induction," Proceedings of the 1992 International Conference on Parallel Processing, Saint Charles, Illinois, August 1992, vol. II, pp. 174-182. (Postscript or PDF)
  • H. G. Dietz and G. Krishnamurthy, "Meta-State Conversion," Proceedings of the 1993 International Conference on Parallel Processing, vol. II, pp. 47-56, Saint Charles, Illinois, August 1993. (Postscript or PDF)
  • H. G. Dietz and B. D. Young, "MIMD Interpretation on a GPU," 22nd International Workshop on Languages and Compilers for Parallel Computing, University of Delaware, October 8, 2009. (local PDF copy pending Springer LNCS publication)
• You Can't Always Get What You Want
  (about Line Associative Registers vs. conventional cache, lots of new work on this)
• See & Hear, Show & Tell
  (about digital imaging using light, sound, etc. - see Digital Imaging Technologies, including new work on computational photography)
• Small Thinking
  (about Nanocontrollers - Kentucky Architecture, Nanoprocessors, & Nanocontrollers)
• Horseshoes and Hand Grenades
  (about precision and accuracy - Numerical Precision, Accuracy, and Range)
• Powerful Beyond Dissipation?
  (about predictive power management - Energy Management)
• Come Together Right Now Over Me
  (about Aggregate Function Networks -- Theory, Software, and Hardware -- the new work here is on-chip AFNs for CMPs)
• Flat Outperformance
  (about Flat Neighborhood Networks - NetWires)

Some related theses from our group:

• Diego A. Rivera Polanco, Collective Communication and Barrier Synchronization on NVIDIA CUDA GPUs (PDF), University of Kentucky MSEE Thesis, September, 2009.
• Bobby Dalton Young, MPI Within A GPU (PDF), University of Kentucky MSEE Thesis, http://hdl.handle.net/10225/1085, July, 2009.
• Soohong P. Kim, Chip Multiprocessors with On-Chip Aggregate Function Network, Purdue University PhD Dissertation, (unofficial local PDF copy pending Purdue posting), August, 2009.
• Akil Kaveti, HDL Implementation and Analysis of a Residual Register for a Floating-Point Arithmetic Unit (PDF), University of Kentucky MSEE Thesis, http://hdl.handle.net/10225/893, 2008.
• Sarojini Priyadarshini Rajachidambaram, Kentucky's Adapter for Parallel Execution and Rapid Synchronization , University of Kentucky MS Thesis, http://hdl.handle.net/10225/709, 2007.
• Swetha Mitta, Kentucky's Adapter for Parallel Execution and Rapid Synchronization , University of Kentucky MSEE Thesis, http://hdl.handle.net/10225/621, 2007.
• Chenxing Wang, Dynamic Voltage Scaling for Priority-Driven Scheduled Distributed Real-Time Systems , University of Kentucky PhD Dissertation, http://hdl.handle.net/10225/747, 2007.
• Timothy Ian Mattox, Exploiting Sparseness of Communication Patterns for the Design of Networks in Massively Parallel Supercomputers (PDF), University of Kentucky PhD Dissertation, http://hdl.handle.net/10225/280, 2006.
• Krishna Melarkode, Line Associative Registers (PDF) University of Kentucky MSEE Thesis, http://hdl.handle.net/10225/255, 2004.

Slow-Update Live View of our Exhibit

The following image was grabbed from a network camera with a 175-degree fisheye lens (door peephole) mounted on the top of our 16-foot-tall sign:

After-Show Stuff

Now that the show is over, here's a little about what we did and what we learned:

• The Theme. Forget about being green (the official theme). This was the year that every machine seemed to have a few Tesla GPUs in it.
• Most surprising thing. The "deep-fried" computing display using oil bath cooling in a rather industrial steel tub (see http://grcooling.com/). It is messy, but possibly a very good idea....
• Research? SC online registration still doesn't understand the concept of student exhibitors. Overall, the quantity of actual researchers in research exhibits continues to go down... which is a very sad thing.

Some photos of our exhibit:

The booth overall (our tall backlit sign was visible from a good distance away)

Our environmentally-friendly print-on-demand handouts

The technical and overview slides (rear projection)

The maze

The student researchers in our booth this year

The only thing set in stone is our name.