I'm a fifth year graduate student at Dept. of Computer Science of Princeton University.   My research interest is Parallel Architectures and Applications, including parallel programming models, performance profiling and modeling, parallel algorithm design, system area network. I  also have great interest in internet communication, especially those related with real-time requirements. Currently I am working with Professor Jaswinder Pal Singh in the PRISM project and CLUSTER project.

• An interesting talk about research from Richard Hamming
• Useful Links
• Recent Publications:
• Hongzhang Shan, Leonid Oliker, Rupak Biswas, Jaswinder Pal Singh, "Message Passing Vs. Shared Address Space on a cluster of SMPs". International Parallel and Distributed Processing Symposium (IPDPS 2001), also Parallel Computing Journal, Volume 29, Issue 2, February 2003. PDF 
• Hongzhang Shan, Leonid Oliker, Rupak Biswas, Jaswinder Pal Singh, "Comapring three programming models for adaptive applications on SGI Origin 2000", Supercomputing'00, Dallas, Texas.

  Best student paper award. [PDF].
• Hongzhang Shan, "Programming FFT on DSM Multiprocessors", 4th International Conference on High Performance Computing in the Asia-Pacific Region (HPC-ASIA 2000). In this paper, we find that if shared memory programs can be properly programmed, they can achieve much better performance than MPI or SHMEM even for such kind of regular, coarse-grained applications like FFT, which is traditionally believed to best fit message passing programming model.[PDF].
• Hongzhang Shan, Jaswinder Pal Singh,  "Parallel Sorting on Cache coherent DSM Multiprocessors", Supercompting'99, Portland, Oregon. [PS, PDF]. This paper investigate the best combination of programming models and algorithms that delivers the best performance on DSM platforms. We also examine the effect of radix size and data distribution.
• Hongzhang Shan, Jaswinder Pal Singh,  "Comparison of Message Passing, SHMEM and Cache- coherent Shared Address Space Programming Models on the SGI Origin 2000", International Conference of Supercomputing, May 1999, Rhodes Island, Greece. [PS, PDF]. Compares the performance of the three programming models on a set of relatively regular applications and sorting. Shows how to improve MPI implementation and what the tradeoffs are. Studies the impact of problem and machine size on the differences.

Invited submission for International Journal on Parallel programming.
• Hongzhang Shan, Jaswinder Pal Singh,  "Tree Building on a Range of Shared Address Space Multiprocessors: Algorithm and Application Performance", International Parallel Processing Synposium, 1998, Orlando, Florida. [PS, PDF]. In this paper, we develops completely new tree-building algorithms that are especially valuable for SVM and compare the performance among all the available shared memory tree-building algorithms to us.
• Dongming Jiang, Hongzhang Shan, Jaswinder Pal Singh,  "Performance Portability of Applications and Optimization Across Shared Address Space Multiprocessors", PPoPP97, Las Vegas. [PS, PDF]. Studies the performance portability of applications across different communication architectures that support a coherent shared address space programming model (from fine-grained hardware-coherent to page-grained SVM). Restructures applications in different classes of optimizations to understand programming issues. 
• Hongzhang Shan, "An Object-Oriented developing system for power-plant simulation", The 11th Symposium of Chinese Simulation Association, Chongqing, China, Oct. 1995.
• Ongoing research:
• I am working on some irregular, dynamic programs that require fine-grained dynamic replication of irregularly communicated remote data, such as N-body problem and Adaptive Meshapplications, which have being widely used in astrophysics, graphics, aerospace and engineering.
• Performance modeling.
• Preparing to release the MPI and SHMEM implementation of SPLASH2 (Shared Memory) benchmarks and the performance comparison of these three programming models on SGI Origin 2000.