CLAS Faculty Service Award 2012

Current Research group

	Rodney James. Rodney James is an Assistant Research Professor with me. He started in February 2011 and has been supported since then by NSF award OCI 1032861. Rodney and I are working on improving the LAPACK software package.
	Henricus Bouwmeester. Henricus is a Professional Research Assistant with me supported by NSF award CNS 0958354. Henricus has been working with me since Fall 2009. Henricus successfully defended his PhD thesis entitled "Tiled Algorithms for Matrix Computations on Multicore Architectures" in November 2012. He has been supported through NSF award CCF 0811520 and NSF award CCF 1054864 during his PhD thesis.
	Bradley Lowery. Brad is a PhD student with me. He is currently supported by the Lynn Bateman Memorial fellowship (Fall 2012-Summer 2013). Brad worked on parallel distributed reduction, in particular he designed an optimal reduction scheme for unidirectional network. Brad also worked on the stability analysis of oblique projections. (Technical reports in progress!)
	Matthew Nabity. Matthew has been working with me since Summer 2008. I supported him during Summer 2008, then he was supported through the NSF GK-12 program for two years (Fall 2008-Spring 2010), then he was supported by the Lynn Bateman Memorial fellowship (Fall 2010-Summer 2011). Since August 2011, Matthew is established at Western Oregon University in the Mathematics Department. His PhD thesis is about accelerating the nonsymmetric eigenvalue problem on multicore architectures. We hope to have Matthew defending his PhD thesis Spring 2013.
	Stephanie Patterson. As of Spring 2013, Stephanie is working as a Graduate Research Assistant with Rodney and I on improving the LAPACK software package. She is supported by NSF award OCI 1032861. She worked with us Summer 2012 as well.

Hardware -- colibri.ucdenver.edu GPU cluster

1	2	3	4

Thanks to NSF award "GPU Cluster for Computing Research," $435 000, CNS 0958354, we own a GPU cluster named "colibri". The cluster is made of 24 nodes and each node is made of 2 Nvidia Tesla M2090 and 2 Intel Sandy Bridge (8 cores each) with an InfiniBand interconnect. The cluster arrived on September 18th 2012 and is up and running and fully tested since mid November 2012.

Pictures

1	2	3	4	5	6

1. Henc and Julien at graduation ceremony (December 2012).
2. Brad and Julien after hiking Bierstdat (14,060ft, on the right, Sawtooth in the middle) (July 2012).
3. Henc, Julien and Cédric hiking in the Rockies (May 2010).
4. Julien and Cédric at work during World Cup 2010 (June 2010).
5. Eugene, Jenya and Julien at the Copper Mountain conference (April 2010).
6. Matthew, Julie and Julien at Red Rocks for Nine Inch Nails (September 2008).

Richer for the experience, moved on to excellent opportunities

	Internship: May 2010 / July 2010 -- Cédric Bourrasset, MS student, Polytech Clermont-Ferrand. Cédric was supported by NSF award CCF 0811520 for three months during Summer 2010.
	Internship: March 2009 / May 2009 -- Lee Rosenberg, undergraduate student, UC Denver Lee was supported by NSF award CCF 0811520 for three month during Summer 2010. Lee developed a tool for simulating the execution of tile algorithms on multicore architectures. Emmanuel Agullo, Henricus Bouwmeester, Jack Dongarra, Jakub Kurzak, Julien Langou, and Lee Rosenberg. Towards an efficient tile matrix inversion of symmetric positive definite matrices on multicore architectures. In the Proceedings of 9th International Meeting on High Performance Computing for Computational Science (VecPar'10), Berkeley, CA, USA, June 22-25, 2010.
	Internship: October 2008 / December 2008 -- Volodymyr Kondratenko, M.S. student, UC Denver Volodymyr was supported by NSF award CCF 0811520 as Research Assistant. He worked on the stability of the tiled LU algorithms.
	Internship: May 2008 / July 2008 -- Eugene Vecharynski, PhD student, UC Denver. I supported Eugene for three months during Summer 2008. Our research work lead to new results in the theory of restarted GMRES. We have two papers, one in SISC and one in Numerical Linear Algebra with Applications Eugene Vecharynski and Julien Langou. Any admissible cycle-convergence behavior is possible for restarted GMRES at its initial cycles. Numerical Linear Algebra with Applications, DOI:10.1002/nla.739, 2010. Eugene Vecharynski and Julien Langou. The cycle-convergence of restarted GMRES for normal matrices is sublinear. SIAM J. Scientific Computing, 32(1):186-196, 2010.
	Internship: May 2008 / July 2008 -- Jenya Kirshtein., PhD student, Denver University. I supported Jenya for three months during Summer 2008. Jenya worked on fast matrix-matrix multiply algorithms.
	March 2006 / June 2006 -- Rémi Delmas, M.S. student. Rémi Delmas obtained his Master degree from INSA Toulouse (option: Applied Mathematics and Modeling), a French engineering School. Now Rémi is enjoying the weather and the food of south-east of France where he works for Thalès IS, an IT consulting company belonging to the Thalès group. Rémi developed on a complete implementation of a C interface to LAPACK, he developed on a profiling layer for LAPACK, and finally he developed the first parallel distributed fault-tolerant matrix-matrix multiply. George Bosilca, Rémi Delmas, Jack Dongarra, and Julien Langou. Algorithm-based fault tolerance applied to high performance computing. J. Parallel Distrib. Comput., 69:410-416, 2009.