Tag gpu

TeraGrid’10 GPU tutorial.

Brandon Hill (Purdue) at TeraGrid'10 GPU tutorial in Pittsburgh

During the first week in August, we helped organize a GPU tutorial for the annual TeraGrid conference, held this year in Pittsburgh. Many thanks to Hyesoon Kim (GT), Brandon Hill (Purdue), Bruce Loftis (UT/NICS), and Tabitha Samuel (UT/NICS) for putting it together.

(UPDATE Aug 19) Slides are available at: http://hpcgarage.org/tg10–gpu-tutorial

PMAA’10 in Basel.

View of the Rhine River (Basel, Switzerland)

Jee and Rich just returned from the Sixth International Workshop on Parallel Matrix Algorithms and Applications (PMAA 2010), which the University of Basel in Switzerland hosted this year. There were many interesting talks, including a keynote by colleague and collaborator, George Biros, sessions on GPUs and autotuning in matrix computations, as well as Jee’s reprise of his PPoPP 2010 talk on sparse matrix-vector multiply on GPUs.

The University of Basel campus abuts the Rhine river. Rich wonders how students and faculty there still manage to get so much great work done…

HPCwire blurb

Our HotPar’10 paper has gotten a couple of views, including this nice write-up in HPCwire by editor Michael Feldman. It’s nice to be noticed! 

HotPar’10 and the *real* HP Garage

Rich just got back from the USENIX Workshop on Hot Topics in Parallelism (HotPar), where he presented the lab’s meta-analysis of the potential and limitations of GPU systems. Slides for the talk should soon appear on the HotPar’10 page if they are not there already.

The workshop itself was great fun! In particular, it had a uniquely lively atmosphere, with highly interactive talk sessions, a prominent and high-quality poster session, great lunch discussions, all at an amazing price! (To reduce cost, we stayed and “workshopped” on Berkeley’s Clark Kerr facilities. Lodging + all meals + registration was only about $600 or so. Amazing!)

As an added bonus, Rich accidentally missed his flight back to ATL and so took a minute to drop by the original Hewlett-Packard Garage, depicted above. It’s a great reminder of the kind of impact the HPC Garage hopes to have some day.

The *real* HP Garage

PPoPP’10 paper on GPU SpMV

Jee’s and Amik’s paper on autotuning sparse matrix-vector multiply for GPUs will appear at PPoPP 2010 in Bangalore, India. Great work, gang!

  • J. W. Choi, A. Singh, R. Vuduc. “Model-driven autotuning of sparse matrix-vector multiply on GPUs.” In Proc. ACM SIGPLAN Symp. Principles and Practice of Parallel Programming (PPoPP), Bangalore, India, January 2010. (accepted)

ICS’09 talk & paper.

Sundar is speaking today at the ACM International Conference on Supercomputing (ICS) on his CPU/GPU stencil kernel work.

Citation & slides: Sundaresan Venkatasubramanian and Richard Vuduc. “Tuned and wildly asynchronous stencil kernels for hybrid CPU/GPU platforms.” In Proc. ACM Int’l. Conf. Supercomputing (ICS), Yorktown Heights, NY, USA, June 2009. [Paper | Slides]