Berkeley Lab - Scientific Computing Seminar

Date:

Wednesday, March 19, 2008

Time:

11:00am-12:00pm

Location:

Building 50F, 1647 Conference Room

Seminar Speaker:

Samuel Williams
University of California, Berkeley

Title:

Autotuning Structured Grid and Sparse Matrix Kernels

Abstract:

We present an auto-tuning approach to optimize application performance on emerging multicore architectures. The methodology extends the idea of search-based performance optimizations, popular in linear algebra and FFT libraries, to application-specific computational kernels. Our work applies this strategy to both a lattice Boltzmann application (LBMHD), as well as the sparse matrix-vector multiplication (SpMV) kernel. Historically, these kernels have made poor use of scalar microprocessors due to their complex data structures and memory access patterns. We explored performance via auto-tuning on a broad set of multicore architectures including the Intel Xeon (clovertown), AMD Opteron (rev.F), Sun Victoria Falls (maramba), and the IBM Cell Broadband Engine. Through this approach we see substantial improvement across all architectures. Additionally, we present detailed analysis of each optimization, which reveal surprising hardware bottlenecks and software challenges for future multicore systems and applications.

The LBMHD portion of this work has been selected as a Best Paper at the upcoming International Parallel and Distributed Processing Symposium (IPDPS) 2008.

Sponsor of Seminar:

David Bailey

Scientific Computing

Contact Esmond G. Ng EGNg@lbl.gov