Scientific Computing Seminar

Date:

Friday, May 7, 2004

Time:

1:00pm-2:00pm

Location:

50A-5132

Seminar Speaker:

Simone S Sbaraglia
Watson Research Center
IBM

Title:

A Simulator Infrastructure to Guide Memory Analysis

Abstract:

Understanding and tuning memory system performance is a critical issue for most scientific programs to achieve good performance on current high performance systems. In this talk we present SIGMA, a new data collection framework and family of cache analysis tools. The SIGMA environment provides detailed cache information by gathering memory reference data using software based binary instrumentation. This data is presented in a application-centric fashion, to help users understand the precise memory references in their program that are causing poor utilization of the memory subsystem. The infrastructure can facilitate quickly probing into the factors that influence application performance by highlighting bottleneck scenarios including: excessive cache/TLB misses and inefficient data layouts. Tools included with SIGMA can assist in perturbation analysis to determine performance variations caused by changes to architecture or program. Some applications of SIGMA to NERSC codes will be presented.

Sponsor of Seminar:

David Skinner

Scientific Computing

Contact Esmond G. Ng EGNg@lbl.gov