Berkeley Lab Scientific Computing Seminar

Date:

Friday, June 16, 2006

Time:

1:00pm-2:00pm

Location:

Building 50A-5132

Seminar Speaker:

Erich Strohmaier
Future Technologies Group
CRD, LBNL

Title:

Performance Complexity: An Execution Time Metric to Characterize Performance Transparency and Programming Complexity

Abstract:

Performance evaluation of code execution focuses on determining performance and efficiency levels for specific application scenarios. However, there is no measure characterizing how complex it is to achieve performance and how transparent performance results are. In this talk we present an execution time metric called /Performance Complexity (PC)/ to capture these important aspects of code development. /PC/ is based on performance results from a set of benchmark experiments and related performance models reflecting the behavior of a programmer. Residual modeling errors are used to derive /PC/ as measure for how transparent program performance is and how complex the performance appears to the programmer. /PC/ is independent from performance (/P/) itself which allows plotting system behavior in a performance-complexity map (P-C Map). We present a detailed description for calculating compatible /P/ and /PC/ values and use results from the parametric benchmark APEX-Map to illustrate the utility of /PC/ for analyzing systems and programming paradigms.

Sponsor of Seminar:

Kathy Yelick

Scientific Computing

Contact Esmond G. Ng EGNg@lbl.gov