Application profiling tools are the instruments used to measure software performance at the function and application levels. Without careful configuration of the tools and the environment, invalid results are readily obtained. The errors may not become obvious if a large, complex application is profiled before more simple validations are attempted. A set of four simple, synthetic reference applications was used to validate configurations for profiling under x86_64 Linux. Results from one validated configuration and examples of observed invalid results are presented. While validation results for specific versions of software quickly lose value, this exercise demonstrates how future configurations can be validated and shows the kinds of errors that may reoccur.
Citation: Technical Note (NIST TN) - 1790
NIST Pub Series: Technical Note (NIST TN)
Pub Type: NIST Pubs
Software, performance, measurement, profiling