Measuring the Execution Time of Fragmented Programs

摘要

A technique for evaluating the execution time of program fragments on superscalar and explicitly parallel processors is described. Rules for the fragmentation and modification of the initial source code in a high-level language are proposed, and examples in C++ are considered. An implementation of the dynamic analysis of programs and the examination of its results with account for side effects caused by specific features of a processor architecture and operating system are considered.