Profiling Methods - Githubissues

Robert mentioned a few possible ways in which performance analysis could be performed on a program to demonstrate TESLA performance overheads etc:

Gem5: architecture simulator that can be used to obtain very detailed insights into where CPU cycles are being spent in the execution of a program
DTrace: scriptable probing tool that can be used to introspect a running program (potentially high overhead)
Callgrind: Valgrind derivative tool that runs the program in a sort of 'virtual machine' - need to check how this affects concurrent executions
GProf: Profiling tool that compiles a program with probing instrumentation added
Wall Clock Time: If a program is performing a meaningful computation, then this can be a good measure of performance overhead

Baltoli / project-docs