RWLock and Mutex Performance Test

gistfile1.txt

Tested on CIMS energon3 server, 8 cores. The code is in branch mutex_test.

CPU Time (by std::clock())
==========================

+ RWLock
Running Concurrent(Producer): Running time: 2.01
PASSED
Running Concurrent(Consumer): Running time: 0.96
PASSED
Running Concurrent(ReadWrite): Running time: 2.33
PASSED
Running Concurrent(Lookup): Running time: 12.95
PASSED

+ Mutex
Running Concurrent(Producer): Running time: 1.95
PASSED
Running Concurrent(Consumer): Running time: 0.53
PASSED
Running Concurrent(ReadWrite): Running time: 0.97
PASSED
Running Concurrent(Lookup): Running time: 15.56
PASSED


Wall-clock Time (by gettimeofday())
===================================

+ RWLock
Running Concurrent(Producer): Running time: 1.48686
PASSED
Running Concurrent(Consumer): Running time: 0.618758
PASSED
Running Concurrent(ReadWrite): Running time: 0.263542
PASSED
Running Concurrent(Lookup): Running time: 1.83095
PASSED

+ Mutex
Running Concurrent(Producer): Running time: 1.48389
PASSED
Running Concurrent(Consumer): Running time: 0.416416
PASSED
Running Concurrent(ReadWrite): Running time: 0.214367
PASSED
Running Concurrent(Lookup): Running time: 13.3602
PASSED


Conclusion
==========

+ Even the two approaches cost similar CPU time, but concurrent lookup test shows 
the performance improvement of parallel execution is significant and the speed-up is 7.3.

+ Another interesting thing worth noting is the overhead of reader/writer lock 
where mutual exclusion is necessary.