apps/testing: Add pthread_mutex_perf created by Anchao

[acassis]

Summary

During the discussion about the impact of adding counter to TCB Mr anchao created a pthread mutex performance example: apache/nuttx#17468 (comment)

Because this example can exercise (not ideally) the pthread mutex I decided to add it to apps/testing/sched, the program name will be called "pmutexp" (because "pmp" is not a good name).

Impact

Another testing tool for NuttX

Testing

nsh> pmutexp
1: diff = 0.278248977 s | avg = 0.278248977 s
2: diff = 0.277865395 s | avg = 0.278057186 s
3: diff = 0.276913045 s | avg = 0.277675805 s
4: diff = 0.277897198 s | avg = 0.277731153 s
5: diff = 0.278829989 s | avg = 0.277950920 s
6: diff = 0.276655238 s | avg = 0.277734973 s
7: diff = 0.278417236 s | avg = 0.277832439 s
8: diff = 0.276738257 s | avg = 0.277695666 s
9: diff = 0.278345268 s | avg = 0.277767844 s
10: diff = 0.279812201 s | avg = 0.277972280 s

===== result =====
count: 10
total: 2.779722804 s
avg: 0.277972280 s
nsh>

[cederom]

Thank you @acassis and @anchao :-)

Can we add some documentation to explain what this standalone tool does and what the results mean - for instance this tool can be used to measure performance of different implementations of mutex in the RTOS, and impact of changes in the mutex related code, the numbers mean ... and the smaller the better :-)

[linguini1]

Looks good to me, but I agree it needs docs too!

[acassis]

@cederom @linguini1 yes I will submit a Documentation/, basically it just try to get (lock) a mutex inside a busy-waiting loop and calculated the elapsed total and average time. This is a simple test for benchmarking specific use case, not a broader/real pthread mutex tester or profiler. There are other tests to mutex on ostest, but I think this is a simple test for validate impact in the kernel, like the TCB counter.

[cederom]

Maybe benchmarks would be better category then? :-)

[acassis]

Maybe benchmarks would be better category then? :-)

I thought it initially, but since it it more related to kernel scheduling performance I opted to add it to testing/sched