tests/benchmarks/timing_info measurements are suddenly higher than previous values on nrf52_pca10040

[arun1joshi]

Describe the bug
tests/benchmarks/timing_info measurements are suddenly higher than previous values on nrf52_pca10040.

To Reproduce

Steps to reproduce the behavior:
run tests/benchmarks/timing_info

Expected behavior
No uptick in measurement

Impact
Its not blocking anything

Screenshots or console output

***** Booting Zephyr OS v1.14.0-rc3-99-g14db4eedff71 (delayed boot 1000ms) *****
starting test - Time Measurement
Timing Results: Clock Frequency: 64 MHz
Context switch                               : 372 cycles ,  5766 ns
Interrupt latency                            : 100 cycles ,  1550 ns
Tick overhead                                :   0 cycles ,     0 ns
Thread Creation                              :1400 cycles , 21700 ns
Thread cancel                                :3388 cycles , 52514 ns
Thread abort                                 :3732 cycles , 57846 ns
Thread Suspend                               :1116 cycles , 17298 ns
Thread Resume                                :1056 cycles , 16368 ns
Thread Yield                                 :1160 cycles , 17980 ns
Thread Sleep                                 :1964 cycles , 30442 ns
Heap Malloc                                  : 792 cycles , 12276 ns
Heap Free                                    : 864 cycles , 13392 ns
Semaphore Take with context switch           :2012 cycles , 31186 ns
Semaphore Give with context switch           :1492 cycles , 23126 ns
Semaphore Take without context switch        : 120 cycles ,  1860 ns
Semaphore Give without context switch        : 624 cycles ,  9672 ns
Mutex lock                                   : 624 cycles ,  9672 ns
Mutex unlock                                 : 736 cycles , 11408 ns
Message Queue Put with context switch        :1384 cycles , 21452 ns
Message Queue Put without context switch     : 304 cycles ,  4712 ns
Message Queue get with context switch        :1652 cycles , 25606 ns
Message Queue get without context switch     : 312 cycles ,  4836 ns
MailBox synchronous put                      :2252 cycles , 34906 ns
MailBox synchronous get                      :1656 cycles , 25668 ns
MailBox asynchronous put                     : 512 cycles ,  7936 ns
MailBox get without context switch           :1228 cycles , 19034 ns
Drop to user mode                            :1336 cycles , 20708 ns
User thread Creation                         :3184 cycles , 49352 ns
Syscall overhead                             : 228 cycles ,  3534 ns
Validation overhead k object init            : 240 cycles ,  3720 ns
Validation overhead k object permission      : 232 cycles ,  3596 ns
Timing Measurement  finished
PASS - main
===================================================================

Environment (please complete the following information):

• OS: Fedora29
• Toolchain: Zephyr SDK v.10.0
• Commit SHA or Version used: 14db4ee

Additional context
Add any other context about the problem here.
I am comparing the values to the last result I have for a 12/mar/19 commit [af4ecea]

[carlescufi]

@pizi-nordic @nashif @arun1joshi could this be a regression with the performance of the compiler?

[nashif]

this can be a lot of things, I would like to see the data from a "good run" first.

[nashif]

no update for a while, not sure this is a bug, numbers change all the time.