native_posix is now deprecated. Let's switch the filter and build
checks to native_sim instead.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
The benchmark sample did not contemplate message subscribers. It adds
the msg subscribers and improves the organization of the project.
Signed-off-by: Rodrigo Peixoto <rodrigopex@gmail.com>
The benchmark duration is in nanoseconds, and it logs the duration to
the console in seconds. The conversion is wrong for scenarios where
the execution lasts less than 100 milliseconds. Fix that by adding
leading zeros while printing duration.
Signed-off-by: Jilay Pandya <jilay.pandya@zeiss.com>
The benchmark sample was not suitable for constrained devices.
With u32, the maximum accumulated time was about 4 seconds.
Some devices take more than 4 seconds to finish the benchmark.
This commit fixes the time counting by replacing the variables
and calls to 64 bits and removes all the dynamic allocation
memory on the sample to keep that as simple as possible. It
adjusts the documentation and the tests as well.
Signed-off-by: Rodrigo Peixoto <rodrigopex@gmail.com>
For the POSIX arch, this test relays on the native_posix RTC.
But that does not exist in the nrf52_bsim.
And without something to measure actual execution time,
it does not make sense to benchmark in the POSIX arch.
So disable this test for this platform,
and add a clear build error to warn users.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
Current implementation has obvious synchronization issue with
global counter variable which is updated (RMW) in multiple threads
without any locking.
Replace regular variable by atomic type. As the maximum possible
value is much less than INT32_MAX it's OK to replace the original
uint64_t by (32-bit) atomic_t.
Signed-off-by: Eugeniy Paltsev <Eugeniy.Paltsev@synopsys.com>
Signed-off-by: Evgeniy Paltsev <PaltsevEvgeniy@gmail.com>
Because posix emulated execution is significantly faster than posix
emulated timer and benchmark finishes before even the first tick occurs.
Signed-off-by: Alp Sayin <alp.sayin@amd.com>
Needed because benchmark finishes faster than a tick, especially on
emulated platforms where tick rate is 100Hz. Output units are also updated
to print numbers with less digits.
Signed-off-by: Alp Sayin <alp.sayin@amd.com>
Stack sizes need to be dependent on something arch specific such as
idle stack size because they're smashing their stacks.
Signed-off-by: Alp Sayin <alp.sayin@amd.com>
