CODEWITHSUNDEEP
c++linuxmultithreading
Ron
Ron

Reputation: 533

C++, two just writers threads without lock

I have 3 threads.

I need to verify there is no data corruptions. If data is from thread A or from thread B, so it is OK.

But, if the result at the variable is data that is no from thread A neither not from thread B so it is corrupted

As i thought because that time_t can be write at one assembly operation so it will be OK.

Is that right?

Thanks

Upvotes: 0

Views: 103

Answers (2)

Angew is no longer proud of SO
Angew is no longer proud of SO

Reputation: 171127

If you have two potentially concurrent writes(1) to a shared location without synchronisation between them, you have Undefined Behaviour and anything can happen. So this is a bad idea.

Note that even if you can't imagine bad execution based on your knowledge of the underlying architecture, the optimiser can get in the way, because it is allowed to assume UB doesn't happen.

(1) Note that even one writer and one reader without synchronisation is exactly the same UB.

Upvotes: 6

Daniel Langr
Daniel Langr

Reputation: 23497

time_t is an arithmetic type, which can basically be integer or floating-point. You can therefore try to define an atomic time_t as (if C++11 is available):

using atomic_time_t = std::atomic<time_t>;

This is portable as long as time_t is trivially copyable, which likely is due to being an arithmetic type. It will work even if not being lock-free. (Floating-point specializations for std::atomic adds C++2a.)

Example:

using atomic_time_t = std::atomic<time_t>;

int main() {
  std::cout << atomic_time_t{}.is_lock_free() << std::endl;
}

This prints out 1 with GCC 7.2 x86_64 compiled with -std=c++11.

If you need to be 100% sure about portability, or don't have C++11, protect accesses (both reads and writes) to your time_t variable by mutex provided by your threading library.

Upvotes: 4

Related Questions