Why does the latency to perform the shared memory operations decreases with increase in the number of iterations performed?

Question

I am using shared memory for transferring data between two process, using boost::interprocess::managed_shared_memory to create and manage shared memory.

I was measuring the latency for the 2 operations (fetch and update) -

update operation - writes to the array in shared memory. fetch operation - reads the shared memory variable.

The array size is 16 bytes. I've created utility functions to update to and fetch the value from a purticular array location of the shared memory variable.

To measure the latency - I get the current time in microseconds and store it a variable. I get the timestamp before the operation and after the operation. Simply subtracting the time variables will the give the latency for the current iteration. Then I calculate the avg of the required number of iterations.

Initially for 1 iteration the latency is around 100 - 300 microseconds and it gradually decreases as I increase the iteraiton to 10, 100, 1000, 10000, .... In the end for 1000000 iterations the latency drops to 30 - 60 microseconds.

Iterations Latency for Update(microseconds) Latency for Fetch(microseconds) 1 254 347 10 84.9 153.5 100 71.16 55.43 1000 37.733 41.47 10000 32.52 37.1141 100000 34.2387 34.2009 1000000 33.4629 34.537

To measure the latency - I get the current time in microseconds and store it a variable. I get the timestamp before the operation and after the operation. Simply subtracting the time variables will the give the latency for the current iteration. Then I calculate the avg of the required number of iterations.

Initially for 1 iteration the latency is around 100 - 300 microseconds and it gradually decreases as I increase the iteraiton to 10, 100, 1000, 10000, .... In the end for 1000000 iterations the latency drops to 30 - 60 microseconds.

This is the results I'm getting, Iterations 1 10 100 1000 10000 100000 1000000 Latency for Update(microseconds) 254 84.9 71.16 37.733 32.52 34.2387 33.4629 Latency for Fetch(microseconds) 347 153.5 55.43 41.47 37.1141 34.2009 34.537

Why does the latency decreases with increase in number of iterations