# Inference the time a process would take depending on the number of threads

I have several datasets consisting on:

• Number of threads n
• Start process time t1
• Stop process time t2
• Operations processed x

So each line of the dataset mean n threads processed x operations in t2-t1 time

When more threads are added, the processing time is reduced, because they run on parallel. However, they have locks between them, so the total time is not (t2 -t1)/n, but a bit more.

I would like to infer the time a process will take to do x operations depending on the thread number, some sort of "parallel factor" so t = x * n * factor gives me the estimated time. How could I achieve this?

-
Refer to Amdahl's law: en.wikipedia.org/wiki/Amdahl's_law – GEL Dec 14 '12 at 19:31