On Tue, 7 May 2013 15:12:43 +0100, Walter Briscoe wrote:

That is interesting - mainly from the ability to time small pieces of
code. How do you do it? I am guessing you get the difference in time
between running the 2 pieces of code a large number of times and infer
that division gets a reasonable approximation to the time for a single
call. That inference might be false. Optimisers can confuse. ;)

I use the high precision event timer (HPET). And I got similar results running just a single instance as well as 100 iterations.


I rarely worry about the time taken to run small pieces of code.
My code usually gets data from the Internet and is IO-bound.
Each round trip to the net takes of the order of a second.
I have yet to work out how to run several transactions in parallel.

I rarely worry about it, but I have had some projects which process hundreds of thousands of rows, so speeding up even short segments can have a significant impact.