We are hoping to use this to test changes that (should) have little impact on the nodes and speed of the program. We don't know if it will help but we wont' know if we don't study it. We probably would not use this for search changes. Recompiling does affect the speed and thus any test involving time introduces the problem we are trying to at least partially solve.Laskos wrote:I think the first rule of thumb is self-testing with identical time use. Then all others.Don wrote:
We do not confine our testing to a single method or time control
Kai
I don't think our computers are consistent either. I can run a timing 3 times and get a different result each time. I could probably help things a lot by customizing the kernel and turning off most services but we use our computers for many things and I don't want to do this just in the chance it might help some.