I had actually two interviews, one at the Technion during their demonstration and one at the company site.

The first one was very theoretical and very easy, few questions about OOP, what is abstract class, how do I use it, describing UML class diagram of my example, why do I need to hold pointers to the abstract class in the list and not objects, where do I use virtual functions, do I know singleton and how to implement it.

The second interview was more difficult and talked mostly about synchronization and client-producer problems. The questions there were very unclear, they wanted me to tell what can be the problems and how I could solve them. During the interview few times they told me wrong things and some times tried to push me in the wrong direction after I did give them correct answers!

I felt that the situation is not under my control, although I do know the material and give correct answers... So, I think the problem was rather with the way of the interview. After all, few times I showed them they don't know all the things and that they say wrong things and this I think wasn't too clever:)

Here I will refer only to the second interview because the first one was really trivial.

You have a plate with areas, some of them must be checked and the order of the check is given. Your machine has three operations: move to the cell, acquire the image, measure the image. How can you do it as efficiently as possible

Suppose, acquisition is much slower, what should be the k? How many threads?

And if vice versa? (here they also told me something wrong, they said that hardware is always slower than software and I told them that it is wrong!!!)

They asked me, based on function int SQRT(int) build function float SQRT2(int) that will return square root with three digits after the point.