G++ optimization beyond -O3/-Ofast

Question

The Problem

We have a mid-sized program for a simulation task, that we need to optimize. We have already done our best optimizing the source to the limi

Answer 1

It is difficult to answer without further detail:

• what type of number crunching?
• what libraries are you using?
• what degree of paralelization?

Can you write down the part of your code which takes the longest? (Typically a tight loop)

If you are CPU bound the answer will be different than if you are IO bound.

Again, please provide further detail.