Hands-On High Performance Programming with Qt 5

Speeding up programs with threads

Now, we have N cores and would naturally expect that our program can be made to run N times faster. However, Amdahl's law sensibly states that this isn't possible, as the total speedup will always be limited by this part of the program that cannot be parallelized. Gustaffson's law somehow refutes this pessimistic outlook, stating that bigger systems will be able to solve bigger problems, while Amdahl's law assumes fixed problem sizes. We will spare the reader the exact formulas, but the conclusion from this theoretical work is that the possible speedups depend on the structures of the problems at hand.

After that theoretical underpinning, let's turn our attention to more practical matters and look at how ...