Without predication, parallelism would be impossible. Instead of waiting for each section of a complex calculation to finish, it is faster if the processor can predict the outcome and proceed on the basis of that prediction. These prediction points are called branches and current processors try to guess which branch to take. If it predicts right, the whole calculation is validated. If it predicts wrong, the string has to be thrown out and the calculation starts over. The Itaniumâ„¢ processor family architecture minimizes wasted calculations not by guessing but by taking both possible paths to the next branch, where it follows both branches again. When it comes to the right result it drops the other branch path that it doesn't need, keeps the branch that it does and it continues on with the calculation. This is truly parallel.