TL;DR
Would it not be great that we change the branch prediction implementation to switch on/off automatically(with every new process)?
This would be difficult to trace or even implement but I'm just guessing.
The CPU stops using branch prediction if a lot of it's predictions are failing but starts again after some time.
Comments
TL;DR Would it not be great that we change the branch prediction implementation to switch on/off automatically(with every new process)? This would be difficult to trace or even implement but I'm just guessing. The CPU stops using branch prediction if a lot of it's predictions are failing but starts again after some time.