shaiko
Advanced Member level 5
Some processor architectures implement branch prediction as a hardware circuit implemented in the device itself - while other rely on the complier to do this.
My question - why not always rely on the compiler?
It saves silicon, saves power and makes the processor less complicated and less expensive.
Sure, it comes at the price of higher compiler complexity - but if the performance is the same...why make the hardware complex when software can do it instead?
My question - why not always rely on the compiler?
It saves silicon, saves power and makes the processor less complicated and less expensive.
Sure, it comes at the price of higher compiler complexity - but if the performance is the same...why make the hardware complex when software can do it instead?