Shrinking the kernel with link-time optimization

The Linux kernel already does all sorts of tricks to encourage the compiler to inline code where it makes sense and to compute static expressions at compile time. So, I would suspect the main performance benefit is going to be from the smaller binary size. That's nothing to sneer at. More memory that is available for other uses is always a benefit. And more compact code is going to help with better utilization of the instruction cache.

In other words, there almost certainly will be a performance improvement. But my gut feeling is that the numbers are going to be so small, most users would not be able to notice.