A simple test to measure the speed of copying data to write-combined memory, regular memory and uncached memory on Windows.
For each type (regular, write-combined and uncached) I test the performance over multiple iterations over the same allocated buffer.
I report the first run timings as well as an average run timing that excludes the first run.
The first run timings are interesting because they show the overhead of first-use of each of the 3 types of buffers, which 
can be substantial for write-combined and uncached buffers.