Currently, we calculate the refresh rate of a detailed timing by
dividing its pixel clock by 1000. This doesn't make any sense.
To get the refresh rate we must calculate the total number of pixels
in one complete frame (active + blanking) and divide the pixel clock by
that value. Thus, this is the correct formula:
As the pixel clock has steps of 10 kHz, this might result in ugly
decimal values. It seems common to use rounding to two digits after the
decimal point.
From the first commit message: