We ignore the bottom 4 bits of the X resolution, so we should
round X resolutions up to the nearest multiple of 16.
Signed-off-by: Russell King <rmk@arm.linux.org.uk>