Use the 32-bit APIC calibration code - it's more mature.
Signed-of-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>