Occasionally, I've discovered that the result of synthesis (default effort) depends on the instances order. I can bring a complete example but in short the illustration is following:
-- 24 LUTs architecture RTL is begin U1: entity .. U2: entity ..
-- 27 LUTs architecture RTL is begin U2: entity .. U1: entity ..
This fact complicates comparition of different design configurations. Is it normal? I've checked Synplify -- it always gives the same results irrespecively to the units order.