If you ignore linear/non-comparison algorithms like radix or bucketsort. Those are much harder to optimize to your particular problem but should result in significantly lower run times when properly adapted.