devkitPro

code is being run from cache - which is fast. if it fits in cache then it will already be fast. if the loop is too big to fit in cache then there will be cache misses/loads. if the the loop has a high iteration then the overhead of repeatedly filling the cache will slow things down.

Thank you, thats so obvious now!!

This is a good article about cache. Its a bit old and it more PC oriented but describes the basics well.
http://arstechnica.com/gadgets/reviews/ ... aching.ars