Sorry for such a late reply. I tested it and the replace array element version is 3-4 times faster. I would also like to know why if I add iteration parallelism to the autoindex version it gets slower although this is exactly the case when it should benefit (I am aware of the possible overhead caused by splitting the array and then joining it together but this couldn't cause it)

Message Edited by ceties on 04-06-2010 04:41 PM
LV 2011, Win7