Peter, this could be good, can you modify your sample to run for say a minute and then we'll ask all compiler users in the group to try their compiler for size and speed.
I agree that it is (still) the case that a human can be smarter than the compiler. I also do agree that in MOST cases the smallest code runs the fastest. However, this seems to be an area where the algorithms for compression differ the most.
If you think this is a good idea, start a new thread for it.
viel spass
Erik