James Mastros sent the following bits through the ether: > Perhaps it's just too late at night, but if this cange makes it better, > tighter code, shouldn't it make the MOPS numbers look better, not worse?
Right: the older code did the average of three ops (eq, add, and branch), and the new code does the average of two ops (add, lt). The branch is the simplest in terms of complexity, so taking it own makes the average slower. That's why the MOPS result doesn't mean much, but the new code is simpler and *faster*, which is why I suggested it. (just not faster per op, you see) Hope that makes it clear, Leon -- Leon Brocard.............................http://www.astray.com/ Nanoware...............................http://www.nanoware.org/ .... Error 404: There is no spoon