Small change causes LE count to change from 108 to 2660

Fixed, in my case. The problem was the line:

LogBitsize(ClkFrequency/baud) was implying an lpm_divide Megafunction. Changed to (LogBitsize(ClkFrequency) - LogBitsize(baud)), and problem resolved. Completely.

I say that the problem resolved completely because two lpm_div megafunctions were being implied, the second by a different line, according to Quartus. But they *both* went away with the changes above? This highlights the fact that there doesn't seem to be any documentation at what 'implies' an lpm_div Megafunction (or any other megafunction). Nor is there any dox on how to *not* imply these explicitly in Quartus. In general, this was huge waste of time because Quartus inferred something that wasn't there and made a 32 bit barrel shifter.

How do I stop Quartus from wasting my time? It would be much useful if Quartus *asked* if that was my intention...