RFA: loop optimization bug

[Dale Johannesen]

I'm seeing a case where the tree optimizers produce a loop with a 
branch into its
middle, like this (CFG at first RTL dump):

Loop begin
7->9
8->9
9->8,10
10->11,13
11->12,13
12->13
13->14,15
14->8
Loop end
15

The ("old") RTL loop optimizer believes that loops either fall through 
into the top,
or start with a branch to the bottom, and is getting confused when 
handed this loop.
Before I get too deep, is it the intent of the tree optimizers to 
produce only CFGs
that conform to the loop optimizer's expectations, or should the RTL 
opts be handling
this?  Thanks.