Comment #11 on issue 576 by tudor.gi...(a)gmail.com: Border of children
depends on another shape in a complex shape
http://code.google.com/p/moose-technology/issues/detail?id=576
I expressed it badly. What I meant to say was that maybe the performance
would not be so bad if the cache with the bounds would still remain a
one-level-of-indirection-cache, and only use the two-level-indirection
(dictionary in dictionary) for the individual properties.
Actually, in VW we only had the cache for the bounds, and the properties
were always computed on the fly, and it was still fine.
So, I come back to the original question, what is the penalty introduced by
your fix?