Looking for more library optimization patterns

[Stanislav Blinov]

Being somewhat involved in aforementioned thread, I second that. 
There seems to be no logical reason to insert a call 
dSliceOpAssignOrWhateverItIsCalled when all the data is actually 
there. At least, in -release -O3 builds :)

Granted, this may be tricky for arbitrarily-typed arrays, but for 
builtins with statically known array sizes? That's a definitive 
win.