[Bug tree-optimization/30911] VRP fails to eliminate range checks in Ada code

["rguenth at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

------- Comment #46 from rguenth at gcc dot gnu dot org  2008-03-28 19:18 -------
Ok, I didn't really expect that ;)

Some new background information.  With the middle-end type-system work we now
omit conversions from sub-types T' to their base-types T.  Thus we have
the three cases

  T' sub;
  T  x;

  x = sub;      (1)
  sub = (T')x;  (2)
  x = VIEW_CONVERT_EXPR <T>(sub);  (3)

where VRP for the simple copy (1) does not restrict x value range based
on the T's TYPE_MIN/MAX_VALUE (but it should).  For (2) the same is
true (though the conversion is _not_ truncating for out of bound values,
so I am not sure if this doesn't break something).  But for both (1)
and (2) holds that a variable of type T' can be assumed to have a
value range restricted by its TYPE_MIN/MAX_VALUE.

Case (3) is special in that it is a propagation barrier, thus x will get
a varying value range.  We could do better here if we knew that subs
value range was not restricted by its TYPE_MIN/MAX_VALUE only.

I don't know if this is really the best setup to optimize Ada range checks,
or if we should always fall back to the base types TYPE_MIN/MAX_VALUE
range and use the type defined range of the sub-types T' only in special
places like for example for the initial value-range of T' variables
default definitions (thus uninitialized values and function parameters
where if I understand correctly in Ada the caller ensures that the
value is in range).  Of course this wouldn't work very well in
combination with inlining.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30911