On Tue, Jul 08, 2025 at 08:15:30AM +0200, Richard Biener wrote:
> > No. Because (most of the) pointer conversions are useless, if you have
> > _2 = &something.fam[0];
> > ...
> > _4 = &something.fam;
> > _5 = .ACCESS_WITH_SIZE (_4, ...);
> > (or whatever else where _4 has the carefully chosen pointer type for the
> > behavior of the IFN), if e.g. SCCVN finds out that _2 and _4 must have the
> > same value, regardless of different type, if useless_type_conversion_p
> > (TREE_TYPE (_4), TREE_TYPE (_2)), it will happily replace _4 with _2, and
> > all of sudden you have a different type.
>
> You could look at the type of the LHS though, unless there are two „same“
> AWS with only different return type and we are able to CSE them.
True. But I'm arguing that instead of the first argument's type/return type
pointee we want to know its original TYPE_SIZE_UNIT (so value by which what
the second argument points to should be multiplied) and a pointer type which
can be used in the dereference of the second argument.
That is I think 4 arguments,
REF_TO_OBJ (ptr), REF_TO_SIZE (ptr), ELTSZ (sizetype), flags (INTEGER_CST,
type original type of REG_TO_SIZE, bitfield of the various flags).
Jakub
