[Patch Ping] [RFC] Alias export patch

[Daniel Berlin]

Richard Guenther wrote:
> On 6/5/06, Daniel Berlin <dberlin@dberlin.org> wrote:
>> Andrey Belevantsev wrote:
>>> Richard Guenther wrote:
>>>> I wonder how/if you deal with the problem that the tree loop optimizers
>>>> (ivopts mostly) generate new pointers as induction variables but we don't
>>>> run may_alias after them, so possibly all interesting (performance wise)
>>>> pointers do not have updated points-to information?  Do your numbers
>>>> (not performance) improve if you specify -fno-tree-loop-optimize?
>>> Since the patch was created, we wanted to insert an additional
>>> pass_may_alias before going out of ssa to cope with this problem.
>>> However, as ivopts does not keep the points-to information updated, this
>>> does not work -- gcc ICEs in verify_ssa.
>> Last time i looked at this issue, it was because none of the aliasing
>> infrastructure (tree-ssa-alias, tree-ssa-structalias) was updated when
>> MEM_REF was introduced, and thus, doesn't treat MEM_REF like a pointer
>> dereference.
>>
>> This causes some incorrect things to happen, and verify_ssa notices.
> 
> Hmm, it should be not too difficult to fix it.  Fixing it to produce useful
> points-to sets for TARGET_MEM_REF may be another thing though.
> 
Probably the best you could do would be to try to get the range
information computed by something else, and use it as an upper bound to
what parts of the object referred to in the TARGET_MEM_REF are being
accessed.

--Dan