On Wed, 2 Nov 2016, Richard Biener wrote:
>
> This fixes the alias check in terminate_all_aliasing_chains -- the
> base we use for the hash table indexing does not constitute an object
> that aliases all stores in the chain (consider for example the MEM_REF
> handling which strips the offset
This fixes the alias check in terminate_all_aliasing_chains -- the
base we use for the hash table indexing does not constitute an object
that aliases all stores in the chain (consider for example the MEM_REF
handling which strips the offset completely).
I've refactored data structures a bit in
