This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: SRA and DECL_ABSTRACT_ORIGIN


"Giovanni Bajo" <giovannibajo@libero.it> writes:

| Gabriel Dos Reis <gdr@integrable-solutions.net> wrote:
| 
| >> Someone else was talking about this the other day on IRC and I
| >> told them how I'd like it solved.
| >
| > I do not follow IRC sessions.  Could you outline your plan again?
| 
| 
| The plan is to store a backpointer to the original decl within the generated
| SRA variable, and use it for diagnostic purposes.

Aha, this is what I was meaning to do with DECL_ABSTRACT_ORIGIN (but
RTH suggested it would not be apporpriate).

| For instance, in the case of
| "a$i" we would be storing a pointer to the original INDIRECT_REF from which the
| scalarized version was created. This is similar to what DannyB did with his
| DECL_DEBUG_ALIAS_OF (http://gcc.gnu.org/ml/gcc-patches/2004-12/msg01883.html)
| but I think vindex isn't free in all the entities SRA generates.

Thanks!

-- Gaby


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]