This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: SSA operand interface change proposal
- From: Andrew MacLeod <amacleod at redhat dot com>
- To: Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>
- Cc: gcc mailing list <gcc at gcc dot gnu dot org>, Diego Novillo <dnovillo at redhat dot com>, Jeff Law <law at redhat dot com>, Richard Henderson <rth at redhat dot com>
- Date: 15 Jun 2004 14:56:10 -0400
- Subject: Re: SSA operand interface change proposal
- References: <1087320259.14876.187.camel@p4> <20040615183848.GA32115@atrey.karlin.mff.cuni.cz>
On Tue, 2004-06-15 at 14:38, Zdenek Dvorak wrote:
> Hello,
>
> > As best I can tell, the 'ivuse' structure used thorughout
> > tree-ssa-loop-ivopts.c contains a pointer to a tree which can be either
> > a use, a def, or an expression of some sort. If we proceed with an
> > interface like this (which I think we ought to), this hunk of code will
> > need to be reworked. I dont want to make this unbearably difficult
> > however. I guess the ivuse structure could have all three pointers
> > (use_operand_p, def_operand_p, and tree *) , and a flag indicating which
> > field it uses, but maybe there is a better solution.
>
> in fact what *op_p refers to is determined by the enum use_type type
> field. So you can just make it an union and updating the code to
> reflect this should be trivial.
Then in fact there may have no issue at all :-) I'll provide an LNO
patch along with the mainline patch which does everything except for
those bits then.
Andrew