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: [tree-ssa] Maintaining/representing def-use/use-def information


On Mon, 2003-12-15 at 16:11, Zdenek Dvorak wrote:
> Hello,
> 
> > > > Ultimately, I dont see much difference other than implementation
> > > > details...
> > > 
> > > ... and about three times more memory spent, I would estimate. For a
> > > single assignment statement, we have:
> > 
> > Which is an implementation detail, not an architecture one.
> 
> on contrary, I see this as a major architectural decision -- identifying the
> variable with the statement that defines it is a very interesting idea.
> 

And did I not say that that one thing was exactly the difference? and
that the rest of the differences, like memory consumption was an
implementation detail?

And we could make that architectural change if we so chose. We could
define an SSA_NAME to have an impllcit MODIFY easily enough, and point
direcdtly to the operational part of the modify. It wouldnt be that big
a deal. Its just not worth doing until we do tree-compaction, whatever
that turns out to be.  For now we work with trees.

Andrew


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