Re: Replace rename_ssa_into_ssa with update_ssa

[Diego Novillo <dnovillo at redhat dot com>]

On Sat, Apr 23, 2005 at 04:52:08PM +0200, Zdenek Dvorak wrote:

> no; sorry, I misunderstood the code a bit.  Just one question -- would
> not it be possible to have update_ssa call directly inside
> tree_duplicate_loop_to_header_edge, instead of having the caller of the
> function do it?  It seems less error-prone to me.
> 
True, though at the time I was trying to minimize calls to
update_ssa.  If you want to experiment with moving update_ssa
into the duplication code, then great.  I agree that it will be
less error prone.  Though, times may go up.  In general we can
withstand calling the duplicator without updating the SSA form on
every iteration.


> Well, but there is not that much of my code left.  It's not question
> of credit for me, I just don't want to risk someone asking me about
> the code, or even wanting me to fix the possible bugs in a code I do
> not understand entirely.
> 
Fair enough.  I'll remove it then.

> Great, that's one of the things I wanted to hear for quite some time.
> I would be even happier to hear this opinion before I was forced to waste
> my time by working on the specialized "hand-code" SSA updating
> (instead of speeding up and improving the generic SSA form updating as
> I wanted to originally), but even this is a progress :-).
> 
Well, remember that update_ssa is not magic.  It's only a
relatively mindless search and replace pass.  Each user has to
tell it exactly what names need to be replaced with new names.
It can only add new PHI nodes for the old names, it will add
PHI nodes needed at the fringes of the region to be updated.

All it does is provide an interface to allow passes to keep the
SSA form up-to-date.  Which has been the goal all along.  We do
seem to have different ways of getting there, though.  The fact
that days only have 24 hours doesn't help either ;)


Diego.