Re: [ast-optimizer-branch] [patch]: SSA for trees

[Daniel Berlin <dan at cgsoftware dot com>]

Diego Novillo <dnovillo@redhat.com> writes:

> On Mon, 23 Jul 2001, Daniel Berlin wrote:
> 
>> I changed the main flag  from -ftree-ssa to -ftree-opt, and
>> -ftree-opt-ssa turns on SSA based tree optimizations (and
>> -ftree-opt-loop turns on the loop optimizations, and
>> -ftree-opt-callgraph does the call graph tracking)
>> 
> Good idea.  Thanks.
> 
>> This is all, if you don't mind, of course. The ChangeLog entry has
>> your name first, mine second, and says that everything not in
>> tree-loop.[ch] and tree-cg.[ch] was done by you.
>> 
> I would rather keep them separate for the time being.  I still
> haven't had a chance of evaluating the patch you sent me over the
> weekend.
Okay.
I'll seperate the changelog entries, etc.

On a note about the patch over the weekend, I noticed that trying to
just add a flag for MU nodes doesn't work. 
You are doing some things that cause you to get lucky because of the
layout of the structs.
(Try inserting an int as the second member of the phi node struct, and
it'll crash in the DFA stuff. This is, for one thing, because it
thinks VARDEF_IMMEDIATE_USES is okay on phi nodes, and it happens to
get lucky hitting the phi_chain member, which is also a varref_list.)


> 
> A related question, how is the evaluation process in the branch?
No clue.
> Should all patches be ultimately approved by Nathan or are we
> going to distribute maintainership for the different sub-projects
> in the ast branch?
Honestly, I think the second might be a better idea than the first.
Especially since they should all be pretty easily seperable.

But it's Nathan's branch, so it's his call what he wants to do.

> 
> 
> Diego.

-- 
"There was a power outage at a department store yesterday.
Twenty people were trapped on the escalators.
"-Steven Wright