Re: RFA: Fix PR38785

[Joern Rennecke <amylaar at spamcop dot net>]

Quoting Paolo Bonzini <bonzini@gnu.org>:

> > +ftree-pre-partial-partial
> > +Common Report Var(flag_tree_pre_partial_partial) Optimization
> > +In SSA-PRE optimization on trees, enable partial-partial   
> > redundancy elimination.
> > +
> > +ftree-pre-partial-partial-obliviously
> > +Common Report Var(flag_tree_pre_partial_partial_obliviously) Optimization
> > +In SSA-PRE optimization on trees, enable partial-partial redundancy
> > +elimination without regard for the cost of the inserted phi nodes.
>
> Something shorter like -ftree-ppre and -ftree-ppre-always?

Fine with me.

> BTW, the PRE at -Os part is a regression, I have
>
>       if (!optimize_size)
>         {
>           /* PRE tends to generate bigger code.  */
>           flag_tree_pre = 1;
>         }

There is:

  if (optimize_size)
    {
      /* Inlining of functions reducing size is a good idea regardless of them
         being declared inline.  */
      flag_inline_functions = 1;

      /* Basic optimization options at -Os are almost the same as -O2.  The
         only difference is that we disable PRE, because it sometimes still
         increases code size.  If the user want to run PRE with -Os, he/she
         will have to indicate so explicitly.  */
      if (optimize > 2)
        optimize = 2;
      flag_tree_pre = 0;

However, tree-ssa-pre.c still performs fre even if not pre, and
I think it still leaves a "*.pre" dump file in this case.