some tree-ssa vs mainline stats

[Dan Nicolaescu]

law@redhat.com writes:

  > In message <200306182344.h5INi34X020009@gremlin.ics.uci.edu>, Dan Nicolaescu writes:

  >  >Total number # of insns in .00.rtl 
  >  >generate-3.4.ii SSA   417674
  >  >generate-3.4.ii HEAD  728202
  >  >combine.i       SSA    41528
  >  >combine.i       HEAD   35524
  > I don't know if the reduction in insns for generate.ii is due to tree-ssa
  > optimizing better or doing something like not inlining as well.  I tend to
  > suspect the latter as most of the time it tends to produce more insns
  > than without tree-ssa.

Part of the reason is that tree-ssa is emitting a LOT less
NOTE_INSN_DELETED insns, which are about 1/3 of the total number of
insns for mainline when compiling the code from PR8361.

More observations, looking at the generate-3.4.ii.t15.optimized dump,
there's a lot of code that looks like this:

  {
    struct GATOM * const * const & __i;
    struct __normal_iterator<const GATOM* const*,std::vector<const GATOM*, std::allocator<const GATOM*> > > * const this;

    this = (struct __normal_iterator<const GATOM* const*,std::vector<const GATOM*, std::allocator<const GATOM*> > > * const)T.9559;
    {
      struct iterator<std::random_access_iterator_tag,const GATOM*,int,const GATOM* const*,const GATOM* const&> * this.9553;

      { // THIS BLOCK CAN BE ELIMINATED
        struct iterator<std::random_access_iterator_tag,const GATOM*,int,const GATOM* const*,const GATOM* const&> * const this;

        (void)0
      };
      this->_M_current = *T.9563
    }
  }


It seems that DCE is getting confused by "this", so it doesn't
eliminate the inner block of dead code. 

Also there's a lot of "if (1)" and "if (0)" code.

        --dan