Re: PATCH: Rest of Intel Atom optimization

[Uros Bizjak <ubizjak at gmail dot com>]

H.J. Lu wrote:
> On Mon, Apr 6, 2009 at 11:14 PM, Uros Bizjak <ubizjak@gmail.com> wrote:
>   
> > > Can extract_insn_cached help here?
> > >       
> > You can look at ix86_attr_length_address_default, there is
> > get_attr_type, used to detect TYPE_LEA.
> >
> >     
>
> Here is the rest of Intel Atom optimization. OK for trunk?
>   

> +static int
> +distance_non_agu_define (rtx op1, rtx op2, rtx insn)
> +{
> +  unsigned int regno1 = REGNO (op1);
> +  unsigned int regno2 = REG_P (op2) ? REGNO (op2) : (unsigned int) -1;

Hm, not true_regnum in the lines above?

(BTW: You can use INVALID_REGNUM for invalid regnumbers).

> +  basic_block bb = BLOCK_FOR_INSN (insn);
> +  int distance = 0;
> +  df_ref *def_rec;

...

> +		  {
> +		    insn_type = get_attr_type (prev);
> +		    if (insn_type != TYPE_LEA)
> +		      goto done;
> +		  }

...

> +done:
> +  /* Restore recog_data which may be modified by get_attr_type.  */
> +  extract_insn_cached (insn);
> +  return distance;


This can't be right. You are restoring recog data on a different insn 
that was processed by get_attr_type. I think that extract_insn_cached 
should immediately follow get_attr_type in order to restore recog data.

Uros.