Re: fix rtl-opt/15289, part 1

[Eric Botcazou <ebotcazou at libertysurf dot fr>]

> This is the beginning of a fix for rtl-opt/15289.
>
> The end to which I am working is to remove gen_realpart/gen_imagpart.

Note that I'm still seeing the execution failures (wrong code) on the SPARC 
that prompted me to post my message about representable subregs, related to 
unaligned complex numbers.  The problem now is that store_bit_field silently 
turns CONCATs into paradoxical subregs here:

  /* If VALUE is a floating-point mode, access it as an integer of the
     corresponding size.  This can occur on a machine with 64 bit registers
     that uses SFmode for float.  This can also occur for unaligned float
     structure fields.  */
  orig_value = value;
  if (GET_MODE_CLASS (GET_MODE (value)) != MODE_INT
      && GET_MODE_CLASS (GET_MODE (value)) != MODE_PARTIAL_INT)
    value = gen_lowpart ((GET_MODE (value) == VOIDmode
			  ? word_mode : int_mode_for_mode (GET_MODE (value))),
			 value);

I'll wait until after your series of patches is in to further investigate.

-- 
Eric Botcazou