Re: [PATCH] Invalidate in cselib sp after processing frame_pointer_needed fp setter (PR rtl-optimization/54921, take 2)

[Richard Henderson]

On 11/19/2012 12:39 PM, Jakub Jelinek wrote:
> +  if (reload_completed
> +      && frame_pointer_needed
> +      && RTX_FRAME_RELATED_P (insn)
> +      && fp_setter_insn (insn))
> +    cselib_invalidate_rtx (stack_pointer_rtx);
...
>                 if (fp_cfa_offset != -1
>                     && hard_frame_pointer_adjustment == -1
>                     && RTX_FRAME_RELATED_P (insn)
> -                   && fp_setter (insn))
> +                   && fp_setter_insn (insn))

I thought we talked about moving the RTX_FRAME_RELATED_P test inside
the function?  Since it's already there anyway...


r~