This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [Committed] signbitl RTL expansion for Darwin's long double


On Mon, Mar 14, 2005 at 09:24:18PM -0700, Roger Sayle wrote:
> *************** const struct real_format c4x_single_form
> *** 4398,4403 ****
> --- 4415,4421 ----
>       -126,
>       128,
>       -1,
> +     -1,
>       false,
>       false,
>       false,
> *************** const struct real_format c4x_extended_fo
> *** 4416,4421 ****
> --- 4434,4440 ----
>       -126,
>       128,
>       -1,
> +     -1,
>       false,
>       false,
>       false,

Here the numbers are 23 and 31 respectively.

> *************** expand_copysign (rtx op0, rtx op1, rtx t
> *** 2866,2872 ****
>     if (fmt == NULL || !fmt->has_signed_zero)
>       return NULL_RTX;
> 
> !   bitpos = fmt->signbit;
>     if (bitpos < 0)
>       return NULL_RTX;
> 
> --- 2866,2872 ----
>     if (fmt == NULL || !fmt->has_signed_zero)
>       return NULL_RTX;
> 
> !   bitpos = fmt->signbit_rw;
>     if (bitpos < 0)
>       return NULL_RTX;

Here we should move this check.  We should use the ro version when
expanding via expand_copysign_absneg and the rw version when using
expand_copysign_bit.  This would then allow copysigntf3 to be deleted
from the rs6000 port, and expand_copysign_absneg made static.



r~


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]