[PATCH] i386: Use fldt instead of fld on e_logl.S

[Adhemerval Zanella Netto]

On 04/08/22 18:35, H.J. Lu wrote:
> On Thu, Aug 4, 2022 at 2:31 PM Fangrui Song <maskray@google.com> wrote:
>>
>> On Thu, Aug 4, 2022 at 2:24 PM H.J. Lu <hjl.tools@gmail.com> wrote:
>>>
>>> On Thu, Aug 4, 2022 at 2:00 PM Adhemerval Zanella
>>> <adhemerval.zanella@linaro.org> wrote:
>>>>
>>>> Clang cannot assemble fldt in the AT&T dialect mode.
>>>> ---
>>>>  sysdeps/i386/i686/fpu/e_logl.S | 4 ++--
>>>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/sysdeps/i386/i686/fpu/e_logl.S b/sysdeps/i386/i686/fpu/e_logl.S
>>>> index 63183ac544..5d53f3bb22 100644
>>>> --- a/sysdeps/i386/i686/fpu/e_logl.S
>>>> +++ b/sysdeps/i386/i686/fpu/e_logl.S
>>>> @@ -43,7 +43,7 @@ ENTRY(__ieee754_logl)
>>>>         fsubl   MO(one)         // x-1 : x : log(2)
>>>>  5:     fld     %st             // x-1 : x-1 : x : log(2)
>>>>         fabs                    // |x-1| : x-1 : x : log(2)
>>>> -       fld     MO(limit)       // 0.29 : |x-1| : x-1 : x : log(2)
>>>> +       fldt    MO(limit)       // 0.29 : |x-1| : x-1 : x : log(2)
>>>>         fcomip  %st(1)          // |x-1| : x-1 : x : log(2)
>>>>         fstp    %st(0)          // x-1 : x : log(2)
>>>>         jc      2f
>>>> @@ -76,7 +76,7 @@ ENTRY(__logl_finite)
>>>>         fsubl   MO(one)         // x-1 : x : log(2)
>>>>         fld     %st             // x-1 : x-1 : x : log(2)
>>>>         fabs                    // |x-1| : x-1 : x : log(2)
>>>> -       fld     MO(limit)       // 0.29 : |x-1| : x-1 : x : log(2)
>>>> +       fldt    MO(limit)       // 0.29 : |x-1| : x-1 : x : log(2)
>>>>         fcomip  %st(1)          // |x-1| : x-1 : x : log(2)
>>>>         fstp    %st(0)          // x-1 : x : log(2)
>>>>         jc      2b
>>>> --
>>>> 2.34.1
>>>>
>>>
>>> I don't think this is correct since there are
>>>
>>>         .type limit,@object
>>> limit:  .double 0.29
>>>
>>> It should be fldl.
>>>
>>> --
>>> H.J.
>>
>> If I am not mistaken, the existing fld has a bug as it is treated as flds
> 
> True.  It doesn't seem to cause any issue.
> 
>> fld limit@GOTOFF(%edx)    # treated as flds
>> flds limit@GOTOFF(%edx)
>> fldl limit@GOTOFF(%edx)
>> fldt limit@GOTOFF(%edx)
>>
>> .section .rodata.cst8,"aM",@progbits,8
>> .p2align 3
>> limit: .double 0.29

fldl works and it see the best option.  Ok with this change?