This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Fix libgomp Fortran tests
On Fri, Sep 08, 2006 at 09:09:15AM -0700, H. J. Lu wrote:
> On Fri, Sep 08, 2006 at 07:16:23AM -0700, Steve Kargl wrote:
>> On Fri, Sep 08, 2006 at 03:27:00AM -0400, Jakub Jelinek wrote:
>>> On Fri, Sep 08, 2006 at 12:20:15AM -0700, Brooks Moses wrote:
>>>>>This can't be right. The test really needs an integer with all bits set,
>>>>>-huge(x) - 1 is Z'80000000'. That's what the OpenMP standard mandates
>>>>>that iand reduction var should be initialized to.
>>>>
>>>> Indeed. This should be x=-1. Or, possibly more clearly, x=not(0).
>>>
>>> Well, -1 only if you know it is two's complement, I believe Fortran standard
>>> doesn't mandate that. So not(0), but won't the FE do the compile time range
>>> checking on it? I.e. don't we have to write
>>> x = 0
>>> x = not(x)
>>
>> The frontend may optimize the above, but it should be legal
>> to do not(0) as long as -huge()-1 <= not <= huge(). If not,
>> you can do it the hard way
>>
>> integer i, x
>> x = 0
>> do i = 1, bit_size(x)
>> x = ibset(x,i)
>> end do
>>
>> BTW, does gcc run on any non-twos-complement hardware?
>>
>
> Can't we just use (-2147483647 - 1)?
>
Read the thread starting at Jakub's post. I replaced
a BOZ constant of the form Z'FFFFFFFFF' with -huge(x)-1
(which is equivalent to your suggestion). Jakub pointed
out that I made my change in haste in that -huge(x)-1
is Z'80000000'.
--
Steve