spec2000 regression

[Jan Hubicka]

> > current tree does not:
> > 
> >         pushl   %ebp
> >         movl    %esp, %ebp
> >         pushl   %eax
> >         pushl   %eax
> >         movl    a, %eax
> >         andl    $-16, %esp
> >         movl    $1, (%eax)
> >         movl    b, %eax
> >         movw    $2, (%eax)
> >         movl    a, %eax
> >         movl    (%eax), %eax
> >         movl    %ebp, %esp
> >         popl    %ebp
> > 
> > so I would guess that something must be wrong... But it can be unrelated to your change of course.
> 
> My guess is that this is not related to aliasing, because if you change that
> long long to long, it will work as you expect (unless -fno-strict-aliasing,
> of course).
It does not for me. I am still getting:
        pushl   %ebp
        movl    %esp, %ebp
        pushl   %eax
        pushl   %eax
        movl    a, %eax
        andl    $-16, %esp
        movl    $1, (%eax)
        movl    b, %eax
        movw    $2, (%eax)
        movl    a, %eax
        movl    (%eax), %eax
        movl    %ebp, %esp
        popl    %ebp

instead of movl $1, %eax I would expect. In fact I first tried long and then I jumped to
conclusion that long may change the pointer value itself (that is incorrect) and I tried
long long.
Perhaps you have older tree.

In 2.95 it works.

Honza
> 
> 	Jakub