Re: [PATCH 3/4] softfloat: fix floatx80 pseudo-denormal comparisons

[Alex Bennée]

Joseph Myers <address@hidden> writes:

> The softfloat floatx80 comparisons fail to allow for pseudo-denormals,
> which should compare equal to corresponding values with biased
> exponent 1 rather than 0.  Add an adjustment for that case when
> comparing numbers with the same sign.
>
> Note that this fix only changes floatx80_compare_internal, not the
> other more specific comparison operations.  That is the only
> comparison function for floatx80 used in the i386 port, which is the
> only supported port with these pseudo-denormal semantics.

Again I can't see anything that triggers this although I noticed
le_quiet has been fixed in the meantime. lt_quiet still fails with:

  ./fp-test -s -l 2 -r all  extF80_lt_quiet
  >> Testing extF80_lt_quiet
  59535872 tests total.
  Errors found in extF80_lt_quiet:
  +0000.0000000000000000  +0000.0000000000000000  => 1 .....  expected 0 .....
  +0000.0000000000000000  -0000.0000000000000000  => 1 .....  expected 0 .....
  +0000.0000000000000001  +0000.0000000000000001  => 1 .....  expected 0 .....
  +0000.0000000000000002  +0000.0000000000000002  => 1 .....  expected 0 .....
  +0000.0000000000000004  +0000.0000000000000004  => 1 .....  expected 0 .....
  +0000.0000000000000008  +0000.0000000000000008  => 1 .....  expected 0 .....
  +0000.0000000000000010  +0000.0000000000000010  => 1 .....  expected 0 .....
  +0000.0000000000000020  +0000.0000000000000020  => 1 .....  expected 0 .....
  +0000.0000000000000040  +0000.0000000000000040  => 1 .....  expected 0 .....
  +0000.0000000000000080  +0000.0000000000000080  => 1 .....  expected 0 .....
  +0000.0000000000000100  +0000.0000000000000100  => 1 .....  expected 0 .....
  +0000.0000000000000200  +0000.0000000000000200  => 1 .....  expected 0 .....
  +0000.0000000000000400  +0000.0000000000000400  => 1 .....  expected 0 .....
  +0000.0000000000000800  +0000.0000000000000800  => 1 .....  expected 0 .....
  +0000.0000000000001000  +0000.0000000000001000  => 1 .....  expected 0 .....
  +0000.0000000000002000  +0000.0000000000002000  => 1 .....  expected 0 .....
  +0000.0000000000004000  +0000.0000000000004000  => 1 .....  expected 0 .....
  +0000.0000000000008000  +0000.0000000000008000  => 1 .....  expected 0 .....
  +0000.0000000000010000  +0000.0000000000010000  => 1 .....  expected 0 .....
  +0000.0000000000020000  +0000.0000000000020000  => 1 .....  expected 0 .....


>
> Signed-off-by: Joseph Myers <address@hidden>
> ---
>  fpu/softfloat.c | 5 +++++
>  1 file changed, 5 insertions(+)
>
> diff --git a/fpu/softfloat.c b/fpu/softfloat.c
> index 6094d267b5..8e9c714e6f 100644
> --- a/fpu/softfloat.c
> +++ b/fpu/softfloat.c
> @@ -7966,6 +7966,11 @@ static inline int floatx80_compare_internal(floatx80 
> a, floatx80 b,
>              return 1 - (2 * aSign);
>          }
>      } else {
> +        /* Normalize pseudo-denormals before comparison.  */
> +        if ((a.high & 0x7fff) == 0 && a.low & UINT64_C(0x8000000000000000))
> +            ++a.high;
> +        if ((b.high & 0x7fff) == 0 && b.low & UINT64_C(0x8000000000000000))
> +            ++b.high;
>          if (a.low == b.low && a.high == b.high) {
>              return float_relation_equal;
>          } else {
> -- 
> 2.17.1


-- 
Alex Bennée