[Bug math/16516] New: spurious underflow exception from erff

[Bug math/16516] New: spurious underflow exception from erff

[zimmerma+gcc at loria dot fr]

https://sourceware.org/bugzilla/show_bug.cgi?id=16516

            Bug ID: 16516
           Summary: spurious underflow exception from erff
           Product: glibc
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: math
          Assignee: unassigned at sourceware dot org
          Reporter: zimmerma+gcc at loria dot fr

Created attachment 7381
  --> https://sourceware.org/bugzilla/attachment.cgi?id=7381&action=edit
test case

on x86_64 in single precision I get with the attached program:

GNU libc version: 2.18.90
GNU libc release: development
x=0x1.c5bf94p-127 (1.041755126e-38)
libc: erff(x)=0x1.000006p-126 (1.175494771e-38) underflow=16
mpfr: erff(x)=0x1.000006p-126 (1.175494771e-38)
ulp difference: 0.000000

See http://webloria.loria.fr/~zimmerma/glibc/

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16516] spurious underflow exception from erff

[zimmerma+gcc at loria dot fr]

https://sourceware.org/bugzilla/show_bug.cgi?id=16516

--- Comment #1 from Paul Zimmermann <zimmerma+gcc at loria dot fr> ---
here is another example with rounding up:

 rounding mode MPFR_RNDU:
      wrong underflow flag for x=3.8b7fa8@-32
      library gives 4.0000b0@-32
      mpfr    gives 4.0000b0@-32
      underflow: mpfr 0, library 16

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16516] spurious underflow exception from erff

[zimmerma+gcc at loria dot fr]

https://sourceware.org/bugzilla/show_bug.cgi?id=16516

--- Comment #2 from Paul Zimmermann <zimmerma+gcc at loria dot fr> ---
and another example in double precision:

Testing function erf for exponent 0 [seed=7626].
 rounding mode MPFR_RNDN:
      wrong underflow flag for x=-3.8b7f12369ded8@-256
      library gives -4.0000000000004@-256
      mpfr    gives -4.0000000000004@-256
      underflow: mpfr 0, library 16

Since the result is larger in magnitude than the smallest normal number
0x1p-1022, we should get no underflow exception.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16516] spurious underflow exception from erff

[zimmerma+gcc at loria dot fr]

https://sourceware.org/bugzilla/show_bug.cgi?id=16516

--- Comment #3 from Paul Zimmermann <zimmerma+gcc at loria dot fr> ---
idem in extended precision:

Testing function erf for exponent 0 [seed=8344].
 rounding mode MPFR_RNDN:
      wrong underflow flag for x=3.8b7f12369ded5518@-4096
      library gives 4.0000000000000028@-4096
      mpfr    gives 4.0000000000000028@-4096
      underflow: mpfr 0, library 16

Note the glibc result is *above* the smallest normal number.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16516] spurious underflow exception from erff

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=16516

--- Comment #4 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "GNU C Library master sources".

The branch, master has been updated
       via  0bf061d3e37e0f72c2378cbded67c58df4f58a4b (commit)
      from  3d8c8bff798a15b6dc0ee94d24fc877db9706589 (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.

- Log -----------------------------------------------------------------
https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=0bf061d3e37e0f72c2378cbded67c58df4f58a4b

commit 0bf061d3e37e0f72c2378cbded67c58df4f58a4b
Author: Joseph Myers <joseph@codesourcery.com>
Date:   Wed May 14 12:34:03 2014 +0000

    Fix erf underflow handling near 0 (bug 16516).

    Bug 16516 reports spurious underflows from erf (for all floating-point
    types), when the result is close to underflowing but does not actually
    underflow.

    erf (x) is about (2/sqrt(pi))*x for x close to 0, so there are
    subnormal arguments for which it does not underflow.  The various
    implementations do (x + efx*x) (for efx = 2/sqrt(pi) - 1), for greater
    accuracy than if just using a single multiplication by an
    approximation to 2/sqrt(pi) (effectively, this way there are a few
    more bits in the approximation to 2/sqrt(pi)).  This can introduce
    underflows when efx*x underflows even though the final result does
    not, so a scaled calculation with 8*efx is done in these cases - but 8
    is not a big enough scale factor to avoid all such underflows.  16 is
    (any underflows with a scale factor of 16 would only occur when the
    final result underflows), so this patch changes the code to use that
    factor.  Rather than recomputing all the values of the efx8 variable,
    it is removed, leaving it to the compiler's constant folding to
    compute 16*efx.  As such scaling can also lose underflows when the
    final scaling down happens to be exact, appropriate checks are added
    to ensure underflow exceptions occur when required in such cases.

    Tested x86_64 and x86; no ulps updates needed.  Also spot-checked for
    powerpc32 and mips64 to verify the changes to the ldbl-128ibm and
    ldbl-128 implementations.

        [BZ #16516]
        * sysdeps/ieee754/dbl-64/s_erf.c (efx8): Remove variable.
        (__erf): Scale by 16 instead of 8 in potentially underflowing
        case.  Ensure exception if result actually underflows.
        * sysdeps/ieee754/flt-32/s_erff.c (efx8): Remove variable.
        (__erff): Scale by 16 instead of 8 in potentially underflowing
        case.  Ensure exception if result actually underflows.
        * sysdeps/ieee754/ldbl-128/s_erfl.c: Include <float.h>.
        (efx8): Remove variable.
        (__erfl): Scale by 16 instead of 8 in potentially underflowing
        case.  Ensure exception if result actually underflows.
        * sysdeps/ieee754/ldbl-128ibm/s_erfl.c: Include <float.h>.
        (efx8): Remove variable.
        (__erfl): Scale by 16 instead of 8 in potentially underflowing
        case.  Ensure exception if result actually underflows.
        * sysdeps/ieee754/ldbl-96/s_erfl.c: Include <float.h>.
        (efx8): Remove variable.
        (__erfl): Scale by 16 instead of 8 in potentially underflowing
        case.  Ensure exception if result actually underflows.
        * math/auto-libm-test-in: Add more tests of erf.
        * math/auto-libm-test-out: Regenerated.

-----------------------------------------------------------------------

Summary of changes:
 ChangeLog                            |   24 ++++
 NEWS                                 |   13 +-
 math/auto-libm-test-in               |    4 +
 math/auto-libm-test-out              |  200 ++++++++++++++++++++++++++++++++++
 sysdeps/ieee754/dbl-64/s_erf.c       |   12 ++-
 sysdeps/ieee754/flt-32/s_erff.c      |   13 ++-
 sysdeps/ieee754/ldbl-128/s_erfl.c    |   16 ++-
 sysdeps/ieee754/ldbl-128ibm/s_erfl.c |   16 ++-
 sysdeps/ieee754/ldbl-96/s_erfl.c     |   14 ++-
 9 files changed, 289 insertions(+), 23 deletions(-)

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16516] spurious underflow exception from erff

[jsm28 at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=16516

Joseph Myers <jsm28 at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #5 from Joseph Myers <jsm28 at gcc dot gnu.org> ---
Fixed for 2.20.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/16516] spurious underflow exception from erff

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=16516

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
              Flags|                            |security-

-- 
You are receiving this mail because:
You are on the CC list for the bug.