[Bug math/19125] New: [powerpc32] llroundf, llround incorrect exceptions

[Bug math/19125] New: [powerpc32] llroundf, llround incorrect exceptions

[jsm28 at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=19125

            Bug ID: 19125
           Summary: [powerpc32] llroundf, llround incorrect exceptions
           Product: glibc
           Version: 2.22
            Status: NEW
          Severity: normal
          Priority: P2
         Component: math
          Assignee: unassigned at sourceware dot org
          Reporter: jsm28 at gcc dot gnu.org
  Target Milestone: ---
              Host: powerpc*-*-*

The powerpc32 implementations of llroundf and llround produce spurious and
missing exceptions (some arising from such exceptions from conversions to long
long, some present even when fctidz is used).  E.g., for float:

Failure: llround (qNaN): Exception "Inexact" set
Failure: llround (-qNaN): Exception "Inexact" set
Failure: llround (max_value): Exception "Inexact" set
Failure: llround (-max_value): Exception "Inexact" set
Failure: llround (0x1p31): Exception "Inexact" set
Failure: llround (0x1p63): Exception "Invalid operation" not set
Failure: llround (0x1p64): Exception "Inexact" set
Failure: llround (0x1p65): Exception "Inexact" set
Failure: llround (-0x1p31): Exception "Inexact" set
Failure: llround (-0x1p64): Exception "Inexact" set
Failure: llround (-0x1p65): Exception "Inexact" set
Failure: llround (0x7fffff80p0): Exception "Inexact" set
Failure: llround (-0x80000100p0): Exception "Inexact" set
Failure: llround (-0x8000010000000000p0): Exception "Invalid operation" not set
Failure: llround (2097151.0): Exception "Inexact" set
Failure: llround (8388608.0): Exception "Inexact" set
Failure: llround (16777216.0): Exception "Inexact" set
Failure: llround (4503601774854144.0): Exception "Inexact" set
Failure: llround (0x1.fffffep+23): Exception "Inexact" set
Failure: llround (-0x1.fffffep+23): Exception "Inexact" set
Failure: llround (0x1.000002p+23): Exception "Inexact" set
Failure: llround (-0x1.000002p+23): Exception "Inexact" set
Failure: llround (0x1.fffffep+23): Exception "Inexact" set
Failure: llround (-0x1.fffffep+23): Exception "Inexact" set

plus similar in other rounding modes.

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

[Bug math/19125] [powerpc32] llroundf, llround incorrect exceptions

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=19125

--- Comment #1 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  d7025badd8bae7ff8bc37551d868a22832a7bc94 (commit)
      from  2caca60d065b659f7d42500a95e50c073d4a33a0 (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=d7025badd8bae7ff8bc37551d868a22832a7bc94

commit d7025badd8bae7ff8bc37551d868a22832a7bc94
Author: Joseph Myers <joseph@codesourcery.com>
Date:   Tue Oct 13 23:30:20 2015 +0000

    Fix powerpc32 llround, llroundf exceptions (bug 19125).

    The powerpc32 implementations of llroundf and llround produce spurious
    and missing exceptions (some arising from such exceptions from
    conversions to long long, some present even when fctidz is used).

    This patch fixes those problems in a similar way to the llrint /
    llrintf fixes.  The spurious exceptions in the fctidz case for large
    arguments arise from a converted value that saturated as LLONG_MAX
    being converted back to float or double (the conversion back being
    inexact, but "inexact" must not be raised together with "invalid"),
    and from the subtraction x - xrf also being inexact for sufficiently
    large arguments (whether the saturation was to LLONG_MAX or
    LLONG_MIN); those are fixed by returning early if the argument is
    large enough that no rounding is needed.

    This code is not used for --with-cpu=power4 builds (I suspect the code
    used in that case may also produce spurious "inexact" exceptions, but
    that's something to investigate later).

    Tested for powerpc.

        [BZ #19125]
        * sysdeps/powerpc/powerpc32/fpu/s_llround.c: Include <limits.h>,
        <math_private.h> and <stdint.h>.
        (__llround): Avoid conversions to and from long long int, and
        subtractions, where those might raise spurious exceptions.
        * sysdeps/powerpc/powerpc32/fpu/s_llroundf.c: Include
        <math_private.h> and <stdint.h>.
        (__llroundf): Avoid conversions to and from long long int, and
        subtractions, where those might raise spurious exceptions.

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

Summary of changes:
 ChangeLog                                  |   12 ++++++++
 NEWS                                       |    2 +-
 sysdeps/powerpc/powerpc32/fpu/s_llround.c  |   40 +++++++++++++++++++++++++++-
 sysdeps/powerpc/powerpc32/fpu/s_llroundf.c |   30 ++++++++++++++++++++-
 4 files changed, 81 insertions(+), 3 deletions(-)

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

[Bug math/19125] [powerpc32] llroundf, llround incorrect exceptions

[jsm28 at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=19125

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED
   Target Milestone|---                         |2.23

--- Comment #2 from Joseph Myers <jsm28 at gcc dot gnu.org> ---
Fixed for 2.23.

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