diff options
author | Joseph Myers <joseph@codesourcery.com> | 2014-06-24 21:00:08 +0000 |
---|---|---|
committer | Joseph Myers <joseph@codesourcery.com> | 2014-06-24 21:00:08 +0000 |
commit | 4060283decf711ed95a6c4993ffeaba39f8cd40a (patch) | |
tree | d303f8bcf7bd9fafbbb5edddc48c7c2be827f84e /math | |
parent | e7dd3c8c1db8c6d293abb995e033893b7328ad19 (diff) | |
download | glibc-4060283decf711ed95a6c4993ffeaba39f8cd40a.tar glibc-4060283decf711ed95a6c4993ffeaba39f8cd40a.tar.gz glibc-4060283decf711ed95a6c4993ffeaba39f8cd40a.tar.bz2 glibc-4060283decf711ed95a6c4993ffeaba39f8cd40a.zip |
Fix x86/x86_64 expm1l spurious underflow exceptions (bug 16539).
This patch fixes bug 16539, spurious underflow exceptions from x86 /
x86-64 expm1l. The problem is that the computation of a base-2
exponent with extra precision involves spurious underflows for
arguments that are small but not subnormal, so a check is added to
just return the argument in those cases. (If the argument *is*
subnormal, underflowing is correct and the existing code will always
underflow, so it suffices to keep using the existing code in that
case; some expm1 implementations have a bug (bug 16353) with missing
underflow exceptions, but I don't think there's such a bug in this
particular version.)
Tested x86_64 and x86; no ulps updates needed.
(auto-libm-test-out diffs omitted below.)
[BZ #16539]
* sysdeps/i386/fpu/e_expl.S (IEEE754_EXPL) [USE_AS_EXPM1L]: Just
return the argument for normal arguments with exponent below -64.
* sysdeps/x86_64/fpu/e_expl.S (IEEE754_EXPL) [USE_AS_EXPM1L]:
Likewise.
* math/auto-libm-test-in: Add another test of expm1.
* math/auto-libm-test-out: Regenerated.
Diffstat (limited to 'math')
-rw-r--r-- | math/auto-libm-test-in | 2 | ||||
-rw-r--r-- | math/auto-libm-test-out | 81 |
2 files changed, 83 insertions, 0 deletions
diff --git a/math/auto-libm-test-in b/math/auto-libm-test-in index 79d276ce99..f36d200df5 100644 --- a/math/auto-libm-test-in +++ b/math/auto-libm-test-in @@ -970,6 +970,8 @@ expm1 0x1p-64 expm1 -0x1p-64 expm1 0x1p-100 expm1 -0x1p-100 +# Bug 16353: underflow exception may be missing +expm1 0x4.0000000000000028p-16384 missing-underflow fma 1.0 2.0 3.0 fma 1.25 0.75 0.0625 diff --git a/math/auto-libm-test-out b/math/auto-libm-test-out index d1ec2be279..ced5283ab2 100644 --- a/math/auto-libm-test-out +++ b/math/auto-libm-test-out @@ -93045,6 +93045,87 @@ expm1 -0x1p-100 = expm1 tonearest ldbl-128ibm -0x1p-100L : -0xf.ffffffffffffffffffffffff8p-104L : inexact-ok = expm1 towardzero ldbl-128ibm -0x1p-100L : -0xf.ffffffffffffffffffffffff8p-104L : inexact-ok = expm1 upward ldbl-128ibm -0x1p-100L : -0xf.ffffffffffffffffffffffff8p-104L : inexact-ok +expm1 0x4.0000000000000028p-16384 missing-underflow += expm1 downward flt-32 0x8p-152f : 0x8p-152f : inexact-ok underflow underflow-ok errno-erange-ok += expm1 tonearest flt-32 0x8p-152f : 0x8p-152f : inexact-ok underflow underflow-ok errno-erange-ok += expm1 towardzero flt-32 0x8p-152f : 0x8p-152f : inexact-ok underflow underflow-ok errno-erange-ok += expm1 upward flt-32 0x8p-152f : 0x1p-148f : inexact-ok underflow underflow-ok errno-erange-ok += expm1 downward dbl-64 0x8p-152 : 0x8p-152 : inexact-ok += expm1 tonearest dbl-64 0x8p-152 : 0x8p-152 : inexact-ok += expm1 towardzero dbl-64 0x8p-152 : 0x8p-152 : inexact-ok += expm1 upward dbl-64 0x8p-152 : 0x8.0000000000008p-152 : inexact-ok += expm1 downward ldbl-96-intel 0x8p-152L : 0x8p-152L : inexact-ok += expm1 tonearest ldbl-96-intel 0x8p-152L : 0x8p-152L : inexact-ok += expm1 towardzero ldbl-96-intel 0x8p-152L : 0x8p-152L : inexact-ok += expm1 upward ldbl-96-intel 0x8p-152L : 0x8.000000000000001p-152L : inexact-ok += expm1 downward ldbl-96-m68k 0x8p-152L : 0x8p-152L : inexact-ok += expm1 tonearest ldbl-96-m68k 0x8p-152L : 0x8p-152L : inexact-ok += expm1 towardzero ldbl-96-m68k 0x8p-152L : 0x8p-152L : inexact-ok += expm1 upward ldbl-96-m68k 0x8p-152L : 0x8.000000000000001p-152L : inexact-ok += expm1 downward ldbl-128 0x8p-152L : 0x8p-152L : inexact-ok += expm1 tonearest ldbl-128 0x8p-152L : 0x8p-152L : inexact-ok += expm1 towardzero ldbl-128 0x8p-152L : 0x8p-152L : inexact-ok += expm1 upward ldbl-128 0x8p-152L : 0x8.0000000000000000000000000008p-152L : inexact-ok += expm1 downward ldbl-128ibm 0x8p-152L : 0x8p-152L : inexact-ok += expm1 tonearest ldbl-128ibm 0x8p-152L : 0x8p-152L : inexact-ok += expm1 towardzero ldbl-128ibm 0x8p-152L : 0x8p-152L : inexact-ok += expm1 upward ldbl-128ibm 0x8p-152L : 0x8.00000000000000000000000004p-152L : inexact-ok += expm1 downward flt-32 0x0p+0f : 0x0p+0f : inexact-ok += expm1 tonearest flt-32 0x0p+0f : 0x0p+0f : inexact-ok += expm1 towardzero flt-32 0x0p+0f : 0x0p+0f : inexact-ok += expm1 upward flt-32 0x0p+0f : 0x0p+0f : inexact-ok += expm1 downward dbl-64 0x0p+0 : 0x0p+0 : inexact-ok += expm1 tonearest dbl-64 0x0p+0 : 0x0p+0 : inexact-ok += expm1 towardzero dbl-64 0x0p+0 : 0x0p+0 : inexact-ok += expm1 upward dbl-64 0x0p+0 : 0x0p+0 : inexact-ok += expm1 downward ldbl-96-intel 0x0p+0L : 0x0p+0L : inexact-ok += expm1 tonearest ldbl-96-intel 0x0p+0L : 0x0p+0L : inexact-ok += expm1 towardzero ldbl-96-intel 0x0p+0L : 0x0p+0L : inexact-ok += expm1 upward ldbl-96-intel 0x0p+0L : 0x0p+0L : inexact-ok += expm1 downward ldbl-96-m68k 0x0p+0L : 0x0p+0L : inexact-ok += expm1 tonearest ldbl-96-m68k 0x0p+0L : 0x0p+0L : inexact-ok += expm1 towardzero ldbl-96-m68k 0x0p+0L : 0x0p+0L : inexact-ok += expm1 upward ldbl-96-m68k 0x0p+0L : 0x0p+0L : inexact-ok += expm1 downward ldbl-128 0x0p+0L : 0x0p+0L : inexact-ok += expm1 tonearest ldbl-128 0x0p+0L : 0x0p+0L : inexact-ok += expm1 towardzero ldbl-128 0x0p+0L : 0x0p+0L : inexact-ok += expm1 upward ldbl-128 0x0p+0L : 0x0p+0L : inexact-ok += expm1 downward ldbl-128ibm 0x0p+0L : 0x0p+0L : inexact-ok += expm1 tonearest ldbl-128ibm 0x0p+0L : 0x0p+0L : inexact-ok += expm1 towardzero ldbl-128ibm 0x0p+0L : 0x0p+0L : inexact-ok += expm1 upward ldbl-128ibm 0x0p+0L : 0x0p+0L : inexact-ok += expm1 downward dbl-64 0x4p-1076 : 0x4p-1076 : inexact-ok underflow underflow-ok errno-erange-ok += expm1 tonearest dbl-64 0x4p-1076 : 0x4p-1076 : inexact-ok underflow underflow-ok errno-erange-ok += expm1 towardzero dbl-64 0x4p-1076 : 0x4p-1076 : inexact-ok underflow underflow-ok errno-erange-ok += expm1 upward dbl-64 0x4p-1076 : 0x8p-1076 : inexact-ok underflow underflow-ok errno-erange-ok += expm1 downward ldbl-96-intel 0x4p-1076L : 0x4p-1076L : inexact-ok += expm1 tonearest ldbl-96-intel 0x4p-1076L : 0x4p-1076L : inexact-ok += expm1 towardzero ldbl-96-intel 0x4p-1076L : 0x4p-1076L : inexact-ok += expm1 upward ldbl-96-intel 0x4p-1076L : 0x4.0000000000000008p-1076L : inexact-ok += expm1 downward ldbl-96-m68k 0x4p-1076L : 0x4p-1076L : inexact-ok += expm1 tonearest ldbl-96-m68k 0x4p-1076L : 0x4p-1076L : inexact-ok += expm1 towardzero ldbl-96-m68k 0x4p-1076L : 0x4p-1076L : inexact-ok += expm1 upward ldbl-96-m68k 0x4p-1076L : 0x4.0000000000000008p-1076L : inexact-ok += expm1 downward ldbl-128 0x4p-1076L : 0x4p-1076L : inexact-ok += expm1 tonearest ldbl-128 0x4p-1076L : 0x4p-1076L : inexact-ok += expm1 towardzero ldbl-128 0x4p-1076L : 0x4p-1076L : inexact-ok += expm1 upward ldbl-128 0x4p-1076L : 0x4.0000000000000000000000000004p-1076L : inexact-ok += expm1 downward ldbl-128ibm 0x4p-1076L : 0x4p-1076L : inexact-ok underflow underflow-ok errno-erange-ok += expm1 tonearest ldbl-128ibm 0x4p-1076L : 0x4p-1076L : inexact-ok underflow underflow-ok errno-erange-ok += expm1 towardzero ldbl-128ibm 0x4p-1076L : 0x4p-1076L : inexact-ok underflow underflow-ok errno-erange-ok += expm1 upward ldbl-128ibm 0x4p-1076L : 0x8p-1076L : inexact-ok underflow underflow-ok errno-erange-ok += expm1 downward ldbl-96-intel 0x4.0000000000000028p-16384L : 0x4.0000000000000028p-16384L : inexact-ok += expm1 tonearest ldbl-96-intel 0x4.0000000000000028p-16384L : 0x4.0000000000000028p-16384L : inexact-ok += expm1 towardzero ldbl-96-intel 0x4.0000000000000028p-16384L : 0x4.0000000000000028p-16384L : inexact-ok += expm1 upward ldbl-96-intel 0x4.0000000000000028p-16384L : 0x4.000000000000003p-16384L : inexact-ok += expm1 downward ldbl-96-m68k 0x4.0000000000000028p-16384L : 0x4.0000000000000028p-16384L : inexact-ok += expm1 tonearest ldbl-96-m68k 0x4.0000000000000028p-16384L : 0x4.0000000000000028p-16384L : inexact-ok += expm1 towardzero ldbl-96-m68k 0x4.0000000000000028p-16384L : 0x4.0000000000000028p-16384L : inexact-ok += expm1 upward ldbl-96-m68k 0x4.0000000000000028p-16384L : 0x4.000000000000003p-16384L : inexact-ok += expm1 downward ldbl-128 0x4.0000000000000028p-16384L : 0x4.0000000000000028p-16384L : inexact-ok += expm1 tonearest ldbl-128 0x4.0000000000000028p-16384L : 0x4.0000000000000028p-16384L : inexact-ok += expm1 towardzero ldbl-128 0x4.0000000000000028p-16384L : 0x4.0000000000000028p-16384L : inexact-ok += expm1 upward ldbl-128 0x4.0000000000000028p-16384L : 0x4.0000000000000028000000000004p-16384L : inexact-ok fma 1.0 2.0 3.0 = fma downward flt-32 0x1p+0f 0x2p+0f 0x3p+0f : 0x5p+0f : = fma tonearest flt-32 0x1p+0f 0x2p+0f 0x3p+0f : 0x5p+0f : |