diff options
| author | Joseph Myers <joseph@codesourcery.com> | 2015-05-15 17:21:08 +0000 |
|---|---|---|
| committer | Joseph Myers <joseph@codesourcery.com> | 2015-05-15 17:21:08 +0000 |
| commit | ff069f024ae8cf15d53429e034d67ddcece0f67a (patch) | |
| tree | d6ad47bf62f47e532cadb862d427eaa2f610c12c /math | |
| parent | b2fb25240813266e4f62f2dd039573f1a042ea2f (diff) | |
| download | glibc-ff069f024ae8cf15d53429e034d67ddcece0f67a.zip glibc-ff069f024ae8cf15d53429e034d67ddcece0f67a.tar.gz glibc-ff069f024ae8cf15d53429e034d67ddcece0f67a.tar.bz2 | |
Fix lgammaf spurious underflows (bug 18220).
The flt-32 implementation of lgammaf produces spurious underflow
exceptions for some large arguments, because of calculations involving
x^-2 multiplied by small constants. This patch fixes this by
adjusting the threshold for a simpler computation to 2**26 (the error
in the simpler computation is on the order of 0.5 * log (x), for a
result on the order of x * log (x)).
Tested for x86_64 and x86.
[BZ #18220]
* sysdeps/ieee754/flt-32/e_lgammaf_r.c (__ieee754_lgammaf_r): Use
2**26 not 2**58 as threshold for returning x * (log (x) - 1).
* math/auto-libm-test-in: Add another test of lgamma.
* math/auto-libm-test-out: Regenerated.
Diffstat (limited to 'math')
| -rw-r--r-- | math/auto-libm-test-in | 1 | ||||
| -rw-r--r-- | math/auto-libm-test-out | 25 |
2 files changed, 26 insertions, 0 deletions
diff --git a/math/auto-libm-test-in b/math/auto-libm-test-in index c4bfe74..2a88403 100644 --- a/math/auto-libm-test-in +++ b/math/auto-libm-test-in @@ -1757,6 +1757,7 @@ lgamma 0.5 lgamma -0.5 lgamma 0.7 lgamma 1.2 +lgamma 0x3.8p56 lgamma 0x1p-5 lgamma -0x1p-5 lgamma 0x1p-10 diff --git a/math/auto-libm-test-out b/math/auto-libm-test-out index d717ea4..0aa7b75 100644 --- a/math/auto-libm-test-out +++ b/math/auto-libm-test-out @@ -139612,6 +139612,31 @@ lgamma 1.2 = lgamma tonearest ldbl-128ibm 0x1.33333333333333333333333333p+0L : -0x1.5db138c7d70c72cb0b57c9e83d8p-4L 1 : inexact-ok = lgamma towardzero ldbl-128ibm 0x1.33333333333333333333333333p+0L : -0x1.5db138c7d70c72cb0b57c9e83dp-4L 1 : inexact-ok = lgamma upward ldbl-128ibm 0x1.33333333333333333333333333p+0L : -0x1.5db138c7d70c72cb0b57c9e83dp-4L 1 : inexact-ok +lgamma 0x3.8p56 += lgamma downward flt-32 0x3.8p+56f : 0x8.8bdd4p+60f 1 : inexact-ok += lgamma tonearest flt-32 0x3.8p+56f : 0x8.8bdd4p+60f 1 : inexact-ok += lgamma towardzero flt-32 0x3.8p+56f : 0x8.8bdd4p+60f 1 : inexact-ok += lgamma upward flt-32 0x3.8p+56f : 0x8.8bdd5p+60f 1 : inexact-ok += lgamma downward dbl-64 0x3.8p+56 : 0x8.8bdd41bf4484p+60 1 : inexact-ok += lgamma tonearest dbl-64 0x3.8p+56 : 0x8.8bdd41bf44848p+60 1 : inexact-ok += lgamma towardzero dbl-64 0x3.8p+56 : 0x8.8bdd41bf4484p+60 1 : inexact-ok += lgamma upward dbl-64 0x3.8p+56 : 0x8.8bdd41bf44848p+60 1 : inexact-ok += lgamma downward ldbl-96-intel 0x3.8p+56L : 0x8.8bdd41bf4484605p+60L 1 : inexact-ok += lgamma tonearest ldbl-96-intel 0x3.8p+56L : 0x8.8bdd41bf4484605p+60L 1 : inexact-ok += lgamma towardzero ldbl-96-intel 0x3.8p+56L : 0x8.8bdd41bf4484605p+60L 1 : inexact-ok += lgamma upward ldbl-96-intel 0x3.8p+56L : 0x8.8bdd41bf4484606p+60L 1 : inexact-ok += lgamma downward ldbl-96-m68k 0x3.8p+56L : 0x8.8bdd41bf4484605p+60L 1 : inexact-ok += lgamma tonearest ldbl-96-m68k 0x3.8p+56L : 0x8.8bdd41bf4484605p+60L 1 : inexact-ok += lgamma towardzero ldbl-96-m68k 0x3.8p+56L : 0x8.8bdd41bf4484605p+60L 1 : inexact-ok += lgamma upward ldbl-96-m68k 0x3.8p+56L : 0x8.8bdd41bf4484606p+60L 1 : inexact-ok += lgamma downward ldbl-128 0x3.8p+56L : 0x8.8bdd41bf44846050819264e2d568p+60L 1 : inexact-ok += lgamma tonearest ldbl-128 0x3.8p+56L : 0x8.8bdd41bf44846050819264e2d57p+60L 1 : inexact-ok += lgamma towardzero ldbl-128 0x3.8p+56L : 0x8.8bdd41bf44846050819264e2d568p+60L 1 : inexact-ok += lgamma upward ldbl-128 0x3.8p+56L : 0x8.8bdd41bf44846050819264e2d57p+60L 1 : inexact-ok += lgamma downward ldbl-128ibm 0x3.8p+56L : 0x8.8bdd41bf44846050819264e2d4p+60L 1 : inexact-ok += lgamma tonearest ldbl-128ibm 0x3.8p+56L : 0x8.8bdd41bf44846050819264e2d4p+60L 1 : inexact-ok += lgamma towardzero ldbl-128ibm 0x3.8p+56L : 0x8.8bdd41bf44846050819264e2d4p+60L 1 : inexact-ok += lgamma upward ldbl-128ibm 0x3.8p+56L : 0x8.8bdd41bf44846050819264e2d8p+60L 1 : inexact-ok lgamma 0x1p-5 = lgamma downward flt-32 0x8p-8f : 0x3.72d02cp+0f 1 : inexact-ok = lgamma tonearest flt-32 0x8p-8f : 0x3.72d03p+0f 1 : inexact-ok |