diff options
author | Joseph Myers <joseph@codesourcery.com> | 2015-05-15 00:16:10 +0000 |
---|---|---|
committer | Joseph Myers <joseph@codesourcery.com> | 2015-05-15 00:16:10 +0000 |
commit | fbc68f03b0e56db6a07d789fd403eb79d726fe96 (patch) | |
tree | a39460f48cef09b5e9577c38b9b061ddfb335fab /sysdeps/ieee754 | |
parent | 9a71f1fcf53615c00b5f9e5da4bba92bccb0efb4 (diff) | |
download | glibc-fbc68f03b0e56db6a07d789fd403eb79d726fe96.tar glibc-fbc68f03b0e56db6a07d789fd403eb79d726fe96.tar.gz glibc-fbc68f03b0e56db6a07d789fd403eb79d726fe96.tar.bz2 glibc-fbc68f03b0e56db6a07d789fd403eb79d726fe96.zip |
Fix erfcf spurious underflows (bug 18217).
The flt-32 implementation of erfcf produces spurious underflow
exceptions for some arguments close to 0, because of calculations
squaring the argument and then multiplying by small constants. This
patch fixes this by adjusting the threshold for arguments for which
the result is so close to 1 that 1 - x will give the right result from
2**-56 to 2**-26. (If 1 - x * 2/sqrt(pi) were used, the errors would be
on the order of x^3 and a much larger threshold could be used.)
Tested for x86_64 and x86.
[BZ #18217]
* sysdeps/ieee754/flt-32/s_erff.c (__erfcf): Use 2**-26 not 2**-56
as threshold for returning 1 - x.
* math/auto-libm-test-in: Add more tests of erfc.
* math/auto-libm-test-out: Regenerated.
Diffstat (limited to 'sysdeps/ieee754')
-rw-r--r-- | sysdeps/ieee754/flt-32/s_erff.c | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/sysdeps/ieee754/flt-32/s_erff.c b/sysdeps/ieee754/flt-32/s_erff.c index d23cff9293..2be44cc40c 100644 --- a/sysdeps/ieee754/flt-32/s_erff.c +++ b/sysdeps/ieee754/flt-32/s_erff.c @@ -169,7 +169,7 @@ float __erfcf(float x) } if(ix < 0x3f580000) { /* |x|<0.84375 */ - if(ix < 0x23800000) /* |x|<2**-56 */ + if(ix < 0x32800000) /* |x|<2**-26 */ return one-x; z = x*x; r = pp0+z*(pp1+z*(pp2+z*(pp3+z*pp4))); |