Bug 16509 - Bad overflowing atan results for directed rounding
Summary: Bad overflowing atan results for directed rounding
Status: RESOLVED DUPLICATE of bug 14759
Alias: None
Product: glibc
Classification: Unclassified
Component: math (show other bugs)
Version: unspecified
: P2 normal
Target Milestone: ---
Assignee: Not yet assigned to anyone
Depends on:
Reported: 2014-01-29 08:19 UTC by Paul Zimmermann
Modified: 2014-06-13 08:49 UTC (History)
0 users

See Also:
Last reconfirmed:
fweimer: security-

test file (479 bytes, text/x-csrc)
2014-01-29 08:19 UTC, Paul Zimmermann

Note You need to log in before you can comment on or make changes to this bug.
Description Paul Zimmermann 2014-01-29 08:19:08 UTC
Created attachment 7375 [details]
test file

in single precision with rounding to zero, atan(0x1p-149) should give zero and
not 0x1p-149:

GNU libc version: 2.18.90
GNU libc release: development
x=0x1p-149 (1.401298464e-45)
libc: atanf(x)=0x1p-149 (1.401298464e-45)
mpfr: atanf(x)=0x0p+0 (0.000000000e+00)
ulp difference: inf

See http://webloria.loria.fr/~zimmerma/glibc/
Comment 1 Joseph Myers 2014-01-29 16:35:24 UTC
For small numbers, ulps are for glibc considered equal to the least subnormal (unless the mathematical result is well under half the least subnormal), so this is counted as a 1ulp error, not a bug.

*** This bug has been marked as a duplicate of bug 14759 ***