[Bug math/18020] New: [ldbl-128ibm] asinhl inaccurate

["jsm28 at gcc dot gnu.org" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=18020

            Bug ID: 18020
           Summary: [ldbl-128ibm] asinhl inaccurate
           Product: glibc
           Version: 2.21
            Status: NEW
          Severity: normal
          Priority: P2
         Component: math
          Assignee: unassigned at sourceware dot org
          Reporter: jsm28 at gcc dot gnu.org

The ldbl-128ibm implementation of asinhl uses cut-offs of 0x1p28 and 0x1p-29 to
determine when to use simpler formulas that avoid possible overflow /
underflow.  Both those cut-offs are inappropriate for this format, resulting in
large errors, e.g.:

Failure: Test: asinh (0x2p+28)
Result:
 is:          2.07944154167983592825e+01   0x1.4cb5ecf0a96504221a8bp+4
 should be:   2.07944154167983592834e+01   0x1.4cb5ecf0a96504231a8bp+4
 difference:  8.67361737988600762432e-19   0x1.00000000004000000000p-60
 ulp       :  2199023255552.5000
 max.ulp   :  1.0000

Failure: Test: asinh (0x4p-32)
Result:
 is:          9.31322574615478515625e-10   0x1.00000000000000000000p-30
 should be:   9.31322574615478515490e-10   0x1.fffffffffffffffaaaabp-31
 difference:  1.34632261157723174603e-28   0x1.55555555556000000000p-93
 ulp       :  5864062014805.5000
 max.ulp   :  1.0000

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