[Bug math/19848] powl(10,n) for n=-4,-5,-6,-7 is off by more than 1 ULP

["joseph at codesourcery dot com" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=19848

--- Comment #1 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
An error exceeding 1ulp for most functions is not itself a bug (does not 
exceed glibc's documented accuracy goals).  However, in some rounding 
modes these show errors exceeding 9ulp, which is the maximum permitted in 
the glibc testsuite, so there is an actual bug in that case that needs to 
be fixed.

testing long double (without inline functions)
Failure: Test: pow_downward (0xap+0, -0x7p+0)
Result:
 is:          9.99999999999999999338e-08   0xd.6bf94d5e57a42b20p-27
 should be:   9.99999999999999999984e-08   0xd.6bf94d5e57a42bc0p-27
 difference:  6.46234853557052870993e-26   0xa.0000000000000000p-87
 ulp       :  10.0000
 max.ulp   :  6.0000
Failure: Test: pow_towardzero (0xap+0, -0x7p+0)
Result:
 is:          9.99999999999999999338e-08   0xd.6bf94d5e57a42b20p-27
 should be:   9.99999999999999999984e-08   0xd.6bf94d5e57a42bc0p-27
 difference:  6.46234853557052870993e-26   0xa.0000000000000000p-87
 ulp       :  10.0000
 max.ulp   :  6.0000

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