This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Fix for logb/logbf/logbl (bz 13954/13955/13956)


From: Adhemerval Zanella <azanella@linux.vnet.ibm.com>
Date: Thu, 03 May 2012 11:24:05 -0300

> Thank you very much for testing it and sorry for the gross
> mistake. The main issue (besides the misplaced ';') is I was using
> 'hx' instead of 'ex' in final calculations. I have corrected it and
> tested on simulated long double with two int64_t. I believe the
> implementation is now correct to test.

Still has failures on 32-bit sparc, see attached math/test-ldouble.out

Strangely (or not), 64-bit sparc is fine.

testing long double (without inline functions)
Failure: Test: logb (0x1p-16400) == -16400
Result:
 is:         -1.63680000000000000000e+04  -0x1.ff800000000000000000p+13
 should be:  -1.64000000000000000000e+04  -0x1.00400000000000000000p+14
 difference:  3.20000000000000000000e+01   0x1.00000000000000000000p+5
 ulp       :  10141204801825835211973625643008.0000
 max.ulp   :  0.0000
Failure: Test: logb (0x.00000000001p-16382) == -16426
Result:
 is:         -1.63940000000000000000e+04  -0x1.00280000000000000000p+14
 should be:  -1.64260000000000000000e+04  -0x1.00a80000000000000000p+14
 difference:  3.20000000000000000000e+01   0x1.00000000000000000000p+5
 ulp       :  10141204801825835211973625643008.0000
 max.ulp   :  0.0000
Maximal error of `logb'
 is      : 10141204801825835211973625643008 ulp
 accepted: 0 ulp

Test suite completed:
  5338 test cases plus 4477 tests for exception flags executed.
  3 errors occurred.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]