This is the mail archive of the libc-alpha@sources.redhat.com 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]

K6-II+, 2.2.3pre3 test results


On Tue, 10 Apr 2001, Matthew Clarke wrote:

....
> Linux ds0.van.maves.ca 2.2.19 #1 mar mar 27 11:12:56 PST 2001 i586 unknown
>
> In this case, the "i586" is an AMD K6-2.  The installed libc is GLIBC CVS
> glibc-2-1-branch from 12 September, 2000.  The compiler is "gcc version
> 2.95.3 20010315 (release)", unpatched.  The binutils are 2.11.90.0.1.  The
> kernel headers used were from Linux 2.2.19.  The GNU gettext is version
> 0.10.36, built & installed a few minutes prior, using the same tools.
>
> Test 1:  built from current CVS head branch as of Sunday morning.
> Test 2:  built from the 2.2.3pre2 tarballs from sourceware.
>
> In both cases, built outside the source tree, with "../<foo>/configure
> --prefix=/usr --enable-add-ons --without-cvs".
....

I run K6-II+ 500 MHZ model 13 stepping 4 at 400 MHZ(66x6, ASUS P/I-P55T2P4
rev. 3.10, 128 MB RAM). Currently I use kernel-2.4.3-ac6, glibc-2.2.2,
gcc-2.95.3(20010315, with recommended patch about atexit),
binutils-2.91.90.0.4(5), etc.

> In both cases, no errors or warnings from configure, no errors and a few
> warnings from "make", and 2 errors and some warnings from "make check".
> Warnings and errors are identical between the two tests.  I have not
> installed or used the built library, just built it & run the test suite.
>
> Errors:
>
> math/tst-ldouble.out:
> ------------------------------------------------------------------------------
> testing long double (without inline functions)
> Failure: Test: tgamma (4) == 6
> Result:
>  is:          6.00000000000000000087e+00   0xc.00000000000000200000p-1
>  should be:   6.00000000000000000000e+00   0xc.00000000000000000000p-1
>  difference:  8.67361737988403547206e-19   0x8.00000000000000000000p-63
>  ulp       :  2.0000
>  max.ulp   :  0.0000
>
> Test suite completed:
>   2480 test cases plus 2272 tests for exception flags executed.
>   1 errors occurred.
> ------------------------------------------------------------------------------
>
> math/tst-ildoubl.out:
> ------------------------------------------------------------------------------
> testing long double (inline functions)
> Failure: Test: tgamma (4) == 6
> Result:
>  is:          6.00000000000000000087e+00   0xc.00000000000000200000p-1
>  should be:   6.00000000000000000000e+00   0xc.00000000000000000000p-1
>  difference:  8.67361737988403547206e-19   0x8.00000000000000000000p-63
>  ulp       :  2.0000
>  max.ulp   :  0.0000
>
> Test suite completed:
>   2418 test cases plus 2210 tests for exception flags executed.
>   1 errors occurred.
....

I met different errors as follows:
(note: test-float emits 2 more erros in case 1 than in case 2.)

1) while using flags such as "-O2 -march=i386 -D__USE_STRING_INLINES
-fstrict-aliasing" ( I recall "-O3" made more errors)

testing float (without inline functions)
Failure: Test: j0 (8.0) == 0.17165080713755390609
Result:
 is:          1.19443722069263458252e-01   0x1.e93dd200000000000000p-4
 should be:   1.71650812029838562012e-01   0x1.5f8a7600000000000000p-3
 difference:  5.22070899605751037598e-02   0x1.abae3400000000000000p-5
 ulp       :  3503558.5000
 max.ulp   :  1.0000
Maximal error of `j0'
 is      :  3503558.5000 ulp
 accepted:  1.0000 ulp
Failure: Test: yn (3, 10.0) == -0.25136265718383732978
Result:
 is:         -1.83247357606887817383e-01  -0x1.774a6400000000000000p-3
 should be:  -2.51362651586532592773e-01  -0x1.01653600000000000000p-2
 difference:  6.81152939796447753906e-02   0x1.17001000000000000000p-4
 ulp       :  2285570.0000
 max.ulp   :  1.0000
Maximal error of `yn'
 is      :  2285570.0000 ulp
 accepted:  3.0000 ulp

Test suite completed:
  2468 test cases plus 2266 tests for exception flags executed.
  4 errors occurred.

testing float (inline functions)
Failure: Test: y0 (8.0) == 0.22352148938756622053
Result:
 is:          2.24330961704254150391e-01   0x1.cb6e0800000000000000p-3
 should be:   2.23521485924720764160e-01   0x1.c9c5a200000000000000p-3
 difference:  8.09475779533386230469e-04   0x1.a8660000000000000000p-11
 ulp       :  54323.0000
 max.ulp   :  1.0000
Maximal error of `y0'
 is      :  54323.0000 ulp
 accepted:  1.0000 ulp

Test suite completed:
  2406 test cases plus 2204 tests for exception flags executed.
  2 errors occurred.


2) while using simply "-O2 -march=i386"

testing float (without inline functions)
Failure: Test: j0 (8.0) == 0.17165080713755390609
Result:
 is:          1.19443722069263458252e-01   0x1.e93dd200000000000000p-4
 should be:   1.71650812029838562012e-01   0x1.5f8a7600000000000000p-3
 difference:  5.22070899605751037598e-02   0x1.abae3400000000000000p-5
 ulp       :  3503558.5000
 max.ulp   :  1.0000
Maximal error of `j0'
 is      :  3503558.5000 ulp
 accepted:  1.0000 ulp

Test suite completed:
  2468 test cases plus 2266 tests for exception flags executed.
  2 errors occurred.

testing float (inline functions)
Failure: Test: y0 (8.0) == 0.22352148938756622053
Result:
 is:          2.24330961704254150391e-01   0x1.cb6e0800000000000000p-3
 should be:   2.23521485924720764160e-01   0x1.c9c5a200000000000000p-3
 difference:  8.09475779533386230469e-04   0x1.a8660000000000000000p-11
 ulp       :  54323.0000
 max.ulp   :  1.0000
Maximal error of `y0'
 is      :  54323.0000 ulp
 accepted:  1.0000 ulp

Test suite completed:
  2406 test cases plus 2204 tests for exception flags executed.
  2 errors occurred.

-- 
 "Where there is a will, there is a way."  jinbo21@hananet.net
  For the future of you and me!            hitel: jinbo21


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