This is the mail archive of the
libc-alpha@sources.redhat.com
mailing list for the glibc project.
Re: Many new math failures
Ulrich Drepper <drepper@redhat.com> writes:
> On Fri, 2002-06-21 at 09:12, Andreas Jaeger wrote:
>
>> I used GCC 3.1 for this. I don't have a problem adding compiler
>> dependent ULPs, just send me the ULPs file and I'll integrate it.
>
> Don't just blindly incorporate them. They have to be complared with the
> old file. I hope you did this already for the files you checked in.
The new file uses different values, it's not easy to compare them.
> gcc 3 had build (actually test) failures forever due to larger errors.
> Instead of adjusting the accepted errors we have to see whether we can
> make the code more stable to reduce the variations in gcc's output.
The old version needed the appended patch to pass the testsuite with
GCC 3.1. I think that should indicate the functions we should look
at.
But I'll do a build with GCC 2.95 now and compare the files,
Andreas
============================================================
Index: sysdeps/i386/fpu/libm-test-ulps
--- sysdeps/i386/fpu/libm-test-ulps 2001/08/27 13:55:13 1.30
+++ sysdeps/i386/fpu/libm-test-ulps 2002/04/02 13:58:00
@@ -96,9 +96,9 @@ ldouble: 329
# cacosh
Test "Real part of: cacosh (-2 - 3 i) == -1.9833870299165354323470769028940395 + 2.1414491111159960199416055713254211 i":
double: 1
-float: 4
+float: 9
idouble: 1
-ifloat: 4
+ifloat: 9
ildouble: 6
ldouble: 6
Test "Imaginary part of: cacosh (-2 - 3 i) == -1.9833870299165354323470769028940395 + 2.1414491111159960199416055713254211 i":
@@ -1234,9 +1234,9 @@ ldouble: 329
Function: Real part of "cacosh":
double: 1
-float: 4
+float: 9
idouble: 1
-ifloat: 4
+ifloat: 9
ildouble: 328
ldouble: 328
@@ -1643,3 +1643,35 @@ ldouble: 7
ildouble: 7
# end of automatic generation
+Test "acos (0.0625) == 1.50825556499840522843072005474337068":
+ldouble: 622
+Test "asin (0.75) == 0.848062078981481008052944338998418080":
+ldouble: 1
+Test "atanh (0.75) == 0.972955074527656652552676371721589865":
+ldouble: 1
+Test "acos (0.0625) == 1.50825556499840522843072005474337068":
+ildouble: 622
+Test "atanh (0.75) == 0.972955074527656652552676371721589865":
+ildouble: 2
+Test "Real part of: ccosh (-2 - 3 i) == -3.7245455049153225654 + 0.5118225699873846088 i":
+float: 1
+Test "Imaginary part of: ctanh (0.7 + 1.2 i) == 1.3472197399061191630 + 0.4778641038326365540 i":
+float: 1
+Test "Real part of: ctanh (-2 - 3 i) == -0.9653858790221331242 + 0.0098843750383224937 i":
+float: 1
+Test "Real part of: ccosh (-2 - 3 i) == -3.7245455049153225654 + 0.5118225699873846088 i":
+ifloat: 1
+Test "Imaginary part of: ctanh (0.7 + 1.2 i) == 1.3472197399061191630 + 0.4778641038326365540 i":
+ifloat: 1
+Test "Real part of: ctanh (-2 - 3 i) == -0.9653858790221331242 + 0.0098843750383224937 i":
+ifloat: 1
+Test "tanh (0.7) == 0.60436777711716349631":
+float: 1
+Test "tanh (-0.7) == -0.60436777711716349631":
+float: 1
+Function: "tanh":
+float: 1
+Test "Imaginary part of: ccosh (0.7 + 1.2 i) == 0.4548202223691477654 + 0.7070296600921537682 i":
+float: 1
+Test "Imaginary part of: ccosh (0.7 + 1.2 i) == 0.4548202223691477654 + 0.7070296600921537682 i":
+ifloat: 1
--
Andreas Jaeger
SuSE Labs aj@suse.de
private aj@arthur.inka.de
http://www.suse.de/~aj