This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug target/32180] Paranoia UCB GSL TestFloat libm tests fail - accuracy of recent gcc math poor
- From: "rob1weld at aol dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 5 Jun 2007 17:22:07 -0000
- Subject: [Bug target/32180] Paranoia UCB GSL TestFloat libm tests fail - accuracy of recent gcc math poor
- References: <bug-32180-13830@http.gcc.gnu.org/bugzilla/>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Comment #11 from rob1weld at aol dot com 2007-06-05 17:22 -------
>> kargl@gcc.gnu.org
>> IEEE 754 does not discuss any of the functions you list above.
>> Comment #4 From Rob
>> That page is a report of the libc6 tests that are ran when the code is built
>> kargl@gcc.gnu.org
>> Compare the ... is similar to a comparison of a Boeing 747 to the Wright brothers' aircraft.
That is why is was good of people who replied to tell us about the version of
libm and the hardware they are using along with any parameters they may have
altered that might affect testing.
It was helpful of them to either "confirm" or "deny" _this_ bug report on
target i686-pc-linux-gnu - but it works on a Cray does not assist this report.
While Boeing _may_ have had pictures of the Wright brother's aircraft on the
wall and considered it's principles when the 747 was designed they are likely
going with modern methods of design and construction.
I made some more builds of GSL using various combinations of -mfpmath=sse
-mfpmath=387 -ffloat-store -fno-float-store and the OS's libm vs. fdlibm:
# grep FAIL: ../gsl-1.9-New-Lib/check_1_log.txt | wc -l
1031
# grep FAIL: ../gsl-1.9-New-Lib_float-store/check_1_log.txt | wc -l
1051
# grep FAIL: ../gsl-1.9-fdlibm/check_1_log.txt | wc -l
1036
# grep FAIL: ../gsl-1.9-New-Lib_SSE/check_1_log.txt | wc -l
1030
# grep FAIL: ../gsl-1.9-Old-Lib/check_1_log.txt | wc -l
1063
grep FAIL: ../gsl-1.9-New-Lib_SSE_float-store/check_1_log.txt | wc -l
1051
Currently, using the new libc6 2.6 libm and -mfpmath=sse is one fail better
than using -mfpmath=387. I'm doing some more testing to see if I can track down
the problem.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32180