[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Help-gsl] test release for GSL 2.5
From: |
maxgacode |
Subject: |
Re: [Help-gsl] test release for GSL 2.5 |
Date: |
Thu, 7 Jun 2018 22:18:52 +0200 |
User-agent: |
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:52.0) Gecko/20100101 Thunderbird/52.8.0 |
Il 06/06/2018 05:34, Patrick Alken ha scritto:
> I would like to release the next version of GSL (2.5) since there
> ftp://alpha.gnu.org/gnu/gsl/gsl-2.4.90.tar.gz
Platform : Windows 7 64 bit latest Service pack
Compiler : GCC 7.3.0 using 64 bit toolchain MINGW64
CFLAGS="-mfpmath=sse -msse2 -O2 -mthreads -DHAVE_INLINE
-DGSL_DISABLE_DEPRECATED -DGSL_RANGE_CHECK=0 -D__USE_MINGW_ANSI_STDIO=1"
Compilation OK.
Make check:
1) Using Microsoft Runtime DLL msvcr110.dll (Visual Studio 2012)
Everything went well!!
2) Using standard runtime MinGW64 linking
FAIL: gsl_sf_bessel_j2_e(1048576.0, &r) [168]
expected: -3.1518539455252412e-007
obtained: -3.1518539455252539e-007 +/- 2.7994086564622246e-022
(rel=8.88178e-016)
fracdiff: 2.0155588470164931e-015
tolerance: 4.5474735088646412e-013
value/expected not consistent within reported error
-3.151853945525253879e-007 2.799408656462224591e-022
FAIL: Bessel Functions [940
This is a known issue related to the error in the trif function due to
loos of precision during the range reduction made by FSIN (used by the
default runtime libraries). The computed value seems correct but it is
not within the expected range. The expected error is "underestimated"
(.ie not compatble with FSIN), I mean that different hardware platforms
or runtime libraries have different trigonometric precision due to range
reduction.
There is also a bug recorded
bug #52927
related to this issue. I don't know whether this issue may be considered
a bug or not. For sure pass/fail conditions depends on both HW and
runtime libs.
I'm also trying to compile verification programs linking with another
static mathematical library (fdlibm 5.3 from Netlib)....
Regards
Max
--
PGP key: wwwkeys.pgp.net: 0EBF4A07