GSL test release 1.12.90

GSL test release 1.12.90

[Brian Gladman]

I have compiled GSL 1.12.90 and the tests on Vista x64 Ultimate using
Visual Studio 2008 Professional (a few minor changes are needed to do
this).

I had two compile errors:

1. I found declarations after executable statements in test_funcs.c
(spring_f)

2. In gsl_const_cgs.h, the MSVC preprocessor does not recognise the
#warning directive.

After removing these compilation errors, all tests except one pass in
both 32 and 64 bit versions of the library.

The test that fails is testrandist, which gives this output in 64-bit mode:

GSL_IEEE_MODE="double-precision,round-to-nearest,mask-all"
FAIL: test gsl_ran_exponential [0,1]
      (0 observed vs 0.393469 expected) [7]

and this in 32-bit mode:

GSL_IEEE_MODE="double-precision,round-to-nearest,mask-all"
FAIL: test gsl_ran_exponential [0,1]
      (0 observed vs 0.393469 expected) [7]
FAIL: test gsl_ran_discrete1 [-0.5,0.5]
      (0.33122 observed vs 0.59 expected) [9]
FAIL: test gsl_ran_discrete1 [0.5,1.5]
      (0.33343 observed vs 0.4 expected) [10]
FAIL: test gsl_ran_discrete1 [1.5,3.5]
      (0.33164 observed vs 0.01 expected) [11]

I have not yet had time to look into this failure any further.

My thanks to all developers for their first class work on GSL.

  Brian Gladman



__________ Information from ESET NOD32 Antivirus, version of virus signature database 4334 (20090814) __________

The message was checked by ESET NOD32 Antivirus.

http://www.eset.com

Re: GSL test release 1.12.90

[Brian Gough]

At Fri, 14 Aug 2009 12:49:56 +0100,
Brian Gladman wrote:
> I had two compile errors:
> 
> 1. I found declarations after executable statements in test_funcs.c
> (spring_f)
> 
> 2. In gsl_const_cgs.h, the MSVC preprocessor does not recognise the
> #warning directive.

Thanks for the testing.

> The test that fails is testrandist, which gives this output in 64-bit mode:
> 
> GSL_IEEE_MODE="double-precision,round-to-nearest,mask-all"
> FAIL: test gsl_ran_exponential [0,1]
>       (0 observed vs 0.393469 expected) [7]
> 

Thanks - it looks like the header file <gsl/gsl_math.h> is missing in
randist/exponential.c for gsl_log1p.

-- 
Brian Gough

Re: GSL test release 1.12.90

[Brian Gladman]

Brian Gough wrote:
> At Fri, 14 Aug 2009 12:49:56 +0100,
> Brian Gladman wrote:
>> I had two compile errors:
>>
>> 1. I found declarations after executable statements in test_funcs.c
>> (spring_f)
>>
>> 2. In gsl_const_cgs.h, the MSVC preprocessor does not recognise the
>> #warning directive.
> 
> Thanks for the testing.
> 
>> The test that fails is testrandist, which gives this output in 64-bit mode:
>>
>> GSL_IEEE_MODE="double-precision,round-to-nearest,mask-all"
>> FAIL: test gsl_ran_exponential [0,1]
>>       (0 observed vs 0.393469 expected) [7]
>>
> 
> Thanks - it looks like the header file <gsl/gsl_math.h> is missing in
> randist/exponential.c for gsl_log1p.

Yes, that fixed it. Thanks Brian.

   Brian Gladman



__________ Information from ESET NOD32 Antivirus, version of virus signature database 4344 (20090818) __________

The message was checked by ESET NOD32 Antivirus.

http://www.eset.com

GNU Scientific Library 1.13 released - report

[maxgacode]

Release 1.13 compilation

Platform Windows XP SP3 + MSys 1.0.11 + MinGW (Gcc 4.4.0 latest 
binutils,winapi,make,runtime)

Using all the following compilation options

1) CFLAGS="-mfpmath=sse -msse2 -O2 -fexceptions"
2) CFLAGS="-mfpmath=sse -msse2 -fexceptions"
3) CFLAGS="-fexceptions"
4) CFLAGS="-O2 -fexceptions"

the library compiled (static library)and passed the make check. 
Especially the option 4) is a pleasant
surprise. I was never able to get a successful make check using OOB 
source code with  -O2 -fexceptions.
Neither integration nor monte routines need hacks (like disabling 
optimization)  to pass the tests

Hope this helps.

Max