The ABINIT Users Mailing List ( CLOSED )

[Masayoshi Mikami <mmikami@rc.m-kagaku.co.jp>]

Dear Francois,

Merci pour votre gentillesse !
In my case, I also just use the internal libraries
available in the ABINIT... So I am puzzled...
I will try to look the difference between v.7.1 and v.8.1,
and other tests in other platforms...
I will let you know if I notice anything relevant.

Bien a vous,
Masayoshi

On 2005/10/10, at 22:25, Francois Jollet wrote:

> Dear Masayoshi,
>
>
> I have tested Test_v4/t04-t08, 94, Tutorial/tpaw* on Linux box (ia32) 
> with IFC v8.0 and v8.1 (Build 20041019) and it works well. v4.6.4 is 
> available on 6 platforms on the web site, and I suppose the tests are 
> ok on these platforms.
>
> So I do not find any compiler dependent problem with these tests.
> I have used the internal librairies available in the ABINIT package. 
> Perhaps you encounter problems due to librairy linkage.
>
> What I can suggest you, as you have no problem with IFC v7.1, is to 
> compare it with IFC v8.1 and to track the bugs if any... ;-)
>
> Best regards
>
> Francois
>
>
> P.S.
> I have no access to Mac OS or Fujitsu, so I cannot run the tests on 
> these platforms
>
>
>
>
>
> Masayoshi Mikami wrote:
> > Dear all, I have noticed something about the tests about the PAW...
> > (Test_v4/t06-08, 94, Tutorial/tpaw*)
> > I am running Test*/Tutorials of v.4.6.4, by typing
> > "make tests_allseq" on my Linux box (ia32) and other platforms.
> > With Intel fortran compiler v.7.1 (Build 20040309Z), it runs fine; 
> > there is no big difference in "fldiff.report" files.
> > But, with IFC v.8.1 (Build 20050520Z, l_fc_pc_8.1.028),
> > quite large numerical differences between the output files and the 
> > attached references appear very large. In passing,
> > The most tutorials (tpaw1_2 to 1_6) fail with errors.
> > (The diff* files is put as "IFCv81_v464_v4.tar.gz" on the Shared web:
> >  http://www.abinit.org/wws/d_read/forum/ (*))
> > In the above tests, the relevant compiler options are like this:
> > FFLAGS=-FR -O3 -xW -tpp7
> > FFLAGS_Src_2psp =-FR -O0 -xW -tpp7
> > FFLAGS_Src_2nonlocal =-FR -O0 -xW -tpp7
> > FFLAGS_Src_3iovars =-FR -O0 -xW -tpp7
> > FFLAGS_Src_9drive =-FR -O0 -xW -tpp7
> > FFLAGS_LIBS= -O3 -xW -tpp7
> > (The only difference is the line "FC="; IFC v.7.1 or v.8.1)
> > On the MacOS X (xlf v.8.1), I do not have any problems with 
> > Test_v4/t04-t06, but I see a sudden death (without any error message, 
> > during "ITER 2" (with the end of writing
> > "Non-SCF iterations: k pt #   2 ...") with Test_v4/t94; in passing, I 
> > had no problems with Test_v4/t94 of v.4.6.3
> > with the same compiler/compiler options...
> > Then, I am now finding some similar numerical differences
> > about the same Tests/Tutotials with the (quite old) Fujitsu vector 
> > machine... (even though the numerical difference seem much 
> > smaller...)
> > So, the PAW part may be dependent on compilers ... ?
> > This is not described in ~Infos/known_problems.4.6.x,
> > so please let me report this ...
> > Best regards,
> > Masayoshi
> > P.S.(*) I intended to include the tpaw*log files, too, but it failed 
> > to transmit the big file (compressed size: 500KB)
> > to the Shared web (and ftp.abinit.org)... I can send the
> > error log files for Tutorial/tpaw*, if requested.
> > P.S.2: on the MacOSX, I had the same problem of "P46.17."
> >  described in ~/Infos/known_problems.4.6.x...