The ABINIT Users Mailing List ( CLOSED )

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

Dear Yusuke,

Thank you for the report for the NEC SX-7 !
We would be fortunate, if you could provide
a bit more detailed information on the system :
From the FAQ,
"- I think there is a bug - making a bug report"
        http://www.abinit.org/ABINIT/Infos/FAQ.html#bug_report
You will reach the bug report format:
http://www.abinit.org/ABINIT/bug_report.html
or "~ABINIT/Infos/problem_report".
-----
1) Version of ABINIT
2) Machine type : vendor + some hardware characteristics (processor)
3) Operating system (including version number)
   "uname -a" will work on most platforms.
   "oslevel" will be preferable on IBM workstations (AIX).
4) Fortran Compiler (including version number)
   For the compiler version number, type man "compiler command" etc.
   e.g.
        pgf90 (PGI compiler on Linux) :  pgf90 -V
        xlf   (on IBM platform)       :  lslpp -l
   Otherwise, ask your system administrator.

  AND ALSO, PLEASE CHECK THE "CPP" VERSION !

5) ... (snip here: your e-mail has already replied to these sections)
-----
Could you tell us about 1), 3), and 4) ? When the system will be
upgraded, such information may be important to know whether
your remedies are needed or not (or modified ...).

Thanks to your post, the "problem_report" format will be modified
(i.e. to check the version of "cpp" (or cc) also).
Arigatoh ("Thanks" in Japanese) !

Best wishes,
Masayoshi

On 2004/04/01, at 14:17, Yusuke Iguchi wrote:

> Dear all
>
> Recently I tried to compile the ABINIT code on NEC SX-7 machine
> at Institute for Molecular Science in Okazaki National Research
> Institute (IMS, www.ims.ac.jp),
> and I encountered several difficulties.
> After modifying the source files, I finally made it.
> In this mail, I will report what I did.
>
> I am happy if the future version of the ABINIT code will
> be free from the difficulties.
>
> If somebody is interested in my modified ABINIT code,
> I can upload the output of diff on my own web directory
> and will show the URL on this mailing list.
>
> Thank you.
> ------------------------------------------------------------
>
> 1. Problem of CPP
>
> Since the preprocessor "cpp" in our environment interpret
> the character of "'" (prime) in a comment
> line as the beginning of a string, an error occurs when the
> number of "'" is odd in a line.
> This problem can be avoided by revising the comments
> in a dozen of source files
> as follows whenever the error reported.
>
> I don't .. -> I do not ..
> G' -> G\prime
>
> This is in the case of the default option of cpp.
>
>
> 2. Compatibility between nec and nolib
>
> In matrginv.f, the ASL (built-in numerical library in SX7)
> routine is linked even if 'nolib' option is selected.
> So I revised the file as follows:
>
> before:
> #elif defined nec
>
> after:
> #elif defined nec && !defined nolib
>
>
> 3. Example of makefile_macros in SX7 at IMS
>
> I compiled the executable without library in SX7
> at IMS by a cross compiler.
>
> I show the makefile_macros.SX7 linked by
> makefile_macros used for the compile as
> unified diff format.
>
> ---------- beginning of the output of diff ------------------
> --- makefile_macros.SX5 Sun May 11 17:52:37 2003
> +++ makefile_macros.SX7 Thu Apr  1 13:29:39 2004
> @@ -20,7 +20,7 @@
>  MACHINE=nec
>
>  # Fortran optimized compilation
> -FC=f90
> +FC=sxf90
>
>  FFLAGS= -f4 -Wf"-pvctl fullmsg noassume nodivloop -L transform 
> fmtlist"
>  #AGGR_VEC:
> @@ -31,8 +31,8 @@
>  #FFLAGS_LIBS= -C hopt -Wf"-pvctl noassume nodivloop"
>
>  # C preprocessor, used to preprocess the fortran source.
> -CPP=/mem/cpp
> -CPP_FLAGS=-P
> +CPP=/lib/cpp
> +CPP_FLAGS=-Dnolib -P
>
>  # C optimized compilation.
>  CC=cc
> @@ -45,15 +45,15 @@
>  MACHINE_DEP_C_SEQ_SUBS_LIST=cclock.o
>
>  # Definition of libraries
> -LAPACK=
> -BLAS=
> +LAPACK=Lib_lapack/lapack.a
> +BLAS=Lib_blas/blas.a
>
>  #Original ASL declaration, from F.Finocchi and C. Roult
>  #ASL_A=ASL_I4R4R8
>  #Modified version, from L. Pizzagalli
> -ASL_A=$$ASL_I4R4R8
> +#ASL_A=$$ASL_I4R4R8
>
> -LIBS= /usr/lib/liblapack.a /usr/lib/libblas.a $(ASL_A)
> +LIBS= $(LAPACK) $(BLAS)
>
>  ####################################################################
>  # For the parallel version
> ------------------ end of the output of diff ---------------------
>
> And I made executable by the command
>
>     make AR_COMMAND=sxar allseq
>
> The compilation was completed after the above work,
> and it executed with no error.
>
> %-----------------------------------------------------------
> % Yusuke IGUCHI : Dept. of Applied Physics, Univ. of Tokyo
> %-----------------------------------------------------------