fadapt

NAME

     fadapt - HPF source-to-source compiler of GMD

SYNOPSIS

     fadapt [options] filename ... filename [-o filename | -od directory]

DESCRIPTION

     fadapt is a tool for the automatic partitioning of data parallel
     High Performance Fortran programs (ADAPTOR).

OPTIONS

           version   print current version of fadapt
          settings   show the options currently set

Options for input source file(s):
   [-on|off:<DIR>]   enable/disable compiler directives starting with DIR
 [-I<include_dir>]   include directories
          -fix <n>   relevant columns in input file (0 for free format)
         [-sp|-dp]   REAL is single / double precision 
         -nostrict   no strict semantic checks of input source

Options for output source file(s):
              -F77   target language = Fortran 77
              -F90   target language = Fortran 90
              -F95   target language = Fortran 95
   -submain=<name>   PROGRAM becomes SUBROUTINE 
        -strip <n>   relevant columns for output file (0 for free format)
        -keep=none   keep nothing of input source
    -keep=comments   keep comments of input source
         -keep=yes   keep unchanged lines of input source
         -keep=all   keep all of input source
       -keep=debug   keep all + debug info of input source
       -incl=never   include never files
    -incl=modified   include modified files
      -incl=always   include always files
         -incl=new   include new files
      -incl=newmod   include new modified files

Options for input/output source file(s):
           [-free]   abbreviation for -fix 0 -strip 0, free source format
         [-extend]   abbreviation for -fix 72 -strip 72, extended format

Translation commands:
          -fortran   fadapt command  = Fortran translation
              -hpf   fadapt command  = HPF translation
              -omp   fadapt command  = OpenMP translation
           -update   fadapt command  = update translation
            -instr   fadapt command  = instrumentation
        -interface   fadapt command  = generate interface file
             -call   fadapt command  = generate call graph

Options for translation:
           -noauto   loop_parallelization = NoAuto 
             -auto   loop_parallelization = Auto
           -dsp=no   generate descriptors = no
         -dsp=auto   generate descriptors = auto
        -dsp=array   generate descriptors = yes
          -dsp=all   generate descriptors = yes + scalar
      -args=single   double arguments     = no
      -args=double   double arguments     = yes
        -pseudo=no   pseudo dynamic array = no
      -pseudo=auto   pseudo dynamic array = auto
       -pseudo=all   pseudo dynamic array = all
    -noprof:<item>   switch off profiling for SUB PAR DATA
      -prof:<item>   switch on profiling for SUB PAR DATA
       [-rif[=<file_name>]  use/def region information file
       [-noopt|-O]   (no) optimization

Options for HPF/OpenMP translation:
               -dm   parallel Processes
               -sm   parallel Threads
               -cb   parallel Cache-Blocks
Options for HPF translation:
              -ddr   default distribution = replicated 
              -ddb   default distribution = block distr
              -dda   default distribution = auto local 
              -ddg   default distribution = auto global
          [-d <n>]   maximal number of distributed dimensions

OPTIONS miscellaneous
                -v   verbose
            -debug   generates intermediate files
             -list   do not delete protocol files
                -w   suppresses warning messages
           -noinfo   suppresses info messages

Abbreviation options:
                -G   -v -debug -list
              -HPF   -off:TRA -off:OMP -off: -on:ADP -on:HPF -hpf
              -OMP   -on:OMP -on: -off:TRA -off:HPF -omp
          -OMP_SIM   -on:OMP -on: -off:TRA -off:HPF -omp -submain=APPL_MAIN
            -INSTR   -update -prof:SUB -prof:PAR -dsp=no -args=single
           -DINSTR   -instr -prof:SUB -prof:DATA -dsp=array -pseudo=no -args=single

FILES

     cube.f           generated parallel node program
     node1.f          generated serial node program  (with flag -1)

     adaptor.def      Protocol of Semantic Phase 1
     adaptor.sem      Protocol of Semantic Phase 2
     adaptor.cf       Protocol of Semantic Phase 3

     adaptor.dist     Protocol of Parallelization Phase 1 (Distribution)
     adaptor.normal   Protocol of Parallelization Phase 2 (Normalizations)
     adaptor.ovl      Protocol of Parallelization Phase 3 (Overlapping)
     adaptor.ipa      Protocol of Parallelization Phase 4 (Interprocedural)
     adaptor.ana      Protocol of Parallelization Phase 5 (Analysis)

     adaptor.args     Protocol of Adaptor Phase 1 (Arguments)
     adaptor.loops    Protocol of Adaptor Phase 2 (Make Loops)
     adaptor.home     Protocol of Adaptor Phase 3 (Find Homes)
     adaptor.local    Protocol of Adaptor Phase 4 (Extract Communication)
     adaptor.class    Protocol of Adaptor Phase 5 (Classification)
     adaptor.opt      Protocol of Adaptor Phase 6 (Optimization)
     adaptor.code     Protocol1 of Adaptor Phase 7 (Code Generation)
     adaptor.final    Protocol2 of Adaptor Phase 7 (Code Generation)

     normal_unparse.f (source program after normalization)
     arg_unparse.f    (source program after arguments)
     loop_unparse.f   (source program after loop generation)
     home_unparse.f   (source program after home determination)
     loc_unparse.f    (source program after localization)
     opt_unparse.f    (optimized source program before code generation)
     code_unparse.f   (generated code with DALIB calls)
     addr_unparse.f   (localized addresses and pseudo dynamic arrays)