10.9;ftn (fortran_compiler), revision 1.0, 3/30/92
ftn (fortran_compiler) - compile a FORTRAN program.
usage: ftn source_file [-ac | -pic]
[-alnchk]
[-b [pathname] | -nb]
[-bounds_violation | -no_bounds_violation]
[-bx pathname]
[-cond | -ncond]
[-config var1...varN]
[-cpu id]
[-db | -dbs | -dba | -ndb]
[-dynm | -save]
[-exp | -nexp]
[-ff]
[-frnd | -nfrnd]
[-i*2 | -i*4]
[-idir dir1...dirN]
[-indexl | -nindexl]
[-info [n] | -ninfo]
[-inlib pathname]
[-inline char]
[-l [pathname] | -nl]
[-l*1 | -l*2 | -l*4]
[-mp]
[-msgs | -nmsgs]
[-natural | -nnatural]
[-nclines]
[-nfpra]
[-opt [n]]
[-overlap arguments]
[-prasm | -nprasm]
[-subchk | -nsubchk]
[-type | -ntype]
[-u]
[-uc | -nuc]
[-version]
[-warn | -nwarn]
[-xref | -nxref]
[-xrs | -nxrs]
[-zero | -nzero]
DESCRIPTION
The FORTRAN compiler translates a FORTRAN source file into one or more
binary object modules, which can be executed or used as input to the
binder. Source files may be a maximum of 32767 lines long. For complete
information, see the Domain FORTRAN Language Reference.
You can typically improve run-time performance if you compile your code
using the -cpu option with a processor-specific id. Use an argument
appropriate to the CPU type of the machine(s) on which you will run the
executable file. The argument mathchip, for example, generates optimal
code for the DSP90, DN330, DN560, DN570, DN580, DN2500, DN3000, DN3500,
DN4000, and DN4500 processors, and for the HP Apollo Series 400s and
400t. Use the cpuhelp utility to help you choose the -cpu argument best
suited to your needs. For information about this utility, type help
cpuhelp.
ARGUMENTS
source_file (required)
Specify the source file to be compiled. If you specify
the pathname without an .ftn suffix, the compiler
automatically appends .ftn, and searches for that
filename. If not found, it searches for the pathname that
you typed. If not found, the compiler reports an error.
OPTIONS
Default options are indicated by "(D)".
-ac (D)
Produce absolute code. This switch only applies when you generate
680x0 code.
-pic Produce position-independent code. The default is to produce
absolute code.
-alnchk
Display messages about alignment of structures. This option is the
default for compilers that generate Series 10000 code.
-b [pathname] (D)
Generate binary file named file.bin. If argument is omitted, binary
file is named source_file.bin.
-nb Suppress binary file.
-bounds_violation
Specify that the program can violate array subscript boundaries
during execution.
-no_bounds_violation (D)
Specify that the program does not access array elements beyond the
declared size of the array.
-bx pathname
Create an object file that does not have the default .bin suffix.
-cond
Compile lines with D or d in column 1.
-ncond (D)
Ignore lines with D or d in column 1.
-config var1...varN
Conditional processing: compile statements associated with var1,
var2, and so forth, via conditional compiler directives.
-cpu id
Generate code for a particular class of processors. The default id
is mathlib_sr10 if you are compiling for a 680x0-based workstation,
and a88k if you are compiling for a Series 10000 workstation.
Values for id are as follows:
mathlib_sr10
Code for workstations with an MC68020 or
MC68030 microprocessor and an MC68881 or
MC68882 floating-point coprocessor, or with
an MC68040 microprocessor.
mathlib Code for MC68020-, MC68030-, and MC68040-
based workstations (425t, 433s, DN5500).
Code compiled with this option runs only on
SR10.3. Use mathlib_sr10 if your code
must also run on SR10.0, SR10.1, or SR10.2.
mathchip Code for MC68020- or MC68030-based
3000 workstations with an MC6888x floating-point
580 coprocessor (400t, 400s, DN4500, DN4000,
570 DN3500, DN3000, DN2500, DN580, DN570,
560 DN560, DN330, DSP90). These seven
330 arguments are synonyms. Use the
90 mathchip argument; the other arguments
may become obsolete at a future release.
a88k Series 10000 code.
FPA1 Code for DN3000, DN4000, or DN4500
workstations with an FPA1 floating-point
accelerator unit.
FPX Code for DN5xx workstations with an FPX
floating-point accelerator unit.
PEB Code for workstations (including DN100, DN320,
DN400, DN600) with a Performance Enhancement
Board.
160 Code for a DSP160, DN460, or DN660 workstation.
460 These three arguments are synonyms.
660
any Series 10000 code, if you are compiling for a
Series 10000 workstation, or generic MC680x0
code, if you are compiling for a MC680x0-based
workstation.
m68k Generic MC680x0-based code.
Using a processor-specific code generation mode can result in
programs that run faster. Use the cpuhelp utility to help you
choose the -cpu argument best suited to your needs. For information
about this utility, type help cpuhelp.
-db (D)
Generate minimal debugging information.
-dbs Generate full runtime debugging information.
-dba Identical to -dbs except that no optimization is performed, even if
you also specify -opt.
-ndb Suppress all debugging information.
-dynm (D)
Allocate local variables on stack if possible.
-save
Allocate all local variables in static space.
-exp Generate expanded code listing (implies -l). You cannot use -exp
and -xref on the same command line.
-nexp (D)
Suppress expanded code listing.
-ff Activate free format: length of source lines may be up to 1024
characters; an ampersand (&) in column 1 specifies a continuation
line.
-frnd
Rounds floating-point numbers at key points during program
execution. Using this option may result in a loss of performance.
-nfrnd (D)
Optimizes execution by computing floating-point expressions in
greater precision than that specified by the program, when the
compiler detects an opportunity to do so.
-i*2 2-byte integer is default integer type.
-i*4 (D)
4-byte integer is default integer type.
-idir dir1...dirN
Add pathname to the search hierarchy of directories for include file
names. The hierarchy applies only to file name strings which do NOT
begin with '.', '~', or '/'. Up to 63 -idir options may be given.
The compiler first tries to open the include file name as given.
Failing that, it prefixes the given include file name with the -idir
pathnames in the same order as supplied on the command line.
-indexl
Disable some optimizations generated for subscript calculations, and
cause all array indexing to use 4-byte integer arithmetic.
-nindexl (D)
Use all the optimizations generated for subscript calculations.
-info [n]
Default is n=2. Display information messages to nth level.
-inlib [pathname]
Load pathname (a PIC binary file) at run time and resolve global
references.
-inline char
Select specified character (char) as in-line comment designator.
-l [pathname]
Generate compilation listing. If pathname is omitted, use
source_file.lst.
-nl (D)
Suppress compilation listing.
-l*1 1-byte logical is default logical type.
-l*2 2-byte logical is default logical type.
-l*4 4-byte logical is default logical type.
-mp For use with the HP Concurrent FORTRAN (HPCF) product to enable
optimization of FORTRAN programs for parallel execution on
multiprocessor Series 10000 workstations. For more information,
refer to the HP Concurrent FORTRAN User's Guide.
-msgs (D)
Generate final error and warning count message.
-nmsgs
Suppress final error and warning count message.
-natural
Use natural alignment for structures and array elements that do not
have alignment attributes.
-nnatural (D)
Use word alignment for structures and array elements that do not
have alignment attributes.
-nclines
Suppress generation of COFF line number tables. Applies only to
680x0 target compilers.
-nfpra
Suppress the reordering of operands in floating-point expressions at
optimization level 4.
-opt [n] (D)
Perform global optimization. n is a single digit integer in the
range 0 to 4, indicating the level of optimization performed. If n
is omitted, or if the entire option is not specified, 3 is the
default level. Increasing the optimization level results in faster
execution of your program, at the expense of longer compilation
time. Special restrictions apply to the use of -opt 4. See the
Domain FORTRAN Language Reference manual for details.
-overlap arguments
Specify the degree of the program's conformity to ANSI FORTRAN
storage association standard. By default, the compiler conforms to
the standard. The default produces the most optimized code.
The option takes one or more of the following arguments:
no_dd (D) Conforms to standard. No dummy argument that is
modified in its subprogram is associated with
another dummy argument to the same subprogram.
exact_dd If two dummy arguments overlap, they must have
the same address.
dd The program does not conform to the standard for
dummy-to-dummy associations.
no_dc (D) Conforms to standard. No dummy argument that is
modified in its subprogram is associated with a
variable in a common block.
exact_dc If a dummy argument is associated with a variable
in a common block, the address of the actual
parameter is identical to the address of the
variable in the common block.
dc The program does not conform to the standard for
dummy-to-common associations.
-prasm (D)
Use Series 10000 assembly language format for the expanded listing
generated by -exp. This option applies only if used with the -exp
option and with a compiler that generates Series 10000 code.
-nprasm
Use an alternate assembly-language format for the expanded listing
generated by -exp. This option applies only if used with the -exp
option and with a compiler that generates Series 10000 code.
-subchk
Perform subscript checking.
-nsubchk (D)
Suppress subscript checking.
-type
Flag all variables and functions not explicitly typed.
-ntype (D)
Suppress type checking.
-u Turn on case sensitivity for identifiers.
-uc Turn on UNIX-compatibility features: appended underscore to
subprogram and common block names; UNIX version of default
filenames; hidden character length variable passed as 4-byte integer
by value.
-nuc Turn off UNIX-compatibility features.
-version
Display version number of compiler.
-warn (D)
Display warning messages.
-nwarn
Suppress warning messages.
-xref
Generate symbol table map and cross reference (implies -l). You can
not use -exp and -xref on the same command line.
-nxref (D)
Suppress symbol table map and cross reference.
-xrs (D)
Save registers across a call to an external procedure or function.
-nxrs
Do not assume that calls to external routines have saved the
registers.
-zero
Initialize static storage to zero.
-nzero (D)
No implicit initialization of static storage to zero.
EXAMPLES
If there are no errors in the source code and the compilation proceeds
normally, the FORTRAN compiler outputs an object file and an optional
listing file. Unless you specify otherwise (with the -b switch), the
object module file has the same name as the source file, except that .bin
replaces .ftn as the suffix. Similarly, .lst replaces .ftn in the
listing file name. Thus, in response to the command
$ ftn graph_line.ftn -l
the FORTRAN compiler reads the file graph_line.ftn and produces an object
file named graph_line.bin and a listing file named graph_line.lst. If
errors occur during compilation, the compiler writes diagnostic messages
to error output and flags the incorrect statements in the listing file.