NAG Fortran Compiler Release 6.1
NAG Fortran Compiler Release 6.1
Table of Contents
- 1 Name
- 2 Usage
- 3 Description
- 4 File Types
- 5 Compiler Options
- 6 Files
- 7 Compilation Messages
- 8 Compiler Limits
- 9 Input/Output Information
- 10 OpenMP Support
- 11 Automatic File Preconnection
- 12 IEEE 754 Arithmetic Support
- 13 Random Number Algorithm
- 14 Automatic Garbage Collection
- 15 Memory Tracing
- 16 Undefined Variable Detection
- 17 Data Types
- 18 Modules
- 19 Runtime Environment Variables
- 20 Debugging
- 21 Producing a Call Graph
- 22 Dependency Analysis
- 23 Generating Interfaces
- 24 Source File Polishing
- 25 Unifying Precision
- 26 See Also
- 27 Supplementary Information
- 28 Bugs
- 29 Author
nagfor — NAG Fortran Compiler Release 6.1
nagfor [mode] [option]... file...
nagfor is the interface to the NAG Fortran Compiler system. The compiler translates programs written in Fortran into executable programs, relocatable binary modules, assembler source files or C source files.
The mode determines the action performed, and can be one of
- Compile (and/or link) the files; this is the default mode if none is specified.
- Produce a callgraph of the Fortran routines in the files (see the Producing a Call Graph section).
- Produce a dependency analysis of the Fortran files (see the Dependency Analysis section).
- Produce a module or INCLUDE file containing procedure interfaces (see the Generating Interfaces section).
- Pretty-print (polish) the Fortran files (see the Source File Polishing section).
- Unify the precision of floating-point and complex entities in the files (see the Unifying Precision section).
Options that do not apply to the current mode of operation (e.g. polish options when the mode is for compilation) are ignored.
A file ending in ‘.f90’ or ‘.f95’ is taken to be a Fortran free-form source file, a file ending in ‘.f’, ‘.for’ or ‘.ftn’ is taken to be a Fortran fixed-form source file; these assumptions can be overridden with the -fixed or -free option. A file ending in ‘.ff90’ or ‘.ff95’ is taken to be a free-form file requiring preprocessing by fpp, and a file ending in ‘.ff’ is taken to be a fixed-form file requiring preprocessing by fpp. A file ending in ‘.F90’ or ‘.F95’ is taken to be a free-form file requiring preprocessing by fpp, and a file ending in ‘.F’ is taken to be a fixed-form files requiring preprocessing by fpp.
If a filename without a suffix is provided nagfor will look for a file with the suffix ‘.f95’, and if that does not exist, the suffix ‘.f90’.
Non-intrinsic modules and INCLUDE files are expected to exist in the current working directory or in a directory named by the -I option.
- Increase the length of each fixed source form input line from 72 characters to 132 characters. This has no effect on free source form input.
- Specify the ABI to compile for, either 32 (the 32-bit ABI),
or one of the 64-bit ABIs: 64, 64c or 64t.
The differences between the three 64-bit ABIs are as follows:
ABI Object size represented in Character length represented in -abi=64 32 bits 32 bits -abi=64t 64 bits 32 bits -abi=64c 64 bits 64 bits
Programs compiled with -abi=32 will run on any x86 Linux system; those compiled with any 64-bit ABI will only run on a 64-bit kernel.
The default is -abi=64t. The -abi=64c option is compatible with the ABI used by Intel Fortran. The -abi=64 option should only be used to generate code compatible with NAGWare Release 5.1(346) or earlier, as it will be withdrawn shortly.
- Specify static or dynamic binding. This only has effect if specified during the link phase. The default is dynamic binding.
- Compile only (produce .o file for each source file), do not link the .o files to produce an executable file. This option is equivalent to -otype=obj.
- Compile with all but the most expensive runtime checks; this omits the -C=dangling, -C=intovf and -C=undefined options.
- Compile checking code according to the value of check,
which must be one of:
all (perform all checks except for -C=undefined), array (check array bounds), bits (check bit intrinsic arguments), calls (check procedure references), dangling (check for dangling pointers), do (check DO loops for zero step values), intovf (check for integer overflow), none (do no checking: this is the default), present (check OPTIONAL references), pointer (check POINTER references), recursion (check for invalid recursion) or undefined (check for undefined variables).
The -C=undefined option is subject to a number of limitations; in particular, it is not binary compatible with Fortran code compiled without that option, and is not compatible with calling C code via a BIND(C) interface. See the Undefined Variable Detection section for further details.
- Colour the message output from the compiler using ANSI escape sequences and the default foreground colouring scheme which is: red for error messages (including fatal errors), blue for warning messages and green for information messages.
- Colour the message output from the compiler according to the specified
This is a comma-separated list of colour specifications, each consisting of a
message category name (“error”, “warn” or “info”) followed
by a colon and the foreground colour name, optionally followed by a plus sign
and the background colour name.
The colouring for unspecified categories will be the default.
Colours are: black, red, green, yellow, blue, magenta, cyan and white.
would be a rather garish colour scheme.
- Make external linkages compatible with other compilers where possible; on Windows this is Microsoft Fortran, on Mac OSX and Linux this is g77, g95 and gfortran, and on other systems this is the operating system vendor's compiler. This affects the naming convention and procedure calling convention (for example, on Windows it causes use of the “STDCALL” calling convention that is commonly used for most DLLs).
- Set the default conversion mode for unformatted files to format.
This format may be overridden by an explicit CONVERT= specifier in the
OPEN statement, or by the environment variable FORT_CONVERTn
(where n is the unit number).
The value of format must be one of the following (not case-sensitive):
Format Description BIG_ENDIAN synonym for BIG_IEEE BIG_IEEE_DD big-endian with IEEE floating-point, quad precision is double-double BIG_IEEE big-endian with IEEE floating-point, including quad precision BIG_NATIVE big-endian with native floating-point format LITTLE_ENDIAN synonym for LITTLE_IEEE LITTLE_IEEE_DD little-endian with IEEE floating-point, quad precision is double-double LITTLE_IEEE little-endian with IEEE floating-point, including quad precision LITTLE_NATIVE little-endian with native floating-point format NATIVE no conversion (the default)
- Defines name to fpp as a preprocessor variable. This only affects files that are being preprocessed by fpp.
- Enable recognition of non-standard double precision complex intrinsic
functions. These act as specific versions of the standard generic intrinsics
Non-standard Equivalent Standard Fortran Generic Intrinsic Function CDABS(A) ABS(A) DCMPLX(X,Y) CMPLX(X,Y,KIND=KIND(0d0)) DCONJG(Z) CONJG(Z) DIMAG(Z) AIMAG(Z) DREAL(Z) REAL(Z) or DBLE(Z)
- Double the size of default INTEGER, LOGICAL, REAL and COMPLEX. Entities specified with explicit kind numbers or byte lengths are unaffected. If quadruple precision REAL is available, the size of DOUBLE PRECISION is also doubled.
- Show but do not execute commands constructed by the compiler driver.
- Allows the compilation and execution of “legacy” software by downgrading the category of common errors found in such software from “Error” to “Warning” (which may then be suppressed entirely with the -w option). This option disables -C=calls, and also enables Hollerith i/o (see the -hollerith_io option).
- Specifies that the encoding system of the Fortran source files is charset, which must be one of ISO_Latin_1, Shift_JIS or UTF_8. If this option is not specified, the default encoding is UTF-8 for Fortran source files that begin with a UTF-8 Byte Order Mark, and ISO Latin-1 (if the language setting is English) or Shift-JIS (if the language setting is Japanese) for other Fortran source files.
- Produce compiler messages in English (default).
- Preprocess only, do not compile. Each file that is preprocessed will produce an output file of the same name with the suffix replaced by .f, .f90 or .f95 according to the suffix of the input file. This option is equivalent to -otype=Fortran.
- Use the Fortran 77/90 version of the SIGN intrinsic instead of the Fortran 95 one (they differ in the treatment of negative zero).
- Specify that the base language is Fortran 95. This only affects extension message generation (Fortran 2003 and 2008 features will be reported as extensions).
- Specify that the base language is Fortran 2003. This only affects extension message generation (Fortran 2008 features will be reported as extensions).
- Specify that the base language is Fortran 2008. This is the default.
- Interpret all Fortran source files according to fixed-form rules.
- Do not store floating-point variables in registers on machines with floating-point registers wider than 64 bits. This can avoid problems with excess precision.
- Preprocess the source files using fpp even if the suffix would normally indicate an ordinary Fortran file.
- Interpret all Fortran source files according to free-form rules.
- Produce information for interactive debugging by the host system debugger.
- Produce debugging information for dbx90, a Fortran 90 aware front-end to the host system debugger. This produces a debug information (.g90) file for each Fortran source file. This option must be specified for both compilation and linking.
- Enables automatic garbage collection of the executable program. This option must be specified for both compilation and linking, and is unavailable on IBM z9 OpenEdition and x64 Mac and Windows. It is incompatible with the -thread_safe and -mtrace options. For more details see the Automatic Garbage Collection section.
- Compile code to produce a traceback when a runtime error message is generated.
Only routines compiled with this option will appear in such a traceback.
This option increases both executable file size and execution time.
It is incompatible with the -thread_safe and
Runtime Error: Invalid input for real editing Program terminated by I/O error on unit 5 (Input_Unit,Formatted,Sequential) main.f90, line 28: Error occurred in READ_DATA main.f90, line 57: Called by READ_COORDS main.f90, line 40: Called by INITIAL main.f90, line 13: Called by $main$
- Display a one-line summary of the options available for the current mode (=compiler, =callgraph, =depend, =interfaces, =polish or =unifyprecision).
- Enable Fortran-66 compatible input/output of character data stored in numeric variables using the A edit descriptor. This was superseded by the CHARACTER datatype in Fortran 77.
- Accept the extensions to Fortran specified by the High Performance Fortran Forum in HPF 1.0. These consist of the EXTRINSIC keyword and a large number of compiler directives. The compiler directives are checked for correctness but have no effect on compilation.
- -I pathname
- Add pathname to the list of directories which are to be searched for module information (.mod) files and INCLUDE files. The current working directory is always searched first, then any directories named in -I options, then the compiler's library directory (see the -Qpath option).
- Set the size of default INTEGER and LOGICAL to 64 bits.
This can be useful for switching between libraries that have 32-bit integer
arguments (on one platform) and 64-bit integer arguments (on another platform),
but which do not provide a named constant with the necessary KIND value.
This has no effect on default REAL and COMPLEX sizes, so the compiler is not standard-conforming in this mode.
- -indirect file
- Read the contents of file as additional arguments to the compiler driver.
This option may also be given by “@file”; note in this case there
is no space between the ‘@’ and the file name.
In an indirect file, arguments may be given on separate lines; on a single line, multiple arguments may be separated by blanks. A blank can be included in an option or file name by putting the whole option or file name in quotes ("); this is the only quoting mechanism. An indirect file may reference other indirect files.
- Set the mode of IEEE arithmetic operation according to mode,
which must be one of full, nonstd or stop.
- enables all IEEE arithmetic facilities including non-stop arithmetic.
- Disables non-stop arithmetic, terminating execution on floating overflow, division by zero or invalid operand. If the hardware supports it, this also disables IEEE gradual underflow, producing zero instead of a denormalised number; this can improve performance on some systems.
- enables all IEEE arithmetic facilities except for non-stop arithmetic; execution will be terminated on floating overflow, division by zero or invalid operand.
The -ieee option must be specified when compiling the main program unit, and its effect is global. The default mode is -ieee=stop. For more details see the IEEE 754 Arithmetic Support section.
- Request output of information messages. The default is to suppress these messages.
- Specify the kind numbering system to be used; option must be one of
byte or sequential.
For -kind=byte, the kind numbers for INTEGER, REAL and LOGICAL will match the number of bytes of storage (e.g., default REAL is 4 and DOUBLE PRECISION is 8). Note that COMPLEX kind numbers are the same as its REAL components, and thus half of the total byte length in the entity.
For -kind=sequential (the default), the kind numbers for all datatypes are numbered sequentially from 1, increasing with precision (e.g., default REAL is 1 and DOUBLE PRECISION is 2).
This option does not affect the interpretation of byte-length specifiers (an extension to Fortran 77).
- Link with library libx.a. The linker will search for this library in the directories specified by -Ldir options followed by the normal system directories (see the ld(1) command).
- Add dir to the list of directories for library files (see the ld(1) command).
- Produce module information files (.mod files) only. This option is equivalent to -otype=mod.
- Set the maximum number of simultaneously active host instances of an internal procedure that is being passed as an actual argument, or assigned to a procedure pointer, to N. The default maximum is normally 30, and increased to 160 if either the -openmp or -thread_safe options are used.
- Set the maximum size of a PARAMETER to N MB (megabytes). N must be in the range 1 to 1048576 (1MB to 1TB); the default is 50 MB.
- Increase the limit on the number of continuation lines from 255 to N. This option will not decrease the limit below the standard number.
- -mdir dir
- Write any module information (.mod) files to directory dir instead of the current working directory.
- Set the encoding scheme for compiler messages to charset, which must be one of ISO_Latin_1, Shift_JIS or UTF_8 (not case-sensitive). The -message_encoding=ISO_Latin_1 option is incompatible with the -nihongo option. The default message encoding is Shift_JIS on Windows and UTF_8 on other systems.
- Downgrade consistency checking of procedure argument lists so that mismatches produce warning messages instead of error messages. This only affects calls to a routine which is not in the current file; calls to a routine in the file being compiled must still be correct. This option disables -C=calls.
- Further downgrade consistency checking of procedure argument lists so that calls to routines in the same file which are incorrect will produce warnings instead of error messages. This option disables -C=calls.
- Trace memory allocation and deallocation. This option is a synonym for -mtrace=on.
- Trace memory allocation and deallocation according to the value of
which must be a comma separated list of one or more of:
address (display addresses), all (all options except for off), line (display file/line info if known), off (disable tracing output), on (enable tracing output), paranoia (protect memory allocator data structures against the user program), size (display size in bytes) or verbose (all options except for off and paranoia ).
This option should be specified during both compilation and linking, and is incompatible with the -gc option. For more details see the Memory Tracing section.
- Initialise REAL and COMPLEX variables to IEEE Signalling NaN, causing a runtime crash if the values are used before being set. This affects local variables, module variables, and INTENT(OUT) dummy arguments only; it does not affect variables in COMMON or EQUIVALENCE.
- Produce compiler messages in Japanese (if necessary, the encoding can be changed by the -message_encoding= option).
- Suppress the warning message that normally appears if a floating-point underflow occurred during execution. This option is only effective if specified during the link phase.
- Do not check for .mod files being out of date.
- Suppress module information (.mod) file production. Combining this with -M will produce no output (other than error and warning messages) at all, equivalent to -otype=none.
- If no licence for the compiler is immediately available, exit with an error instead of queueing for it.
- -o output
- Name the output file output instead of the default. If an executable is being produced the default is a.out; otherwise it is file.o with the -c option, file.c with the -S option, and file.f, file.f90 or file.f95 with the -F option, where file is the base part of the source file (i.e. with the suffix removed).
- Normal optimisation, equivalent to -O2.
- Set the optimisation level to N.
The optimisation levels are:
- No optimisation. This is the default, and is recommended when debugging.
- Minimal quick optimisation.
- Normal optimisation.
- Further optimisation.
- Maximal optimisation.
- This is a synonym for -Oassumed=contig.
- Optimises assumed-shape array dummy arguments according to the value of
which must be one of
- Optimised for contiguous actual arguments. If the actual argument is not contiguous a runtime error will occur (the compiler is not standard-conforming under this option).
- Optimised for contiguous actual arguments; if the actual argument is not contiguous (i.e. it is an array section) a contiguous local copy is made. This may speed up array section accessing if a sufficiently large number of array element or array operations is performed (i.e. if the cost of making the local copy is less than the overhead of discontiguous array accesses), but usually makes such accesses slower. Note that this option does not affect dummy arguments with the TARGET attribute; these are always accessed via the dope vector.
- Optimised for low-moderate accesses to array section (discontiguous) actual arguments. This is the default.
Note that CHARACTER arrays are not affected by these options.
- Specify the dimension of the blocks used for evaluating the MATMUL intrinsic. The default value (only for -O1 and above) is system and datatype dependent.
- Disable the optimisation of constant propagation. This is the default for -O1 and lower.
- Enable the optimisation of constant propagation. This is the default for -O2 and higher.
- Specify that the program does not alter the default rounding mode. This enables the use of faster code for the ANINT intrinsic.
- Specify the depth to which simple loops and array operations should be unrolled. The default is no unrolling (i.e. a depth of 1) for -O0 and -O1, and a depth of 2 for -O and higher optimisation levels. It can be advantageous to disable the Fortran compiler's loop unrolling if the C compiler normally does a very good job itself — this can be accomplished with -Ounroll=1.
- Perform possibly unsafe optimisations that may depend on the numerical stability of the program.
- Recognise OpenMP directives and link with the OpenMP support library. For more details see the OpenMP Support section.
- Specify the type of output file required to filetype, which must be one of
c (C source file), exe (executable file), fortran (Fortran source file), mod (module information file), none (no output file), obj (object file).
The -c, -F and -M options are equivalent to -otype=obj, -otype=Fortran and -otype=mod respectively.
- Compile code to generate profiling information which is written at run-time to an implementation-dependent file (usually gmon.out or mon.out). An execution profile may then be generated using gprof. This option must be specified for compilation and linking and may be unavailable on some implementations.
- Produce position-independent code (small model), for use in a shared library. If the shared library is too big for the small model, use -PIC.
- Produce position-independent code (large model), for use in a shared library.
- -Qpath pathname
- Change the compiler library pathname from its default location to pathname. (The default location on Unix is usually ‘/usr/local/lib/NAG_Fortran’.)
- Double the size of default REAL and COMPLEX, and on machines for
which quadruple-precision floating-point arithmetic is available, double the
size of DOUBLE PRECISION (and the non-standard DOUBLE COMPLEX).
REAL or COMPLEX specified with explicit KIND numbers or
byte lengths are unaffected — but since the KIND intrinsic returns the
correct values, COMPLEX(KIND(0d0)) on a machine with quad-precision
floating-point will correctly select quad-precision COMPLEX.
This has no effect on INTEGER sizes, and so the compiler is not standard-conforming in this mode.
Note: This option has been superseded by the -double option which doubles the size of all numeric data types.
- Strip symbol table information from the executable file. This option is only effective if specified during the link phase.
- Produce assembler (actually C source code). The resulting .c file should be compiled with the NAG Fortran compiler, not with the C compiler directly. This option is equivalent to -otype=c.
- This is equivalent to inserting the SAVE statement in all subprograms which are not pure and not declared RECURSIVE, thus causing all non-automatic local variables in such subprograms to be statically allocated. It has no effect on variables in BLOCK constructs.
- Produce obsolescence warning messages for use of ‘CHARACTER*’ syntax. This message is not produced by default since many programs contain this syntax.
- Specify the machine for which code should be generated and optimised.
For x86-32 (x86-compatible 32-bit mode compilation on Linux, MacOSX and
Windows), machine may be one of
i486, i586, i686, pentium2, pentium3, pentium4, prescott the specified Intel processor, k6, k6-2, k6-3, k6-4, athlon, athlon-4, athlon-xp, athlon-mp the specified AMD processor, pentium (equivalent to i586) or pentiumpro (equivalent to i686).
The default is to compile for pentium4 on Linux and MacOSX, and prescott on Windows.
For x86-64 (x86-compatible 64-bit mode compilation on Linux, MacOSX and Windows), machine may be either athlon64 or nocona.
- -tempdir directory
- Set the directory used for the compiler's temporary files to directory. The default is to use the directory named by the TMPDIR environment variable, or if that is not set, /tmp on Unix-like systems and the Windows temporary folder on Windows.
- Compile code for safe execution in a multi-threaded environment. This must be specified when compiling and also during the link phase. It is incompatible with the -gc and -gline options.
- Report execution times for the various compilation phases.
- Specify that IMPLICIT NONE is in effect by default, unless overridden by explicit IMPLICIT statements.
- Specify default sharing of NONE in OpenMP PARALLEL and TASK constructs (including in combined constructs such as PARALLELDO). This has the same effect as the DEFAULT(NONE) clause, unless overridden by an explicit DEFAULT(...) directive.
- Bind with the unshared (static) version of the Fortran runtime system; this allows a dynamically linked executable to be run on systems where the NAG Fortran Compiler is not installed. This option is only effective if specified during the link phase.
- Verbose. Print the name of each file as it is compiled.
- Print version information about the compiler.
- Suppress all warning messages. This option is a synonym for -w=all.
- Suppress the warning messages specified by class, which must be one of
all, alloctr, obs, ques, uda, uei, uep,
uip, ulv, unreffed, unused, uparam, usf,
usy, x77 or x95.
- suppresses all warning messages;
- suppresses warning messages about the use of allocatable components, dummy arguments and functions;
- suppresses warning messages about the use of obsolescent features;
- suppresses warning messages about questionable usage;
- suppresses warning messages about unused dummy arguments;
- suppresses warning messages about unused explicit imports;
- suppresses warning messages about unused external procedures;
- suppresses warning messages about unused intrinsic procedures;
- suppresses warning messages about unused local variables;
- suppresses warning messages about variables set but never referenced;
- suppresses warning messages about unused entities — this is equivalent to ‘-w=uda -w=uei -w=uep -w=uip -w=ulv -w=uparam -w=usf -w=usy’;
- suppresses warning messages about unused PARAMETERs;
- suppresses warning messages about unused statement functions;
- suppresses warning messages about unused symbols;
- suppresses extension warnings for obsolete but common extensions to Fortran 77 — these are TAB format, byte-length specifiers and Hollerith constants;
- suppresses extension warnings for extensions to modern Fortran (not just Fortran 95) that are not part of any Fortran standard.
- The -W
option can be used to specify the path to use for a compilation component or
to pass an option directly to such a component.
The possible combinations are:
- Specify the path used for the Fortran Compiler front-end. Note that this does not affect the library directory; the -Qpath option should be used to specify that.
- Specify the path to use for invoking the C compiler; this is used both for the final stage of compilation and for linking.
- Pass option directly to the host C compiler when compiling (producing the .o file). Multiple options may be specified in a single -Wc, option by separating them with commas.
- Specify the path to use for invoking the linker (producing the executable).
- Pass option directly to the host C compiler when linking (producing the executable). Multiple options may be specified in a single -Wl, option by separating them with commas. A comma may be included in an option by repeating it, e.g. -Wl,-filelist=file1,,file2,,file3 becomes the linker option -filelist=file1,file2,file3.
- Specify the path to use for invoking the fpp preprocessor.
- Pass option directly to fpp when preprocessing.
- Specify a list of external procedures for which to suppress argument data type and arrayness consistency checking. The procedure names should be separated by commas, e.g. -wmismatch=p_one,p2. Unlike the -mismatch option, this only affects data type and arrayness checking, and no warning messages are produced.
- Report on the availability of licences for the compiler instead of compiling anything.
|/opt/NAG_Fortran/lib||Default NAG Fortran Compiler library directory on Sun Solaris (see -Qpath); referred to as library hereafter.|
|/usr/local/lib/NAG_Fortran||Default NAG Fortran Compiler library directory on other Unix-based operating systems.|
|C:\Program Files\NAG\EFBuilder 6.1\nagfor\lib||Default NAG Fortran Compiler library directory on 32-bit Windows.|
|C:\Program Files (x86)\NAG\EFBuilder 6.1\nagfor\lib||Default NAG Fortran Compiler library directory on 64-bit Windows.|
The messages produced by the NAG Fortran Compiler itself during compilation are intended to be self-explanatory. The linker, or more rarely the host C compiler, may produce occasional messages.
Messages produced by the compiler are classified by severity level; these levels are:
|Info||informational message, noting an aspect of the source code in which the user may be interested.|
|Warning||the source code appears likely to be in error.|
|Questionable||some questionable usage has been found in the source code which may indicate a programming error. This has the same severity as “warning”.|
|Extension||some non-standard-conforming source code has been detected but has successfully been compiled as an extension to the language. This has the same severity as “warning”.|
|Obsolescent||some archaic source code has been detected which although standard-conforming was classified as obsolescent by the Fortran standard (selected according to the -f95, -f2003 and -f2008 options). This has the same severity as “warning”.|
|Deleted feature used||a feature that was present in an older Fortran standard but deleted from the Fortran standard selected by a -fN option was used. This has the same severity as “warning”.|
|Error||the source code does not conform to the Fortran standard or does not make sense. Compilation continues after recovery.|
|Fatal||a serious error in the user's program from which the compiler cannot recover, the compilation is immediately terminated.|
|Panic||an internal inconsistency is found by one of the compiler's self-checks; this is a bug in the compiler itself and NAG should be notified.|
|Maximum INCLUDE file nesting||20|
|Maximum number of INCLUDE file references per compilation||2047|
|Maximum DATA-implied-DO loop nesting||99|
|Maximum array-constructor-implied-DO loop nesting||99|
|Maximum number of dummy arguments||32767|
|Maximum number of arguments to MIN and MAX||100|
|Maximum character length (except as below)||2147483647|
|Maximum character length (64-bit Windows and -abi=64c Linux)||1099511627775 (240-1)|
|Maximum array size (32-bit systems)||2147483647 bytes|
|Maximum array size (64-bit systems)||1 TiB|
|Maximum unit number||2147483647|
|Maximum input/output record length||2147483647 bytes|
|Standard error (stderr) unit number||0|
|Standard input (stdin) unit number||5|
|Standard output (stdout) unit number||6|
|Default maximum record length for formatted output||1024 characters|
|Default maximum record length for unformatted output||2147483647 bytes|
The default directory used for files opened with STATUS='SCRATCH' is ‘/tmp’ on Unix and the Windows temporary directory on Windows. This default may be overridden with the TMPDIR environment variable.
OpenMP 3.1 is supported.
When using the IEEE arithmetic support modules, the IEEE modes (rounding, halting and underflow) are propagated into spawned OpenMP threads at the beginning of a PARALLEL construct, and any IEEE flag that are set by an OpenMP thread is passed back to the parent thread at the end of the PARALLEL construct.
The following table lists the OpenMP environment variables with their default values and, if applicable, their limits.
|OMP_NUM_THREADS||number of cores||1-32768|
|OMP_DYNAMIC||False||true or false|
|OMP_NESTED||False||true or false|
|OMP_STACKSIZE||0||<1GB (32-bit) or 16GB (64-bit)|
|OMP_WAIT_POLICY||None||active or passive|
Note that although the NAG runtime supports up to 32768 threads, operating system limits may prevent usage of so many.
OpenMP is not compatible with the -C=undefined and -gline options.
All logical unit numbers are automatically preconnected to specific files. These files need not exist and will only be opened or created if they are accessed with READ or WRITE without an explicit OPEN. By default the specific filename for unit n is fort.n; however if the environment variable FORTnn exists its value is used as the filename. Note that there are two digits in this variable name, e.g. the variable controlling unit 1 is FORT01 whereas the default filename is ‘fort.1’ (unless the prefix has been changed, see the description of module F90_PRECONN_IO).
A file preconnected in this manner is opened with ACCESS='SEQUENTIAL'. If the initial READ or WRITE is an unformatted i/o statement, it is opened with FORM='UNFORMATTED' otherwise it is opened with FORM='FORMATTED'. By default a formatted connection is opened with BLANK='NULL' and POSITION='REWIND' (see module F90_PRECONN_IO).
Automatic preconnection applies only to the initial use of a logical unit; once CLOSEd the unit will not be reconnected automatically but must be explicitly OPENed.
Note that this facility means that it is possible for a READ or WRITE statement with an IOSTAT= clause to receive an i/o error code associated with the implicit OPEN.
If no floating-point option is specified, any floating divide-by-zero, overflow or invalid operand exception will cause the execution of the program to be terminated (with an informative message and usually a core dump). Occurrence of floating underflow may be reported on normal termination of the program. On hardware supporting IEEE 754 standard arithmetic gradual underflow with denormalised numbers will be enabled. Note that this mode of operation is the only one available on hardware which does not support IEEE 754.
If the -ieee=full option is specified, non-stop arithmetic is enabled; thus REAL variables may take on the values +Infinity, −Infinity and NaN (Not-a-Number). If any of the floating exceptions listed above are detected by the hardware during execution, this fact will be reported on normal termination. The -ieee=full option must be specified when compiling the main program and has global effect.
If the -ieee=nonstd option is specified, floating-point exceptions are handled in the default manner (i.e. execution is terminated). However, gradual underflow is not enabled, so results which would have produced a denormalised number produce zero instead. This option can only be used on hardware for which this mode of operation is faster. Like -ieee=full, the -ieee=nonstd option must be specified when compiling the main program and has global effect.
The random number generator supplied as the intrinsic subroutine RANDOM_NUMBER is the “Mersenne Twister”.
Note that this generator has a large state (630 32-bit integers) and an extremely long period (approx 106000), and therefore it is strongly recommended that the RANDOM_SEED routine only be used with a PUT argument that is the value returned by a previous call with GET; i.e., only to repeat a previous sequence. This is because if a user-specified seed has low entropy (likely since there are 630 values to be supplied), it is highly likely to set the generator to an apparently-low-entropy part of the sequence.
If you do want to provide your own seed (and thus entropy), you should store your values in the initial elements of the seed array and set all the remaining elements to zero — trailing zero elements will be ignored and not used to initialise the generator. Note that the seed is a random bitstream, and is therefore expected to have approximately half of its bits nonzero (thus providing many small integer values will likely result in a low-entropy part of the Mersenne Twister sequence being reached).
The -gc option enables use of the runtime garbage collector. It is necessary to use this option during the link phase for it to have effect; specifying it additionally during the compilation phase can result in improved performance.
The supplied Technical Information note (TECHINFO) lists whether garbage collection is available for your system. If it is available, there will be a file ‘gc.o’ in the compiler's library directory.
The collector used is based on version 5.3 of the publicly available general purpose garbage collecting storage allocator of Hans-J Boehm, Alan J Demers and Xerox Corporation, described in “Garbage Collection in an Uncooperative Environment” (H Boehm and M Weiser, Software Practice and Experience, September 1988, pp 807-820).
The copyright notice attached to their latest version is as follows:
Copyright 1988, 1989 Hans-J. Boehm, Alan J. Demers Copyright (c) 1991-1995 by Xerox Corporation. All rights reserved. Copyright 1996-1999 by Silicon Graphics. All rights reserved. Copyright 1999 by Hewlett-Packard Company. All rights reserved. THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY EXPRESSED OR IMPLIED. ANY USE IS AT YOUR OWN RISK. Permission is hereby granted to use or copy this program for any purpose, provided the above notices are retained on all copies. Permission to modify the code and to distribute modified code is granted, provided the above notices are retained, and a notice that the code was modified is included with the above copyright notice.
Note that the “NO WARRANTY” disclaimer refers to the original copyright holders Boehm, Demers, Xerox Corporation, Silicon Graphics and Hewlett-Packard Company. The modified collector distributed in binary form with the NAG Fortran Compiler is subject to the same warranty and conditions as the rest of the NAG Fortran compilation system.
The module F90_GC is provided; it contains functions and variables that can control the behaviour of the garbage collector.
Tracing of memory allocation and deallocation is provided by the -mtrace option. Control is provided over whether the address, size, and line number of each allocation is displayed, or the tracing output can be suppressed entirely. A “paranoia” mode is provided where the memory allocator protects its data structures against inadvertent modification by the user program.
Runtime environment variables may be used to override the tracing options a program was built with, and to specify where to write the tracing output. These are only operative if the program was built with some tracing option; -mtrace=off will build a program with the tracing-capable memory allocator.
If -mtrace=off is not specified, use of any -mtrace option will implicitly do a -mtrace=on.
Basic tracing produces a message to the memory tracing file (normally standard error) for each allocation and deallocation, including those for automatic variables, i/o buffers and compiler-generated temporaries. Each allocation is numbered sequentially; the first three items are the i/o buffers for units 0, 5 and 6 (standard error, standard input and standard output).
All -mtrace= suboptions may be overridden at run time by the NAGFORTRAN_MTRACE_OPTIONS environment variable, which should be set to the required trace_opt_list (e.g. ‘on,size’). The memory tracing file may be specified at run time by the NAGFORTRAN_MTRACE_FILE environment variable.
The NAGFORTRAN_MTRACE_OPTIONS variable can also contain an option to limit the total amount of memory that may be allocated. The ‘limit=N’ option limits the maximum memory allocated to N MiB (mebibytes), but only if the program was built with a tracing option (minimally, -mtrace=off). Exceeding the memory limit will result in a normal “out of memory” condition, which if it occurs in an ALLOCATE statement, can be captured by a STAT= clause. Note that the memory limit applies to the overall memory usage including automatic variables and compiler-generated array temporaries.
The -mtrace option must be specified when linking, and is incompatible with -gc. Additionally, line number information is only available for those files compiled with -mtrace=line.
The nagfmcheck program can be used to check the output from the -mtrace option. It is designed to be used as a filter. Any lines that do not look like memory tracing output are ignored. It reports to standard output any errors it detects such as deallocating something twice, deallocating something that was never allocated, or deallocating something with a size different from that with which it was allocated. It also reports any apparent memory leaks, though this is less useful if the program terminated prematurely.
Use of undefined variables can be detected with the -C=undefined option. Program units compiled with this option use a different ABI, which means that they are incompatible with program units compiled without this option, and not interoperable with C; thus the whole program must be Fortran code and compiled the same way. For this reason, -C=undefined is not part of -C or -C=all.
Currently, there are a number of other limitations on the use of -C=undefined.
- It is incompatible with pointers in an initialised COMMON.
- All intrinsic modules are available, but the ISO_C_BINDING module can only be used with all-Fortran programs as the option makes changes to the ABI.
- Internal READ from a CHARACTER array requires the entire specified array subobject to be “defined”, even those elements corresponding to records not actually read.
- Internal WRITE to a CHARACTER array is considered to define the entire specified array subobject, even those elements corresponding to records not actually written.
- Certain intrinsic functions require the entirety of their arguments to be defined, even if some portions are not actually required for the value of the function. For example, the PAD argument to RESHAPE when no padding is actually required, and elements of the ARRAY argument to PACK that correspond to false elements of the MASK.
- It is incompatible with the use of OpenMP directives.
- It cannot be used on types with length type parameters.
- It cannot be used when CLASS(*) variables are allocated using the MOLD= specifier.
The table below lists the data types provided by the NAG Fortran Compiler together with their kind numbers. There are two possibilities for the KIND numbers: the default mode of operation (which may be specified explicitly by the -kind=sequential option) and the “byte” numbering scheme (specified by the -kind=byte option).
|Type||KIND Number||KIND Number||Description|
|REAL||1||4||Single precision floating-point|
|REAL||2||8||Double precision floating-point|
|REAL||3||16||Quadruple precision floating-point|
|COMPLEX||1||4||Single precision complex|
|COMPLEX||2||8||Double precision complex|
|COMPLEX||3||16||Quadruple precision complex|
|LOGICAL||1||1||Single byte logical|
|LOGICAL||2||2||Double byte logical|
|LOGICAL||4||8||Eight byte logical|
|INTEGER||3||4||32-bit (default) integer|
|CHARACTER||1||1||ASCII or ISO 8859-1 character|
|CHARACTER||2||2||JIS X 0213 character|
|CHARACTER||3||3||Unicode (UCS-2) character|
|CHARACTER||4||4||ISO 10646 (UCS-4) character|
Note that on all machines except Sun Solaris with the SunPro C compiler, quadruple precision is actually “double double” precision; this provides nearly twice the precision of Double precision but with a reduced exponent range.
The F90_KIND module contains named parameters useful for specifying which kind you want regardless of whether the numbering system is “sequential” or “byte”.
Additional intrinsic modules built into the NAG Fortran Compiler are described in the “nag_modules” document.
To use a module it must be previously compiled, or must be defined in the file prior to its use. When separately compiling a module the -c option should be specified.
Compiling a module creates a ‘.mod’ file and a ‘.o’ file. The ‘.mod’ file is used by the compiler at compile time to provide information about module contents, the ‘.o’ file (if generated) contains the code of any module procedures and must be specified when creating an executable file.
Note that the name of the ‘.mod’ file will be the name of the module, the ‘.o’ file will be named after the original source file.
When a previously compiled module is USEd the NAG Fortran Compiler attempts to find its source file and, if that is successful, checks the modification times producing a warning message if the ‘.mod’ file is out of date.
The following variables control the runtime environment for programs compiled with the NAG Fortran Compiler.
- Programs compiled using any -mtrace= option will write the memory trace to this file. The default is standard error.
- Changes the memory tracing options for programs compiled using any -mtrace= option.
- Runtime error messages will be written to this file. The default is standard error.
- Controls the language used for runtime error messages. This may be ‘English’ or ‘Japanese’ (not case-sensitive); the default is English.
- Controls the directory used for scratch files (the default is system-dependent).
For operating systems other than Windows a Modern Fortran-aware debugger might be available as dbx90; see TECHINFO.txt for details.
In general, host system debuggers, such as dbx or gdb, may be used successfully on Fortran code as the names of the original source files, plus line numbers, are passed through to the intermediate C files. In using such debuggers it should be noted that most local variables have an underscore appended to their names. It may be useful to look at the intermediate C code when debugging; this is produced by the -S option.
The call graph generator takes a set of Fortran source files and produces a call graph with optional index and called-by tables. C files and fpp-processed files are not handled.
The call graph generator understands the following compiler options with the same meaning: -132, -dcfuns, -double, -dryrun, -dusty, -encoding, -english, -f2003, -f2008, -f95, -fixed, -free, -help, -hpf, -I, -i8, -indirect, -info, -kind, -max_parameter_size, -maxcontin, -mismatch, -mismatch_all, -nihongo, -nocheck_modtime, -nomod, -noqueue, -o, -openmp, -Qpath, -r8, -strict95, -thread_safe, -u, -u=sharing, -v, -V, -w and -xlicinfo.
The “@filename” syntax may also be used, with the same effect as the “-indirect filename” option.
The call graph is written to the file specified by the -o option, or to standard output if no -o option is specified.
The following additional options control the output produced.
- Produce a “called-by” table showing, for each routine, the routines that call it directly or indirectly. This is produced at the end of the output.
- Indent by N for each level in the graph, up to the maximum. The default is -indent=4.
- The maximum indentation is N. The default is -indent_max=70.
- Produce an alphabetic index listing, for each routine, the line of the call graph where the routine first appears. This follows the call graph itself and precedes the called-by table (when the -calledby option is used).
- Show ENTRY point names in the call graph; without this option, calls to an ENTRY point are shown as calls to the containing subprogram.
- If a call is via a generic identifier, show the generic identifier in the call graph.
- Show the host scope names for calls to internal and module procedures.
- Show the class of each procedure (e.g. ‘module’, ‘internal’, ...).
- If a called procedure was renamed on a USE statement, show the renaming.
The dependency analyser takes a set of Fortran source files and produces dependency information in the form specified. C files and fpp-processed files are not handled.
The dependency analyser understands the following compiler options with the same meaning: -132, -dryrun, -english, -fixed, -free, -help, -hpf, -I, -indirect, -nihongo, -o, -Qpath, -tempdir, -v and -V. The “@filename” syntax may also be used with the same effect as the “-indirect filename” option.
The following additional options control the operation of the dependency analyser:
- This option controls the output form, type must be one of:
blist (the filenames as an ordered build list), dfile (the dependencies in Makefile format, written to separate file.d files), info (the dependencies as English descriptions) or make (the dependencies in Makefile format).
The default is -otype=info. If -otype=dfile is specified, no -o option is permitted; otherwise, the result is written to the file specified by the -o option or to standard output if no -o option is specified.
- Specifes the form to use for dependency paths; pathtype must be either absolute or relative. With -paths=absolute, paths for INCLUDE files that are relative specifications will be prefixed by the current working directory.
The interface generator takes a set of Fortran source files and produces interfaces for the procedures therein. The output is either a module (in a new source file), or an INCLUDE file.
The interfaces are written either to the file specified by the -o option, or if module output is being produced to the file with the same name as the module and extension ‘.f90’, or otherwise (an INCLUDE file is being produced) to ‘interfaces.inc’. In each case the interfaces are all within a single INTERFACE block.
The interface generator understands the following compiler options with the same meaning: -132, -dcfuns, -double, -dryrun, -dusty, -encoding, -english, -f2003, -f2008, -f95, -fixed, -free, -help, -hpf, -I, -i8, -indirect, -info, -kind, -max_parameter_size, -maxcontin, -mismatch, -mismatch_all, -nihongo, -nocheck_modtime, -noqueue, -o, -openmp, -Qpath, -r8, -strict95, -tempdir, -thread_safe, -u, -u=sharing, -v, -V, -w and -xlicinfo.
The interface generator understands the following polish options with the same meaning: -align_right_continuation, -array_constructor_brackets, -break_long_comment_word, -character_decl, -idcase, -indent, -indent_comment_marker, -indent_comments, -indent_continuation, -indent_max, -kwcase, -margin, -name_scopes, -noalign_right_continuation, -nobreak_long_comment_word, -noindent_comment_marker, -noindent_comments, -width. Note that unlike nagfor =polish, -name_scopes=Asis acts as if it were -name_scopes=Keywords, and is the default. Also, -array_constructor_brackets=Asis acts as if it were -array_constructor_brackets=ParenSlash, and is the default.
The following additional options control the operation of the interface generator:
- Add a comment before the INTERFACE statement, giving the date and time that the file was generated. This is the default.
- Add a comment after each procedure heading (SUBROUTINE or FUNCTION statement) indicating the source of the procedure.
- Specifies the name of the module to generate containing procedure interfaces. The default is ‘interfaces’.
- Specify the type of output file required to type, which must be one of
include (INCLUDE file), module (Fortran module in a new source file).
The default is -otype=module.
- Do not add any comment before the INTERFACE statement.
- Do not add any comment after each procedure heading. This is the default.
The polisher takes a set of Fortran source files, which may be in fixed or free form, and produces a free form “polished” version of each file. C files and fpp-processed files are not handled.
The polisher understands the following compiler options with the same meaning: -132, -encoding, -english, -fixed, -free, -help, -I, -indirect, -info, -nihongo, -noqueue, -o, -Qpath, -tempdir, -u, -v, -V, -w and -xlicinfo.
The polished output is written to the file specified by the -o option, or to the same filename with the extension replaced by ‘.f90_pol’ if no -o option is specified. The output file cannot have the same name as the input file.
The following additional options control the operation of the polisher:
- Align the continuation markers (ampersands) at the end of a continued line to column N+2, where N is the normal line width (specified by the -width= option). This only affects lines that do not end with an inline comment.
- Enable options to alter comments; without this option, any options that would otherwise alter the comments are ignored.
- Specify the form to use for array constructor delimiters; X must be one of Asis (same as the input file), ParenSlash (use parentheses+slash pairs, i.e. ‘(/ ... /)’) or Square (use square brackets, i.e. ‘[ ... ]’). The default is -array_constructor_brackets=Asis.
- Turn comment lines that have no text (other than the comment-initiating character) into plain blank lines; this is the default if the -alter_comments option is set.
- Ensure that there is a blank line after the declarations and before the first executable statement; this is the default.
- Specify whether to write a Unicode Byte-Order Mark at the beginning of the output file; X must be one of Asis (same as the input file), Insert (insert a byte-order mark) or Remove (remove any byte-order mark). This option only has effect if the input file is known to be in UTF-8 encoding, either because it begins with a byte-order mark or the -encoding=UTF8 option was used. The default is -bom=Asis.
- If a comment line will be split into two lines, the comment may be broken in the middle of a long word.
- Specify the style to be used for CHARACTER type declaration statements;
style must be one of the following (not case-sensitive):
Asis (same as the input statement, but obey any -kind_keyword= option), Keywords (use LEN= and KIND=), Kind_Keyword_Only (use KIND= but not LEN=) or No_Keywords (use modern style with no keywords).
The default is Asis; with any other style, the obsolescent “CHARACTER*length” form will be changed to the modern “CHARACTER(length)” form. When both the length and kind appear in the input statement, the length will appear first in the output statement.
- Specify whether to add optional commas in FORMAT statements; X must be one of Asis (use the same comma scheme as the input), Insert or Remove. The default is -commas_in_formats=Insert.
- Delete all comments (if the -alter_comments option is set).
- Delete blank lines and comment lines that have no text (other than the comment-initiating character), if the -alter_comments option is set.
- Delete labels that are never referenced; this is the default.
- If renumbering FORMAT statements in a separate sequence, the first FORMAT statement will be N; the default is -format_start=90000.
- If renumbering FORMAT statements in a separate sequence, the step from one label to the next will be N; the default is -format_step=10. Note that this may be negative (but not zero).
- Set the case to use for identifiers; X must be one of C (for Capitalised), L (for lowercase) or U (for UPPERCASE); the default is -idcase=L.
- Indent statements within a construct by N spaces from the current indentation level; the default is -indent=2.
- When indenting comments, the comment-initiating character should be indented to the indentation level; this is the default.
- Indent comments; this is the default if the -alter_comments option is set. The result is also affected by the -indent_comment_marker option.
- Indent continuation lines by an additional N spaces; the default is -indent_continuation=2.
- Set the maximum indentation level to N spaces; the default is -indent_max=60. The value must be at least 10 less than the output line length (-width=).
- Set the indentation level for inline comments to column N; the default is -inline_comment_index=35.
- Do not delete blank lines or comment lines with no text; this is the opposite of -delete_blank_lines and is the default.
- Do not delete non-blank comment lines; this is the opposite of -delete_comments and is the default.
- Do not delete unused (unreferenced) labels; this is the opposite of -delete_unused_labels.
- Specifies how to handle the KIND= specifier in declarations; X must be one of Asis (take no action but preserve the input status), Insert (insert KIND= if not present), or Remove (remove KIND= if present); the default is -kind_keyword=Asis.
- Set the case to use for language keywords; X must be one of C (for Capitalised), L (for lowercase) or U (for UPPERCASE); the default is -kwcase=C.
- Indent labels; this is the opposite to -label_before_indent.
- Output the statement label, if any, before indenting the statement; this is the default.
- Leave FORMAT statements in the same position as they are in the input file; this is the opposite of -move_formats_to_end, and is the default.
- Set the left margin (initial indent) to N. The value must be at least 10 less than the output line length (-width=). The default value for the left margin is 4.
- Move FORMAT statements to the end of the subprogram or program unit, immediately before the CONTAINS or END statement.
- Specify whether to add optional keywords and scope names to the END or END TYPE statement for a scope; X must be one of Asis (leave as is), Insert (insert keywords and/or names), Keywords (insert keywords but remove names) or Remove (remove optional keywords and names). This option also applies to the END INTERFACE statement. The default is -name_scopes=Insert.
- Do not align the continuation markers (ampersands) at the end of continued lines; this is the default.
- Do not alter comments in any way; this is the default.
- Do not turn blank comments to blank lines.
- Do not insert a blank line between the last declaration and the first executable statement.
- If a comment line will be split into two lines, do not break the comment in the middle of a long word; this is the default.
- Place the comment-initiating character for a comment line in column 1.
- Do not indent the text of a comment line.
- Do not renumber statement labels.
- When renumbering statement labels, use a single sequence for both FORMAT and non-FORMAT statements; this is the default.
- Do not change DO loop terminating statements.
- Do not wrap long comment lines (they will still be indented if comments are being indented).
- Specifies the form to use for relational operators, X must be either F77- (use .EQ., .LE., etc.) or F90+ (use ==, <=, etc.); the default is -relational=F90+.
- Renumber statement labels; this is the default.
- When renumbering statement labels, the first label will be N; the default is -renumber_start=100.
- When renumbering statement labels, the step from one label to the next will be N; the default value is -renumber_step=10.
- When renumbering statement labels, renumber FORMAT statements in a separate sequence from non-FORMAT statements.
- Change the terminating statements of all DO loops so that each loop ends with an ENDDO statement; this is the default.
- Set the maximum length of the text on each output line to N; the default is -width=78. Note that in the case of continuation lines, an additional two characters (‘ &’) will be produced after the last text on a line and this may take the line length over the limit. The width must be at least 10 more than the left margin (-margin=) and the maximum indent (-indent_max=). The maximum width setting is 1024, however values higher than 130 will produce output that does not conform to the Fortran standard.
- Wrap long comment lines that would otherwise exceed the maximum line length. This is the default.
The precision unifier standardises floating-point and complex variable declarations, floating-point and complex literal constants, and some specific (non-generic) intrinsic procedures in a set of Fortran source files in order to unify the precision of these entities.
Standardisation to quadruple precision is only available on machines for which quadruple-precision floating-point arithmetic is available.
The tool attempts to make a standardising precision parameter accessible in program units (and interface blocks) via a use statement. You can control the form of this statement: the -pp_name= option controls the name of the precision parameter, and the -pp_module= option supplies the name of its host module (which is known as the ‘precision module’). The default form for the use statement (when no options are specified) is USE WORKING_PRECISION, ONLY: WP.
The -precision= option (whose default value is Double) can be supplied to set the desired unifying precision. The tool will use this setting when performing a number of checks of the validity of the standardisation process on the input files.
The precision module can be created by the tool, but otherwise does not itself undergo precision unification. A warning is issued if the tool encounters this module. A message is also emitted if no definition for the precision parameter is found in the module, or otherwise if the defined precision parameter specifies a different kind to the desired precision as provided or implied by the -precision= option.
The tool searches each program unit and interface block in the input source and determines whether the precision parameter is already accessible. If it is not, then a use statement, in the form given above, is inserted in the last allowable position for its statement type. For an internal or module procedure this statement is placed in the host. If the precision parameter is already declared in the form INTEGER, PARAMETER :: wp = constant_expression, then this declaration is deleted and a new use statement added, as previously described. (This PARAMETER form of statement is only recognised as declaring the precision parameter if it precedes all declarations of floating-point or complex entities in the scoping unit.) Any other form of definition or import of the precision parameter will not be modified, and the tool issues a warning that the standardised use statement could not be inserted.
Type declarations for floating-point and complex entities are standardised to include the precision parameter as kind parameter. Entities that are implicitly typed to be floating-point or complex are explicitly declared, in the same form. In the case when a function is defined with a floating-point or complex type specification on the function statement, this specification is deleted and a distinct type declaration statement for the function result is inserted into the function's declaration section.
Floating-point and complex literal constants are standardised to use the precision parameter as their kind.
The following non-generic procedures are standardised to the listed form:
|Specific Name||Generic Replacement||Specific Name (cont.)||Generic Replacement (cont.)|
(See also the description of -dcfuns.)
Furthermore, DBLE is converted to REAL. Following that, the KIND= argument is added to calls to REAL and CMPLX, when appropriate.
In cases where unifying the precision of the input source may lead in the generated output to undesirable side effects, or even invalid Fortran, the tool will attempt to issue a warning alerting you to the possibility. Here is a non-exhaustive list of situations where it may be inappropriate to apply this tool.
- Your source intentionally uses a mix of floating-point and complex precisions.
- You are employing Fortran language features for generic programming (such as generic interface blocks or parameterised derived types).
- You have floating-point or complex data in EQUIVALENCE statements or in references to the TRANSFER intrinsic.
- You have explicitly-typed intrinsic functions, or are passing intrinsic functions as procedure arguments.
- You are using the DPROD intrinsic (perhaps as a means of performing higher- (double-) precision computations in a single-precision program unit).
For procedures spread across several files clearly it is desirable to make sure this tool is applied to all files consistently. This will ensure, for example, that procedure references and the corresponding procedure definitions do not become inconsistent with respect to the type standardisation.
The precision unifier understands the following compiler options with the same meaning: -132, -abi, -dcfuns, -double, -dryrun, -dusty, -encoding, -english, -f2003, -f2008, -f95, -fixed, -free, -help, -hpf, -I, -i8, -indirect, -info, -kind, -max_parameter_size, -maxcontin, -mismatch, -mismatch_all, -nihongo, -nocheck_modtime, -nomod, -noqueue, -o, -openmp, -Qpath, -r8, -strict95, -tempdir, -thread_safe, -u, -u=sharing, -v, -V, -w and -xlicinfo.
Note that using the -double or -r8 option affects the meaning of the -precision= option; see the description of the latter, below.
The standardised output is written to the file specified by the -o option, or to the same filename with the extension replaced by ‘.f90_prs’ if no -o option is specified. The output file cannot have the same name as the input file.
The precision unifier understands all the polish options with the same meaning.
The following additional options control the operation of this tool:
- If creating the precision module, do not add a comment saying when it was generated.
- Automatically create the precision module, in the file whose name is the name
of the module, converted to lower case, with file type ‘.f90’; thus the
default filename is working_precision.f90.
If the file already exists it will not be overwritten by this option.
The created module will contain only the definition of the precision parameter, and unless the -nocmt_generation option is given, a comment identifying when the module was created.
- Specifies the name of the precision parameter to use in the standardised output, which must be a legal identifier that does not conflict with existing names in the input source. The default is ‘WP’.
- Specifies the name of the precision module from which the precision parameter is to be imported. This module name must be a legal identifier that does not conflict with existing names in the input source. The default is ‘WORKING_PRECISION’.
- Do not create the precision module. This is the default.
- Specifies the desired target unifying precision in the output; X must
be one of Single (i.e., same precision as default REAL), Double
(i.e., same precision as default DOUBLE PRECISION) or Quadruple.
The default is -precision=Double.
Note that, since -double and -r8 double the size of default REAL (and possibly default DOUBLE PRECISION), specifying -double or -r8 will likewise modify the meaning of this -precision= option too.
Please check the web page http://www.nag.co.uk/doc/inun/np61/supplementary.html for details of any new information related to the applicability or usage of this product.
Please report any bugs found to ‘firstname.lastname@example.org’ or ‘email@example.com’, along with any suggestions for improvements.
Malcolm Cohen, Nihon Numerical Algorithms Group KK, Tokyo, Japan.