HP OpenVMS Systems Documentation

HP OpenVMS RTL Library (LIB$) Manual

The Convert Floating-Point Data Type routine provides a simplified options-interface for converting a floating-point data type to another supported floating-point data type.

Format

CVT$CONVERT_FLOAT input_value, input_type_code, output_value, output_type_code, options

RETURNS

OpenVMS usage:	cond_value
type:	longword (unsigned)
access:	write only
mechanism:	by value

Arguments

input_value

OpenVMS usage:	varying_arg
type:	unspecified
access:	read only
mechanism:	by reference

The address of a data area containing a floating-point number that is to be converted. The input_value argument may contain floating-point data in F_Floating, D_Floating, G_Floating, H_Floating, IEEE_S_Floating, IEEE_T_Floating, IEEE_X_Floating, IBM_Long_Floating, IBM_Short_Floating, or CRAY_Floating format. The value of the input_type_code argument determines the format and size of the input_value argument.

input_type_code

OpenVMS usage:	longword_unsigned
type:	longword (unsigned)
access:	read only
mechanism:	by value

The value of a longword bit mask specifying the type of floating-point data being passed in the input_value argument. Valid type codes are:

input_type_code	Format	Size in Bytes
CVT$K_VAX_F	F_Floating	4
CVT$K_VAX_D	D_Floating	8
CVT$K_VAX_G	G_Floating	8
CVT$K_VAX_H	H_Floating	16
CVT$K_IEEE_S	IEEE_S_Floating	4
CVT$K_IEEE_T	IEEE_T_Floating	8
CVT$K_IEEE_X	IEEE_X_Floating	16
CVT$K_IBM_LONG	IBM_Long_Floating	8
CVT$K_IBM_SHORT	IBM_Short_Floating	4
CVT$K_CRAY	CRAY_Floating	8

Declarations for the input_type_code argument are in the $CVTDEF module found in the system symbol libraries.

output_value

OpenVMS usage:	varying_arg
type:	unspecified
access:	write only
mechanism:	by reference

The address of a data area that receives the converted floating-point number. The output_value argument can contain floating-point data in F_Floating, D_Floating, G_Floating, H_Floating, IEEE_S_Floating, IEEE_T_Floating, IEEE_X_Floating, IBM_Long_Floating, IBM_Short_Floating, or CRAY_Floating format. The value of the output_type_code argument determines the size and format of the data placed into the output_value argument.

output_type_code

OpenVMS usage:	longword_unsigned
type:	longword (unsigned)
access:	read only
mechanism:	by value

The value of a longword bit mask specifying the type of floating-point data that the input_value argument will be converted into and returned in the output_value argument. Valid type codes are:

output_type_code	Format	Size in Bytes
CVT$K_VAX_F	F_Floating	4
CVT$K_VAX_D	D_Floating	8
CVT$K_VAX_G	G_Floating	8
CVT$K_VAX_H	H_Floating	16
CVT$K_IEEE_S	IEEE_S_Floating	4
CVT$K_IEEE_T	IEEE_T_Floating	8
CVT$K_IEEE_X	IEEE_X_Floating	16
CVT$K_IBM_LONG	IBM_Long_Floating	8
CVT$K_IBM_SHORT	IBM_Short_Floating	4
CVT$K_CRAY	CRAY_Floating	8

Declarations for the output_type_code argument are in the $CVTDEF module found in the system symbol libraries.

options

OpenVMS usage:	mask_longword
type:	longword (unsigned)
access:	read only
mechanism:	by value

Conversion option specifier. The options argument is the address of a longword bit mask in which each option bit set causes the corresponding option to be used during the conversion.

The following options can be specified using the options argument:

Option	Description
CVT$M_ROUND_TO_NEAREST	The default rounding option for conversions to IEEE data types. This IEEE Std. 754 rounding mode results in the representable output value nearest to the infinitely precise result. If the two nearest representable values are equally near, the one whose least significant bit is 0 is the result.
CVT$M_VAX_ROUNDING	The default rounding option for conversions to non-IEEE data types. Performs "traditional" style rounding. This mode results in the representable output value nearest to the infinitely precise result. If the two nearest representable values are equally near, the output value is the closest to either positive infinity or negative infinity, depending on the sign of the input value.
CVT$M_TRUNCATE	Round the output value toward zero (truncate).
CVT$M_ROUND_TO_POS	Round the output value toward positive infinity.
CVT$M_ROUND_TO_NEG	Round the output value toward negative infinity.
CVT$M_BIG_ENDIAN	Interprets IEEE data types as Big Endian.
CVT$M_ERR_UNDERFLOW	Report underflow conditions as errors.

Declarations for the options argument are in the $CVTDEF module found in the system symbol libraries.

Description

CVT$CONVERT_FLOAT is a general-purpose, floating-point conversion routine that converts any input_type_code floating-point data type into any output_type_code floating-point data type. The conversion is subject to the options specified in the options argument.

Note OpenVMS compilers do not support arithmetic operations for all of the above floating-point data types. Additional floating-point data types are supported by this routine for data conversion purposes only.

Condition Values Returned

CVT$_NORMAL	Normal successful completion.
CVT$_INPCONERR	Input conversion error.
CVT$_INVINPTYP	Invalid input type code.
CVT$_INVOPT	Invalid option argument.
CVT$_INVOUTTYP	Invalid output type code.
CVT$_INVVAL	Input value was an invalid number or NaN.
CVT$_NEGINF	Input value was negative infinity.
CVT$_OUTCONERR	Output conversion error.
CVT$_OVERFLOW	Overflow detected during conversion.
CVT$_POSINF	Input value was positive infinity.
CVT$_UNDERFLOW	Underflow detected during conversion.

Return status values are in the $CVTMSG module found in the system symbol libraries.

Example

/*
** ===================================================================
**
**  Example of CVT$CONVERT_FLOAT
**
** -------------------------------------------------------------------
**
**  This example program reads IEEE T floating-point numbers from an
**  input file, converts them to VAX D floating-point numbers and
**  writes the result to an output file.
**
**  The input and output file names can be specified as the first and
**  second arguments on the command line as follows:
**
**     $ EXAMPLE IEEE_T_INPUT_FILE.DAT VAX_D_OUTPUT_FILE.DAT
**
**  If the input or output files are not included on the command
**  line then the program prompts the user for them.
**
** ===================================================================
*/
#include <stdio.h>

unsigned long CVT$CONVERT_FLOAT(void    *input_value,
                                unsigned long input_type,
                                void    *output_value,
                                unsigned long output_type,
                                unsigned long options);

globalvalue CVT$K_VAX_D;
globalvalue CVT$K_IEEE_T;
globalvalue CVT$M_ROUND_TO_NEAREST;
globalvalue CVT$_NORMAL;

main(int argc, char *argv[])

{

  double        D_Float_number;
  unsigned long IEEE_Double_number[2];
  unsigned long options;
  char          in_filename[80];
  char          out_filename[80];
  FILE          *in_file, *out_file;
  unsigned long ret_status;

/*
** Find out where we are going to get the data from.
** First look at the first argument of the command line.
** If nothing is there, then attempt to use IEEE_T_IN.DAT.
** =================================================================
*/
if (argc == 1)
{
  printf("Enter input data file: [IEEE_T_IN.DAT]: ");
  if (gets(in_filename) == NULL)
    exit(1);

  if (strlen(in_filename) == 0)
    strcpy(in_filename, "IEEE_T_IN.DAT");
}
else
  strcpy(in_filename, argv[1]);

/*
** Find out where we are going to put the output data.
** First look at the second argument of the command line.
** If nothing is there, then put it in VAX_D_OUT.DAT
** =================================================================
*/
if (argc <= 2)
{
  printf("Enter output data file: [VAX_D_OUT.DAT]: ");
  if (gets(out_filename) == NULL)
    exit(1);

  if (strlen(out_filename) == 0)
    strcpy(out_filename, "VAX_D_OUT.DAT");
}
else
  strcpy(out_filename, argv[2]);

/*
** Open the input and output files.
** -----------------------------------------------------------------
*/
if ((in_file = fopen(in_filename, "r")) == NULL)
{
  fprintf(stderr, "%s couldn't open file %s\n", argv[0], in_filename);
  exit(1);
}

out_file = fopen(out_filename, "wb");

options     = CVT$M_ROUND_TO_NEAREST;
ret_status  = CVT$_NORMAL;

/*
** Read in each number from the file, convert it, and write it to
** the output file.
** =================================================================
*/
while ((fread (&IEEE_Double_number[0],
               sizeof(IEEE_Double_number),
               1,
               in_file) == 1) &&
       (ret_status == CVT$_NORMAL))
{
  ret_status = CVT$CONVERT_FLOAT(&IEEE_Double_number[0], CVT$K_IEEE_T,
                                 &D_Float_number, CVT$K_VAX_D,
                                 options);

  if (ret_status == CVT$NORMAL)
  {
    fwrite(&D_Float_number, sizeof(D_Float_number), 1, out_file);
    printf("Converted data: %lf.\n", D_Float_number);
  }
}
fclose(in_file);
fclose(out_file);

if (ret_status == CVT$_NORMAL)
  exit(1);
else
  exit(ret_status);
}
      

The Convert Floating-Point Data Type routine converts floating-point data types to other supported floating-point data types and allows additional control over the converted results. CVT$FTOF functionality is also available on other platforms supported by HP.

Format

status = CVT$FTOF input_value, input_type_code, output_value, output_type_code, options

RETURNS

OpenVMS usage:	mask_longword
type:	longword (unsigned)
access:	write only
mechanism:	by value

The status return value is an unsigned longword bit mask containing the condition codes raised by the function. CVT$FTOF returns CVT$K_NORMAL; otherwise, it sets one or more recoverable and unrecoverable conditions. Use the following condition names to determine which conditions are set:

Condition Name	Condition (always reported by default)
CVT$K_NORMAL	Normal successful completion.
CVT$M_INVALID_INPUT_TYPE	Invalid input type code.
CVT$M_INVALID_OUTPUT_TYPE	Invalid output type code.
CVT$M_INVALID_OPTION	Invalid option argument.

Condition Name	Condition (reported only if the CVT$M_REPORT_ALL option is selected)
CVT$M_RESULT_INFINITE	Conversion produced an infinite result. 1
CVT$M_RESULT_DENORMALIZED	Conversion produced a denormalized result. 1
CVT$M_RESULT_OVERFLOW_RANGE	Conversion yielded an exponent greater than 60000 (8). 2
CVT$M_RESULT_UNDERFLOW_RANGE	Conversion yielded an exponent less than 20000 (8). 2
CVT$M_RESULT_UNNORMALIZED	Conversion produced an unnormalized result. 3
CVT$M_RESULT_INVALID	Conversion result is either ROP (reserved operand), NaN (not a number), or closest equivalent. CRAY and IBM data types return 0. 4
CVT$M_RESULT_OVERFLOW	Conversion resulted in overflow. 4
CVT$M_RESULT_UNDERFLOW	Conversion resulted in underflow. 4
CVT$M_RESULT_INEXACT	Conversion resulted in a loss of precision. 4

Return status values are in the $CVTDEF module in the system symbol libraries.

Arguments

input_value

OpenVMS usage:	varying_arg
type:	unspecified
access:	read only
mechanism:	by reference

The address of a data area containing a floating-point number to be converted. The number can be floating-point data in one of the following formats:

F_Floating	Big_Endian_IEEE_S_Floating
D_Floating	Big_Endian_IEEE_T_Floating
G_Floating	Big_Endian_IEEE_X_Floating
H_Floating	IBM_Long_Floating
IEEE_S_Floating	IBM_Short_Floating
IEEE_T_Floating	CRAY_Floating_Single
IEEE_X_Floating

The value of the input_type_code argument determines the format and size of the input_value argument.

input_type_code

OpenVMS usage:	longword_unsigned
type:	longword (unsigned)
access:	read only
mechanism:	by value

The value of a longword bit mask specifying the type of floating-point data being passed in the input_value argument. Valid type codes are:

Input_type_code	Format	Size in Bytes
CVT$K_VAX_F	F_Floating	4
CVT$K_VAX_D	D_Floating	8
CVT$K_VAX_G	G_Floating	8
CVT$K_VAX_H	H_Floating	16
CVT$K_IEEE_S	IEEE_S_Floating	4
CVT$K_IEEE_T	IEEE_T_Floating	8
CVT$K_IEEE_X	IEEE_X_Floating	16
CVT$K_BIG_ENDIAN_IEEE_S	Big_Endian_IEEE_S_Floating	4
CVT$K_BIG_ENDIAN_IEEE_T	Big_Endian_IEEE_T_Floating	8
CVT$K_BIG_ENDIAN_IEEE_X	Big_Endian_IEEE_X_Floating	16
CVT$K_IBM_LONG	IBM_Long_Floating	8
CVT$K_IBM_SHORT	IBM_Short_Floating	4
CVT$K_CRAY_SINGLE	CRAY_Floating	8

Declarations for the input_type_code argument are in the $CVTDEF module found in the system symbol libraries.

output_value

OpenVMS usage:	varying_arg
type:	unspecified
access:	write only
mechanism:	by reference

The address of a data area that receives the converted floating-point number. The number can be floating-point data in F_Floating, D_Floating, G_Floating, H_Floating, IEEE_S_Floating, IEEE_T_Floating, IEEE_X_Floating, Big_Endian_IEEE_S_Floating, Big_Endian_IEEE_T_Floating, Big_Endian_IEEE_X_Floating, IBM_Long_Floating, IBM_Short_Floating, or CRAY_Floating_Single format. The value of the output_type_code argument determines the size and format of the converted floating-point number.

output_type_code

OpenVMS usage:	longword_unsigned
type:	longword (unsigned)
access:	read only
mechanism:	by value

The value of a longword bit mask specifying the type of floating-point data that the input_value argument will be converted into and returned in the output_value argument. Valid type codes are:

Output_type_code	Format	Size in Bytes
CVT$K_VAX_F	F_Floating	4
CVT$K_VAX_D	D_Floating	8
CVT$K_VAX_G	G_Floating	8
CVT$K_VAX_H	H_Floating	16
CVT$K_IEEE_S	IEEE_S_Floating	4
CVT$K_IEEE_T	IEEE_T_Floating	8
CVT$K_IEEE_X	IEEE_X_Floating	16
CVT$K_BIG_ENDIAN_IEEE_S	Big_Endian_IEEE_S_Floating	4
CVT$K_BIG_ENDIAN_IEEE_T	Big_Endian_IEEE_T_Floating	8
CVT$K_BIG_ENDIAN_IEEE_X	Big_Endian_IEEE_X_Floating	16
CVT$K_IBM_LONG	IBM_Long_Floating	8
CVT$K_IBM_SHORT	IBM_Short_Floating	4
CVT$K_CRAY_SINGLE	CRAY_Floating	8

Declarations for the output_type_code argument are in the $CVTDEF module found in the system symbol libraries.

options

OpenVMS usage:	mask_longword
type:	longword (unsigned)
access:	read only
mechanism:	by value

Conversion option specifier. The options argument is the address of a longword bit mask in which each option bit set causes the corresponding option to be used during the conversion. Provide a zero (0) value to the options argument to select default behavior or choose one or more options (status condition option, rounding options, "FORCE" options, CRAY and IBM options) from the following tables. Specify only the options that apply to your conversion. A conflicting or incompatible options argument is reported as an error (CVT$M_INVALID_OPTION).

Applicable Conversion	Option	Description
	Status Condition Option
All	CVT$M_REPORT_ALL	Report all applicable status conditions as the default. The reporting of recoverable status conditions is disabled by default when this option is not used.
	Rounding Options
All	CVT$M_ROUND_TO_NEAREST	The default rounding option for conversions to IEEE data types. This IEEE Std. 754 rounding mode results in the representable output value nearest to the infinitely precise result. If the two nearest representable values are equally near, the one whose least significant bit is 0 is the result.
All	CVT$M_BIASED_ROUNDING	The default rounding option for conversions to non-IEEE data types. Performs "traditional" style rounding. This mode results in the representable output value nearest to the infinitely precise result. If the two nearest representable values are equally near, the output value is the closest to either positive infinity or negative infinity depending on the sign of the input value.
All	CVT$M_ROUND_TO_ZERO	Round the output value toward zero (truncate).
All	CVT$M_ROUND_TO_POS	Round the output value toward positive infinity.
All	CVT$M_ROUND_TO_NEG	Round the output value toward negative infinity.
	"FORCE" Options
All	CVT$M_FORCE_ALL_SPECIAL_VALUES	Apply all applicable "FORCE" options for the current conversion.
IEEE	CVT$M_FORCE_DENORM_TO_ZERO 1	Force a denormalized IEEE output value to zero.
IEEE	CVT$M_FORCE_INF_TO_MAX_FLOAT 1	Force a positive IEEE infinite output value to +max_float and force a negative IEEE infinite output value to --max_float.
IEEE or VAX	CVT$M_FORCE_INVALID_TO_ZERO 2	Force an invalid IEEE NaN (not a number) output value or a VAX ROP (reserved operand) output value to zero.
	CRAY Format Conversion Options
CRAY	CVT$M_ALLOW_OVRFLW_RANGE_VALUES	Allow an input/output exponent value > 60000 (8).
CRAY	CVT$M_ALLOW_UDRFLW_RANGE_VALUES	Allow an input/output exponent value < 20000 (8).
IBM Format Conversion Option
IBM	CVT$M_ALLOW_UNNORMALIZED_VALUES	Allow unnormalized input arguments. Allow an unnormalized output value for a small value that would normalize to zero.