Skip to content

ctor

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History
 
 

ctor

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
benchmark
benchmark
 
 
docs
docs
 
 
examples
examples
 
 
include/stdlib/number/float16
include/stdlib/number/float16
 
 
lib
lib
 
 
src
src
 
 
test
test
 
 
README.md
README.md
 
 
manifest.json
manifest.json
 
 
package.json
package.json
 
 

README.md

Float16

16-bit half-precision floating-point number.

Usage

var Float16 = require( '@stdlib/number/float16/ctor' );

Float16( value )

16-bit half-precision floating-point number constructor.

var x = new Float16( 5.0 );
// returns <Float16>

Properties

Float16.name

Static property returning the constructor name.

var str = Float16.name;
// returns 'Float16'

Float16.BYTES_PER_ELEMENT

Size (in bytes) of the underlying value.

var nbytes = Float16.BYTES_PER_ELEMENT;
// returns 2

Float16.prototype.BYTES_PER_ELEMENT

Size (in bytes) of the underlying value.

var x = new Float16( 5.0 );

var nbytes = x.BYTES_PER_ELEMENT;
// returns 2

Instance

A Float16 instance has the following properties...

value

A read-only property returning the underlying value as a number.

var x = new Float16( 5.0 );

var v = x.value;
// returns 5.0

Methods

Accessor Methods

These methods do not mutate a Float16 instance and, instead return a half-precision floating-point number representation.

Float16.prototype.toString()

Returns a string representation of a Float16 instance.

var x = new Float16( 5.0 );
var str = x.toString();
// returns '5'

x = new Float16( -3.14 );
str = x.toString();
// returns '-3.140625'

Float16.prototype.toJSON()

Returns a JSON representation of a Float16 instance. JSON.stringify() implicitly calls this method when stringifying a Float16 instance.

var x = new Float16( 5.0 );

var o = x.toJSON();
/*
  {
    "type": "Float16",
    "value": 5.0
  }
*/

To revive a Float16 number from a JSON string, see @stdlib/number/float16/reviver.

Float16.prototype.valueOf()

Converts a Float16 instance to a primitive value.

var x = new Float16( 5.0 );
var v = x.valueOf();
// returns 5.0

x = new Float16( 3.14 );
v = x.valueOf();
// returns 3.140625

Notes

  • The underlying value is stored as a half-precision floating-point number IEEE 754 with 1 sign bit, 5 exponent bits, and 10 significand bits.
  • A half-precision floating-point number has a range of approximately ±6.55e4 and a precision of about 3-4 decimal digits.

Examples

var Float16 = require( '@stdlib/number/float16/ctor' );

var x = new Float16( 3.14 );

console.log( 'type: %s', typeof x );
// => 'type: object'

console.log( 'str: %s', x );
// => 'str: 3.140625'

console.log( 'value: %d', x.value );
// => 'value: 3.140625'

console.log( 'JSON: %s', JSON.stringify( x ) );
// => 'JSON: {"type":"Float16","value":3.140625}'

C APIs

Usage

#include "stdlib/number/float16/ctor.h"

stdlib_float16_t

An opaque type definition for a half-precision floating-point number.

stdlib_float16_t v = stdlib_float16_from_bits( 51648 );

stdlib_float16_bits_t

An opaque type definition for a union for accessing the underlying binary representation of a half-precision floating-point number.

#include <stdint.h>

stdlib_float16_t x = stdlib_float16_from_bits( 51648 );

stdlib_float16_bits_t y;
y.value = x;

uint16_t bits = y.bits;
// returns 51648

The union has the following members:

  • value: stdlib_float16_t half-precision floating-point number.
  • bits: uint16_t binary representation.

The union allows "type punning"; however, while (more or less) defined in C99, behavior is implementation-defined in C++. For more robust conversion, prefer using explicit helpers for converting to and from binary representation.

stdlib_float16_from_bits( bits )

Converts a 16-bit binary representation to a half-precision floating-point number.

stdlib_float16_t v = stdlib_float16_from_bits( 51648 ); // => -11.5

The function accepts the following arguments:

  • bits: [in] uint16_t 16-bit integer corresponding to a binary representation.

stdlib_float16_to_bits( x )

Converts a half-precision floating-point number to a 16-bit binary representation.

#include <stdint.h>

stdlib_float16_t v = stdlib_float16_from_bits( 51648 ); // => -11.5

uint16_t bits = stdlib_float16_to_bits( v );

The function accepts the following arguments:

  • x: [in] stdlib_float16_t half-precision floating-point number.

Notes

  • The stdlib_float16_t type should be treated as a storage and interchange type. Native hardware support for mathematical functions operating on half-precision floating-point numbers varies. As a consequence, for most operations, one should first promote to single-precision (i.e., float), perform the desired operation, and then downcast back to half-precision.

Examples

#include "stdlib/number/float16/ctor.h"
#include <stdint.h>
#include <stdio.h>

int main( void ) {
  const stdlib_float16_t x[] = {
    stdlib_float16_from_bits( 51648 ), // -11.5
    stdlib_float16_from_bits( 18880 )  // 11.5
  };

  int i;
  for ( i = 0; i < 2; i++ ) {
    printf( "%d\n", stdlib_float16_to_bits( x[ i ] ) );
  }
}