Bit Casting

Bril has an extension for bit-casting between ints and floats. Thus, this extension depends on the float extension.

Operations

There are two new instructions, both of them value operations:

• float2bits: Takes as input a single argument of type float and bitcasts the float to an int.
• bits2float: Takes as input a single argument of type int and bitcasts the integer to a float.

Syntax

The following JSON represents a bitcast from the float <input variable> to an integer, storing the result in <destination variable>:

{
    "op": "float2bits",
    "dest": "<destination variable>",
    "type": "int",
    "args": ["<input variable>"]
}

The following JSON represents a bitcast from the integer <input variable> to a float, storing the result in <destination variable>:

{
    "op": "bits2float",
    "dest": "<destination variable>",
    "type": "float",
    "args": ["<input variable>"]
}

Examples

The following textual excerpt of Bril converts an integer into a floating-point number (in this case, $0.1$):

as_int: int = const 4591870180066957722;
as_float: float = bits2float as_int;
print as_float;