org.graalvm.compiler.core.common.calc

Class ReinterpretUtils

java.lang.Object

org.graalvm.compiler.core.common.calc.ReinterpretUtils

public class ReinterpretUtils
extends Object

Utilities for bitwise reinterpret operations on primitive types.

Constructor Summary

Constructors

Constructor and Description
ReinterpretUtils()

Method Summary

Modifier and Type	Method and Description
static IntegerStamp	floatToInt(FloatStamp stamp) Compute the IntegerStamp from a FloatStamp, losing as little information as possible.
static Constant	foldConstant(Stamp resultStamp, Constant constant) Reinterpret the constant as a constant matching the given stamp.
static FloatStamp	intToFloat(IntegerStamp stamp) Compute the IntegerStamp from a FloatStamp, losing as little information as possible.

Methods inherited from class java.lang.Object

clone, equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

ReinterpretUtils

public ReinterpretUtils()

Method Detail

foldConstant

public static Constant foldConstant(Stamp resultStamp,
                                    Constant constant)

Reinterpret the constant as a constant matching the given stamp.

floatToInt

public static IntegerStamp floatToInt(FloatStamp stamp)

Compute the IntegerStamp from a FloatStamp, losing as little information as possible. Sorting by their bit pattern reinterpreted as signed integers gives the following order of floating point numbers: -0 | negative numbers | -Inf | NaNs | 0 | positive numbers | +Inf | NaNs So we can compute a better integer range if we know that the input is positive, negative, finite, non-zero and/or not NaN.

intToFloat

public static FloatStamp intToFloat(IntegerStamp stamp)

Compute the IntegerStamp from a FloatStamp, losing as little information as possible. Sorting by their bit pattern reinterpreted as signed integers gives the following order of floating point numbers: -0 | negative numbers | -Inf | NaNs | 0 | positive numbers | +Inf | NaNs So from certain integer ranges we may be able to infer something about the sign, finiteness or NaN-ness of the result.