< prev index next >

src/java.base/share/classes/jdk/internal/math/FloatConsts.java

Print this page

        

@@ -1,7 +1,7 @@
 /*
- * Copyright (c) 2003, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2003, 2016, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
  * under the terms of the GNU General Public License version 2 only, as
  * published by the Free Software Foundation.  Oracle designates this

@@ -25,80 +25,53 @@
 
 package jdk.internal.math;
 
 /**
  * This class contains additional constants documenting limits of the
- * <code>float</code> type.
+ * {@code float} type.
  *
  * @author Joseph D. Darcy
  */
 
 public class FloatConsts {
     /**
      * Don't let anyone instantiate this class.
      */
     private FloatConsts() {}
 
-    public static final float POSITIVE_INFINITY = java.lang.Float.POSITIVE_INFINITY;
-    public static final float NEGATIVE_INFINITY = java.lang.Float.NEGATIVE_INFINITY;
-    public static final float NaN = java.lang.Float.NaN;
-    public static final float MAX_VALUE = java.lang.Float.MAX_VALUE;
-    public static final float MIN_VALUE = java.lang.Float.MIN_VALUE;
-
-    /**
-     * A constant holding the smallest positive normal value of type
-     * <code>float</code>, 2<sup>-126</sup>.  It is equal to the value
-     * returned by <code>Float.intBitsToFloat(0x00800000)</code>.
-     */
-    public static final float   MIN_NORMAL      = 1.17549435E-38f;
-
     /**
      * The number of logical bits in the significand of a
-     * <code>float</code> number, including the implicit bit.
+     * {@code float} number, including the implicit bit.
      */
     public static final int SIGNIFICAND_WIDTH   = 24;
 
     /**
-     * Maximum exponent a finite <code>float</code> number may have.
-     * It is equal to the value returned by
-     * <code>Math.ilogb(Float.MAX_VALUE)</code>.
-     */
-    public static final int     MAX_EXPONENT    = 127;
-
-    /**
-     * Minimum exponent a normalized <code>float</code> number may
-     * have.  It is equal to the value returned by
-     * <code>Math.ilogb(Float.MIN_NORMAL)</code>.
-     */
-    public static final int     MIN_EXPONENT    = -126;
-
-    /**
-     * The exponent the smallest positive <code>float</code> subnormal
+     * The exponent the smallest positive {@code float} subnormal
      * value would have if it could be normalized.
      */
-    public static final int     MIN_SUB_EXPONENT = MIN_EXPONENT -
+    public static final int     MIN_SUB_EXPONENT = Float.MIN_EXPONENT -
                                                    (SIGNIFICAND_WIDTH - 1);
 
     /**
-     * Bias used in representing a <code>float</code> exponent.
+     * Bias used in representing a {@code float} exponent.
      */
     public static final int     EXP_BIAS        = 127;
 
     /**
-     * Bit mask to isolate the sign bit of a <code>float</code>.
+     * Bit mask to isolate the sign bit of a {@code float}.
      */
     public static final int     SIGN_BIT_MASK   = 0x80000000;
 
     /**
      * Bit mask to isolate the exponent field of a
-     * <code>float</code>.
+     * {@code float}.
      */
     public static final int     EXP_BIT_MASK    = 0x7F800000;
 
     /**
      * Bit mask to isolate the significand field of a
-     * <code>float</code>.
+     * {@code float}.
      */
     public static final int     SIGNIF_BIT_MASK = 0x007FFFFF;
 
     static {
         // verify bit masks cover all bit positions and that the bit
< prev index next >