欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页

lucene3.5之SmallFloat

程序员文章站 2022-05-15 11:23:45
...
package org.apache.lucene.util;
/**
 * Copyright 2005 The Apache Software Foundation
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */


/** Floating point numbers smaller than 32 bits.
 *
 * @lucene.internal
 */
public class SmallFloat {

  /** Converts a 32 bit float to an 8 bit float.
   * <br>Values less than zero are all mapped to zero.
   * <br>Values are truncated (rounded down) to the nearest 8 bit value.
   * <br>Values between zero and the smallest representable value
   *  are rounded up.
   *
   * @param f the 32 bit float to be converted to an 8 bit float (byte)
   * @param numMantissaBits the number of mantissa bits to use in the byte, with the remainder to be used in the exponent
   * @param zeroExp the zero-point in the range of exponent values
   * @return the 8 bit float representation
   */
  public static byte floatToByte(float f, int numMantissaBits, int zeroExp) {
    // Adjustment from a float zero exponent to our zero exponent,
    // shifted over to our exponent position.
    int fzero = (63-zeroExp)<<numMantissaBits;
    int bits = Float.floatToRawIntBits(f);
    int smallfloat = bits >> (24-numMantissaBits);
    if (smallfloat <= fzero) {
      return (bits<=0) ?
        (byte)0   // negative numbers and zero both map to 0 byte
       :(byte)1;  // underflow is mapped to smallest non-zero number.
    } else if (smallfloat >= fzero + 0x100) {
      return -1;  // overflow maps to largest number
    } else {
      return (byte)(smallfloat - fzero);
    }
  }

  /** Converts an 8 bit float to a 32 bit float. */
  public static float byteToFloat(byte b, int numMantissaBits, int zeroExp) {
    // on Java1.5 & 1.6 JVMs, prebuilding a decoding array and doing a lookup
    // is only a little bit faster (anywhere from 0% to 7%)
    if (b == 0) return 0.0f;
    int bits = (b&0xff) << (24-numMantissaBits);
    bits += (63-zeroExp) << 24;
    return Float.intBitsToFloat(bits);
  }


  //
  // Some specializations of the generic functions follow.
  // The generic functions are just as fast with current (1.5)
  // -server JVMs, but still slower with client JVMs.
  //

  /** floatToByte(b, mantissaBits=3, zeroExponent=15)
   * <br>smallest non-zero value = 5.820766E-10
   * <br>largest value = 7.5161928E9
   * <br>epsilon = 0.125
   */
  public static byte floatToByte315(float f) {
    int bits = Float.floatToRawIntBits(f);
    int smallfloat = bits >> (24-3);
    if (smallfloat <= ((63-15)<<3)) {
      return (bits<=0) ? (byte)0 : (byte)1;
    }
    if (smallfloat >= ((63-15)<<3) + 0x100) {
      return -1;
    }
    return (byte)(smallfloat - ((63-15)<<3));
 }

  /** byteToFloat(b, mantissaBits=3, zeroExponent=15) */
  public static float byte315ToFloat(byte b) {
    // on Java1.5 & 1.6 JVMs, prebuilding a decoding array and doing a lookup
    // is only a little bit faster (anywhere from 0% to 7%)
    if (b == 0) return 0.0f;
    int bits = (b&0xff) << (24-3);
    bits += (63-15) << 24;
    return Float.intBitsToFloat(bits);
  }


  /** floatToByte(b, mantissaBits=5, zeroExponent=2)
   * <br>smallest nonzero value = 0.033203125
   * <br>largest value = 1984.0
   * <br>epsilon = 0.03125
   */
  public static byte floatToByte52(float f) {
    int bits = Float.floatToRawIntBits(f);
    int smallfloat = bits >> (24-5);
    if (smallfloat <= (63-2)<<5) {
      return (bits<=0) ? (byte)0 : (byte)1;
    }
    if (smallfloat >= ((63-2)<<5) + 0x100) {
      return -1;
    }
    return (byte)(smallfloat - ((63-2)<<5));
  }

  /** byteToFloat(b, mantissaBits=5, zeroExponent=2) */
  public static float byte52ToFloat(byte b) {
    // on Java1.5 & 1.6 JVMs, prebuilding a decoding array and doing a lookup
    // is only a little bit faster (anywhere from 0% to 7%)
    if (b == 0) return 0.0f;
    int bits = (b&0xff) << (24-5);
    bits += (63-2) << 24;
    return Float.intBitsToFloat(bits);
  }
}


这个类的功能是:

1、对于小于0的置为0

2、对于0与小的可表示值之间的值向上舍入

3、其余的向下舍入到8位值

首先,我们看看什么是符点数

浮点数是属于有理数中某特定子集的数的数字表示,在计算机中用以近似表示任意某个实数。具体的说,这个实数由一个整数或定点数(即尾数)乘以某个基数(计算机中通常是2)的整数次幂得到,这种表示方法类似于基数为10的科学记数法。

然后,我们来看一下IEEE754标准是如何描述符点数

IEEE754标准

  IEEE754代码
  标准表示法
  为便于软件的移植,浮点数的表示格式应该有统一标准(定义)。1985年IEEE(Institute of Electrical and Electronics Engineers)提出了IEEE754标准。该标准规定基数为2,阶码E用移码表示,尾数M用原码表示,根据二进制的规格化方法,最高数字位总是1,该标准将这个1缺省存储,使得尾数表示范围比实际存储的多一位。实数 的IEEE754标准的浮点数格式为:
  具体有三种形式:
  IEEE754三种浮点数的格式参数
  
类型 存储位数


偏移值

数符(s) 阶码(E) 尾数(M) 总位数 十六进制 十进制
短实数(Single,Float) 1位 8位 23位 32位 0x7FH +127
长实数(Double) 1位 11 位 52位 64位 0x3FFH +1023
临时实数(延伸双精确度,不常用) 1位 15位 64位 80位 0x3FFFH +16383
  对于阶码为0或为255(2047)的情况,IEEE有特殊的规定:
  如果 E 是0 并且 M 是0,这个数±0(和符号位相关) 如果 E = 2 − 1 并且 M 是0,这个数是 ±无穷大(同样和符号位相关) 如果 E = 2 − 1 并且 M 非0,这个数表示为不是一个数(NaN)。
  标准浮点数的存储在尾数中隐含存储着一个1,因此在计算尾数的真值时比一般形式要多一个整数1。对于阶码E的存储形式因为是127的偏移,所以在计算其移码时与人们熟悉的128偏移不一样,正数的值比用128偏移求得的少1,负数的值多1,为避免计算错误,方便理解,常将E当成二进制真值进行存储。例如:将数值-0.5按IEEE754单精度格式存储,先将-0.5换成二进制并写成标准形式:-0.5(10进制)=-0.1(2进制)=-1.0×2-1(2进制,-1是指数),这里s=1,M为全0,E-127=-1,E=126(10进制)=01111110(2进制),则存储形式为:
  1 01111110 000000000000000000000000=BF000000(16进制) 
 
  public static byte floatToByte(float f, int numMantissaBits, int zeroExp)完成将float转化为byte的任务,可以按指定尾数位数和指数值范围的零点位置
其中调用了Float.floatToRawIntBits(float value)
其功能是:根据 IEEE 754 的浮点“单一形式”中的位布局,返回指定浮点值的表示形式,并保留非数字 (NaN) 值。
numMantissaBits:在byte中使用的尾数位数

zeroExp:在指数值范围内的零点
f:需要转化的32位float
返回8位float,对精度有影响

  

 

public static float byteToFloat(byte b, int numMantissaBits, int zeroExp)

8位到32位flloat

 

public static byte floatToByte315(float f)

相当于:floatToByte(b, mantissaBits=3, zeroExponent=15)

 

  public static float byte315ToFloat(byte b)

相当于: byteToFloat(b, mantissaBits=3, zeroExponent=15) 

 

 public static byte floatToByte52(float f)

相当于 :floatToByte(b, mantissaBits=5, zeroExponent=2)

public static float byte52ToFloat(byte b)

相当于:byteToFloat(b, mantissaBits=5, zeroExponent=2)

相关标签: lucene java