• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1#!/bin/sh
2#
3# intgamma.sh
4#
5# Last changed in libpng 1.6.0 [February 14, 2013]
6#
7# COPYRIGHT: Written by John Cunningham Bowler, 2013.
8# To the extent possible under law, the author has waived all copyright and
9# related or neighboring rights to this work.  This work is published from:
10# United States.
11#
12# Shell script to generate png.c 8-bit and 16-bit log tables (see the code in
13# png.c for details).
14#
15# This script uses the "bc" arbitrary precision calculator to calculate 32-bit
16# fixed point values of logarithms appropriate to finding the log of an 8-bit
17# (0..255) value and a similar table for the exponent calculation.
18#
19# "bc" must be on the path when the script is executed, and the math library
20# (-lm) must be available
21#
22# function to print out a list of numbers as integers; the function truncates
23# the integers which must be one-per-line
24function print(){
25   awk 'BEGIN{
26      str = ""
27   }
28   {
29      sub("\\.[0-9]*$", "")
30      if ($0 == "")
31         $0 = "0"
32
33      if (str == "")
34         t = "   " $0 "U"
35      else
36         t = str ", " $0 "U"
37
38      if (length(t) >= 80) {
39         print str ","
40         str = "   " $0 "U"
41      } else
42         str = t
43   }
44   END{
45      print str
46   }'
47}
48#
49# The logarithm table.
50cat <<END
51/* 8-bit log table: png_8bit_l2[128]
52 * This is a table of -log(value/255)/log(2) for 'value' in the range 128 to
53 * 255, so it's the base 2 logarithm of a normalized 8-bit floating point
54 * mantissa.  The numbers are 32-bit fractions.
55 */
56static const png_uint_32
57png_8bit_l2[128] =
58{
59END
60#
61bc -lqws <<END | print
62f=65536*65536/l(2)
63for (i=128;i<256;++i) { .5 - l(i/255)*f; }
64END
65echo '};'
66echo
67#
68# The exponent table.
69cat <<END
70/* The 'exp()' case must invert the above, taking a 20-bit fixed point
71 * logarithmic value and returning a 16 or 8-bit number as appropriate.  In
72 * each case only the low 16 bits are relevant - the fraction - since the
73 * integer bits (the top 4) simply determine a shift.
74 *
75 * The worst case is the 16-bit distinction between 65535 and 65534; this
76 * requires perhaps spurious accuracy in the decoding of the logarithm to
77 * distinguish log2(65535/65534.5) - 10^-5 or 17 bits.  There is little chance
78 * of getting this accuracy in practice.
79 *
80 * To deal with this the following exp() function works out the exponent of the
81 * frational part of the logarithm by using an accurate 32-bit value from the
82 * top four fractional bits then multiplying in the remaining bits.
83 */
84static const png_uint_32
85png_32bit_exp[16] =
86{
87END
88#
89bc -lqws <<END | print
90f=l(2)/16
91for (i=0;i<16;++i) {
92   x = .5 + e(-i*f)*2^32;
93   if (x >= 2^32) x = 2^32-1;
94   x;
95}
96END
97echo '};'
98echo
99#
100# And the table of adjustment values.
101cat <<END
102/* Adjustment table; provided to explain the numbers in the code below. */
103#if 0
104END
105bc -lqws <<END | awk '{ printf "%5d %s\n", 12-NR, $0 }'
106for (i=11;i>=0;--i){
107   (1 - e(-(2^i)/65536*l(2))) * 2^(32-i)
108}
109END
110echo '#endif'
111