FreeSWITCH API Documentation  1.7.0
g711.h
Go to the documentation of this file.
1 /*
2  * SpanDSP - a series of DSP components for telephony
3  *
4  * g711.h - In line A-law and u-law conversion routines
5  *
6  * Written by Steve Underwood <steveu@coppice.org>
7  *
8  * Copyright (C) 2001 Steve Underwood
9  *
10  * Despite my general liking of the GPL, I place this code in the
11  * public domain for the benefit of all mankind - even the slimy
12  * ones who might try to proprietize my work and use it to my
13  * detriment.
14  *
15  * $Id: g711.h,v 1.1 2006/06/07 15:46:39 steveu Exp $
16  */
17 
18 /*! \file */
19 
20 /*! \page g711_page A-law and mu-law handling
21 Lookup tables for A-law and u-law look attractive, until you consider the impact
22 on the CPU cache. If it causes a substantial area of your processor cache to get
23 hit too often, cache sloshing will severely slow things down. The main reason
24 these routines are slow in C, is the lack of direct access to the CPU's "find
25 the first 1" instruction. A little in-line assembler fixes that, and the
26 conversion routines can be faster than lookup tables, in most real world usage.
27 A "find the first 1" instruction is available on most modern CPUs, and is a
28 much underused feature.
29 
30 If an assembly language method of bit searching is not available, these routines
31 revert to a method that can be a little slow, so the cache thrashing might not
32 seem so bad :(
33 
34 Feel free to submit patches to add fast "find the first 1" support for your own
35 favorite processor.
36 
37 Look up tables are used for transcoding between A-law and u-law, since it is
38 difficult to achieve the precise transcoding procedure laid down in the G.711
39 specification by other means.
40 */
41 
42 #if !defined(FREESWITCH_G711_H)
43 #define FREESWITCH_G711_H
44 
45 #ifdef __cplusplus
46 extern "C" {
47 #endif
48 
49 #ifdef _MSC_VER
50 #ifndef __inline__
51 #define __inline__ __inline
52 #endif
53 #if !defined(_STDINT) && !defined(uint32_t)
54  typedef unsigned __int8 uint8_t;
55  typedef __int16 int16_t;
56  typedef __int32 int32_t;
57  typedef unsigned __int16 uint16_t;
58 #endif
59 #endif
60 
61 #if defined(__i386__)
62 /*! \brief Find the bit position of the highest set bit in a word
63  \param bits The word to be searched
64  \return The bit number of the highest set bit, or -1 if the word is zero. */
65  static __inline__ int top_bit(unsigned int bits) {
66  int res;
67 
68  __asm__ __volatile__(" movl $-1,%%edx;\n" " bsrl %%eax,%%edx;\n":"=d"(res)
69  :"a" (bits));
70  return res;
71  }
72  /*- End of function --------------------------------------------------------*//*! \brief Find the bit position of the lowest set bit in a word
73  \param bits The word to be searched
74  \return The bit number of the lowest set bit, or -1 if the word is zero. */ static __inline__ int bottom_bit(unsigned int bits) {
75  int res;
76 
77  __asm__ __volatile__(" movl $-1,%%edx;\n" " bsfl %%eax,%%edx;\n":"=d"(res)
78  :"a" (bits));
79  return res;
80  }
81 /*- End of function --------------------------------------------------------*/
82 #elif defined(__x86_64__)
83  static __inline__ int top_bit(unsigned int bits) {
84  int res;
85 
86  __asm__ __volatile__(" movq $-1,%%rdx;\n" " bsrq %%rax,%%rdx;\n":"=d"(res)
87  :"a" (bits));
88  return res;
89  }
90 /*- End of function --------------------------------------------------------*/ static __inline__ int bottom_bit(unsigned int bits) {
91  int res;
92 
93  __asm__ __volatile__(" movq $-1,%%rdx;\n" " bsfq %%rax,%%rdx;\n":"=d"(res)
94  :"a" (bits));
95  return res;
96  }
97 /*- End of function --------------------------------------------------------*/
98 #else
99  static __inline__ int top_bit(unsigned int bits) {
100  int i;
101 
102  if (bits == 0)
103  return -1;
104  i = 0;
105  if (bits & 0xFFFF0000) {
106  bits &= 0xFFFF0000;
107  i += 16;
108  }
109  if (bits & 0xFF00FF00) {
110  bits &= 0xFF00FF00;
111  i += 8;
112  }
113  if (bits & 0xF0F0F0F0) {
114  bits &= 0xF0F0F0F0;
115  i += 4;
116  }
117  if (bits & 0xCCCCCCCC) {
118  bits &= 0xCCCCCCCC;
119  i += 2;
120  }
121  if (bits & 0xAAAAAAAA) {
122  bits &= 0xAAAAAAAA;
123  i += 1;
124  }
125  return i;
126  }
127 /*- End of function --------------------------------------------------------*/
128 
129  static __inline__ int bottom_bit(unsigned int bits) {
130  int i;
131 
132  if (bits == 0)
133  return -1;
134  i = 32;
135  if (bits & 0x0000FFFF) {
136  bits &= 0x0000FFFF;
137  i -= 16;
138  }
139  if (bits & 0x00FF00FF) {
140  bits &= 0x00FF00FF;
141  i -= 8;
142  }
143  if (bits & 0x0F0F0F0F) {
144  bits &= 0x0F0F0F0F;
145  i -= 4;
146  }
147  if (bits & 0x33333333) {
148  bits &= 0x33333333;
149  i -= 2;
150  }
151  if (bits & 0x55555555) {
152  bits &= 0x55555555;
153  i -= 1;
154  }
155  return i;
156  }
157 /*- End of function --------------------------------------------------------*/
158 #endif
159 
160 /* N.B. It is tempting to use look-up tables for A-law and u-law conversion.
161  * However, you should consider the cache footprint.
162  *
163  * A 64K byte table for linear to x-law and a 512 byte table for x-law to
164  * linear sound like peanuts these days, and shouldn't an array lookup be
165  * real fast? No! When the cache sloshes as badly as this one will, a tight
166  * calculation may be better. The messiest part is normally finding the
167  * segment, but a little inline assembly can fix that on an i386, x86_64 and
168  * many other modern processors.
169  */
170 
171 /*
172  * Mu-law is basically as follows:
173  *
174  * Biased Linear Input Code Compressed Code
175  * ------------------------ ---------------
176  * 00000001wxyza 000wxyz
177  * 0000001wxyzab 001wxyz
178  * 000001wxyzabc 010wxyz
179  * 00001wxyzabcd 011wxyz
180  * 0001wxyzabcde 100wxyz
181  * 001wxyzabcdef 101wxyz
182  * 01wxyzabcdefg 110wxyz
183  * 1wxyzabcdefgh 111wxyz
184  *
185  * Each biased linear code has a leading 1 which identifies the segment
186  * number. The value of the segment number is equal to 7 minus the number
187  * of leading 0's. The quantization interval is directly available as the
188  * four bits wxyz. * The trailing bits (a - h) are ignored.
189  *
190  * Ordinarily the complement of the resulting code word is used for
191  * transmission, and so the code word is complemented before it is returned.
192  *
193  * For further information see John C. Bellamy's Digital Telephony, 1982,
194  * John Wiley & Sons, pps 98-111 and 472-476.
195  */
196 
197 //#define ULAW_ZEROTRAP /* turn on the trap as per the MIL-STD */
198 #define ULAW_BIAS 0x84 /* Bias for linear code. */
199 
200 /*! \brief Encode a linear sample to u-law
201  \param linear The sample to encode.
202  \return The u-law value.
203 */
204  static __inline__ uint8_t linear_to_ulaw(int linear) {
205  uint8_t u_val;
206  int mask;
207  int seg;
208 
209  /* Get the sign and the magnitude of the value. */
210  if (linear < 0) {
211  linear = ULAW_BIAS - linear;
212  mask = 0x7F;
213  } else {
214  linear = ULAW_BIAS + linear;
215  mask = 0xFF;
216  }
217 
218  seg = top_bit(linear | 0xFF) - 7;
219 
220  /*
221  * Combine the sign, segment, quantization bits,
222  * and complement the code word.
223  */
224  if (seg >= 8)
225  u_val = (uint8_t) (0x7F ^ mask);
226  else
227  u_val = (uint8_t) (((seg << 4) | ((linear >> (seg + 3)) & 0xF)) ^ mask);
228 #ifdef ULAW_ZEROTRAP
229  /* Optional ITU trap */
230  if (u_val == 0)
231  u_val = 0x02;
232 #endif
233  return u_val;
234  }
235 /*- End of function --------------------------------------------------------*/
236 
237 /*! \brief Decode an u-law sample to a linear value.
238  \param ulaw The u-law sample to decode.
239  \return The linear value.
240 */
241  static __inline__ int16_t ulaw_to_linear(uint8_t ulaw) {
242  int t;
243 
244  /* Complement to obtain normal u-law value. */
245  ulaw = ~ulaw;
246  /*
247  * Extract and bias the quantization bits. Then
248  * shift up by the segment number and subtract out the bias.
249  */
250  t = (((ulaw & 0x0F) << 3) + ULAW_BIAS) << (((int) ulaw & 0x70) >> 4);
251  return (int16_t) ((ulaw & 0x80) ? (ULAW_BIAS - t) : (t - ULAW_BIAS));
252  }
253 /*- End of function --------------------------------------------------------*/
254 
255 /*
256  * A-law is basically as follows:
257  *
258  * Linear Input Code Compressed Code
259  * ----------------- ---------------
260  * 0000000wxyza 000wxyz
261  * 0000001wxyza 001wxyz
262  * 000001wxyzab 010wxyz
263  * 00001wxyzabc 011wxyz
264  * 0001wxyzabcd 100wxyz
265  * 001wxyzabcde 101wxyz
266  * 01wxyzabcdef 110wxyz
267  * 1wxyzabcdefg 111wxyz
268  *
269  * For further information see John C. Bellamy's Digital Telephony, 1982,
270  * John Wiley & Sons, pps 98-111 and 472-476.
271  */
272 
273 #define ALAW_AMI_MASK 0x55
274 
275 /*! \brief Encode a linear sample to A-law
276  \param linear The sample to encode.
277  \return The A-law value.
278 */
279  static __inline__ uint8_t linear_to_alaw(int linear) {
280  int mask;
281  int seg;
282 
283  if (linear >= 0) {
284  /* Sign (bit 7) bit = 1 */
285  mask = ALAW_AMI_MASK | 0x80;
286  } else {
287  /* Sign (bit 7) bit = 0 */
288  mask = ALAW_AMI_MASK;
289  linear = -linear - 8;
290  }
291 
292  /* Convert the scaled magnitude to segment number. */
293  seg = top_bit(linear | 0xFF) - 7;
294  if (seg >= 8) {
295  if (linear >= 0) {
296  /* Out of range. Return maximum value. */
297  return (uint8_t) (0x7F ^ mask);
298  }
299  /* We must be just a tiny step below zero */
300  return (uint8_t) (0x00 ^ mask);
301  }
302  /* Combine the sign, segment, and quantization bits. */
303  return (uint8_t) (((seg << 4) | ((linear >> ((seg) ? (seg + 3) : 4)) & 0x0F)) ^ mask);
304  }
305 /*- End of function --------------------------------------------------------*/
306 
307 /*! \brief Decode an A-law sample to a linear value.
308  \param alaw The A-law sample to decode.
309  \return The linear value.
310 */
311  static __inline__ int16_t alaw_to_linear(uint8_t alaw) {
312  int i;
313  int seg;
314 
315  alaw ^= ALAW_AMI_MASK;
316  i = ((alaw & 0x0F) << 4);
317  seg = (((int) alaw & 0x70) >> 4);
318  if (seg)
319  i = (i + 0x108) << (seg - 1);
320  else
321  i += 8;
322  return (int16_t) ((alaw & 0x80) ? i : -i);
323  }
324 /*- End of function --------------------------------------------------------*/
325 
326 /*! \brief Transcode from A-law to u-law, using the procedure defined in G.711.
327  \param alaw The A-law sample to transcode.
328  \return The best matching u-law value.
329 */
330  uint8_t alaw_to_ulaw(uint8_t alaw);
331 
332 /*! \brief Transcode from u-law to A-law, using the procedure defined in G.711.
333  \param ulaw The u-law sample to transcode.
334  \return The best matching A-law value.
335 */
336  uint8_t ulaw_to_alaw(uint8_t ulaw);
337 
338 #ifdef __cplusplus
339 }
340 #endif
341 
342 #endif
343 /*- End of file ------------------------------------------------------------*/
uint8_t ulaw_to_alaw(uint8_t ulaw)
Transcode from u-law to A-law, using the procedure defined in G.711.
Definition: g711.c:86
static __inline__ int bottom_bit(unsigned int bits)
Definition: g711.h:129
static __inline__ uint8_t linear_to_alaw(int linear)
Encode a linear sample to A-law.
Definition: g711.h:279
uint8_t alaw_to_ulaw(uint8_t alaw)
Transcode from A-law to u-law, using the procedure defined in G.711.
Definition: g711.c:79
static __inline__ int16_t alaw_to_linear(uint8_t alaw)
Decode an A-law sample to a linear value.
Definition: g711.h:311
#define ALAW_AMI_MASK
Definition: g711.h:273
static __inline__ int16_t ulaw_to_linear(uint8_t ulaw)
Decode an u-law sample to a linear value.
Definition: g711.h:241
static __inline__ uint8_t linear_to_ulaw(int linear)
Encode a linear sample to u-law.
Definition: g711.h:204
static __inline__ int top_bit(unsigned int bits)
Definition: g711.h:99
#define ULAW_BIAS
Definition: g711.h:198