source: golgotha/src/i4/loaders/jpg/jidctint.cc @ 80

Last change on this file since 80 was 80, checked in by Sam Hocevar, 11 years ago
  • Adding the Golgotha source code. Not sure what's going to be interesting in there, but since it's all public domain, there's certainly stuff to pick up.
File size: 15.3 KB
Line 
1/********************************************************************** <BR>
2  This file is part of Crack dot Com's free source code release of
3  Golgotha. <a href="http://www.crack.com/golgotha_release"> <BR> for
4  information about compiling & licensing issues visit this URL</a>
5  <PRE> If that doesn't help, contact Jonathan Clark at
6  golgotha_source@usa.net (Subject should have "GOLG" in it)
7***********************************************************************/
8
9/*
10 * jidctint.c
11 *
12 * Copyright (C) 1991-1996, Thomas G. Lane.
13 * This file is part of the Independent JPEG Group's software.
14 * For conditions of distribution and use, see the accompanying README file.
15 *
16 * This file contains a slow-but-accurate integer implementation of the
17 * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
18 * must also perform dequantization of the input coefficients.
19 *
20 * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
21 * on each row (or vice versa, but it's more convenient to emit a row at
22 * a time).  Direct algorithms are also available, but they are much more
23 * complex and seem not to be any faster when reduced to code.
24 *
25 * This implementation is based on an algorithm described in
26 *   C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT
27 *   Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics,
28 *   Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991.
29 * The primary algorithm described there uses 11 multiplies and 29 adds.
30 * We use their alternate method with 12 multiplies and 32 adds.
31 * The advantage of this method is that no data path contains more than one
32 * multiplication; this allows a very simple and accurate implementation in
33 * scaled fixed-point arithmetic, with a minimal number of shifts.
34 */
35
36#define JPEG_INTERNALS
37#include "loaders/jpg/jinclude.h"
38#include "loaders/jpg/jpeglib.h"
39#include "loaders/jpg/jdct.h"           /* Private declarations for DCT subsystem */
40
41#ifdef DCT_ISLOW_SUPPORTED
42
43
44/*
45 * This module is specialized to the case DCTSIZE = 8.
46 */
47
48#if DCTSIZE != 8
49  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
50#endif
51
52
53/*
54 * The poop on this scaling stuff is as follows:
55 *
56 * Each 1-D IDCT step produces outputs which are a factor of sqrt(N)
57 * larger than the true IDCT outputs.  The final outputs are therefore
58 * a factor of N larger than desired; since N=8 this can be cured by
59 * a simple right shift at the end of the algorithm.  The advantage of
60 * this arrangement is that we save two multiplications per 1-D IDCT,
61 * because the y0 and y4 inputs need not be divided by sqrt(N).
62 *
63 * We have to do addition and subtraction of the integer inputs, which
64 * is no problem, and multiplication by fractional constants, which is
65 * a problem to do in integer arithmetic.  We multiply all the constants
66 * by CONST_SCALE and convert them to integer constants (thus retaining
67 * CONST_BITS bits of precision in the constants).  After doing a
68 * multiplication we have to divide the product by CONST_SCALE, with proper
69 * rounding, to produce the correct output.  This division can be done
70 * cheaply as a right shift of CONST_BITS bits.  We postpone shifting
71 * as long as possible so that partial sums can be added together with
72 * full fractional precision.
73 *
74 * The outputs of the first pass are scaled up by PASS1_BITS bits so that
75 * they are represented to better-than-integral precision.  These outputs
76 * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
77 * with the recommended scaling.  (To scale up 12-bit sample data further, an
78 * intermediate INT32 array would be needed.)
79 *
80 * To avoid overflow of the 32-bit intermediate results in pass 2, we must
81 * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
82 * shows that the values given below are the most effective.
83 */
84
85#if BITS_IN_JSAMPLE == 8
86#define CONST_BITS  13
87#define PASS1_BITS  2
88#else
89#define CONST_BITS  13
90#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
91#endif
92
93/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
94 * causing a lot of useless floating-point operations at run time.
95 * To get around this we use the following pre-calculated constants.
96 * If you change CONST_BITS you may want to add appropriate values.
97 * (With a reasonable C compiler, you can just rely on the FIX() macro...)
98 */
99
100#if CONST_BITS == 13
101#define FIX_0_298631336  ((INT32)  2446)        /* FIX(0.298631336) */
102#define FIX_0_390180644  ((INT32)  3196)        /* FIX(0.390180644) */
103#define FIX_0_541196100  ((INT32)  4433)        /* FIX(0.541196100) */
104#define FIX_0_765366865  ((INT32)  6270)        /* FIX(0.765366865) */
105#define FIX_0_899976223  ((INT32)  7373)        /* FIX(0.899976223) */
106#define FIX_1_175875602  ((INT32)  9633)        /* FIX(1.175875602) */
107#define FIX_1_501321110  ((INT32)  12299)       /* FIX(1.501321110) */
108#define FIX_1_847759065  ((INT32)  15137)       /* FIX(1.847759065) */
109#define FIX_1_961570560  ((INT32)  16069)       /* FIX(1.961570560) */
110#define FIX_2_053119869  ((INT32)  16819)       /* FIX(2.053119869) */
111#define FIX_2_562915447  ((INT32)  20995)       /* FIX(2.562915447) */
112#define FIX_3_072711026  ((INT32)  25172)       /* FIX(3.072711026) */
113#else
114#define FIX_0_298631336  FIX(0.298631336)
115#define FIX_0_390180644  FIX(0.390180644)
116#define FIX_0_541196100  FIX(0.541196100)
117#define FIX_0_765366865  FIX(0.765366865)
118#define FIX_0_899976223  FIX(0.899976223)
119#define FIX_1_175875602  FIX(1.175875602)
120#define FIX_1_501321110  FIX(1.501321110)
121#define FIX_1_847759065  FIX(1.847759065)
122#define FIX_1_961570560  FIX(1.961570560)
123#define FIX_2_053119869  FIX(2.053119869)
124#define FIX_2_562915447  FIX(2.562915447)
125#define FIX_3_072711026  FIX(3.072711026)
126#endif
127
128
129/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
130 * For 8-bit samples with the recommended scaling, all the variable
131 * and constant values involved are no more than 16 bits wide, so a
132 * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
133 * For 12-bit samples, a full 32-bit multiplication will be needed.
134 */
135
136#if BITS_IN_JSAMPLE == 8
137#define MULTIPLY(var,const)  MULTIPLY16C16(var,const)
138#else
139#define MULTIPLY(var,const)  ((var) * (const))
140#endif
141
142
143/* Dequantize a coefficient by multiplying it by the multiplier-table
144 * entry; produce an int result.  In this module, both inputs and result
145 * are 16 bits or less, so either int or short multiply will work.
146 */
147
148#define DEQUANTIZE(coef,quantval)  (((ISLOW_MULT_TYPE) (coef)) * (quantval))
149
150
151/*
152 * Perform dequantization and inverse DCT on one block of coefficients.
153 */
154
155GLOBAL(void)
156jpeg_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr,
157                 JCOEFPTR coef_block,
158                 JSAMPARRAY output_buf, JDIMENSION output_col)
159{
160  INT32 tmp0, tmp1, tmp2, tmp3;
161  INT32 tmp10, tmp11, tmp12, tmp13;
162  INT32 z1, z2, z3, z4, z5;
163  JCOEFPTR inptr;
164  ISLOW_MULT_TYPE * quantptr;
165  int * wsptr;
166  JSAMPROW outptr;
167  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
168  int ctr;
169  int workspace[DCTSIZE2];      /* buffers data between passes */
170  SHIFT_TEMPS
171
172  /* Pass 1: process columns from input, store into work array. */
173  /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
174  /* furthermore, we scale the results by 2**PASS1_BITS. */
175
176  inptr = coef_block;
177  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
178  wsptr = workspace;
179  for (ctr = DCTSIZE; ctr > 0; ctr--) {
180    /* Due to quantization, we will usually find that many of the input
181     * coefficients are zero, especially the AC terms.  We can exploit this
182     * by short-circuiting the IDCT calculation for any column in which all
183     * the AC terms are zero.  In that case each output is equal to the
184     * DC coefficient (with scale factor as needed).
185     * With typical images and quantization tables, half or more of the
186     * column DCT calculations can be simplified this way.
187     */
188   
189    if ((inptr[DCTSIZE*1] | inptr[DCTSIZE*2] | inptr[DCTSIZE*3] |
190         inptr[DCTSIZE*4] | inptr[DCTSIZE*5] | inptr[DCTSIZE*6] |
191         inptr[DCTSIZE*7]) == 0) {
192      /* AC terms all zero */
193      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
194     
195      wsptr[DCTSIZE*0] = dcval;
196      wsptr[DCTSIZE*1] = dcval;
197      wsptr[DCTSIZE*2] = dcval;
198      wsptr[DCTSIZE*3] = dcval;
199      wsptr[DCTSIZE*4] = dcval;
200      wsptr[DCTSIZE*5] = dcval;
201      wsptr[DCTSIZE*6] = dcval;
202      wsptr[DCTSIZE*7] = dcval;
203     
204      inptr++;                  /* advance pointers to next column */
205      quantptr++;
206      wsptr++;
207      continue;
208    }
209   
210    /* Even part: reverse the even part of the forward DCT. */
211    /* The rotator is sqrt(2)*c(-6). */
212   
213    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
214    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
215   
216    z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
217    tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065);
218    tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
219   
220    z2 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
221    z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
222
223    tmp0 = (z2 + z3) << CONST_BITS;
224    tmp1 = (z2 - z3) << CONST_BITS;
225   
226    tmp10 = tmp0 + tmp3;
227    tmp13 = tmp0 - tmp3;
228    tmp11 = tmp1 + tmp2;
229    tmp12 = tmp1 - tmp2;
230   
231    /* Odd part per figure 8; the matrix is unitary and hence its
232     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
233     */
234   
235    tmp0 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
236    tmp1 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
237    tmp2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
238    tmp3 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
239   
240    z1 = tmp0 + tmp3;
241    z2 = tmp1 + tmp2;
242    z3 = tmp0 + tmp2;
243    z4 = tmp1 + tmp3;
244    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
245   
246    tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
247    tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
248    tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
249    tmp3 = MULTIPLY(tmp3, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
250    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
251    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
252    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
253    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
254   
255    z3 += z5;
256    z4 += z5;
257   
258    tmp0 += z1 + z3;
259    tmp1 += z2 + z4;
260    tmp2 += z2 + z3;
261    tmp3 += z1 + z4;
262   
263    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
264   
265    wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
266    wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
267    wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
268    wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS);
269    wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS);
270    wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
271    wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
272    wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
273   
274    inptr++;                    /* advance pointers to next column */
275    quantptr++;
276    wsptr++;
277  }
278 
279  /* Pass 2: process rows from work array, store into output array. */
280  /* Note that we must descale the results by a factor of 8 == 2**3, */
281  /* and also undo the PASS1_BITS scaling. */
282
283  wsptr = workspace;
284  for (ctr = 0; ctr < DCTSIZE; ctr++) {
285    outptr = output_buf[ctr] + output_col;
286    /* Rows of zeroes can be exploited in the same way as we did with columns.
287     * However, the column calculation has created many nonzero AC terms, so
288     * the simplification applies less often (typically 5% to 10% of the time).
289     * On machines with very fast multiplication, it's possible that the
290     * test takes more time than it's worth.  In that case this section
291     * may be commented out.
292     */
293   
294#ifndef NO_ZERO_ROW_TEST
295    if ((wsptr[1] | wsptr[2] | wsptr[3] | wsptr[4] | wsptr[5] | wsptr[6] |
296         wsptr[7]) == 0) {
297      /* AC terms all zero */
298      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
299                                  & RANGE_MASK];
300     
301      outptr[0] = dcval;
302      outptr[1] = dcval;
303      outptr[2] = dcval;
304      outptr[3] = dcval;
305      outptr[4] = dcval;
306      outptr[5] = dcval;
307      outptr[6] = dcval;
308      outptr[7] = dcval;
309
310      wsptr += DCTSIZE;         /* advance pointer to next row */
311      continue;
312    }
313#endif
314   
315    /* Even part: reverse the even part of the forward DCT. */
316    /* The rotator is sqrt(2)*c(-6). */
317   
318    z2 = (INT32) wsptr[2];
319    z3 = (INT32) wsptr[6];
320   
321    z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
322    tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065);
323    tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
324   
325    tmp0 = ((INT32) wsptr[0] + (INT32) wsptr[4]) << CONST_BITS;
326    tmp1 = ((INT32) wsptr[0] - (INT32) wsptr[4]) << CONST_BITS;
327   
328    tmp10 = tmp0 + tmp3;
329    tmp13 = tmp0 - tmp3;
330    tmp11 = tmp1 + tmp2;
331    tmp12 = tmp1 - tmp2;
332   
333    /* Odd part per figure 8; the matrix is unitary and hence its
334     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
335     */
336   
337    tmp0 = (INT32) wsptr[7];
338    tmp1 = (INT32) wsptr[5];
339    tmp2 = (INT32) wsptr[3];
340    tmp3 = (INT32) wsptr[1];
341   
342    z1 = tmp0 + tmp3;
343    z2 = tmp1 + tmp2;
344    z3 = tmp0 + tmp2;
345    z4 = tmp1 + tmp3;
346    z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
347   
348    tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
349    tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
350    tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
351    tmp3 = MULTIPLY(tmp3, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
352    z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
353    z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
354    z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
355    z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
356   
357    z3 += z5;
358    z4 += z5;
359   
360    tmp0 += z1 + z3;
361    tmp1 += z2 + z4;
362    tmp2 += z2 + z3;
363    tmp3 += z1 + z4;
364   
365    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
366   
367    outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp3,
368                                          CONST_BITS+PASS1_BITS+3)
369                            & RANGE_MASK];
370    outptr[7] = range_limit[(int) DESCALE(tmp10 - tmp3,
371                                          CONST_BITS+PASS1_BITS+3)
372                            & RANGE_MASK];
373    outptr[1] = range_limit[(int) DESCALE(tmp11 + tmp2,
374                                          CONST_BITS+PASS1_BITS+3)
375                            & RANGE_MASK];
376    outptr[6] = range_limit[(int) DESCALE(tmp11 - tmp2,
377                                          CONST_BITS+PASS1_BITS+3)
378                            & RANGE_MASK];
379    outptr[2] = range_limit[(int) DESCALE(tmp12 + tmp1,
380                                          CONST_BITS+PASS1_BITS+3)
381                            & RANGE_MASK];
382    outptr[5] = range_limit[(int) DESCALE(tmp12 - tmp1,
383                                          CONST_BITS+PASS1_BITS+3)
384                            & RANGE_MASK];
385    outptr[3] = range_limit[(int) DESCALE(tmp13 + tmp0,
386                                          CONST_BITS+PASS1_BITS+3)
387                            & RANGE_MASK];
388    outptr[4] = range_limit[(int) DESCALE(tmp13 - tmp0,
389                                          CONST_BITS+PASS1_BITS+3)
390                            & RANGE_MASK];
391   
392    wsptr += DCTSIZE;           /* advance pointer to next row */
393  }
394}
395
396#endif /* DCT_ISLOW_SUPPORTED */
Note: See TracBrowser for help on using the repository browser.