1 /*****************************************************************************
2 * idctclassic.c : Classic IDCT module
3 *****************************************************************************
4 * Copyright (C) 1999-2001 VideoLAN
5 * $Id: idctclassic.c,v 1.25 2002/06/01 12:31:59 sam Exp $
7 * Authors: Gaƫl Hendryckx <jimmy@via.ecp.fr>
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
19 * You should have received a copy of the GNU General Public License
20 * along with this program; if not, write to the Free Software
21 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
22 *****************************************************************************/
24 /*****************************************************************************
26 *****************************************************************************/
35 /*****************************************************************************
36 * Local and extern prototypes.
37 *****************************************************************************/
38 static void idct_getfunctions( function_list_t * p_function_list );
40 /*****************************************************************************
41 * Build configuration tree.
42 *****************************************************************************/
47 SET_DESCRIPTION( _("classic IDCT module") )
48 ADD_CAPABILITY( IDCT, 100 )
49 ADD_SHORTCUT( "classic" )
53 idct_getfunctions( &p_module->p_functions->idct );
56 MODULE_DEACTIVATE_START
57 MODULE_DEACTIVATE_STOP
59 /* Following functions are local */
61 /*****************************************************************************
62 * NormScan : Unused in this IDCT
63 *****************************************************************************/
64 static void NormScan( u8 ppi_scan[2][64] )
68 /*****************************************************************************
69 * IDCT : IDCT function for normal matrices
70 *****************************************************************************/
71 static inline void IDCT( dctelem_t * p_block )
73 s32 tmp0, tmp1, tmp2, tmp3;
74 s32 tmp10, tmp11, tmp12, tmp13;
75 s32 z1, z2, z3, z4, z5;
80 /* Pass 1: process rows. */
81 /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
82 /* furthermore, we scale the results by 2**PASS1_BITS. */
85 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
87 /* Due to quantization, we will usually find that many of the input
88 * coefficients are zero, especially the AC terms. We can exploit this
89 * by short-circuiting the IDCT calculation for any row in which all
90 * the AC terms are zero. In that case each output is equal to the
91 * DC coefficient (with scale factor as needed).
92 * With typical images and quantization tables, half or more of the
93 * row DCT calculations can be simplified this way.
96 if ((dataptr[1] | dataptr[2] | dataptr[3] | dataptr[4] |
97 dataptr[5] | dataptr[6] | dataptr[7]) == 0)
99 /* AC terms all zero */
100 dctelem_t dcval = (dctelem_t) (dataptr[0] << PASS1_BITS);
111 dataptr += DCTSIZE; /* advance pointer to next row */
115 /* Even part: reverse the even part of the forward DCT. */
116 /* The rotator is sqrt(2)*c(-6). */
118 z2 = (s32) dataptr[2];
119 z3 = (s32) dataptr[6];
121 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
122 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
123 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
125 tmp0 = ((s32) dataptr[0] + (s32) dataptr[4]) << CONST_BITS;
126 tmp1 = ((s32) dataptr[0] - (s32) dataptr[4]) << CONST_BITS;
133 /* Odd part per figure 8; the matrix is unitary and hence its
134 * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively.
137 tmp0 = (s32) dataptr[7];
138 tmp1 = (s32) dataptr[5];
139 tmp2 = (s32) dataptr[3];
140 tmp3 = (s32) dataptr[1];
146 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
148 tmp0 = MULTIPLY(tmp0, FIX(0.298631336)); /* sqrt(2) * (-c1+c3+c5-c7) */
149 tmp1 = MULTIPLY(tmp1, FIX(2.053119869)); /* sqrt(2) * ( c1+c3-c5+c7) */
150 tmp2 = MULTIPLY(tmp2, FIX(3.072711026)); /* sqrt(2) * ( c1+c3+c5-c7) */
151 tmp3 = MULTIPLY(tmp3, FIX(1.501321110)); /* sqrt(2) * ( c1+c3-c5-c7) */
152 z1 = MULTIPLY(z1, - FIX(0.899976223)); /* sqrt(2) * (c7-c3) */
153 z2 = MULTIPLY(z2, - FIX(2.562915447)); /* sqrt(2) * (-c1-c3) */
154 z3 = MULTIPLY(z3, - FIX(1.961570560)); /* sqrt(2) * (-c3-c5) */
155 z4 = MULTIPLY(z4, - FIX(0.390180644)); /* sqrt(2) * (c5-c3) */
165 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
167 dataptr[0] = (dctelem_t) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
168 dataptr[7] = (dctelem_t) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
169 dataptr[1] = (dctelem_t) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
170 dataptr[6] = (dctelem_t) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS);
171 dataptr[2] = (dctelem_t) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS);
172 dataptr[5] = (dctelem_t) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
173 dataptr[3] = (dctelem_t) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
174 dataptr[4] = (dctelem_t) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
176 dataptr += DCTSIZE; /* advance pointer to next row */
179 /* Pass 2: process columns. */
180 /* Note that we must descale the results by a factor of 8 == 2**3, */
181 /* and also undo the PASS1_BITS scaling. */
184 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
186 /* Columns of zeroes can be exploited in the same way as we did with rows.
187 * However, the row calculation has created many nonzero AC terms, so the
188 * simplification applies less often (typically 5% to 10% of the time).
189 * On machines with very fast multiplication, it's possible that the
190 * test takes more time than it's worth. In that case this section
191 * may be commented out.
194 #ifndef NO_ZERO_COLUMN_TEST /* Adds a test but avoids calculus */
195 if ((dataptr[DCTSIZE*1] | dataptr[DCTSIZE*2] | dataptr[DCTSIZE*3] |
196 dataptr[DCTSIZE*4] | dataptr[DCTSIZE*5] | dataptr[DCTSIZE*6] |
197 dataptr[DCTSIZE*7]) == 0)
199 /* AC terms all zero */
200 dctelem_t dcval = (dctelem_t) DESCALE((s32) dataptr[0], PASS1_BITS+3);
202 dataptr[DCTSIZE*0] = dcval;
203 dataptr[DCTSIZE*1] = dcval;
204 dataptr[DCTSIZE*2] = dcval;
205 dataptr[DCTSIZE*3] = dcval;
206 dataptr[DCTSIZE*4] = dcval;
207 dataptr[DCTSIZE*5] = dcval;
208 dataptr[DCTSIZE*6] = dcval;
209 dataptr[DCTSIZE*7] = dcval;
211 dataptr++; /* advance pointer to next column */
216 /* Even part: reverse the even part of the forward DCT. */
217 /* The rotator is sqrt(2)*c(-6). */
219 z2 = (s32) dataptr[DCTSIZE*2];
220 z3 = (s32) dataptr[DCTSIZE*6];
222 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
223 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
224 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
226 tmp0 = ((s32) dataptr[DCTSIZE*0] + (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
227 tmp1 = ((s32) dataptr[DCTSIZE*0] - (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
234 /* Odd part per figure 8; the matrix is unitary and hence its
235 * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively.
238 tmp0 = (s32) dataptr[DCTSIZE*7];
239 tmp1 = (s32) dataptr[DCTSIZE*5];
240 tmp2 = (s32) dataptr[DCTSIZE*3];
241 tmp3 = (s32) dataptr[DCTSIZE*1];
247 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
249 tmp0 = MULTIPLY(tmp0, FIX(0.298631336)); /* sqrt(2) * (-c1+c3+c5-c7) */
250 tmp1 = MULTIPLY(tmp1, FIX(2.053119869)); /* sqrt(2) * ( c1+c3-c5+c7) */
251 tmp2 = MULTIPLY(tmp2, FIX(3.072711026)); /* sqrt(2) * ( c1+c3+c5-c7) */
252 tmp3 = MULTIPLY(tmp3, FIX(1.501321110)); /* sqrt(2) * ( c1+c3-c5-c7) */
253 z1 = MULTIPLY(z1, - FIX(0.899976223)); /* sqrt(2) * (c7-c3) */
254 z2 = MULTIPLY(z2, - FIX(2.562915447)); /* sqrt(2) * (-c1-c3) */
255 z3 = MULTIPLY(z3, - FIX(1.961570560)); /* sqrt(2) * (-c3-c5) */
256 z4 = MULTIPLY(z4, - FIX(0.390180644)); /* sqrt(2) * (c5-c3) */
266 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
268 dataptr[DCTSIZE*0] = (dctelem_t) DESCALE(tmp10 + tmp3,
269 CONST_BITS+PASS1_BITS+3);
270 dataptr[DCTSIZE*7] = (dctelem_t) DESCALE(tmp10 - tmp3,
271 CONST_BITS+PASS1_BITS+3);
272 dataptr[DCTSIZE*1] = (dctelem_t) DESCALE(tmp11 + tmp2,
273 CONST_BITS+PASS1_BITS+3);
274 dataptr[DCTSIZE*6] = (dctelem_t) DESCALE(tmp11 - tmp2,
275 CONST_BITS+PASS1_BITS+3);
276 dataptr[DCTSIZE*2] = (dctelem_t) DESCALE(tmp12 + tmp1,
277 CONST_BITS+PASS1_BITS+3);
278 dataptr[DCTSIZE*5] = (dctelem_t) DESCALE(tmp12 - tmp1,
279 CONST_BITS+PASS1_BITS+3);
280 dataptr[DCTSIZE*3] = (dctelem_t) DESCALE(tmp13 + tmp0,
281 CONST_BITS+PASS1_BITS+3);
282 dataptr[DCTSIZE*4] = (dctelem_t) DESCALE(tmp13 - tmp0,
283 CONST_BITS+PASS1_BITS+3);
285 dataptr++; /* advance pointer to next column */
289 static inline void RestoreCPUState( )
294 #include "idct_sparse.h"
295 #include "idct_decl.h"