1 /*****************************************************************************
2 * idctclassic.c : Classic IDCT module
3 *****************************************************************************
4 * Copyright (C) 1999, 2000 VideoLAN
5 * $Id: idctclassic.c,v 1.15 2001/09/05 16:07:49 massiot 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 #define MODULE_NAME idctclassic
25 #include "modules_inner.h"
27 /*****************************************************************************
29 *****************************************************************************/
44 #include "modules_export.h"
46 /*****************************************************************************
47 * Local and extern prototypes.
48 *****************************************************************************/
49 static void idct_getfunctions( function_list_t * p_function_list );
51 /*****************************************************************************
52 * Build configuration tree.
53 *****************************************************************************/
55 ADD_WINDOW( "Configuration for classic IDCT module" )
56 ADD_COMMENT( "Ha, ha -- nothing to configure yet" )
60 p_module->i_capabilities = MODULE_CAPABILITY_NULL
61 | MODULE_CAPABILITY_IDCT;
62 p_module->psz_longname = "classic IDCT module";
66 idct_getfunctions( &p_module->p_functions->idct );
69 MODULE_DEACTIVATE_START
70 MODULE_DEACTIVATE_STOP
72 /* Following functions are local */
74 /*****************************************************************************
75 * idct_Probe: returns a preference score
76 *****************************************************************************/
77 static int idct_Probe( probedata_t *p_data )
79 if( TestMethod( IDCT_METHOD_VAR, "idctclassic" )
80 || TestMethod( IDCT_METHOD_VAR, "classic" ) )
85 /* This plugin always works */
89 /*****************************************************************************
90 * NormScan : Unused in this IDCT
91 *****************************************************************************/
92 static void NormScan( u8 ppi_scan[2][64] )
96 /*****************************************************************************
97 * IDCT : IDCT function for normal matrices
98 *****************************************************************************/
99 static __inline__ void IDCT( dctelem_t * p_block )
101 s32 tmp0, tmp1, tmp2, tmp3;
102 s32 tmp10, tmp11, tmp12, tmp13;
103 s32 z1, z2, z3, z4, z5;
108 /* Pass 1: process rows. */
109 /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
110 /* furthermore, we scale the results by 2**PASS1_BITS. */
113 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
115 /* Due to quantization, we will usually find that many of the input
116 * coefficients are zero, especially the AC terms. We can exploit this
117 * by short-circuiting the IDCT calculation for any row in which all
118 * the AC terms are zero. In that case each output is equal to the
119 * DC coefficient (with scale factor as needed).
120 * With typical images and quantization tables, half or more of the
121 * row DCT calculations can be simplified this way.
124 if ((dataptr[1] | dataptr[2] | dataptr[3] | dataptr[4] |
125 dataptr[5] | dataptr[6] | dataptr[7]) == 0)
127 /* AC terms all zero */
128 dctelem_t dcval = (dctelem_t) (dataptr[0] << PASS1_BITS);
139 dataptr += DCTSIZE; /* advance pointer to next row */
143 /* Even part: reverse the even part of the forward DCT. */
144 /* The rotator is sqrt(2)*c(-6). */
146 z2 = (s32) dataptr[2];
147 z3 = (s32) dataptr[6];
149 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
150 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
151 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
153 tmp0 = ((s32) dataptr[0] + (s32) dataptr[4]) << CONST_BITS;
154 tmp1 = ((s32) dataptr[0] - (s32) dataptr[4]) << CONST_BITS;
161 /* Odd part per figure 8; the matrix is unitary and hence its
162 * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively.
165 tmp0 = (s32) dataptr[7];
166 tmp1 = (s32) dataptr[5];
167 tmp2 = (s32) dataptr[3];
168 tmp3 = (s32) dataptr[1];
174 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
176 tmp0 = MULTIPLY(tmp0, FIX(0.298631336)); /* sqrt(2) * (-c1+c3+c5-c7) */
177 tmp1 = MULTIPLY(tmp1, FIX(2.053119869)); /* sqrt(2) * ( c1+c3-c5+c7) */
178 tmp2 = MULTIPLY(tmp2, FIX(3.072711026)); /* sqrt(2) * ( c1+c3+c5-c7) */
179 tmp3 = MULTIPLY(tmp3, FIX(1.501321110)); /* sqrt(2) * ( c1+c3-c5-c7) */
180 z1 = MULTIPLY(z1, - FIX(0.899976223)); /* sqrt(2) * (c7-c3) */
181 z2 = MULTIPLY(z2, - FIX(2.562915447)); /* sqrt(2) * (-c1-c3) */
182 z3 = MULTIPLY(z3, - FIX(1.961570560)); /* sqrt(2) * (-c3-c5) */
183 z4 = MULTIPLY(z4, - FIX(0.390180644)); /* sqrt(2) * (c5-c3) */
193 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
195 dataptr[0] = (dctelem_t) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
196 dataptr[7] = (dctelem_t) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
197 dataptr[1] = (dctelem_t) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
198 dataptr[6] = (dctelem_t) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS);
199 dataptr[2] = (dctelem_t) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS);
200 dataptr[5] = (dctelem_t) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
201 dataptr[3] = (dctelem_t) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
202 dataptr[4] = (dctelem_t) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
204 dataptr += DCTSIZE; /* advance pointer to next row */
207 /* Pass 2: process columns. */
208 /* Note that we must descale the results by a factor of 8 == 2**3, */
209 /* and also undo the PASS1_BITS scaling. */
212 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
214 /* Columns of zeroes can be exploited in the same way as we did with rows.
215 * However, the row calculation has created many nonzero AC terms, so the
216 * simplification applies less often (typically 5% to 10% of the time).
217 * On machines with very fast multiplication, it's possible that the
218 * test takes more time than it's worth. In that case this section
219 * may be commented out.
222 #ifndef NO_ZERO_COLUMN_TEST /* Adds a test but avoids calculus */
223 if ((dataptr[DCTSIZE*1] | dataptr[DCTSIZE*2] | dataptr[DCTSIZE*3] |
224 dataptr[DCTSIZE*4] | dataptr[DCTSIZE*5] | dataptr[DCTSIZE*6] |
225 dataptr[DCTSIZE*7]) == 0)
227 /* AC terms all zero */
228 dctelem_t dcval = (dctelem_t) DESCALE((s32) dataptr[0], PASS1_BITS+3);
230 dataptr[DCTSIZE*0] = dcval;
231 dataptr[DCTSIZE*1] = dcval;
232 dataptr[DCTSIZE*2] = dcval;
233 dataptr[DCTSIZE*3] = dcval;
234 dataptr[DCTSIZE*4] = dcval;
235 dataptr[DCTSIZE*5] = dcval;
236 dataptr[DCTSIZE*6] = dcval;
237 dataptr[DCTSIZE*7] = dcval;
239 dataptr++; /* advance pointer to next column */
244 /* Even part: reverse the even part of the forward DCT. */
245 /* The rotator is sqrt(2)*c(-6). */
247 z2 = (s32) dataptr[DCTSIZE*2];
248 z3 = (s32) dataptr[DCTSIZE*6];
250 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
251 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
252 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
254 tmp0 = ((s32) dataptr[DCTSIZE*0] + (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
255 tmp1 = ((s32) dataptr[DCTSIZE*0] - (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
262 /* Odd part per figure 8; the matrix is unitary and hence its
263 * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively.
266 tmp0 = (s32) dataptr[DCTSIZE*7];
267 tmp1 = (s32) dataptr[DCTSIZE*5];
268 tmp2 = (s32) dataptr[DCTSIZE*3];
269 tmp3 = (s32) dataptr[DCTSIZE*1];
275 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
277 tmp0 = MULTIPLY(tmp0, FIX(0.298631336)); /* sqrt(2) * (-c1+c3+c5-c7) */
278 tmp1 = MULTIPLY(tmp1, FIX(2.053119869)); /* sqrt(2) * ( c1+c3-c5+c7) */
279 tmp2 = MULTIPLY(tmp2, FIX(3.072711026)); /* sqrt(2) * ( c1+c3+c5-c7) */
280 tmp3 = MULTIPLY(tmp3, FIX(1.501321110)); /* sqrt(2) * ( c1+c3-c5-c7) */
281 z1 = MULTIPLY(z1, - FIX(0.899976223)); /* sqrt(2) * (c7-c3) */
282 z2 = MULTIPLY(z2, - FIX(2.562915447)); /* sqrt(2) * (-c1-c3) */
283 z3 = MULTIPLY(z3, - FIX(1.961570560)); /* sqrt(2) * (-c3-c5) */
284 z4 = MULTIPLY(z4, - FIX(0.390180644)); /* sqrt(2) * (c5-c3) */
294 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
296 dataptr[DCTSIZE*0] = (dctelem_t) DESCALE(tmp10 + tmp3,
297 CONST_BITS+PASS1_BITS+3);
298 dataptr[DCTSIZE*7] = (dctelem_t) DESCALE(tmp10 - tmp3,
299 CONST_BITS+PASS1_BITS+3);
300 dataptr[DCTSIZE*1] = (dctelem_t) DESCALE(tmp11 + tmp2,
301 CONST_BITS+PASS1_BITS+3);
302 dataptr[DCTSIZE*6] = (dctelem_t) DESCALE(tmp11 - tmp2,
303 CONST_BITS+PASS1_BITS+3);
304 dataptr[DCTSIZE*2] = (dctelem_t) DESCALE(tmp12 + tmp1,
305 CONST_BITS+PASS1_BITS+3);
306 dataptr[DCTSIZE*5] = (dctelem_t) DESCALE(tmp12 - tmp1,
307 CONST_BITS+PASS1_BITS+3);
308 dataptr[DCTSIZE*3] = (dctelem_t) DESCALE(tmp13 + tmp0,
309 CONST_BITS+PASS1_BITS+3);
310 dataptr[DCTSIZE*4] = (dctelem_t) DESCALE(tmp13 - tmp0,
311 CONST_BITS+PASS1_BITS+3);
313 dataptr++; /* advance pointer to next column */
317 #include "idct_sparse.h"
318 #include "idct_decl.h"