1 /*****************************************************************************
2 * idctclassic.c : Classic IDCT module
3 *****************************************************************************
4 * Copyright (C) 1999, 2000 VideoLAN
5 * $Id: idctclassic.c,v 1.5 2001/01/17 18:17:30 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
26 /*****************************************************************************
28 *****************************************************************************/
40 #include "video_output.h"
42 #include "video_decoder.h"
45 #include "modules_inner.h"
49 /*****************************************************************************
50 * Local and extern prototypes.
51 *****************************************************************************/
52 static void idct_getfunctions( function_list_t * p_function_list );
53 static int idct_Probe ( probedata_t *p_data );
54 static void vdec_NormScan ( u8 ppi_scan[2][64] );
57 /*****************************************************************************
58 * Build configuration tree.
59 *****************************************************************************/
61 ADD_WINDOW( "Configuration for classic IDCT module" )
62 ADD_COMMENT( "Ha, ha -- nothing to configure yet" )
65 /*****************************************************************************
66 * InitModule: get the module structure and configuration.
67 *****************************************************************************
68 * We have to fill psz_name, psz_longname and psz_version. These variables
69 * will be strdup()ed later by the main application because the module can
70 * be unloaded later to save memory, and we want to be able to access this
71 * data even after the module has been unloaded.
72 *****************************************************************************/
73 int InitModule( module_t * p_module )
75 p_module->psz_name = MODULE_STRING;
76 p_module->psz_longname = "classic C IDCT module";
77 p_module->psz_version = VERSION;
79 p_module->i_capabilities = MODULE_CAPABILITY_NULL
80 | MODULE_CAPABILITY_IDCT;
85 /*****************************************************************************
86 * ActivateModule: set the module to an usable state.
87 *****************************************************************************
88 * This function fills the capability functions and the configuration
89 * structure. Once ActivateModule() has been called, the i_usage can
90 * be set to 0 and calls to NeedModule() be made to increment it. To unload
91 * the module, one has to wait until i_usage == 0 and call DeactivateModule().
92 *****************************************************************************/
93 int ActivateModule( module_t * p_module )
95 p_module->p_functions = malloc( sizeof( module_functions_t ) );
96 if( p_module->p_functions == NULL )
101 idct_getfunctions( &p_module->p_functions->idct );
103 p_module->p_config = p_config;
108 /*****************************************************************************
109 * DeactivateModule: make sure the module can be unloaded.
110 *****************************************************************************
111 * This function must only be called when i_usage == 0. If it successfully
112 * returns, i_usage can be set to -1 and the module unloaded. Be careful to
113 * lock usage_lock during the whole process.
114 *****************************************************************************/
115 int DeactivateModule( module_t * p_module )
117 free( p_module->p_functions );
122 /* Following functions are local */
124 /*****************************************************************************
125 * Functions exported as capabilities. They are declared as static so that
126 * we don't pollute the namespace too much.
127 *****************************************************************************/
128 static void idct_getfunctions( function_list_t * p_function_list )
130 p_function_list->pf_probe = idct_Probe;
131 p_function_list->functions.idct.pf_init = vdec_InitIDCT;
132 p_function_list->functions.idct.pf_sparse_idct = vdec_SparseIDCT;
133 p_function_list->functions.idct.pf_idct = vdec_IDCT;
134 p_function_list->functions.idct.pf_norm_scan = vdec_NormScan;
137 /*****************************************************************************
138 * idct_Probe: returns a preference score
139 *****************************************************************************/
140 static int idct_Probe( probedata_t *p_data )
142 if( TestMethod( IDCT_METHOD_VAR, "idctclassic" ) )
147 /* This plugin always works */
151 /*****************************************************************************
152 * vdec_NormScan : Unused in this IDCT
153 *****************************************************************************/
154 static void vdec_NormScan( u8 ppi_scan[2][64] )
158 /*****************************************************************************
159 * vdec_IDCT : IDCT function for normal matrices
160 *****************************************************************************/
161 void vdec_IDCT( vdec_thread_t * p_vdec, dctelem_t * p_block,
164 /* dct classique: pour tester la meilleure entre la classique et la */
166 s32 tmp0, tmp1, tmp2, tmp3;
167 s32 tmp10, tmp11, tmp12, tmp13;
168 s32 z1, z2, z3, z4, z5;
173 /* Pass 1: process rows. */
174 /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
175 /* furthermore, we scale the results by 2**PASS1_BITS. */
178 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
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 row 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 * row DCT calculations can be simplified this way.
189 if ((dataptr[1] | dataptr[2] | dataptr[3] | dataptr[4] |
190 dataptr[5] | dataptr[6] | dataptr[7]) == 0)
192 /* AC terms all zero */
193 dctelem_t dcval = (dctelem_t) (dataptr[0] << PASS1_BITS);
204 dataptr += DCTSIZE; /* advance pointer to next row */
208 /* Even part: reverse the even part of the forward DCT. */
209 /* The rotator is sqrt(2)*c(-6). */
211 z2 = (s32) dataptr[2];
212 z3 = (s32) dataptr[6];
214 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
215 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
216 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
218 tmp0 = ((s32) dataptr[0] + (s32) dataptr[4]) << CONST_BITS;
219 tmp1 = ((s32) dataptr[0] - (s32) dataptr[4]) << CONST_BITS;
226 /* Odd part per figure 8; the matrix is unitary and hence its
227 * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively.
230 tmp0 = (s32) dataptr[7];
231 tmp1 = (s32) dataptr[5];
232 tmp2 = (s32) dataptr[3];
233 tmp3 = (s32) dataptr[1];
239 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
241 tmp0 = MULTIPLY(tmp0, FIX(0.298631336)); /* sqrt(2) * (-c1+c3+c5-c7) */
242 tmp1 = MULTIPLY(tmp1, FIX(2.053119869)); /* sqrt(2) * ( c1+c3-c5+c7) */
243 tmp2 = MULTIPLY(tmp2, FIX(3.072711026)); /* sqrt(2) * ( c1+c3+c5-c7) */
244 tmp3 = MULTIPLY(tmp3, FIX(1.501321110)); /* sqrt(2) * ( c1+c3-c5-c7) */
245 z1 = MULTIPLY(z1, - FIX(0.899976223)); /* sqrt(2) * (c7-c3) */
246 z2 = MULTIPLY(z2, - FIX(2.562915447)); /* sqrt(2) * (-c1-c3) */
247 z3 = MULTIPLY(z3, - FIX(1.961570560)); /* sqrt(2) * (-c3-c5) */
248 z4 = MULTIPLY(z4, - FIX(0.390180644)); /* sqrt(2) * (c5-c3) */
258 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
260 dataptr[0] = (dctelem_t) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
261 dataptr[7] = (dctelem_t) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
262 dataptr[1] = (dctelem_t) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
263 dataptr[6] = (dctelem_t) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS);
264 dataptr[2] = (dctelem_t) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS);
265 dataptr[5] = (dctelem_t) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
266 dataptr[3] = (dctelem_t) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
267 dataptr[4] = (dctelem_t) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
269 dataptr += DCTSIZE; /* advance pointer to next row */
272 /* Pass 2: process columns. */
273 /* Note that we must descale the results by a factor of 8 == 2**3, */
274 /* and also undo the PASS1_BITS scaling. */
277 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
279 /* Columns of zeroes can be exploited in the same way as we did with rows.
280 * However, the row calculation has created many nonzero AC terms, so the
281 * simplification applies less often (typically 5% to 10% of the time).
282 * On machines with very fast multiplication, it's possible that the
283 * test takes more time than it's worth. In that case this section
284 * may be commented out.
287 #ifndef NO_ZERO_COLUMN_TEST /*ajoute un test mais evite des calculs */
288 if ((dataptr[DCTSIZE*1] | dataptr[DCTSIZE*2] | dataptr[DCTSIZE*3] |
289 dataptr[DCTSIZE*4] | dataptr[DCTSIZE*5] | dataptr[DCTSIZE*6] |
290 dataptr[DCTSIZE*7]) == 0)
292 /* AC terms all zero */
293 dctelem_t dcval = (dctelem_t) DESCALE((s32) dataptr[0], PASS1_BITS+3);
295 dataptr[DCTSIZE*0] = dcval;
296 dataptr[DCTSIZE*1] = dcval;
297 dataptr[DCTSIZE*2] = dcval;
298 dataptr[DCTSIZE*3] = dcval;
299 dataptr[DCTSIZE*4] = dcval;
300 dataptr[DCTSIZE*5] = dcval;
301 dataptr[DCTSIZE*6] = dcval;
302 dataptr[DCTSIZE*7] = dcval;
304 dataptr++; /* advance pointer to next column */
309 /* Even part: reverse the even part of the forward DCT. */
310 /* The rotator is sqrt(2)*c(-6). */
312 z2 = (s32) dataptr[DCTSIZE*2];
313 z3 = (s32) dataptr[DCTSIZE*6];
315 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
316 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
317 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
319 tmp0 = ((s32) dataptr[DCTSIZE*0] + (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
320 tmp1 = ((s32) dataptr[DCTSIZE*0] - (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
327 /* Odd part per figure 8; the matrix is unitary and hence its
328 * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively.
331 tmp0 = (s32) dataptr[DCTSIZE*7];
332 tmp1 = (s32) dataptr[DCTSIZE*5];
333 tmp2 = (s32) dataptr[DCTSIZE*3];
334 tmp3 = (s32) dataptr[DCTSIZE*1];
340 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
342 tmp0 = MULTIPLY(tmp0, FIX(0.298631336)); /* sqrt(2) * (-c1+c3+c5-c7) */
343 tmp1 = MULTIPLY(tmp1, FIX(2.053119869)); /* sqrt(2) * ( c1+c3-c5+c7) */
344 tmp2 = MULTIPLY(tmp2, FIX(3.072711026)); /* sqrt(2) * ( c1+c3+c5-c7) */
345 tmp3 = MULTIPLY(tmp3, FIX(1.501321110)); /* sqrt(2) * ( c1+c3-c5-c7) */
346 z1 = MULTIPLY(z1, - FIX(0.899976223)); /* sqrt(2) * (c7-c3) */
347 z2 = MULTIPLY(z2, - FIX(2.562915447)); /* sqrt(2) * (-c1-c3) */
348 z3 = MULTIPLY(z3, - FIX(1.961570560)); /* sqrt(2) * (-c3-c5) */
349 z4 = MULTIPLY(z4, - FIX(0.390180644)); /* sqrt(2) * (c5-c3) */
359 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
361 dataptr[DCTSIZE*0] = (dctelem_t) DESCALE(tmp10 + tmp3,
362 CONST_BITS+PASS1_BITS+3);
363 dataptr[DCTSIZE*7] = (dctelem_t) DESCALE(tmp10 - tmp3,
364 CONST_BITS+PASS1_BITS+3);
365 dataptr[DCTSIZE*1] = (dctelem_t) DESCALE(tmp11 + tmp2,
366 CONST_BITS+PASS1_BITS+3);
367 dataptr[DCTSIZE*6] = (dctelem_t) DESCALE(tmp11 - tmp2,
368 CONST_BITS+PASS1_BITS+3);
369 dataptr[DCTSIZE*2] = (dctelem_t) DESCALE(tmp12 + tmp1,
370 CONST_BITS+PASS1_BITS+3);
371 dataptr[DCTSIZE*5] = (dctelem_t) DESCALE(tmp12 - tmp1,
372 CONST_BITS+PASS1_BITS+3);
373 dataptr[DCTSIZE*3] = (dctelem_t) DESCALE(tmp13 + tmp0,
374 CONST_BITS+PASS1_BITS+3);
375 dataptr[DCTSIZE*4] = (dctelem_t) DESCALE(tmp13 - tmp0,
376 CONST_BITS+PASS1_BITS+3);
378 dataptr++; /* advance pointer to next column */