1 /*****************************************************************************
2 * idctclassic.c : Classic IDCT module
3 *****************************************************************************
4 * Copyright (C) 1999, 2000 VideoLAN
5 * $Id: idctclassic.c,v 1.9 2001/05/06 04:32:02 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 #define MODULE_NAME idctclassic
26 /*****************************************************************************
28 *****************************************************************************/
40 #include "video_output.h"
42 #include "video_decoder.h"
45 #include "modules_inner.h"
47 #include "vdec_block.h"
48 #include "vdec_idct.h"
50 /*****************************************************************************
51 * Local and extern prototypes.
52 *****************************************************************************/
53 static void idct_getfunctions( function_list_t * p_function_list );
54 static int idct_Probe ( probedata_t *p_data );
55 static void vdec_NormScan ( u8 ppi_scan[2][64] );
58 /*****************************************************************************
59 * Build configuration tree.
60 *****************************************************************************/
62 ADD_WINDOW( "Configuration for classic IDCT module" )
63 ADD_COMMENT( "Ha, ha -- nothing to configure yet" )
66 /*****************************************************************************
67 * InitModule: get the module structure and configuration.
68 *****************************************************************************
69 * We have to fill psz_name, psz_longname and psz_version. These variables
70 * will be strdup()ed later by the main application because the module can
71 * be unloaded later to save memory, and we want to be able to access this
72 * data even after the module has been unloaded.
73 *****************************************************************************/
76 p_module->psz_name = MODULE_STRING;
77 p_module->psz_longname = "classic IDCT module";
78 p_module->psz_version = VERSION;
80 p_module->i_capabilities = MODULE_CAPABILITY_NULL
81 | MODULE_CAPABILITY_IDCT;
86 /*****************************************************************************
87 * ActivateModule: set the module to an usable state.
88 *****************************************************************************
89 * This function fills the capability functions and the configuration
90 * structure. Once ActivateModule() has been called, the i_usage can
91 * be set to 0 and calls to NeedModule() be made to increment it. To unload
92 * the module, one has to wait until i_usage == 0 and call DeactivateModule().
93 *****************************************************************************/
96 p_module->p_functions = malloc( sizeof( module_functions_t ) );
97 if( p_module->p_functions == NULL )
102 idct_getfunctions( &p_module->p_functions->idct );
104 p_module->p_config = p_config;
109 /*****************************************************************************
110 * DeactivateModule: make sure the module can be unloaded.
111 *****************************************************************************
112 * This function must only be called when i_usage == 0. If it successfully
113 * returns, i_usage can be set to -1 and the module unloaded. Be careful to
114 * lock usage_lock during the whole process.
115 *****************************************************************************/
118 free( p_module->p_functions );
123 /* Following functions are local */
125 /*****************************************************************************
126 * Functions exported as capabilities. They are declared as static so that
127 * we don't pollute the namespace too much.
128 *****************************************************************************/
129 static void idct_getfunctions( function_list_t * p_function_list )
131 p_function_list->pf_probe = idct_Probe;
132 #define F p_function_list->functions.idct
133 F.pf_idct_init = _M( vdec_InitIDCT );
134 F.pf_sparse_idct = _M( vdec_SparseIDCT );
135 F.pf_idct = _M( vdec_IDCT );
136 F.pf_norm_scan = vdec_NormScan;
137 F.pf_vdec_init = _M( vdec_Init );
138 F.pf_decode_mb_c = _M( vdec_DecodeMacroblockC );
139 F.pf_decode_mb_bw = _M( vdec_DecodeMacroblockBW );
143 /*****************************************************************************
144 * idct_Probe: returns a preference score
145 *****************************************************************************/
146 static int idct_Probe( probedata_t *p_data )
148 if( TestMethod( IDCT_METHOD_VAR, "idctclassic" ) )
153 /* This plugin always works */
157 /*****************************************************************************
158 * vdec_NormScan : Unused in this IDCT
159 *****************************************************************************/
160 static void vdec_NormScan( u8 ppi_scan[2][64] )
164 /*****************************************************************************
165 * vdec_IDCT : IDCT function for normal matrices
166 *****************************************************************************/
167 void _M( vdec_IDCT )( vdec_thread_t * p_vdec, dctelem_t * p_block,
170 /* dct classique: pour tester la meilleure entre la classique et la */
172 s32 tmp0, tmp1, tmp2, tmp3;
173 s32 tmp10, tmp11, tmp12, tmp13;
174 s32 z1, z2, z3, z4, z5;
179 /* Pass 1: process rows. */
180 /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
181 /* furthermore, we scale the results by 2**PASS1_BITS. */
184 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
186 /* Due to quantization, we will usually find that many of the input
187 * coefficients are zero, especially the AC terms. We can exploit this
188 * by short-circuiting the IDCT calculation for any row in which all
189 * the AC terms are zero. In that case each output is equal to the
190 * DC coefficient (with scale factor as needed).
191 * With typical images and quantization tables, half or more of the
192 * row DCT calculations can be simplified this way.
195 if ((dataptr[1] | dataptr[2] | dataptr[3] | dataptr[4] |
196 dataptr[5] | dataptr[6] | dataptr[7]) == 0)
198 /* AC terms all zero */
199 dctelem_t dcval = (dctelem_t) (dataptr[0] << PASS1_BITS);
210 dataptr += DCTSIZE; /* advance pointer to next row */
214 /* Even part: reverse the even part of the forward DCT. */
215 /* The rotator is sqrt(2)*c(-6). */
217 z2 = (s32) dataptr[2];
218 z3 = (s32) dataptr[6];
220 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
221 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
222 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
224 tmp0 = ((s32) dataptr[0] + (s32) dataptr[4]) << CONST_BITS;
225 tmp1 = ((s32) dataptr[0] - (s32) dataptr[4]) << CONST_BITS;
232 /* Odd part per figure 8; the matrix is unitary and hence its
233 * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively.
236 tmp0 = (s32) dataptr[7];
237 tmp1 = (s32) dataptr[5];
238 tmp2 = (s32) dataptr[3];
239 tmp3 = (s32) dataptr[1];
245 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
247 tmp0 = MULTIPLY(tmp0, FIX(0.298631336)); /* sqrt(2) * (-c1+c3+c5-c7) */
248 tmp1 = MULTIPLY(tmp1, FIX(2.053119869)); /* sqrt(2) * ( c1+c3-c5+c7) */
249 tmp2 = MULTIPLY(tmp2, FIX(3.072711026)); /* sqrt(2) * ( c1+c3+c5-c7) */
250 tmp3 = MULTIPLY(tmp3, FIX(1.501321110)); /* sqrt(2) * ( c1+c3-c5-c7) */
251 z1 = MULTIPLY(z1, - FIX(0.899976223)); /* sqrt(2) * (c7-c3) */
252 z2 = MULTIPLY(z2, - FIX(2.562915447)); /* sqrt(2) * (-c1-c3) */
253 z3 = MULTIPLY(z3, - FIX(1.961570560)); /* sqrt(2) * (-c3-c5) */
254 z4 = MULTIPLY(z4, - FIX(0.390180644)); /* sqrt(2) * (c5-c3) */
264 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
266 dataptr[0] = (dctelem_t) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
267 dataptr[7] = (dctelem_t) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
268 dataptr[1] = (dctelem_t) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
269 dataptr[6] = (dctelem_t) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS);
270 dataptr[2] = (dctelem_t) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS);
271 dataptr[5] = (dctelem_t) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
272 dataptr[3] = (dctelem_t) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
273 dataptr[4] = (dctelem_t) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
275 dataptr += DCTSIZE; /* advance pointer to next row */
278 /* Pass 2: process columns. */
279 /* Note that we must descale the results by a factor of 8 == 2**3, */
280 /* and also undo the PASS1_BITS scaling. */
283 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
285 /* Columns of zeroes can be exploited in the same way as we did with rows.
286 * However, the row calculation has created many nonzero AC terms, so the
287 * simplification applies less often (typically 5% to 10% of the time).
288 * On machines with very fast multiplication, it's possible that the
289 * test takes more time than it's worth. In that case this section
290 * may be commented out.
293 #ifndef NO_ZERO_COLUMN_TEST /*ajoute un test mais evite des calculs */
294 if ((dataptr[DCTSIZE*1] | dataptr[DCTSIZE*2] | dataptr[DCTSIZE*3] |
295 dataptr[DCTSIZE*4] | dataptr[DCTSIZE*5] | dataptr[DCTSIZE*6] |
296 dataptr[DCTSIZE*7]) == 0)
298 /* AC terms all zero */
299 dctelem_t dcval = (dctelem_t) DESCALE((s32) dataptr[0], PASS1_BITS+3);
301 dataptr[DCTSIZE*0] = dcval;
302 dataptr[DCTSIZE*1] = dcval;
303 dataptr[DCTSIZE*2] = dcval;
304 dataptr[DCTSIZE*3] = dcval;
305 dataptr[DCTSIZE*4] = dcval;
306 dataptr[DCTSIZE*5] = dcval;
307 dataptr[DCTSIZE*6] = dcval;
308 dataptr[DCTSIZE*7] = dcval;
310 dataptr++; /* advance pointer to next column */
315 /* Even part: reverse the even part of the forward DCT. */
316 /* The rotator is sqrt(2)*c(-6). */
318 z2 = (s32) dataptr[DCTSIZE*2];
319 z3 = (s32) dataptr[DCTSIZE*6];
321 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
322 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
323 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
325 tmp0 = ((s32) dataptr[DCTSIZE*0] + (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
326 tmp1 = ((s32) dataptr[DCTSIZE*0] - (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
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.
337 tmp0 = (s32) dataptr[DCTSIZE*7];
338 tmp1 = (s32) dataptr[DCTSIZE*5];
339 tmp2 = (s32) dataptr[DCTSIZE*3];
340 tmp3 = (s32) dataptr[DCTSIZE*1];
346 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
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) */
365 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
367 dataptr[DCTSIZE*0] = (dctelem_t) DESCALE(tmp10 + tmp3,
368 CONST_BITS+PASS1_BITS+3);
369 dataptr[DCTSIZE*7] = (dctelem_t) DESCALE(tmp10 - tmp3,
370 CONST_BITS+PASS1_BITS+3);
371 dataptr[DCTSIZE*1] = (dctelem_t) DESCALE(tmp11 + tmp2,
372 CONST_BITS+PASS1_BITS+3);
373 dataptr[DCTSIZE*6] = (dctelem_t) DESCALE(tmp11 - tmp2,
374 CONST_BITS+PASS1_BITS+3);
375 dataptr[DCTSIZE*2] = (dctelem_t) DESCALE(tmp12 + tmp1,
376 CONST_BITS+PASS1_BITS+3);
377 dataptr[DCTSIZE*5] = (dctelem_t) DESCALE(tmp12 - tmp1,
378 CONST_BITS+PASS1_BITS+3);
379 dataptr[DCTSIZE*3] = (dctelem_t) DESCALE(tmp13 + tmp0,
380 CONST_BITS+PASS1_BITS+3);
381 dataptr[DCTSIZE*4] = (dctelem_t) DESCALE(tmp13 - tmp0,
382 CONST_BITS+PASS1_BITS+3);
384 dataptr++; /* advance pointer to next column */