1 /*****************************************************************************
2 * idctclassic.c : Classic IDCT module
3 *****************************************************************************
4 * Copyright (C) 1999, 2000 VideoLAN
5 * $Id: idctclassic.c,v 1.10 2001/05/30 17:03:12 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" )
67 p_module->i_capabilities = MODULE_CAPABILITY_NULL
68 | MODULE_CAPABILITY_IDCT;
69 p_module->psz_longname = "classic IDCT module";
73 idct_getfunctions( &p_module->p_functions->idct );
76 MODULE_DEACTIVATE_START
77 MODULE_DEACTIVATE_STOP
79 /* Following functions are local */
81 /*****************************************************************************
82 * Functions exported as capabilities. They are declared as static so that
83 * we don't pollute the namespace too much.
84 *****************************************************************************/
85 static void idct_getfunctions( function_list_t * p_function_list )
87 p_function_list->pf_probe = idct_Probe;
88 #define F p_function_list->functions.idct
89 F.pf_idct_init = _M( vdec_InitIDCT );
90 F.pf_sparse_idct = _M( vdec_SparseIDCT );
91 F.pf_idct = _M( vdec_IDCT );
92 F.pf_norm_scan = vdec_NormScan;
93 F.pf_decode_init = _M( vdec_InitDecode );
94 F.pf_decode_mb_c = _M( vdec_DecodeMacroblockC );
95 F.pf_decode_mb_bw = _M( vdec_DecodeMacroblockBW );
99 /*****************************************************************************
100 * idct_Probe: returns a preference score
101 *****************************************************************************/
102 static int idct_Probe( probedata_t *p_data )
104 if( TestMethod( IDCT_METHOD_VAR, "idctclassic" ) )
109 /* This plugin always works */
113 /*****************************************************************************
114 * vdec_NormScan : Unused in this IDCT
115 *****************************************************************************/
116 static void vdec_NormScan( u8 ppi_scan[2][64] )
120 /*****************************************************************************
121 * vdec_IDCT : IDCT function for normal matrices
122 *****************************************************************************/
123 void _M( vdec_IDCT )( vdec_thread_t * p_vdec, dctelem_t * p_block,
126 /* dct classique: pour tester la meilleure entre la classique et la */
128 s32 tmp0, tmp1, tmp2, tmp3;
129 s32 tmp10, tmp11, tmp12, tmp13;
130 s32 z1, z2, z3, z4, z5;
135 /* Pass 1: process rows. */
136 /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
137 /* furthermore, we scale the results by 2**PASS1_BITS. */
140 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
142 /* Due to quantization, we will usually find that many of the input
143 * coefficients are zero, especially the AC terms. We can exploit this
144 * by short-circuiting the IDCT calculation for any row in which all
145 * the AC terms are zero. In that case each output is equal to the
146 * DC coefficient (with scale factor as needed).
147 * With typical images and quantization tables, half or more of the
148 * row DCT calculations can be simplified this way.
151 if ((dataptr[1] | dataptr[2] | dataptr[3] | dataptr[4] |
152 dataptr[5] | dataptr[6] | dataptr[7]) == 0)
154 /* AC terms all zero */
155 dctelem_t dcval = (dctelem_t) (dataptr[0] << PASS1_BITS);
166 dataptr += DCTSIZE; /* advance pointer to next row */
170 /* Even part: reverse the even part of the forward DCT. */
171 /* The rotator is sqrt(2)*c(-6). */
173 z2 = (s32) dataptr[2];
174 z3 = (s32) dataptr[6];
176 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
177 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
178 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
180 tmp0 = ((s32) dataptr[0] + (s32) dataptr[4]) << CONST_BITS;
181 tmp1 = ((s32) dataptr[0] - (s32) dataptr[4]) << CONST_BITS;
188 /* Odd part per figure 8; the matrix is unitary and hence its
189 * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively.
192 tmp0 = (s32) dataptr[7];
193 tmp1 = (s32) dataptr[5];
194 tmp2 = (s32) dataptr[3];
195 tmp3 = (s32) dataptr[1];
201 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
203 tmp0 = MULTIPLY(tmp0, FIX(0.298631336)); /* sqrt(2) * (-c1+c3+c5-c7) */
204 tmp1 = MULTIPLY(tmp1, FIX(2.053119869)); /* sqrt(2) * ( c1+c3-c5+c7) */
205 tmp2 = MULTIPLY(tmp2, FIX(3.072711026)); /* sqrt(2) * ( c1+c3+c5-c7) */
206 tmp3 = MULTIPLY(tmp3, FIX(1.501321110)); /* sqrt(2) * ( c1+c3-c5-c7) */
207 z1 = MULTIPLY(z1, - FIX(0.899976223)); /* sqrt(2) * (c7-c3) */
208 z2 = MULTIPLY(z2, - FIX(2.562915447)); /* sqrt(2) * (-c1-c3) */
209 z3 = MULTIPLY(z3, - FIX(1.961570560)); /* sqrt(2) * (-c3-c5) */
210 z4 = MULTIPLY(z4, - FIX(0.390180644)); /* sqrt(2) * (c5-c3) */
220 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
222 dataptr[0] = (dctelem_t) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
223 dataptr[7] = (dctelem_t) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
224 dataptr[1] = (dctelem_t) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
225 dataptr[6] = (dctelem_t) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS);
226 dataptr[2] = (dctelem_t) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS);
227 dataptr[5] = (dctelem_t) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
228 dataptr[3] = (dctelem_t) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
229 dataptr[4] = (dctelem_t) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
231 dataptr += DCTSIZE; /* advance pointer to next row */
234 /* Pass 2: process columns. */
235 /* Note that we must descale the results by a factor of 8 == 2**3, */
236 /* and also undo the PASS1_BITS scaling. */
239 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
241 /* Columns of zeroes can be exploited in the same way as we did with rows.
242 * However, the row calculation has created many nonzero AC terms, so the
243 * simplification applies less often (typically 5% to 10% of the time).
244 * On machines with very fast multiplication, it's possible that the
245 * test takes more time than it's worth. In that case this section
246 * may be commented out.
249 #ifndef NO_ZERO_COLUMN_TEST /*ajoute un test mais evite des calculs */
250 if ((dataptr[DCTSIZE*1] | dataptr[DCTSIZE*2] | dataptr[DCTSIZE*3] |
251 dataptr[DCTSIZE*4] | dataptr[DCTSIZE*5] | dataptr[DCTSIZE*6] |
252 dataptr[DCTSIZE*7]) == 0)
254 /* AC terms all zero */
255 dctelem_t dcval = (dctelem_t) DESCALE((s32) dataptr[0], PASS1_BITS+3);
257 dataptr[DCTSIZE*0] = dcval;
258 dataptr[DCTSIZE*1] = dcval;
259 dataptr[DCTSIZE*2] = dcval;
260 dataptr[DCTSIZE*3] = dcval;
261 dataptr[DCTSIZE*4] = dcval;
262 dataptr[DCTSIZE*5] = dcval;
263 dataptr[DCTSIZE*6] = dcval;
264 dataptr[DCTSIZE*7] = dcval;
266 dataptr++; /* advance pointer to next column */
271 /* Even part: reverse the even part of the forward DCT. */
272 /* The rotator is sqrt(2)*c(-6). */
274 z2 = (s32) dataptr[DCTSIZE*2];
275 z3 = (s32) dataptr[DCTSIZE*6];
277 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
278 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
279 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
281 tmp0 = ((s32) dataptr[DCTSIZE*0] + (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
282 tmp1 = ((s32) dataptr[DCTSIZE*0] - (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
289 /* Odd part per figure 8; the matrix is unitary and hence its
290 * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively.
293 tmp0 = (s32) dataptr[DCTSIZE*7];
294 tmp1 = (s32) dataptr[DCTSIZE*5];
295 tmp2 = (s32) dataptr[DCTSIZE*3];
296 tmp3 = (s32) dataptr[DCTSIZE*1];
302 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
304 tmp0 = MULTIPLY(tmp0, FIX(0.298631336)); /* sqrt(2) * (-c1+c3+c5-c7) */
305 tmp1 = MULTIPLY(tmp1, FIX(2.053119869)); /* sqrt(2) * ( c1+c3-c5+c7) */
306 tmp2 = MULTIPLY(tmp2, FIX(3.072711026)); /* sqrt(2) * ( c1+c3+c5-c7) */
307 tmp3 = MULTIPLY(tmp3, FIX(1.501321110)); /* sqrt(2) * ( c1+c3-c5-c7) */
308 z1 = MULTIPLY(z1, - FIX(0.899976223)); /* sqrt(2) * (c7-c3) */
309 z2 = MULTIPLY(z2, - FIX(2.562915447)); /* sqrt(2) * (-c1-c3) */
310 z3 = MULTIPLY(z3, - FIX(1.961570560)); /* sqrt(2) * (-c3-c5) */
311 z4 = MULTIPLY(z4, - FIX(0.390180644)); /* sqrt(2) * (c5-c3) */
321 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
323 dataptr[DCTSIZE*0] = (dctelem_t) DESCALE(tmp10 + tmp3,
324 CONST_BITS+PASS1_BITS+3);
325 dataptr[DCTSIZE*7] = (dctelem_t) DESCALE(tmp10 - tmp3,
326 CONST_BITS+PASS1_BITS+3);
327 dataptr[DCTSIZE*1] = (dctelem_t) DESCALE(tmp11 + tmp2,
328 CONST_BITS+PASS1_BITS+3);
329 dataptr[DCTSIZE*6] = (dctelem_t) DESCALE(tmp11 - tmp2,
330 CONST_BITS+PASS1_BITS+3);
331 dataptr[DCTSIZE*2] = (dctelem_t) DESCALE(tmp12 + tmp1,
332 CONST_BITS+PASS1_BITS+3);
333 dataptr[DCTSIZE*5] = (dctelem_t) DESCALE(tmp12 - tmp1,
334 CONST_BITS+PASS1_BITS+3);
335 dataptr[DCTSIZE*3] = (dctelem_t) DESCALE(tmp13 + tmp0,
336 CONST_BITS+PASS1_BITS+3);
337 dataptr[DCTSIZE*4] = (dctelem_t) DESCALE(tmp13 - tmp0,
338 CONST_BITS+PASS1_BITS+3);
340 dataptr++; /* advance pointer to next column */