1 /*****************************************************************************
2 * idctclassic.c : Classic IDCT module
3 *****************************************************************************
4 * Copyright (C) 1999, 2000 VideoLAN
5 * $Id: idctclassic.c,v 1.14 2001/08/22 17:21:45 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 *****************************************************************************/
40 #include "vdec_idct.h"
43 #include "modules_export.h"
45 /*****************************************************************************
46 * Local and extern prototypes.
47 *****************************************************************************/
48 static void idct_getfunctions( function_list_t * p_function_list );
49 static int idct_Probe ( probedata_t *p_data );
50 static void vdec_NormScan ( u8 ppi_scan[2][64] );
53 /*****************************************************************************
54 * Build configuration tree.
55 *****************************************************************************/
57 ADD_WINDOW( "Configuration for classic IDCT module" )
58 ADD_COMMENT( "Ha, ha -- nothing to configure yet" )
62 p_module->i_capabilities = MODULE_CAPABILITY_NULL
63 | MODULE_CAPABILITY_IDCT;
64 p_module->psz_longname = "classic IDCT module";
68 idct_getfunctions( &p_module->p_functions->idct );
71 MODULE_DEACTIVATE_START
72 MODULE_DEACTIVATE_STOP
74 /* Following functions are local */
76 /*****************************************************************************
77 * Functions exported as capabilities. They are declared as static so that
78 * we don't pollute the namespace too much.
79 *****************************************************************************/
80 static void idct_getfunctions( function_list_t * p_function_list )
82 p_function_list->pf_probe = idct_Probe;
83 #define F p_function_list->functions.idct
84 F.pf_idct_init = _M( vdec_InitIDCT );
85 F.pf_sparse_idct = _M( vdec_SparseIDCT );
86 F.pf_idct = _M( vdec_IDCT );
87 F.pf_norm_scan = vdec_NormScan;
88 F.pf_decode_init = _M( vdec_InitDecode );
89 F.pf_addblock = _M( vdec_AddBlock );
90 F.pf_copyblock = _M( vdec_CopyBlock );
94 /*****************************************************************************
95 * idct_Probe: returns a preference score
96 *****************************************************************************/
97 static int idct_Probe( probedata_t *p_data )
99 if( TestMethod( IDCT_METHOD_VAR, "idctclassic" )
100 || TestMethod( IDCT_METHOD_VAR, "classic" ) )
105 /* This plugin always works */
109 /*****************************************************************************
110 * vdec_NormScan : Unused in this IDCT
111 *****************************************************************************/
112 static void vdec_NormScan( u8 ppi_scan[2][64] )
116 /*****************************************************************************
117 * vdec_IDCT : IDCT function for normal matrices
118 *****************************************************************************/
119 void _M( vdec_IDCT )( void * p_unused_data, dctelem_t * p_block,
122 /* dct classique: pour tester la meilleure entre la classique et la */
124 s32 tmp0, tmp1, tmp2, tmp3;
125 s32 tmp10, tmp11, tmp12, tmp13;
126 s32 z1, z2, z3, z4, z5;
131 /* Pass 1: process rows. */
132 /* Note results are scaled up by sqrt(8) compared to a true IDCT; */
133 /* furthermore, we scale the results by 2**PASS1_BITS. */
136 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
138 /* Due to quantization, we will usually find that many of the input
139 * coefficients are zero, especially the AC terms. We can exploit this
140 * by short-circuiting the IDCT calculation for any row in which all
141 * the AC terms are zero. In that case each output is equal to the
142 * DC coefficient (with scale factor as needed).
143 * With typical images and quantization tables, half or more of the
144 * row DCT calculations can be simplified this way.
147 if ((dataptr[1] | dataptr[2] | dataptr[3] | dataptr[4] |
148 dataptr[5] | dataptr[6] | dataptr[7]) == 0)
150 /* AC terms all zero */
151 dctelem_t dcval = (dctelem_t) (dataptr[0] << PASS1_BITS);
162 dataptr += DCTSIZE; /* advance pointer to next row */
166 /* Even part: reverse the even part of the forward DCT. */
167 /* The rotator is sqrt(2)*c(-6). */
169 z2 = (s32) dataptr[2];
170 z3 = (s32) dataptr[6];
172 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
173 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
174 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
176 tmp0 = ((s32) dataptr[0] + (s32) dataptr[4]) << CONST_BITS;
177 tmp1 = ((s32) dataptr[0] - (s32) dataptr[4]) << CONST_BITS;
184 /* Odd part per figure 8; the matrix is unitary and hence its
185 * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively.
188 tmp0 = (s32) dataptr[7];
189 tmp1 = (s32) dataptr[5];
190 tmp2 = (s32) dataptr[3];
191 tmp3 = (s32) dataptr[1];
197 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
199 tmp0 = MULTIPLY(tmp0, FIX(0.298631336)); /* sqrt(2) * (-c1+c3+c5-c7) */
200 tmp1 = MULTIPLY(tmp1, FIX(2.053119869)); /* sqrt(2) * ( c1+c3-c5+c7) */
201 tmp2 = MULTIPLY(tmp2, FIX(3.072711026)); /* sqrt(2) * ( c1+c3+c5-c7) */
202 tmp3 = MULTIPLY(tmp3, FIX(1.501321110)); /* sqrt(2) * ( c1+c3-c5-c7) */
203 z1 = MULTIPLY(z1, - FIX(0.899976223)); /* sqrt(2) * (c7-c3) */
204 z2 = MULTIPLY(z2, - FIX(2.562915447)); /* sqrt(2) * (-c1-c3) */
205 z3 = MULTIPLY(z3, - FIX(1.961570560)); /* sqrt(2) * (-c3-c5) */
206 z4 = MULTIPLY(z4, - FIX(0.390180644)); /* sqrt(2) * (c5-c3) */
216 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
218 dataptr[0] = (dctelem_t) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
219 dataptr[7] = (dctelem_t) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
220 dataptr[1] = (dctelem_t) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
221 dataptr[6] = (dctelem_t) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS);
222 dataptr[2] = (dctelem_t) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS);
223 dataptr[5] = (dctelem_t) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
224 dataptr[3] = (dctelem_t) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
225 dataptr[4] = (dctelem_t) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
227 dataptr += DCTSIZE; /* advance pointer to next row */
230 /* Pass 2: process columns. */
231 /* Note that we must descale the results by a factor of 8 == 2**3, */
232 /* and also undo the PASS1_BITS scaling. */
235 for (rowctr = DCTSIZE-1; rowctr >= 0; rowctr--)
237 /* Columns of zeroes can be exploited in the same way as we did with rows.
238 * However, the row calculation has created many nonzero AC terms, so the
239 * simplification applies less often (typically 5% to 10% of the time).
240 * On machines with very fast multiplication, it's possible that the
241 * test takes more time than it's worth. In that case this section
242 * may be commented out.
245 #ifndef NO_ZERO_COLUMN_TEST /* Adds a test but avoids calculus */
246 if ((dataptr[DCTSIZE*1] | dataptr[DCTSIZE*2] | dataptr[DCTSIZE*3] |
247 dataptr[DCTSIZE*4] | dataptr[DCTSIZE*5] | dataptr[DCTSIZE*6] |
248 dataptr[DCTSIZE*7]) == 0)
250 /* AC terms all zero */
251 dctelem_t dcval = (dctelem_t) DESCALE((s32) dataptr[0], PASS1_BITS+3);
253 dataptr[DCTSIZE*0] = dcval;
254 dataptr[DCTSIZE*1] = dcval;
255 dataptr[DCTSIZE*2] = dcval;
256 dataptr[DCTSIZE*3] = dcval;
257 dataptr[DCTSIZE*4] = dcval;
258 dataptr[DCTSIZE*5] = dcval;
259 dataptr[DCTSIZE*6] = dcval;
260 dataptr[DCTSIZE*7] = dcval;
262 dataptr++; /* advance pointer to next column */
267 /* Even part: reverse the even part of the forward DCT. */
268 /* The rotator is sqrt(2)*c(-6). */
270 z2 = (s32) dataptr[DCTSIZE*2];
271 z3 = (s32) dataptr[DCTSIZE*6];
273 z1 = MULTIPLY(z2 + z3, FIX(0.541196100));
274 tmp2 = z1 + MULTIPLY(z3, - FIX(1.847759065));
275 tmp3 = z1 + MULTIPLY(z2, FIX(0.765366865));
277 tmp0 = ((s32) dataptr[DCTSIZE*0] + (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
278 tmp1 = ((s32) dataptr[DCTSIZE*0] - (s32) dataptr[DCTSIZE*4]) << CONST_BITS;
285 /* Odd part per figure 8; the matrix is unitary and hence its
286 * transpose is its inverse. i0..i3 are y7,y5,y3,y1 respectively.
289 tmp0 = (s32) dataptr[DCTSIZE*7];
290 tmp1 = (s32) dataptr[DCTSIZE*5];
291 tmp2 = (s32) dataptr[DCTSIZE*3];
292 tmp3 = (s32) dataptr[DCTSIZE*1];
298 z5 = MULTIPLY(z3 + z4, FIX(1.175875602)); /* sqrt(2) * c3 */
300 tmp0 = MULTIPLY(tmp0, FIX(0.298631336)); /* sqrt(2) * (-c1+c3+c5-c7) */
301 tmp1 = MULTIPLY(tmp1, FIX(2.053119869)); /* sqrt(2) * ( c1+c3-c5+c7) */
302 tmp2 = MULTIPLY(tmp2, FIX(3.072711026)); /* sqrt(2) * ( c1+c3+c5-c7) */
303 tmp3 = MULTIPLY(tmp3, FIX(1.501321110)); /* sqrt(2) * ( c1+c3-c5-c7) */
304 z1 = MULTIPLY(z1, - FIX(0.899976223)); /* sqrt(2) * (c7-c3) */
305 z2 = MULTIPLY(z2, - FIX(2.562915447)); /* sqrt(2) * (-c1-c3) */
306 z3 = MULTIPLY(z3, - FIX(1.961570560)); /* sqrt(2) * (-c3-c5) */
307 z4 = MULTIPLY(z4, - FIX(0.390180644)); /* sqrt(2) * (c5-c3) */
317 /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
319 dataptr[DCTSIZE*0] = (dctelem_t) DESCALE(tmp10 + tmp3,
320 CONST_BITS+PASS1_BITS+3);
321 dataptr[DCTSIZE*7] = (dctelem_t) DESCALE(tmp10 - tmp3,
322 CONST_BITS+PASS1_BITS+3);
323 dataptr[DCTSIZE*1] = (dctelem_t) DESCALE(tmp11 + tmp2,
324 CONST_BITS+PASS1_BITS+3);
325 dataptr[DCTSIZE*6] = (dctelem_t) DESCALE(tmp11 - tmp2,
326 CONST_BITS+PASS1_BITS+3);
327 dataptr[DCTSIZE*2] = (dctelem_t) DESCALE(tmp12 + tmp1,
328 CONST_BITS+PASS1_BITS+3);
329 dataptr[DCTSIZE*5] = (dctelem_t) DESCALE(tmp12 - tmp1,
330 CONST_BITS+PASS1_BITS+3);
331 dataptr[DCTSIZE*3] = (dctelem_t) DESCALE(tmp13 + tmp0,
332 CONST_BITS+PASS1_BITS+3);
333 dataptr[DCTSIZE*4] = (dctelem_t) DESCALE(tmp13 - tmp0,
334 CONST_BITS+PASS1_BITS+3);
336 dataptr++; /* advance pointer to next column */