src/dct/simple_idct.c

/*****************************************************************************
 *
 *  XVID MPEG-4 VIDEO CODEC
 *  - Inverse DCT (More precise version)  -
 *
 *  Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at>
 *
 *  Originally distributed under the GNU LGPL License (ffmpeg).
 *  It is licensed under the GNU GPL for the Xvid tree.
 *
 *  This program is free software ; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation ; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY ; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program ; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
 *
 * $Id: simple_idct.c,v 1.5 2004/04/05 20:36:36 edgomez Exp $
 *
 ****************************************************************************/

/*
  based upon some outcommented c code from mpeg2dec (idct_mmx.c
  written by Aaron Holtzman <aholtzma@ess.engr.uvic.ca>)
 */

#include "../portab.h"
#include "idct.h"

#if 0
#define W1 2841 /* 2048*sqrt (2)*cos (1*pi/16) */
#define W2 2676 /* 2048*sqrt (2)*cos (2*pi/16) */
#define W3 2408 /* 2048*sqrt (2)*cos (3*pi/16) */
#define W4 2048 /* 2048*sqrt (2)*cos (4*pi/16) */
#define W5 1609 /* 2048*sqrt (2)*cos (5*pi/16) */
#define W6 1108 /* 2048*sqrt (2)*cos (6*pi/16) */
#define W7 565  /* 2048*sqrt (2)*cos (7*pi/16) */
#define ROW_SHIFT 8
#define COL_SHIFT 17
#else
#define W1  22725  /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
#define W2  21407  /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
#define W3  19266  /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
#define W4  16383  /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
#define W5  12873  /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
#define W6  8867   /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
#define W7  4520   /* cos(i*M_PI/16)*sqrt(2)*(1<<14) + 0.5 */
#define ROW_SHIFT 11
#define COL_SHIFT 20 /* 6 */
#endif

/*
 PPC mac operation.  Causes compile problems on newer ppc targets
 
 Was originally: #if defined(ARCH_IS_PPC)
 */
#if 0

/* signed 16x16 -> 32 multiply add accumulate */
#define MAC16(rt, ra, rb) \
    asm ("maclhw %0, %2, %3" : "=r" (rt) : "0" (rt), "r" (ra), "r" (rb));

/* signed 16x16 -> 32 multiply */
#define MUL16(rt, ra, rb) \
    asm ("mullhw %0, %1, %2" : "=r" (rt) : "r" (ra), "r" (rb));

#else

/* signed 16x16 -> 32 multiply add accumulate */
#define MAC16(rt, ra, rb) rt += (ra) * (rb)

/* signed 16x16 -> 32 multiply */
#define MUL16(rt, ra, rb) rt = (ra) * (rb)

#endif

static __inline void idctRowCondDC (int16_t * const row)
{
        int a0, a1, a2, a3, b0, b1, b2, b3;
#ifdef FAST_64BIT
        uint64_t temp;
#else
        uint32_t temp;
#endif

#ifdef FAST_64BIT
#ifdef ARCH_IS_BIG_ENDIAN
#define ROW0_MASK 0xffff000000000000LL
#else
#define ROW0_MASK 0xffffLL
#endif
        if ( ((((uint64_t *)row)[0] & ~ROW0_MASK) |
              ((uint64_t *)row)[1]) == 0) {
            temp = (row[0] << 3) & 0xffff;
            temp += temp << 16;
            temp += temp << 32;
            ((uint64_t *)row)[0] = temp;
            ((uint64_t *)row)[1] = temp;
            return;
        }
#else
        if (!(((uint32_t*)row)[1] |
              ((uint32_t*)row)[2] |
              ((uint32_t*)row)[3] |
              row[1])) {
            temp = (row[0] << 3) & 0xffff;
            temp += temp << 16;
            ((uint32_t*)row)[0]=((uint32_t*)row)[1] =
                ((uint32_t*)row)[2]=((uint32_t*)row)[3] = temp;
                return;
        }
#endif

        a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1));
        a1 = a0;
        a2 = a0;
        a3 = a0;

        /* no need to optimize : gcc does it */
        a0 += W2 * row[2];
        a1 += W6 * row[2];
        a2 -= W6 * row[2];
        a3 -= W2 * row[2];

        MUL16(b0, W1, row[1]);
        MAC16(b0, W3, row[3]);
        MUL16(b1, W3, row[1]);
        MAC16(b1, -W7, row[3]);
        MUL16(b2, W5, row[1]);
        MAC16(b2, -W1, row[3]);
        MUL16(b3, W7, row[1]);
        MAC16(b3, -W5, row[3]);

#ifdef FAST_64BIT
        temp = ((uint64_t*)row)[1];
#else
        temp = ((uint32_t*)row)[2] | ((uint32_t*)row)[3];
#endif
        if (temp != 0) {
            a0 += W4*row[4] + W6*row[6];
            a1 += - W4*row[4] - W2*row[6];
            a2 += - W4*row[4] + W2*row[6];
            a3 += W4*row[4] - W6*row[6];

            MAC16(b0, W5, row[5]);
            MAC16(b0, W7, row[7]);

            MAC16(b1, -W1, row[5]);
            MAC16(b1, -W5, row[7]);

            MAC16(b2, W7, row[5]);
            MAC16(b2, W3, row[7]);

            MAC16(b3, W3, row[5]);
            MAC16(b3, -W1, row[7]);
        }

        row[0] = (a0 + b0) >> ROW_SHIFT;
        row[7] = (a0 - b0) >> ROW_SHIFT;
        row[1] = (a1 + b1) >> ROW_SHIFT;
        row[6] = (a1 - b1) >> ROW_SHIFT;
        row[2] = (a2 + b2) >> ROW_SHIFT;
        row[5] = (a2 - b2) >> ROW_SHIFT;
        row[3] = (a3 + b3) >> ROW_SHIFT;
        row[4] = (a3 - b3) >> ROW_SHIFT;
}


static __inline void idctSparseCol (int16_t * const col)
{
        int a0, a1, a2, a3, b0, b1, b2, b3;

        /* XXX: I did that only to give same values as previous code */
        a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4));
        a1 = a0;
        a2 = a0;
        a3 = a0;

        a0 +=  + W2*col[8*2];
        a1 +=  + W6*col[8*2];
        a2 +=  - W6*col[8*2];
        a3 +=  - W2*col[8*2];

        MUL16(b0, W1, col[8*1]);
        MUL16(b1, W3, col[8*1]);
        MUL16(b2, W5, col[8*1]);
        MUL16(b3, W7, col[8*1]);

        MAC16(b0, + W3, col[8*3]);
        MAC16(b1, - W7, col[8*3]);
        MAC16(b2, - W1, col[8*3]);
        MAC16(b3, - W5, col[8*3]);

        if(col[8*4]){
            a0 += + W4*col[8*4];
            a1 += - W4*col[8*4];
            a2 += - W4*col[8*4];
            a3 += + W4*col[8*4];
        }

        if (col[8*5]) {
            MAC16(b0, + W5, col[8*5]);
            MAC16(b1, - W1, col[8*5]);
            MAC16(b2, + W7, col[8*5]);
            MAC16(b3, + W3, col[8*5]);
        }

        if(col[8*6]){
            a0 += + W6*col[8*6];
            a1 += - W2*col[8*6];
            a2 += + W2*col[8*6];
            a3 += - W6*col[8*6];
        }

        if (col[8*7]) {
            MAC16(b0, + W7, col[8*7]);
            MAC16(b1, - W5, col[8*7]);
            MAC16(b2, + W3, col[8*7]);
            MAC16(b3, - W1, col[8*7]);
        }

        col[0 ] = ((a0 + b0) >> COL_SHIFT);
        col[8 ] = ((a1 + b1) >> COL_SHIFT);
        col[16] = ((a2 + b2) >> COL_SHIFT);
        col[24] = ((a3 + b3) >> COL_SHIFT);
        col[32] = ((a3 - b3) >> COL_SHIFT);
        col[40] = ((a2 - b2) >> COL_SHIFT);
        col[48] = ((a1 - b1) >> COL_SHIFT);
        col[56] = ((a0 - b0) >> COL_SHIFT);
}

void simple_idct_c(int16_t * const block)
{
    int i;
    for(i=0; i<8; i++)
        idctRowCondDC(block + i*8);

    for(i=0; i<8; i++)
        idctSparseCol(block + i);
}
Revision:	1.6
Committed:	Tue Mar 9 10:00:14 2010 UTC (14 years, 7 months ago) by Isibaar
Content type:	text/plain
Branch:	MAIN
CVS Tags:	release-1_3_1, release-1_3_0, rc1_1_3_0, tag-branching-1_3_0, HEAD
Branch point for:	release-1_3-branch
Changes since 1.5:	+2 -2 lines
Log Message:	app-level multi-threading for xvid_encraw
#	Content
1	/*****************************************************************************
2	*
3	* XVID MPEG-4 VIDEO CODEC
4	* - Inverse DCT (More precise version) -
5	*
6	* Copyright (c) 2001 Michael Niedermayer <michaelni@gmx.at>
7	*
8	* Originally distributed under the GNU LGPL License (ffmpeg).
9	* It is licensed under the GNU GPL for the Xvid tree.
10	*
11	* This program is free software ; you can redistribute it and/or modify
12	* it under the terms of the GNU General Public License as published by
13	* the Free Software Foundation ; either version 2 of the License, or
14	* (at your option) any later version.
15	*
16	* This program is distributed in the hope that it will be useful,
17	* but WITHOUT ANY WARRANTY ; without even the implied warranty of
18	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19	* GNU General Public License for more details.
20	*
21	* You should have received a copy of the GNU General Public License
22	* along with this program ; if not, write to the Free Software
23	* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
24	*
25	* $Id: simple_idct.c,v 1.5 2004/04/05 20:36:36 edgomez Exp $
26	*
27	****************************************************************************/
28
29	/*
30	based upon some outcommented c code from mpeg2dec (idct_mmx.c
31	written by Aaron Holtzman <aholtzma@ess.engr.uvic.ca>)
32	*/
33
34	#include "../portab.h"
35	#include "idct.h"
36
37	#if 0
38	#define W1 2841 /* 2048sqrt (2)cos (1pi/16) /
39	#define W2 2676 /* 2048sqrt (2)cos (2pi/16) /
40	#define W3 2408 /* 2048sqrt (2)cos (3pi/16) /
41	#define W4 2048 /* 2048sqrt (2)cos (4pi/16) /
42	#define W5 1609 /* 2048sqrt (2)cos (5pi/16) /
43	#define W6 1108 /* 2048sqrt (2)cos (6pi/16) /
44	#define W7 565 /* 2048sqrt (2)cos (7pi/16) /
45	#define ROW_SHIFT 8
46	#define COL_SHIFT 17
47	#else
48	#define W1 22725 /* cos(iM_PI/16)sqrt(2)(1<<14) + 0.5 /
49	#define W2 21407 /* cos(iM_PI/16)sqrt(2)(1<<14) + 0.5 /
50	#define W3 19266 /* cos(iM_PI/16)sqrt(2)(1<<14) + 0.5 /
51	#define W4 16383 /* cos(iM_PI/16)sqrt(2)(1<<14) + 0.5 /
52	#define W5 12873 /* cos(iM_PI/16)sqrt(2)(1<<14) + 0.5 /
53	#define W6 8867 /* cos(iM_PI/16)sqrt(2)(1<<14) + 0.5 /
54	#define W7 4520 /* cos(iM_PI/16)sqrt(2)(1<<14) + 0.5 /
55	#define ROW_SHIFT 11
56	#define COL_SHIFT 20 /* 6 */
57	#endif
58
59	/*
60	PPC mac operation. Causes compile problems on newer ppc targets
61
62	Was originally: #if defined(ARCH_IS_PPC)
63	*/
64	#if 0
65
66	/* signed 16x16 -> 32 multiply add accumulate */
67	#define MAC16(rt, ra, rb) \
68	asm ("maclhw %0, %2, %3" : "=r" (rt) : "0" (rt), "r" (ra), "r" (rb));
69
70	/* signed 16x16 -> 32 multiply */
71	#define MUL16(rt, ra, rb) \
72	asm ("mullhw %0, %1, %2" : "=r" (rt) : "r" (ra), "r" (rb));
73
74	#else
75
76	/* signed 16x16 -> 32 multiply add accumulate */
77	#define MAC16(rt, ra, rb) rt += (ra) * (rb)
78
79	/* signed 16x16 -> 32 multiply */
80	#define MUL16(rt, ra, rb) rt = (ra) * (rb)
81
82	#endif
83
84	static __inline void idctRowCondDC (int16_t * const row)
85	{
86	int a0, a1, a2, a3, b0, b1, b2, b3;
87	#ifdef FAST_64BIT
88	uint64_t temp;
89	#else
90	uint32_t temp;
91	#endif
92
93	#ifdef FAST_64BIT
94	#ifdef ARCH_IS_BIG_ENDIAN
95	#define ROW0_MASK 0xffff000000000000LL
96	#else
97	#define ROW0_MASK 0xffffLL
98	#endif
99	if ( ((((uint64_t *)row)[0] & ~ROW0_MASK) \|
100	((uint64_t *)row)[1]) == 0) {
101	temp = (row[0] << 3) & 0xffff;
102	temp += temp << 16;
103	temp += temp << 32;
104	((uint64_t *)row)[0] = temp;
105	((uint64_t *)row)[1] = temp;
106	return;
107	}
108	#else
109	if (!(((uint32_t*)row)[1] \|
110	((uint32_t*)row)[2] \|
111	((uint32_t*)row)[3] \|
112	row[1])) {
113	temp = (row[0] << 3) & 0xffff;
114	temp += temp << 16;
115	((uint32_t)row)[0]=((uint32_t)row)[1] =
116	((uint32_t)row)[2]=((uint32_t)row)[3] = temp;
117	return;
118	}
119	#endif
120
121	a0 = (W4 * row[0]) + (1 << (ROW_SHIFT - 1));
122	a1 = a0;
123	a2 = a0;
124	a3 = a0;
125
126	/* no need to optimize : gcc does it */
127	a0 += W2 * row[2];
128	a1 += W6 * row[2];
129	a2 -= W6 * row[2];
130	a3 -= W2 * row[2];
131
132	MUL16(b0, W1, row[1]);
133	MAC16(b0, W3, row[3]);
134	MUL16(b1, W3, row[1]);
135	MAC16(b1, -W7, row[3]);
136	MUL16(b2, W5, row[1]);
137	MAC16(b2, -W1, row[3]);
138	MUL16(b3, W7, row[1]);
139	MAC16(b3, -W5, row[3]);
140
141	#ifdef FAST_64BIT
142	temp = ((uint64_t*)row)[1];
143	#else
144	temp = ((uint32_t)row)[2] \| ((uint32_t)row)[3];
145	#endif
146	if (temp != 0) {
147	a0 += W4row[4] + W6row[6];
148	a1 += - W4row[4] - W2row[6];
149	a2 += - W4row[4] + W2row[6];
150	a3 += W4row[4] - W6row[6];
151
152	MAC16(b0, W5, row[5]);
153	MAC16(b0, W7, row[7]);
154
155	MAC16(b1, -W1, row[5]);
156	MAC16(b1, -W5, row[7]);
157
158	MAC16(b2, W7, row[5]);
159	MAC16(b2, W3, row[7]);
160
161	MAC16(b3, W3, row[5]);
162	MAC16(b3, -W1, row[7]);
163	}
164
165	row[0] = (a0 + b0) >> ROW_SHIFT;
166	row[7] = (a0 - b0) >> ROW_SHIFT;
167	row[1] = (a1 + b1) >> ROW_SHIFT;
168	row[6] = (a1 - b1) >> ROW_SHIFT;
169	row[2] = (a2 + b2) >> ROW_SHIFT;
170	row[5] = (a2 - b2) >> ROW_SHIFT;
171	row[3] = (a3 + b3) >> ROW_SHIFT;
172	row[4] = (a3 - b3) >> ROW_SHIFT;
173	}
174
175
176	static __inline void idctSparseCol (int16_t * const col)
177	{
178	int a0, a1, a2, a3, b0, b1, b2, b3;
179
180	/* XXX: I did that only to give same values as previous code */
181	a0 = W4 * (col[8*0] + ((1<<(COL_SHIFT-1))/W4));
182	a1 = a0;
183	a2 = a0;
184	a3 = a0;
185
186	a0 += + W2col[82];
187	a1 += + W6col[82];
188	a2 += - W6col[82];
189	a3 += - W2col[82];
190
191	MUL16(b0, W1, col[8*1]);
192	MUL16(b1, W3, col[8*1]);
193	MUL16(b2, W5, col[8*1]);
194	MUL16(b3, W7, col[8*1]);
195
196	MAC16(b0, + W3, col[8*3]);
197	MAC16(b1, - W7, col[8*3]);
198	MAC16(b2, - W1, col[8*3]);
199	MAC16(b3, - W5, col[8*3]);
200
201	if(col[8*4]){
202	a0 += + W4col[84];
203	a1 += - W4col[84];
204	a2 += - W4col[84];
205	a3 += + W4col[84];
206	}
207
208	if (col[8*5]) {
209	MAC16(b0, + W5, col[8*5]);
210	MAC16(b1, - W1, col[8*5]);
211	MAC16(b2, + W7, col[8*5]);
212	MAC16(b3, + W3, col[8*5]);
213	}
214
215	if(col[8*6]){
216	a0 += + W6col[86];
217	a1 += - W2col[86];
218	a2 += + W2col[86];
219	a3 += - W6col[86];
220	}
221
222	if (col[8*7]) {
223	MAC16(b0, + W7, col[8*7]);
224	MAC16(b1, - W5, col[8*7]);
225	MAC16(b2, + W3, col[8*7]);
226	MAC16(b3, - W1, col[8*7]);
227	}
228
229	col[0 ] = ((a0 + b0) >> COL_SHIFT);
230	col[8 ] = ((a1 + b1) >> COL_SHIFT);
231	col[16] = ((a2 + b2) >> COL_SHIFT);
232	col[24] = ((a3 + b3) >> COL_SHIFT);
233	col[32] = ((a3 - b3) >> COL_SHIFT);
234	col[40] = ((a2 - b2) >> COL_SHIFT);
235	col[48] = ((a1 - b1) >> COL_SHIFT);
236	col[56] = ((a0 - b0) >> COL_SHIFT);
237	}
238
239	void simple_idct_c(int16_t * const block)
240	{
241	int i;
242	for(i=0; i<8; i++)
243	idctRowCondDC(block + i*8);
244
245	for(i=0; i<8; i++)
246	idctSparseCol(block + i);
247	}