vp3dsp.c 9.44 KB
Newer Older
1 2 3
/*
 * Copyright (C) 2004 the ffmpeg project
 *
4 5 6
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
7 8
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
9
 * version 2.1 of the License, or (at your option) any later version.
10
 *
11
 * FFmpeg is distributed in the hope that it will be useful,
12 13 14 15 16
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with FFmpeg; if not, write to the Free Software
18
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 20 21
 */

/**
22
 * @file
23
 * Standard C DSP-oriented functions cribbed from the original VP3
24 25 26
 * source code.
 */

27
#include "libavutil/attributes.h"
28
#include "libavutil/common.h"
29
#include "avcodec.h"
30
#include "dsputil.h"
31
#include "rnd_avg.h"
32
#include "vp3dsp.h"
33 34 35 36 37 38 39 40 41 42

#define IdctAdjustBeforeShift 8
#define xC1S7 64277
#define xC2S6 60547
#define xC3S5 54491
#define xC4S4 46341
#define xC5S3 36410
#define xC6S2 25080
#define xC7S1 12785

43 44
#define M(a,b) (((a) * (b))>>16)

45
static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int type)
46
{
47
    int16_t *ip = input;
48

49 50
    int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
    int Ed, Gd, Add, Bdd, Fd, Hd;
51

52
    int i;
53

54 55 56
    /* Inverse DCT on the rows now */
    for (i = 0; i < 8; i++) {
        /* Check for non-zero values */
57 58 59 60 61 62
        if ( ip[0 * 8] | ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
             ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) {
            A = M(xC1S7, ip[1 * 8]) + M(xC7S1, ip[7 * 8]);
            B = M(xC7S1, ip[1 * 8]) - M(xC1S7, ip[7 * 8]);
            C = M(xC3S5, ip[3 * 8]) + M(xC5S3, ip[5 * 8]);
            D = M(xC3S5, ip[5 * 8]) - M(xC5S3, ip[3 * 8]);
63

64 65
            Ad = M(xC4S4, (A - C));
            Bd = M(xC4S4, (B - D));
66

67 68
            Cd = A + C;
            Dd = B + D;
69

70 71
            E = M(xC4S4, (ip[0 * 8] + ip[4 * 8]));
            F = M(xC4S4, (ip[0 * 8] - ip[4 * 8]));
72

73 74
            G = M(xC2S6, ip[2 * 8]) + M(xC6S2, ip[6 * 8]);
            H = M(xC6S2, ip[2 * 8]) - M(xC2S6, ip[6 * 8]);
75

76 77
            Ed = E - G;
            Gd = E + G;
78

79 80
            Add = F + Ad;
            Bdd = Bd - H;
81

82 83
            Fd = F - Ad;
            Hd = Bd + H;
84 85

            /*  Final sequence of operations over-write original inputs. */
86 87
            ip[0 * 8] = Gd + Cd ;
            ip[7 * 8] = Gd - Cd ;
88

89 90
            ip[1 * 8] = Add + Hd;
            ip[2 * 8] = Add - Hd;
91

92 93
            ip[3 * 8] = Ed + Dd ;
            ip[4 * 8] = Ed - Dd ;
94

95 96
            ip[5 * 8] = Fd + Bdd;
            ip[6 * 8] = Fd - Bdd;
97 98
        }

99
        ip += 1;            /* next row */
100
    }
101

102
    ip = input;
103 104 105

    for ( i = 0; i < 8; i++) {
        /* Check for non-zero values (bitwise or faster than ||) */
106 107
        if ( ip[1] | ip[2] | ip[3] |
             ip[4] | ip[5] | ip[6] | ip[7] ) {
108

109 110 111 112
            A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]);
            B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]);
            C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]);
            D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]);
113

114 115
            Ad = M(xC4S4, (A - C));
            Bd = M(xC4S4, (B - D));
116

117 118
            Cd = A + C;
            Dd = B + D;
119

120 121
            E = M(xC4S4, (ip[0] + ip[4])) + 8;
            F = M(xC4S4, (ip[0] - ip[4])) + 8;
Michael Niedermayer's avatar
Michael Niedermayer committed
122 123 124 125 126

            if(type==1){  //HACK
                E += 16*128;
                F += 16*128;
            }
127

128 129
            G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]);
            H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]);
130

131 132
            Ed = E - G;
            Gd = E + G;
133

134 135
            Add = F + Ad;
            Bdd = Bd - H;
136

137 138
            Fd = F - Ad;
            Hd = Bd + H;
139 140

            /* Final sequence of operations over-write original inputs. */
141
            if (type == 1) {
142 143
                dst[0*stride] = av_clip_uint8((Gd + Cd )  >> 4);
                dst[7*stride] = av_clip_uint8((Gd - Cd )  >> 4);
144

145 146
                dst[1*stride] = av_clip_uint8((Add + Hd ) >> 4);
                dst[2*stride] = av_clip_uint8((Add - Hd ) >> 4);
147

148 149
                dst[3*stride] = av_clip_uint8((Ed + Dd )  >> 4);
                dst[4*stride] = av_clip_uint8((Ed - Dd )  >> 4);
150

151 152
                dst[5*stride] = av_clip_uint8((Fd + Bdd ) >> 4);
                dst[6*stride] = av_clip_uint8((Fd - Bdd ) >> 4);
153
            }else{
154 155
                dst[0*stride] = av_clip_uint8(dst[0*stride] + ((Gd + Cd )  >> 4));
                dst[7*stride] = av_clip_uint8(dst[7*stride] + ((Gd - Cd )  >> 4));
156

157 158
                dst[1*stride] = av_clip_uint8(dst[1*stride] + ((Add + Hd ) >> 4));
                dst[2*stride] = av_clip_uint8(dst[2*stride] + ((Add - Hd ) >> 4));
159

160 161
                dst[3*stride] = av_clip_uint8(dst[3*stride] + ((Ed + Dd )  >> 4));
                dst[4*stride] = av_clip_uint8(dst[4*stride] + ((Ed - Dd )  >> 4));
162

163 164
                dst[5*stride] = av_clip_uint8(dst[5*stride] + ((Fd + Bdd ) >> 4));
                dst[6*stride] = av_clip_uint8(dst[6*stride] + ((Fd - Bdd ) >> 4));
165
            }
166 167

        } else {
168
            if (type == 1) {
169 170 171 172 173 174 175
                dst[0*stride]=
                dst[1*stride]=
                dst[2*stride]=
                dst[3*stride]=
                dst[4*stride]=
                dst[5*stride]=
                dst[6*stride]=
176
                dst[7*stride]= av_clip_uint8(128 + ((xC4S4 * ip[0] + (IdctAdjustBeforeShift<<16))>>20));
177
            }else{
178 179
                if(ip[0]){
                    int v= ((xC4S4 * ip[0] + (IdctAdjustBeforeShift<<16))>>20);
180 181 182 183 184 185 186 187
                    dst[0*stride] = av_clip_uint8(dst[0*stride] + v);
                    dst[1*stride] = av_clip_uint8(dst[1*stride] + v);
                    dst[2*stride] = av_clip_uint8(dst[2*stride] + v);
                    dst[3*stride] = av_clip_uint8(dst[3*stride] + v);
                    dst[4*stride] = av_clip_uint8(dst[4*stride] + v);
                    dst[5*stride] = av_clip_uint8(dst[5*stride] + v);
                    dst[6*stride] = av_clip_uint8(dst[6*stride] + v);
                    dst[7*stride] = av_clip_uint8(dst[7*stride] + v);
188 189
                }
            }
190 191
        }

192
        ip += 8;            /* next column */
193
        dst++;
194 195
    }
}
196

Diego Biurrun's avatar
Diego Biurrun committed
197 198 199
static void vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size,
                           int16_t *block/*align 16*/)
{
200
    idct(dest, line_size, block, 1);
201
    memset(block, 0, sizeof(*block) * 64);
202 203
}

Diego Biurrun's avatar
Diego Biurrun committed
204 205 206
static void vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size,
                           int16_t *block/*align 16*/)
{
207
    idct(dest, line_size, block, 2);
208
    memset(block, 0, sizeof(*block) * 64);
209
}
210

211
static void vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size,
Diego Biurrun's avatar
Diego Biurrun committed
212 213
                              int16_t *block/*align 16*/)
{
214
    int i, dc = (block[0] + 15) >> 5;
David Conrad's avatar
David Conrad committed
215 216

    for(i = 0; i < 8; i++){
217 218 219 220 221 222 223 224
        dest[0] = av_clip_uint8(dest[0] + dc);
        dest[1] = av_clip_uint8(dest[1] + dc);
        dest[2] = av_clip_uint8(dest[2] + dc);
        dest[3] = av_clip_uint8(dest[3] + dc);
        dest[4] = av_clip_uint8(dest[4] + dc);
        dest[5] = av_clip_uint8(dest[5] + dc);
        dest[6] = av_clip_uint8(dest[6] + dc);
        dest[7] = av_clip_uint8(dest[7] + dc);
David Conrad's avatar
David Conrad committed
225 226
        dest += line_size;
    }
227
    block[0] = 0;
David Conrad's avatar
David Conrad committed
228 229
}

230 231
static void vp3_v_loop_filter_c(uint8_t *first_pixel, int stride,
                                int *bounding_values)
232 233 234 235 236 237 238 239 240 241 242 243 244 245 246
{
    unsigned char *end;
    int filter_value;
    const int nstride= -stride;

    for (end= first_pixel + 8; first_pixel < end; first_pixel++) {
        filter_value =
            (first_pixel[2 * nstride] - first_pixel[ stride])
         +3*(first_pixel[0          ] - first_pixel[nstride]);
        filter_value = bounding_values[(filter_value + 4) >> 3];
        first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
        first_pixel[0] = av_clip_uint8(first_pixel[0] - filter_value);
    }
}

247 248
static void vp3_h_loop_filter_c(uint8_t *first_pixel, int stride,
                                int *bounding_values)
249 250 251 252 253 254 255 256 257 258 259 260 261
{
    unsigned char *end;
    int filter_value;

    for (end= first_pixel + 8*stride; first_pixel != end; first_pixel += stride) {
        filter_value =
            (first_pixel[-2] - first_pixel[ 1])
         +3*(first_pixel[ 0] - first_pixel[-1]);
        filter_value = bounding_values[(filter_value + 4) >> 3];
        first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
        first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
    }
}
262

263 264 265 266 267 268 269 270
static void put_no_rnd_pixels_l2(uint8_t *dst, const uint8_t *src1,
                                 const uint8_t *src2, ptrdiff_t stride, int h)
{
    int i;

    for (i = 0; i < h; i++) {
        uint32_t a, b;

271 272
        a = AV_RN32(&src1[i * stride]);
        b = AV_RN32(&src2[i * stride]);
273
        AV_WN32A(&dst[i * stride], no_rnd_avg32(a, b));
274 275
        a = AV_RN32(&src1[i * stride + 4]);
        b = AV_RN32(&src2[i * stride + 4]);
276 277 278 279
        AV_WN32A(&dst[i * stride + 4], no_rnd_avg32(a, b));
    }
}

280 281
av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags)
{
282 283
    c->put_no_rnd_pixels_l2 = put_no_rnd_pixels_l2;

284 285 286 287 288 289 290 291
    c->idct_put      = vp3_idct_put_c;
    c->idct_add      = vp3_idct_add_c;
    c->idct_dc_add   = vp3_idct_dc_add_c;
    c->v_loop_filter = vp3_v_loop_filter_c;
    c->h_loop_filter = vp3_h_loop_filter_c;

    if (ARCH_ARM)
        ff_vp3dsp_init_arm(c, flags);
292 293
    if (ARCH_BFIN)
        ff_vp3dsp_init_bfin(c, flags);
294 295 296 297 298
    if (ARCH_PPC)
        ff_vp3dsp_init_ppc(c, flags);
    if (ARCH_X86)
        ff_vp3dsp_init_x86(c, flags);
}