GCC Code Coverage Report
Directory: ../../../ffmpeg/ Exec Total Coverage
File: src/libavcodec/rv40dsp.c Lines: 197 200 98.5 %
Date: 2020-08-14 10:39:37 Branches: 92 92 100.0 %

Line Branch Exec Source
1
/*
2
 * RV40 decoder motion compensation functions
3
 * Copyright (c) 2008 Konstantin Shishkov
4
 *
5
 * This file is part of FFmpeg.
6
 *
7
 * FFmpeg is free software; you can redistribute it and/or
8
 * modify it under the terms of the GNU Lesser General Public
9
 * License as published by the Free Software Foundation; either
10
 * version 2.1 of the License, or (at your option) any later version.
11
 *
12
 * FFmpeg is distributed in the hope that it will be useful,
13
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15
 * Lesser General Public License for more details.
16
 *
17
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with FFmpeg; if not, write to the Free Software
19
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20
 */
21
22
/**
23
 * @file
24
 * RV40 decoder motion compensation functions
25
 */
26
27
#include "libavutil/common.h"
28
#include "libavutil/intreadwrite.h"
29
#include "avcodec.h"
30
#include "h264qpel.h"
31
#include "mathops.h"
32
#include "pixels.h"
33
#include "rnd_avg.h"
34
#include "rv34dsp.h"
35
#include "libavutil/avassert.h"
36
37
#define RV40_LOWPASS(OPNAME, OP) \
38
static void OPNAME ## rv40_qpel8_h_lowpass(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride,\
39
                                                     const int h, const int C1, const int C2, const int SHIFT){\
40
    const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;\
41
    int i;\
42
    for(i = 0; i < h; i++)\
43
    {\
44
        OP(dst[0], (src[-2] + src[ 3] - 5*(src[-1]+src[2]) + src[0]*C1 + src[1]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
45
        OP(dst[1], (src[-1] + src[ 4] - 5*(src[ 0]+src[3]) + src[1]*C1 + src[2]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
46
        OP(dst[2], (src[ 0] + src[ 5] - 5*(src[ 1]+src[4]) + src[2]*C1 + src[3]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
47
        OP(dst[3], (src[ 1] + src[ 6] - 5*(src[ 2]+src[5]) + src[3]*C1 + src[4]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
48
        OP(dst[4], (src[ 2] + src[ 7] - 5*(src[ 3]+src[6]) + src[4]*C1 + src[5]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
49
        OP(dst[5], (src[ 3] + src[ 8] - 5*(src[ 4]+src[7]) + src[5]*C1 + src[6]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
50
        OP(dst[6], (src[ 4] + src[ 9] - 5*(src[ 5]+src[8]) + src[6]*C1 + src[7]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
51
        OP(dst[7], (src[ 5] + src[10] - 5*(src[ 6]+src[9]) + src[7]*C1 + src[8]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
52
        dst += dstStride;\
53
        src += srcStride;\
54
    }\
55
}\
56
\
57
static void OPNAME ## rv40_qpel8_v_lowpass(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride,\
58
                                           const int w, const int C1, const int C2, const int SHIFT){\
59
    const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;\
60
    int i;\
61
    for(i = 0; i < w; i++)\
62
    {\
63
        const int srcB  = src[-2*srcStride];\
64
        const int srcA  = src[-1*srcStride];\
65
        const int src0  = src[0 *srcStride];\
66
        const int src1  = src[1 *srcStride];\
67
        const int src2  = src[2 *srcStride];\
68
        const int src3  = src[3 *srcStride];\
69
        const int src4  = src[4 *srcStride];\
70
        const int src5  = src[5 *srcStride];\
71
        const int src6  = src[6 *srcStride];\
72
        const int src7  = src[7 *srcStride];\
73
        const int src8  = src[8 *srcStride];\
74
        const int src9  = src[9 *srcStride];\
75
        const int src10 = src[10*srcStride];\
76
        OP(dst[0*dstStride], (srcB + src3  - 5*(srcA+src2) + src0*C1 + src1*C2 + (1<<(SHIFT-1))) >> SHIFT);\
77
        OP(dst[1*dstStride], (srcA + src4  - 5*(src0+src3) + src1*C1 + src2*C2 + (1<<(SHIFT-1))) >> SHIFT);\
78
        OP(dst[2*dstStride], (src0 + src5  - 5*(src1+src4) + src2*C1 + src3*C2 + (1<<(SHIFT-1))) >> SHIFT);\
79
        OP(dst[3*dstStride], (src1 + src6  - 5*(src2+src5) + src3*C1 + src4*C2 + (1<<(SHIFT-1))) >> SHIFT);\
80
        OP(dst[4*dstStride], (src2 + src7  - 5*(src3+src6) + src4*C1 + src5*C2 + (1<<(SHIFT-1))) >> SHIFT);\
81
        OP(dst[5*dstStride], (src3 + src8  - 5*(src4+src7) + src5*C1 + src6*C2 + (1<<(SHIFT-1))) >> SHIFT);\
82
        OP(dst[6*dstStride], (src4 + src9  - 5*(src5+src8) + src6*C1 + src7*C2 + (1<<(SHIFT-1))) >> SHIFT);\
83
        OP(dst[7*dstStride], (src5 + src10 - 5*(src6+src9) + src7*C1 + src8*C2 + (1<<(SHIFT-1))) >> SHIFT);\
84
        dst++;\
85
        src++;\
86
    }\
87
}\
88
\
89
static void OPNAME ## rv40_qpel16_v_lowpass(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride,\
90
                                            const int w, const int C1, const int C2, const int SHIFT){\
91
    OPNAME ## rv40_qpel8_v_lowpass(dst  , src  , dstStride, srcStride, 8, C1, C2, SHIFT);\
92
    OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
93
    src += 8*srcStride;\
94
    dst += 8*dstStride;\
95
    OPNAME ## rv40_qpel8_v_lowpass(dst  , src  , dstStride, srcStride, w-8, C1, C2, SHIFT);\
96
    OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, w-8, C1, C2, SHIFT);\
97
}\
98
\
99
static void OPNAME ## rv40_qpel16_h_lowpass(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride,\
100
                                            const int h, const int C1, const int C2, const int SHIFT){\
101
    OPNAME ## rv40_qpel8_h_lowpass(dst  , src  , dstStride, srcStride, 8, C1, C2, SHIFT);\
102
    OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
103
    src += 8*srcStride;\
104
    dst += 8*dstStride;\
105
    OPNAME ## rv40_qpel8_h_lowpass(dst  , src  , dstStride, srcStride, h-8, C1, C2, SHIFT);\
106
    OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, h-8, C1, C2, SHIFT);\
107
}\
108
\
109
110
#define RV40_MC(OPNAME, SIZE) \
111
static void OPNAME ## rv40_qpel ## SIZE ## _mc10_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
112
{\
113
    OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
114
}\
115
\
116
static void OPNAME ## rv40_qpel ## SIZE ## _mc30_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
117
{\
118
    OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
119
}\
120
\
121
static void OPNAME ## rv40_qpel ## SIZE ## _mc01_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
122
{\
123
    OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
124
}\
125
\
126
static void OPNAME ## rv40_qpel ## SIZE ## _mc11_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
127
{\
128
    uint8_t full[SIZE*(SIZE+5)];\
129
    uint8_t * const full_mid = full + SIZE*2;\
130
    put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
131
    OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
132
}\
133
\
134
static void OPNAME ## rv40_qpel ## SIZE ## _mc21_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
135
{\
136
    uint8_t full[SIZE*(SIZE+5)];\
137
    uint8_t * const full_mid = full + SIZE*2;\
138
    put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
139
    OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
140
}\
141
\
142
static void OPNAME ## rv40_qpel ## SIZE ## _mc31_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
143
{\
144
    uint8_t full[SIZE*(SIZE+5)];\
145
    uint8_t * const full_mid = full + SIZE*2;\
146
    put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\
147
    OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
148
}\
149
\
150
static void OPNAME ## rv40_qpel ## SIZE ## _mc12_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
151
{\
152
    uint8_t full[SIZE*(SIZE+5)];\
153
    uint8_t * const full_mid = full + SIZE*2;\
154
    put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
155
    OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
156
}\
157
\
158
static void OPNAME ## rv40_qpel ## SIZE ## _mc22_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
159
{\
160
    uint8_t full[SIZE*(SIZE+5)];\
161
    uint8_t * const full_mid = full + SIZE*2;\
162
    put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
163
    OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
164
}\
165
\
166
static void OPNAME ## rv40_qpel ## SIZE ## _mc32_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
167
{\
168
    uint8_t full[SIZE*(SIZE+5)];\
169
    uint8_t * const full_mid = full + SIZE*2;\
170
    put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\
171
    OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
172
}\
173
\
174
static void OPNAME ## rv40_qpel ## SIZE ## _mc03_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
175
{\
176
    OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
177
}\
178
\
179
static void OPNAME ## rv40_qpel ## SIZE ## _mc13_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
180
{\
181
    uint8_t full[SIZE*(SIZE+5)];\
182
    uint8_t * const full_mid = full + SIZE*2;\
183
    put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
184
    OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
185
}\
186
\
187
static void OPNAME ## rv40_qpel ## SIZE ## _mc23_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
188
{\
189
    uint8_t full[SIZE*(SIZE+5)];\
190
    uint8_t * const full_mid = full + SIZE*2;\
191
    put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
192
    OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
193
}\
194
\
195
196
#define op_avg(a, b)  a = (((a)+cm[b]+1)>>1)
197
#define op_put(a, b)  a = cm[b]
198
199
5250836
RV40_LOWPASS(put_       , op_put)
200
141007
RV40_LOWPASS(avg_       , op_avg)
201
202
#undef op_avg
203
#undef op_put
204
205
117202
RV40_MC(put_, 8)
206
23297
RV40_MC(put_, 16)
207
RV40_MC(avg_, 8)
208
3811
RV40_MC(avg_, 16)
209
210
#define PIXOP2(OPNAME, OP)                                              \
211
static inline void OPNAME ## _pixels8_xy2_8_c(uint8_t *block,           \
212
                                              const uint8_t *pixels,    \
213
                                              ptrdiff_t line_size,      \
214
                                              int h)                    \
215
{                                                                       \
216
    /* FIXME HIGH BIT DEPTH */                                          \
217
    int j;                                                              \
218
                                                                        \
219
    for (j = 0; j < 2; j++) {                                           \
220
        int i;                                                          \
221
        const uint32_t a = AV_RN32(pixels);                             \
222
        const uint32_t b = AV_RN32(pixels + 1);                         \
223
        uint32_t l0 = (a & 0x03030303UL) +                              \
224
                      (b & 0x03030303UL) +                              \
225
                           0x02020202UL;                                \
226
        uint32_t h0 = ((a & 0xFCFCFCFCUL) >> 2) +                       \
227
                      ((b & 0xFCFCFCFCUL) >> 2);                        \
228
        uint32_t l1, h1;                                                \
229
                                                                        \
230
        pixels += line_size;                                            \
231
        for (i = 0; i < h; i += 2) {                                    \
232
            uint32_t a = AV_RN32(pixels);                               \
233
            uint32_t b = AV_RN32(pixels + 1);                           \
234
            l1 = (a & 0x03030303UL) +                                   \
235
                 (b & 0x03030303UL);                                    \
236
            h1 = ((a & 0xFCFCFCFCUL) >> 2) +                            \
237
                 ((b & 0xFCFCFCFCUL) >> 2);                             \
238
            OP(*((uint32_t *) block),                                   \
239
               h0 + h1 + (((l0 + l1) >> 2) & 0x0F0F0F0FUL));            \
240
            pixels += line_size;                                        \
241
            block  += line_size;                                        \
242
            a = AV_RN32(pixels);                                        \
243
            b = AV_RN32(pixels + 1);                                    \
244
            l0 = (a & 0x03030303UL) +                                   \
245
                 (b & 0x03030303UL) +                                   \
246
                      0x02020202UL;                                     \
247
            h0 = ((a & 0xFCFCFCFCUL) >> 2) +                            \
248
                 ((b & 0xFCFCFCFCUL) >> 2);                             \
249
            OP(*((uint32_t *) block),                                   \
250
               h0 + h1 + (((l0 + l1) >> 2) & 0x0F0F0F0FUL));            \
251
            pixels += line_size;                                        \
252
            block  += line_size;                                        \
253
        }                                                               \
254
        pixels += 4 - line_size * (h + 1);                              \
255
        block  += 4 - line_size * h;                                    \
256
    }                                                                   \
257
}                                                                       \
258
                                                                        \
259
CALL_2X_PIXELS(OPNAME ## _pixels16_xy2_8_c,                             \
260
               OPNAME ## _pixels8_xy2_8_c,                              \
261
               8)                                                       \
262
263
#define op_avg(a, b) a = rnd_avg32(a, b)
264
#define op_put(a, b) a = b
265

376584
PIXOP2(avg, op_avg)
266

1002794
PIXOP2(put, op_put)
267
#undef op_avg
268
#undef op_put
269
270
20935
static void put_rv40_qpel16_mc33_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
271
{
272
20935
    put_pixels16_xy2_8_c(dst, src, stride, 16);
273
20935
}
274
4828
static void avg_rv40_qpel16_mc33_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
275
{
276
4828
    avg_pixels16_xy2_8_c(dst, src, stride, 16);
277
4828
}
278
16939
static void put_rv40_qpel8_mc33_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
279
{
280
16939
    put_pixels8_xy2_8_c(dst, src, stride, 8);
281
16939
}
282
static void avg_rv40_qpel8_mc33_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
283
{
284
    avg_pixels8_xy2_8_c(dst, src, stride, 8);
285
}
286
287
static const int rv40_bias[4][4] = {
288
    {  0, 16, 32, 16 },
289
    { 32, 28, 32, 28 },
290
    {  0, 32, 16, 32 },
291
    { 32, 28, 32, 28 }
292
};
293
294
#define RV40_CHROMA_MC(OPNAME, OP)\
295
static void OPNAME ## rv40_chroma_mc4_c(uint8_t *dst /*align 8*/,\
296
                                        uint8_t *src /*align 1*/,\
297
                                        ptrdiff_t stride, int h, int x, int y)\
298
{\
299
    const int A = (8-x) * (8-y);\
300
    const int B = (  x) * (8-y);\
301
    const int C = (8-x) * (  y);\
302
    const int D = (  x) * (  y);\
303
    int i;\
304
    int bias = rv40_bias[y>>1][x>>1];\
305
    \
306
    av_assert2(x<8 && y<8 && x>=0 && y>=0);\
307
\
308
    if(D){\
309
        for(i = 0; i < h; i++){\
310
            OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
311
            OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
312
            OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
313
            OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
314
            dst += stride;\
315
            src += stride;\
316
        }\
317
    }else{\
318
        const int E = B + C;\
319
        const ptrdiff_t step = C ? stride : 1;\
320
        for(i = 0; i < h; i++){\
321
            OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
322
            OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
323
            OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
324
            OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
325
            dst += stride;\
326
            src += stride;\
327
        }\
328
    }\
329
}\
330
\
331
static void OPNAME ## rv40_chroma_mc8_c(uint8_t *dst/*align 8*/,\
332
                                        uint8_t *src/*align 1*/,\
333
                                        ptrdiff_t stride, int h, int x, int y)\
334
{\
335
    const int A = (8-x) * (8-y);\
336
    const int B = (  x) * (8-y);\
337
    const int C = (8-x) * (  y);\
338
    const int D = (  x) * (  y);\
339
    int i;\
340
    int bias = rv40_bias[y>>1][x>>1];\
341
    \
342
    av_assert2(x<8 && y<8 && x>=0 && y>=0);\
343
\
344
    if(D){\
345
        for(i = 0; i < h; i++){\
346
            OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
347
            OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
348
            OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
349
            OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
350
            OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + bias));\
351
            OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + bias));\
352
            OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + bias));\
353
            OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + bias));\
354
            dst += stride;\
355
            src += stride;\
356
        }\
357
    }else{\
358
        const int E = B + C;\
359
        const ptrdiff_t step = C ? stride : 1;\
360
        for(i = 0; i < h; i++){\
361
            OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
362
            OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
363
            OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
364
            OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
365
            OP(dst[4], (A*src[4] + E*src[step+4] + bias));\
366
            OP(dst[5], (A*src[5] + E*src[step+5] + bias));\
367
            OP(dst[6], (A*src[6] + E*src[step+6] + bias));\
368
            OP(dst[7], (A*src[7] + E*src[step+7] + bias));\
369
            dst += stride;\
370
            src += stride;\
371
        }\
372
    }\
373
}
374
375
#define op_avg(a, b) a = (((a)+((b)>>6)+1)>>1)
376
#define op_put(a, b) a = ((b)>>6)
377
378


12207120
RV40_CHROMA_MC(put_, op_put)
379


166734
RV40_CHROMA_MC(avg_, op_avg)
380
381
#define RV40_WEIGHT_FUNC(size) \
382
static void rv40_weight_func_rnd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\
383
{\
384
    int i, j;\
385
\
386
    for (j = 0; j < size; j++) {\
387
        for (i = 0; i < size; i++)\
388
            dst[i] = ((((unsigned)w2 * src1[i]) >> 9) + (((unsigned)w1 * src2[i]) >> 9) + 0x10) >> 5;\
389
        src1 += stride;\
390
        src2 += stride;\
391
        dst  += stride;\
392
    }\
393
}\
394
static void rv40_weight_func_nornd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\
395
{\
396
    int i, j;\
397
\
398
    for (j = 0; j < size; j++) {\
399
        for (i = 0; i < size; i++)\
400
            dst[i] = ((unsigned)w2 * src1[i] + (unsigned)w1 * src2[i] + 0x10) >> 5;\
401
        src1 += stride;\
402
        src2 += stride;\
403
        dst  += stride;\
404
    }\
405
}
406
407

63316344
RV40_WEIGHT_FUNC(16)
408

16930744
RV40_WEIGHT_FUNC(8)
409
410
/**
411
 * dither values for deblocking filter - left/top values
412
 */
413
static const uint8_t rv40_dither_l[16] = {
414
    0x40, 0x50, 0x20, 0x60, 0x30, 0x50, 0x40, 0x30,
415
    0x50, 0x40, 0x50, 0x30, 0x60, 0x20, 0x50, 0x40
416
};
417
418
/**
419
 * dither values for deblocking filter - right/bottom values
420
 */
421
static const uint8_t rv40_dither_r[16] = {
422
    0x40, 0x30, 0x60, 0x20, 0x50, 0x30, 0x30, 0x40,
423
    0x40, 0x40, 0x50, 0x30, 0x20, 0x60, 0x30, 0x40
424
};
425
426
#define CLIP_SYMM(a, b) av_clip(a, -(b), b)
427
/**
428
 * weaker deblocking very similar to the one described in 4.4.2 of JVT-A003r1
429
 */
430
1488979
static av_always_inline void rv40_weak_loop_filter(uint8_t *src,
431
                                                   const int step,
432
                                                   const ptrdiff_t stride,
433
                                                   const int filter_p1,
434
                                                   const int filter_q1,
435
                                                   const int alpha,
436
                                                   const int beta,
437
                                                   const int lim_p0q0,
438
                                                   const int lim_q1,
439
                                                   const int lim_p1)
440
{
441
1488979
    const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
442
    int i, t, u, diff;
443
444
7444895
    for (i = 0; i < 4; i++, src += stride) {
445
5955916
        int diff_p1p0 = src[-2*step] - src[-1*step];
446
5955916
        int diff_q1q0 = src[ 1*step] - src[ 0*step];
447
5955916
        int diff_p1p2 = src[-2*step] - src[-3*step];
448
5955916
        int diff_q1q2 = src[ 1*step] - src[ 2*step];
449
450
5955916
        t = src[0*step] - src[-1*step];
451
5955916
        if (!t)
452
1305988
            continue;
453
454
4649928
        u = (alpha * FFABS(t)) >> 7;
455

4649928
        if (u > 3 - (filter_p1 && filter_q1))
456
515619
            continue;
457
458
4134309
        t *= 1 << 2;
459

4134309
        if (filter_p1 && filter_q1)
460
3431142
            t += src[-2*step] - src[1*step];
461
462
4134309
        diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0);
463
4134309
        src[-1*step] = cm[src[-1*step] + diff];
464
4134309
        src[ 0*step] = cm[src[ 0*step] - diff];
465
466

4134309
        if (filter_p1 && FFABS(diff_p1p2) <= beta) {
467
3566828
            t = (diff_p1p0 + diff_p1p2 - diff) >> 1;
468
3566828
            src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)];
469
        }
470
471

4134309
        if (filter_q1 && FFABS(diff_q1q2) <= beta) {
472
3569006
            t = (diff_q1q0 + diff_q1q2 + diff) >> 1;
473
3569006
            src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)];
474
        }
475
    }
476
1488979
}
477
478
747005
static void rv40_h_weak_loop_filter(uint8_t *src, const ptrdiff_t stride,
479
                                    const int filter_p1, const int filter_q1,
480
                                    const int alpha, const int beta,
481
                                    const int lim_p0q0, const int lim_q1,
482
                                    const int lim_p1)
483
{
484
747005
    rv40_weak_loop_filter(src, stride, 1, filter_p1, filter_q1,
485
                          alpha, beta, lim_p0q0, lim_q1, lim_p1);
486
747005
}
487
488
741974
static void rv40_v_weak_loop_filter(uint8_t *src, const ptrdiff_t stride,
489
                                    const int filter_p1, const int filter_q1,
490
                                    const int alpha, const int beta,
491
                                    const int lim_p0q0, const int lim_q1,
492
                                    const int lim_p1)
493
{
494
741974
    rv40_weak_loop_filter(src, 1, stride, filter_p1, filter_q1,
495
                          alpha, beta, lim_p0q0, lim_q1, lim_p1);
496
741974
}
497
498
328913
static av_always_inline void rv40_strong_loop_filter(uint8_t *src,
499
                                                     const int step,
500
                                                     const ptrdiff_t stride,
501
                                                     const int alpha,
502
                                                     const int lims,
503
                                                     const int dmode,
504
                                                     const int chroma)
505
{
506
    int i;
507
508
1644565
    for(i = 0; i < 4; i++, src += stride){
509
        int sflag, p0, q0, p1, q1;
510
1315652
        int t = src[0*step] - src[-1*step];
511
512
1315652
        if (!t)
513
443419
            continue;
514
515
872233
        sflag = (alpha * FFABS(t)) >> 7;
516
872233
        if (sflag > 1)
517
128838
            continue;
518
519
743395
        p0 = (25*src[-3*step] + 26*src[-2*step] + 26*src[-1*step] +
520
743395
              26*src[ 0*step] + 25*src[ 1*step] +
521
743395
              rv40_dither_l[dmode + i]) >> 7;
522
523
743395
        q0 = (25*src[-2*step] + 26*src[-1*step] + 26*src[ 0*step] +
524
743395
              26*src[ 1*step] + 25*src[ 2*step] +
525
743395
              rv40_dither_r[dmode + i]) >> 7;
526
527
743395
        if (sflag) {
528
193774
            p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims);
529
193774
            q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims);
530
        }
531
532
743395
        p1 = (25*src[-4*step] + 26*src[-3*step] + 26*src[-2*step] + 26*p0 +
533
743395
              25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;
534
743395
        q1 = (25*src[-1*step] + 26*q0 + 26*src[ 1*step] + 26*src[ 2*step] +
535
743395
              25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;
536
537
743395
        if (sflag) {
538
193774
            p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims);
539
193774
            q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims);
540
        }
541
542
743395
        src[-2*step] = p1;
543
743395
        src[-1*step] = p0;
544
743395
        src[ 0*step] = q0;
545
743395
        src[ 1*step] = q1;
546
547
743395
        if(!chroma){
548
332200
            src[-3*step] = (25*src[-1*step] + 26*src[-2*step] +
549
332200
                            51*src[-3*step] + 26*src[-4*step] + 64) >> 7;
550
332200
            src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] +
551
332200
                            51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;
552
        }
553
    }
554
328913
}
555
556
173906
static void rv40_h_strong_loop_filter(uint8_t *src, const ptrdiff_t stride,
557
                                      const int alpha, const int lims,
558
                                      const int dmode, const int chroma)
559
{
560
173906
    rv40_strong_loop_filter(src, stride, 1, alpha, lims, dmode, chroma);
561
173906
}
562
563
155007
static void rv40_v_strong_loop_filter(uint8_t *src, const ptrdiff_t stride,
564
                                      const int alpha, const int lims,
565
                                      const int dmode, const int chroma)
566
{
567
155007
    rv40_strong_loop_filter(src, 1, stride, alpha, lims, dmode, chroma);
568
155007
}
569
570
2315053
static av_always_inline int rv40_loop_filter_strength(uint8_t *src,
571
                                                      int step, ptrdiff_t stride,
572
                                                      int beta, int beta2,
573
                                                      int edge,
574
                                                      int *p1, int *q1)
575
{
576
2315053
    int sum_p1p0 = 0, sum_q1q0 = 0, sum_p1p2 = 0, sum_q1q2 = 0;
577
2315053
    int strong0 = 0, strong1 = 0;
578
    uint8_t *ptr;
579
    int i;
580
581
11575265
    for (i = 0, ptr = src; i < 4; i++, ptr += stride) {
582
9260212
        sum_p1p0 += ptr[-2*step] - ptr[-1*step];
583
9260212
        sum_q1q0 += ptr[ 1*step] - ptr[ 0*step];
584
    }
585
586
2315053
    *p1 = FFABS(sum_p1p0) < (beta << 2);
587
2315053
    *q1 = FFABS(sum_q1q0) < (beta << 2);
588
589

2315053
    if(!*p1 && !*q1)
590
497161
        return 0;
591
592
1817892
    if (!edge)
593
1423014
        return 0;
594
595
1974390
    for (i = 0, ptr = src; i < 4; i++, ptr += stride) {
596
1579512
        sum_p1p2 += ptr[-2*step] - ptr[-3*step];
597
1579512
        sum_q1q2 += ptr[ 1*step] - ptr[ 2*step];
598
    }
599
600

394878
    strong0 = *p1 && (FFABS(sum_p1p2) < beta2);
601

394878
    strong1 = *q1 && (FFABS(sum_q1q2) < beta2);
602
603

394878
    return strong0 && strong1;
604
}
605
606
1170505
static int rv40_h_loop_filter_strength(uint8_t *src, ptrdiff_t stride,
607
                                       int beta, int beta2, int edge,
608
                                       int *p1, int *q1)
609
{
610
1170505
    return rv40_loop_filter_strength(src, stride, 1, beta, beta2, edge, p1, q1);
611
}
612
613
1144548
static int rv40_v_loop_filter_strength(uint8_t *src, ptrdiff_t stride,
614
                                       int beta, int beta2, int edge,
615
                                       int *p1, int *q1)
616
{
617
1144548
    return rv40_loop_filter_strength(src, 1, stride, beta, beta2, edge, p1, q1);
618
}
619
620
4
av_cold void ff_rv40dsp_init(RV34DSPContext *c)
621
{
622
    H264QpelContext qpel;
623
624
4
    ff_rv34dsp_init(c);
625
4
    ff_h264qpel_init(&qpel, 8);
626
627
4
    c->put_pixels_tab[0][ 0] = qpel.put_h264_qpel_pixels_tab[0][0];
628
4
    c->put_pixels_tab[0][ 1] = put_rv40_qpel16_mc10_c;
629
4
    c->put_pixels_tab[0][ 2] = qpel.put_h264_qpel_pixels_tab[0][2];
630
4
    c->put_pixels_tab[0][ 3] = put_rv40_qpel16_mc30_c;
631
4
    c->put_pixels_tab[0][ 4] = put_rv40_qpel16_mc01_c;
632
4
    c->put_pixels_tab[0][ 5] = put_rv40_qpel16_mc11_c;
633
4
    c->put_pixels_tab[0][ 6] = put_rv40_qpel16_mc21_c;
634
4
    c->put_pixels_tab[0][ 7] = put_rv40_qpel16_mc31_c;
635
4
    c->put_pixels_tab[0][ 8] = qpel.put_h264_qpel_pixels_tab[0][8];
636
4
    c->put_pixels_tab[0][ 9] = put_rv40_qpel16_mc12_c;
637
4
    c->put_pixels_tab[0][10] = put_rv40_qpel16_mc22_c;
638
4
    c->put_pixels_tab[0][11] = put_rv40_qpel16_mc32_c;
639
4
    c->put_pixels_tab[0][12] = put_rv40_qpel16_mc03_c;
640
4
    c->put_pixels_tab[0][13] = put_rv40_qpel16_mc13_c;
641
4
    c->put_pixels_tab[0][14] = put_rv40_qpel16_mc23_c;
642
4
    c->put_pixels_tab[0][15] = put_rv40_qpel16_mc33_c;
643
4
    c->avg_pixels_tab[0][ 0] = qpel.avg_h264_qpel_pixels_tab[0][0];
644
4
    c->avg_pixels_tab[0][ 1] = avg_rv40_qpel16_mc10_c;
645
4
    c->avg_pixels_tab[0][ 2] = qpel.avg_h264_qpel_pixels_tab[0][2];
646
4
    c->avg_pixels_tab[0][ 3] = avg_rv40_qpel16_mc30_c;
647
4
    c->avg_pixels_tab[0][ 4] = avg_rv40_qpel16_mc01_c;
648
4
    c->avg_pixels_tab[0][ 5] = avg_rv40_qpel16_mc11_c;
649
4
    c->avg_pixels_tab[0][ 6] = avg_rv40_qpel16_mc21_c;
650
4
    c->avg_pixels_tab[0][ 7] = avg_rv40_qpel16_mc31_c;
651
4
    c->avg_pixels_tab[0][ 8] = qpel.avg_h264_qpel_pixels_tab[0][8];
652
4
    c->avg_pixels_tab[0][ 9] = avg_rv40_qpel16_mc12_c;
653
4
    c->avg_pixels_tab[0][10] = avg_rv40_qpel16_mc22_c;
654
4
    c->avg_pixels_tab[0][11] = avg_rv40_qpel16_mc32_c;
655
4
    c->avg_pixels_tab[0][12] = avg_rv40_qpel16_mc03_c;
656
4
    c->avg_pixels_tab[0][13] = avg_rv40_qpel16_mc13_c;
657
4
    c->avg_pixels_tab[0][14] = avg_rv40_qpel16_mc23_c;
658
4
    c->avg_pixels_tab[0][15] = avg_rv40_qpel16_mc33_c;
659
4
    c->put_pixels_tab[1][ 0] = qpel.put_h264_qpel_pixels_tab[1][0];
660
4
    c->put_pixels_tab[1][ 1] = put_rv40_qpel8_mc10_c;
661
4
    c->put_pixels_tab[1][ 2] = qpel.put_h264_qpel_pixels_tab[1][2];
662
4
    c->put_pixels_tab[1][ 3] = put_rv40_qpel8_mc30_c;
663
4
    c->put_pixels_tab[1][ 4] = put_rv40_qpel8_mc01_c;
664
4
    c->put_pixels_tab[1][ 5] = put_rv40_qpel8_mc11_c;
665
4
    c->put_pixels_tab[1][ 6] = put_rv40_qpel8_mc21_c;
666
4
    c->put_pixels_tab[1][ 7] = put_rv40_qpel8_mc31_c;
667
4
    c->put_pixels_tab[1][ 8] = qpel.put_h264_qpel_pixels_tab[1][8];
668
4
    c->put_pixels_tab[1][ 9] = put_rv40_qpel8_mc12_c;
669
4
    c->put_pixels_tab[1][10] = put_rv40_qpel8_mc22_c;
670
4
    c->put_pixels_tab[1][11] = put_rv40_qpel8_mc32_c;
671
4
    c->put_pixels_tab[1][12] = put_rv40_qpel8_mc03_c;
672
4
    c->put_pixels_tab[1][13] = put_rv40_qpel8_mc13_c;
673
4
    c->put_pixels_tab[1][14] = put_rv40_qpel8_mc23_c;
674
4
    c->put_pixels_tab[1][15] = put_rv40_qpel8_mc33_c;
675
4
    c->avg_pixels_tab[1][ 0] = qpel.avg_h264_qpel_pixels_tab[1][0];
676
4
    c->avg_pixels_tab[1][ 1] = avg_rv40_qpel8_mc10_c;
677
4
    c->avg_pixels_tab[1][ 2] = qpel.avg_h264_qpel_pixels_tab[1][2];
678
4
    c->avg_pixels_tab[1][ 3] = avg_rv40_qpel8_mc30_c;
679
4
    c->avg_pixels_tab[1][ 4] = avg_rv40_qpel8_mc01_c;
680
4
    c->avg_pixels_tab[1][ 5] = avg_rv40_qpel8_mc11_c;
681
4
    c->avg_pixels_tab[1][ 6] = avg_rv40_qpel8_mc21_c;
682
4
    c->avg_pixels_tab[1][ 7] = avg_rv40_qpel8_mc31_c;
683
4
    c->avg_pixels_tab[1][ 8] = qpel.avg_h264_qpel_pixels_tab[1][8];
684
4
    c->avg_pixels_tab[1][ 9] = avg_rv40_qpel8_mc12_c;
685
4
    c->avg_pixels_tab[1][10] = avg_rv40_qpel8_mc22_c;
686
4
    c->avg_pixels_tab[1][11] = avg_rv40_qpel8_mc32_c;
687
4
    c->avg_pixels_tab[1][12] = avg_rv40_qpel8_mc03_c;
688
4
    c->avg_pixels_tab[1][13] = avg_rv40_qpel8_mc13_c;
689
4
    c->avg_pixels_tab[1][14] = avg_rv40_qpel8_mc23_c;
690
4
    c->avg_pixels_tab[1][15] = avg_rv40_qpel8_mc33_c;
691
692
4
    c->put_chroma_pixels_tab[0] = put_rv40_chroma_mc8_c;
693
4
    c->put_chroma_pixels_tab[1] = put_rv40_chroma_mc4_c;
694
4
    c->avg_chroma_pixels_tab[0] = avg_rv40_chroma_mc8_c;
695
4
    c->avg_chroma_pixels_tab[1] = avg_rv40_chroma_mc4_c;
696
697
4
    c->rv40_weight_pixels_tab[0][0] = rv40_weight_func_rnd_16;
698
4
    c->rv40_weight_pixels_tab[0][1] = rv40_weight_func_rnd_8;
699
4
    c->rv40_weight_pixels_tab[1][0] = rv40_weight_func_nornd_16;
700
4
    c->rv40_weight_pixels_tab[1][1] = rv40_weight_func_nornd_8;
701
702
4
    c->rv40_weak_loop_filter[0]     = rv40_h_weak_loop_filter;
703
4
    c->rv40_weak_loop_filter[1]     = rv40_v_weak_loop_filter;
704
4
    c->rv40_strong_loop_filter[0]   = rv40_h_strong_loop_filter;
705
4
    c->rv40_strong_loop_filter[1]   = rv40_v_strong_loop_filter;
706
4
    c->rv40_loop_filter_strength[0] = rv40_h_loop_filter_strength;
707
4
    c->rv40_loop_filter_strength[1] = rv40_v_loop_filter_strength;
708
709
    if (ARCH_AARCH64)
710
        ff_rv40dsp_init_aarch64(c);
711
    if (ARCH_ARM)
712
        ff_rv40dsp_init_arm(c);
713
    if (ARCH_X86)
714
4
        ff_rv40dsp_init_x86(c);
715
4
}