2 * RV40 decoder motion compensation functions
3 * Copyright (c) 2008 Konstantin Shishkov
5 * This file is part of FFmpeg.
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.
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.
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
24 * RV40 decoder motion compensation functions
31 #define RV40_LOWPASS(OPNAME, OP) \
32 static av_unused void OPNAME ## rv40_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
33 const int h, const int C1, const int C2, const int SHIFT){\
34 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
36 for(i = 0; i < h; i++)\
38 OP(dst[0], (src[-2] + src[ 3] - 5*(src[-1]+src[2]) + src[0]*C1 + src[1]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
39 OP(dst[1], (src[-1] + src[ 4] - 5*(src[ 0]+src[3]) + src[1]*C1 + src[2]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
40 OP(dst[2], (src[ 0] + src[ 5] - 5*(src[ 1]+src[4]) + src[2]*C1 + src[3]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
41 OP(dst[3], (src[ 1] + src[ 6] - 5*(src[ 2]+src[5]) + src[3]*C1 + src[4]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
42 OP(dst[4], (src[ 2] + src[ 7] - 5*(src[ 3]+src[6]) + src[4]*C1 + src[5]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
43 OP(dst[5], (src[ 3] + src[ 8] - 5*(src[ 4]+src[7]) + src[5]*C1 + src[6]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
44 OP(dst[6], (src[ 4] + src[ 9] - 5*(src[ 5]+src[8]) + src[6]*C1 + src[7]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
45 OP(dst[7], (src[ 5] + src[10] - 5*(src[ 6]+src[9]) + src[7]*C1 + src[8]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
51 static void OPNAME ## rv40_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
52 const int w, const int C1, const int C2, const int SHIFT){\
53 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
55 for(i = 0; i < w; i++)\
57 const int srcB = src[-2*srcStride];\
58 const int srcA = src[-1*srcStride];\
59 const int src0 = src[0 *srcStride];\
60 const int src1 = src[1 *srcStride];\
61 const int src2 = src[2 *srcStride];\
62 const int src3 = src[3 *srcStride];\
63 const int src4 = src[4 *srcStride];\
64 const int src5 = src[5 *srcStride];\
65 const int src6 = src[6 *srcStride];\
66 const int src7 = src[7 *srcStride];\
67 const int src8 = src[8 *srcStride];\
68 const int src9 = src[9 *srcStride];\
69 const int src10 = src[10*srcStride];\
70 OP(dst[0*dstStride], (srcB + src3 - 5*(srcA+src2) + src0*C1 + src1*C2 + (1<<(SHIFT-1))) >> SHIFT);\
71 OP(dst[1*dstStride], (srcA + src4 - 5*(src0+src3) + src1*C1 + src2*C2 + (1<<(SHIFT-1))) >> SHIFT);\
72 OP(dst[2*dstStride], (src0 + src5 - 5*(src1+src4) + src2*C1 + src3*C2 + (1<<(SHIFT-1))) >> SHIFT);\
73 OP(dst[3*dstStride], (src1 + src6 - 5*(src2+src5) + src3*C1 + src4*C2 + (1<<(SHIFT-1))) >> SHIFT);\
74 OP(dst[4*dstStride], (src2 + src7 - 5*(src3+src6) + src4*C1 + src5*C2 + (1<<(SHIFT-1))) >> SHIFT);\
75 OP(dst[5*dstStride], (src3 + src8 - 5*(src4+src7) + src5*C1 + src6*C2 + (1<<(SHIFT-1))) >> SHIFT);\
76 OP(dst[6*dstStride], (src4 + src9 - 5*(src5+src8) + src6*C1 + src7*C2 + (1<<(SHIFT-1))) >> SHIFT);\
77 OP(dst[7*dstStride], (src5 + src10 - 5*(src6+src9) + src7*C1 + src8*C2 + (1<<(SHIFT-1))) >> SHIFT);\
83 static void OPNAME ## rv40_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
84 const int w, const int C1, const int C2, const int SHIFT){\
85 OPNAME ## rv40_qpel8_v_lowpass(dst , src , dstStride, srcStride, 8, C1, C2, SHIFT);\
86 OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
89 OPNAME ## rv40_qpel8_v_lowpass(dst , src , dstStride, srcStride, w-8, C1, C2, SHIFT);\
90 OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, w-8, C1, C2, SHIFT);\
93 static void OPNAME ## rv40_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
94 const int h, const int C1, const int C2, const int SHIFT){\
95 OPNAME ## rv40_qpel8_h_lowpass(dst , src , dstStride, srcStride, 8, C1, C2, SHIFT);\
96 OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
99 OPNAME ## rv40_qpel8_h_lowpass(dst , src , dstStride, srcStride, h-8, C1, C2, SHIFT);\
100 OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, h-8, C1, C2, SHIFT);\
104 #define RV40_MC(OPNAME, SIZE) \
105 static void OPNAME ## rv40_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
106 OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
109 static void OPNAME ## rv40_qpel ## SIZE ## _mc20_c(uint8_t *dst, uint8_t *src, int stride){\
110 OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 20, 20, 5);\
113 static void OPNAME ## rv40_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\
114 OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
117 static void OPNAME ## rv40_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\
118 OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
121 static void OPNAME ## rv40_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
122 uint8_t full[SIZE*(SIZE+5)];\
123 uint8_t * const full_mid = full + SIZE*2;\
124 put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
125 OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
128 static void OPNAME ## rv40_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
129 uint8_t full[SIZE*(SIZE+5)];\
130 uint8_t * const full_mid = full + SIZE*2;\
131 put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
132 OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
135 static void OPNAME ## rv40_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\
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, 52, 6);\
139 OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
142 static void OPNAME ## rv40_qpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\
143 OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 20, 20, 5);\
146 static void OPNAME ## rv40_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
147 uint8_t full[SIZE*(SIZE+5)];\
148 uint8_t * const full_mid = full + SIZE*2;\
149 put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
150 OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
153 static void OPNAME ## rv40_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
154 uint8_t full[SIZE*(SIZE+5)];\
155 uint8_t * const full_mid = full + SIZE*2;\
156 put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
157 OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
160 static void OPNAME ## rv40_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\
161 uint8_t full[SIZE*(SIZE+5)];\
162 uint8_t * const full_mid = full + SIZE*2;\
163 put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\
164 OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
167 static void OPNAME ## rv40_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\
168 OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
171 static void OPNAME ## rv40_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\
172 uint8_t full[SIZE*(SIZE+5)];\
173 uint8_t * const full_mid = full + SIZE*2;\
174 put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
175 OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
178 static void OPNAME ## rv40_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\
179 uint8_t full[SIZE*(SIZE+5)];\
180 uint8_t * const full_mid = full + SIZE*2;\
181 put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
182 OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
186 #define op_avg(a, b) a = (((a)+cm[b]+1)>>1)
187 #define op_put(a, b) a = cm[b]
189 RV40_LOWPASS(put_ , op_put)
190 RV40_LOWPASS(avg_ , op_avg)
200 static const int rv40_bias[4][4] = {
207 #define RV40_CHROMA_MC(OPNAME, OP)\
208 static void OPNAME ## rv40_chroma_mc4_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
209 const int A = (8-x) * (8-y);\
210 const int B = ( x) * (8-y);\
211 const int C = (8-x) * ( y);\
212 const int D = ( x) * ( y);\
214 int bias = rv40_bias[y>>1][x>>1];\
216 assert(x<8 && y<8 && x>=0 && y>=0);\
219 for(i = 0; i < h; i++){\
220 OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
221 OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
222 OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
223 OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
228 const int E = B + C;\
229 const int step = C ? stride : 1;\
230 for(i = 0; i < h; i++){\
231 OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
232 OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
233 OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
234 OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
241 static void OPNAME ## rv40_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
242 const int A = (8-x) * (8-y);\
243 const int B = ( x) * (8-y);\
244 const int C = (8-x) * ( y);\
245 const int D = ( x) * ( y);\
247 int bias = rv40_bias[y>>1][x>>1];\
249 assert(x<8 && y<8 && x>=0 && y>=0);\
252 for(i = 0; i < h; i++){\
253 OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
254 OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
255 OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
256 OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
257 OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + bias));\
258 OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + bias));\
259 OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + bias));\
260 OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + bias));\
265 const int E = B + C;\
266 const int step = C ? stride : 1;\
267 for(i = 0; i < h; i++){\
268 OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
269 OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
270 OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
271 OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
272 OP(dst[4], (A*src[4] + E*src[step+4] + bias));\
273 OP(dst[5], (A*src[5] + E*src[step+5] + bias));\
274 OP(dst[6], (A*src[6] + E*src[step+6] + bias));\
275 OP(dst[7], (A*src[7] + E*src[step+7] + bias));\
282 #define op_avg(a, b) a = (((a)+((b)>>6)+1)>>1)
283 #define op_put(a, b) a = ((b)>>6)
285 RV40_CHROMA_MC(put_, op_put)
286 RV40_CHROMA_MC(avg_, op_avg)
288 #define RV40_WEIGHT_FUNC(size) \
289 static void rv40_weight_func_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, int stride)\
293 for (j = 0; j < size; j++) {\
294 for (i = 0; i < size; i++)\
295 dst[i] = (((w2 * src1[i]) >> 9) + ((w1 * src2[i]) >> 9) + 0x10) >> 5;\
305 av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp) {
306 c->put_pixels_tab[0][ 0] = dsp->put_h264_qpel_pixels_tab[0][0];
307 c->put_pixels_tab[0][ 1] = put_rv40_qpel16_mc10_c;
308 c->put_pixels_tab[0][ 2] = put_rv40_qpel16_mc20_c;
309 c->put_pixels_tab[0][ 3] = put_rv40_qpel16_mc30_c;
310 c->put_pixels_tab[0][ 4] = put_rv40_qpel16_mc01_c;
311 c->put_pixels_tab[0][ 5] = put_rv40_qpel16_mc11_c;
312 c->put_pixels_tab[0][ 6] = put_rv40_qpel16_mc21_c;
313 c->put_pixels_tab[0][ 7] = put_rv40_qpel16_mc31_c;
314 c->put_pixels_tab[0][ 8] = put_rv40_qpel16_mc02_c;
315 c->put_pixels_tab[0][ 9] = put_rv40_qpel16_mc12_c;
316 c->put_pixels_tab[0][10] = put_rv40_qpel16_mc22_c;
317 c->put_pixels_tab[0][11] = put_rv40_qpel16_mc32_c;
318 c->put_pixels_tab[0][12] = put_rv40_qpel16_mc03_c;
319 c->put_pixels_tab[0][13] = put_rv40_qpel16_mc13_c;
320 c->put_pixels_tab[0][14] = put_rv40_qpel16_mc23_c;
321 c->put_pixels_tab[0][15] = ff_put_rv40_qpel16_mc33_c;
322 c->avg_pixels_tab[0][ 0] = dsp->avg_h264_qpel_pixels_tab[0][0];
323 c->avg_pixels_tab[0][ 1] = avg_rv40_qpel16_mc10_c;
324 c->avg_pixels_tab[0][ 2] = avg_rv40_qpel16_mc20_c;
325 c->avg_pixels_tab[0][ 3] = avg_rv40_qpel16_mc30_c;
326 c->avg_pixels_tab[0][ 4] = avg_rv40_qpel16_mc01_c;
327 c->avg_pixels_tab[0][ 5] = avg_rv40_qpel16_mc11_c;
328 c->avg_pixels_tab[0][ 6] = avg_rv40_qpel16_mc21_c;
329 c->avg_pixels_tab[0][ 7] = avg_rv40_qpel16_mc31_c;
330 c->avg_pixels_tab[0][ 8] = avg_rv40_qpel16_mc02_c;
331 c->avg_pixels_tab[0][ 9] = avg_rv40_qpel16_mc12_c;
332 c->avg_pixels_tab[0][10] = avg_rv40_qpel16_mc22_c;
333 c->avg_pixels_tab[0][11] = avg_rv40_qpel16_mc32_c;
334 c->avg_pixels_tab[0][12] = avg_rv40_qpel16_mc03_c;
335 c->avg_pixels_tab[0][13] = avg_rv40_qpel16_mc13_c;
336 c->avg_pixels_tab[0][14] = avg_rv40_qpel16_mc23_c;
337 c->avg_pixels_tab[0][15] = ff_avg_rv40_qpel16_mc33_c;
338 c->put_pixels_tab[1][ 0] = dsp->put_h264_qpel_pixels_tab[1][0];
339 c->put_pixels_tab[1][ 1] = put_rv40_qpel8_mc10_c;
340 c->put_pixels_tab[1][ 2] = put_rv40_qpel8_mc20_c;
341 c->put_pixels_tab[1][ 3] = put_rv40_qpel8_mc30_c;
342 c->put_pixels_tab[1][ 4] = put_rv40_qpel8_mc01_c;
343 c->put_pixels_tab[1][ 5] = put_rv40_qpel8_mc11_c;
344 c->put_pixels_tab[1][ 6] = put_rv40_qpel8_mc21_c;
345 c->put_pixels_tab[1][ 7] = put_rv40_qpel8_mc31_c;
346 c->put_pixels_tab[1][ 8] = put_rv40_qpel8_mc02_c;
347 c->put_pixels_tab[1][ 9] = put_rv40_qpel8_mc12_c;
348 c->put_pixels_tab[1][10] = put_rv40_qpel8_mc22_c;
349 c->put_pixels_tab[1][11] = put_rv40_qpel8_mc32_c;
350 c->put_pixels_tab[1][12] = put_rv40_qpel8_mc03_c;
351 c->put_pixels_tab[1][13] = put_rv40_qpel8_mc13_c;
352 c->put_pixels_tab[1][14] = put_rv40_qpel8_mc23_c;
353 c->put_pixels_tab[1][15] = ff_put_rv40_qpel8_mc33_c;
354 c->avg_pixels_tab[1][ 0] = dsp->avg_h264_qpel_pixels_tab[1][0];
355 c->avg_pixels_tab[1][ 1] = avg_rv40_qpel8_mc10_c;
356 c->avg_pixels_tab[1][ 2] = avg_rv40_qpel8_mc20_c;
357 c->avg_pixels_tab[1][ 3] = avg_rv40_qpel8_mc30_c;
358 c->avg_pixels_tab[1][ 4] = avg_rv40_qpel8_mc01_c;
359 c->avg_pixels_tab[1][ 5] = avg_rv40_qpel8_mc11_c;
360 c->avg_pixels_tab[1][ 6] = avg_rv40_qpel8_mc21_c;
361 c->avg_pixels_tab[1][ 7] = avg_rv40_qpel8_mc31_c;
362 c->avg_pixels_tab[1][ 8] = avg_rv40_qpel8_mc02_c;
363 c->avg_pixels_tab[1][ 9] = avg_rv40_qpel8_mc12_c;
364 c->avg_pixels_tab[1][10] = avg_rv40_qpel8_mc22_c;
365 c->avg_pixels_tab[1][11] = avg_rv40_qpel8_mc32_c;
366 c->avg_pixels_tab[1][12] = avg_rv40_qpel8_mc03_c;
367 c->avg_pixels_tab[1][13] = avg_rv40_qpel8_mc13_c;
368 c->avg_pixels_tab[1][14] = avg_rv40_qpel8_mc23_c;
369 c->avg_pixels_tab[1][15] = ff_avg_rv40_qpel8_mc33_c;
371 c->put_chroma_pixels_tab[0] = put_rv40_chroma_mc8_c;
372 c->put_chroma_pixels_tab[1] = put_rv40_chroma_mc4_c;
373 c->avg_chroma_pixels_tab[0] = avg_rv40_chroma_mc8_c;
374 c->avg_chroma_pixels_tab[1] = avg_rv40_chroma_mc4_c;
376 c->rv40_weight_pixels_tab[0] = rv40_weight_func_16;
377 c->rv40_weight_pixels_tab[1] = rv40_weight_func_8;
380 ff_rv40dsp_init_x86(c, dsp);