FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
h264qpel.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2004 Romain Dolbeau <romain@dolbeau.org>
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * 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  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14  * Lesser General Public License for more details.
15  *
16  * 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 #include "config.h"
22 #include "libavutil/attributes.h"
23 #include "libavutil/cpu.h"
24 #include "libavutil/intreadwrite.h"
27 #include "libavcodec/h264qpel.h"
28 #include "dsputil_altivec.h"
29 
30 #if HAVE_ALTIVEC
31 
32 #define PUT_OP_U8_ALTIVEC(d, s, dst) d = s
33 #define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s)
34 
35 #define OP_U8_ALTIVEC PUT_OP_U8_ALTIVEC
36 #define PREFIX_h264_qpel16_h_lowpass_altivec put_h264_qpel16_h_lowpass_altivec
37 #define PREFIX_h264_qpel16_h_lowpass_num altivec_put_h264_qpel16_h_lowpass_num
38 #define PREFIX_h264_qpel16_v_lowpass_altivec put_h264_qpel16_v_lowpass_altivec
39 #define PREFIX_h264_qpel16_v_lowpass_num altivec_put_h264_qpel16_v_lowpass_num
40 #define PREFIX_h264_qpel16_hv_lowpass_altivec put_h264_qpel16_hv_lowpass_altivec
41 #define PREFIX_h264_qpel16_hv_lowpass_num altivec_put_h264_qpel16_hv_lowpass_num
42 #include "h264qpel_template.c"
43 #undef OP_U8_ALTIVEC
44 #undef PREFIX_h264_qpel16_h_lowpass_altivec
45 #undef PREFIX_h264_qpel16_h_lowpass_num
46 #undef PREFIX_h264_qpel16_v_lowpass_altivec
47 #undef PREFIX_h264_qpel16_v_lowpass_num
48 #undef PREFIX_h264_qpel16_hv_lowpass_altivec
49 #undef PREFIX_h264_qpel16_hv_lowpass_num
50 
51 #define OP_U8_ALTIVEC AVG_OP_U8_ALTIVEC
52 #define PREFIX_h264_qpel16_h_lowpass_altivec avg_h264_qpel16_h_lowpass_altivec
53 #define PREFIX_h264_qpel16_h_lowpass_num altivec_avg_h264_qpel16_h_lowpass_num
54 #define PREFIX_h264_qpel16_v_lowpass_altivec avg_h264_qpel16_v_lowpass_altivec
55 #define PREFIX_h264_qpel16_v_lowpass_num altivec_avg_h264_qpel16_v_lowpass_num
56 #define PREFIX_h264_qpel16_hv_lowpass_altivec avg_h264_qpel16_hv_lowpass_altivec
57 #define PREFIX_h264_qpel16_hv_lowpass_num altivec_avg_h264_qpel16_hv_lowpass_num
58 #include "h264qpel_template.c"
59 #undef OP_U8_ALTIVEC
60 #undef PREFIX_h264_qpel16_h_lowpass_altivec
61 #undef PREFIX_h264_qpel16_h_lowpass_num
62 #undef PREFIX_h264_qpel16_v_lowpass_altivec
63 #undef PREFIX_h264_qpel16_v_lowpass_num
64 #undef PREFIX_h264_qpel16_hv_lowpass_altivec
65 #undef PREFIX_h264_qpel16_hv_lowpass_num
66 
67 #define H264_MC(OPNAME, SIZE, CODETYPE) \
68 static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## CODETYPE (uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
69 {\
70  ff_ ## OPNAME ## pixels ## SIZE ## _ ## CODETYPE(dst, src, stride, SIZE);\
71 }\
72 \
73 static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
74 { \
75  DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
76  put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
77  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\
78 }\
79 \
80 static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
81 {\
82  OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(dst, src, stride, stride);\
83 }\
84 \
85 static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
86 {\
87  DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
88  put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
89  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+1, half, stride, stride, SIZE);\
90 }\
91 \
92 static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
93 {\
94  DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
95  put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
96  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src, half, stride, stride, SIZE);\
97 }\
98 \
99 static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
100 {\
101  OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(dst, src, stride, stride);\
102 }\
103 \
104 static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
105 {\
106  DECLARE_ALIGNED(16, uint8_t, half)[SIZE*SIZE];\
107  put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(half, src, SIZE, stride);\
108  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, src+stride, half, stride, stride, SIZE);\
109 }\
110 \
111 static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
112 {\
113  DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
114  DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
115  put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
116  put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
117  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
118 }\
119 \
120 static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
121 {\
122  DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
123  DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
124  put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
125  put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
126  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
127 }\
128 \
129 static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
130 {\
131  DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
132  DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
133  put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
134  put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
135  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
136 }\
137 \
138 static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
139 {\
140  DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
141  DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
142  put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
143  put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
144  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfV, stride, SIZE, SIZE);\
145 }\
146 \
147 static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
148 {\
149  DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
150  OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(dst, tmp, src, stride, SIZE, stride);\
151 }\
152 \
153 static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
154 {\
155  DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
156  DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
157  DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
158  put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src, SIZE, stride);\
159  put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
160  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\
161 }\
162 \
163 static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
164 {\
165  DECLARE_ALIGNED(16, uint8_t, halfH)[SIZE*SIZE];\
166  DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
167  DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
168  put_h264_qpel ## SIZE ## _h_lowpass_ ## CODETYPE(halfH, src + stride, SIZE, stride);\
169  put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
170  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfH, halfHV, stride, SIZE, SIZE);\
171 }\
172 \
173 static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
174 {\
175  DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
176  DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
177  DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
178  put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src, SIZE, stride);\
179  put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
180  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\
181 }\
182 \
183 static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## CODETYPE(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\
184 {\
185  DECLARE_ALIGNED(16, uint8_t, halfV)[SIZE*SIZE];\
186  DECLARE_ALIGNED(16, uint8_t, halfHV)[SIZE*SIZE];\
187  DECLARE_ALIGNED(16, int16_t, tmp)[SIZE*(SIZE+8)];\
188  put_h264_qpel ## SIZE ## _v_lowpass_ ## CODETYPE(halfV, src+1, SIZE, stride);\
189  put_h264_qpel ## SIZE ## _hv_lowpass_ ## CODETYPE(halfHV, tmp, src, SIZE, SIZE, stride);\
190  OPNAME ## pixels ## SIZE ## _l2_ ## CODETYPE(dst, halfV, halfHV, stride, SIZE, SIZE);\
191 }\
192 
193 static inline void put_pixels16_l2_altivec( uint8_t * dst, const uint8_t * src1,
194  const uint8_t * src2, int dst_stride,
195  int src_stride1, int h)
196 {
197  int i;
198  vec_u8 a, b, d, tmp1, tmp2, mask, mask_, edges, align;
199 
200  mask_ = vec_lvsl(0, src2);
201 
202  for (i = 0; i < h; i++) {
203 
204  tmp1 = vec_ld(i * src_stride1, src1);
205  mask = vec_lvsl(i * src_stride1, src1);
206  tmp2 = vec_ld(i * src_stride1 + 15, src1);
207 
208  a = vec_perm(tmp1, tmp2, mask);
209 
210  tmp1 = vec_ld(i * 16, src2);
211  tmp2 = vec_ld(i * 16 + 15, src2);
212 
213  b = vec_perm(tmp1, tmp2, mask_);
214 
215  tmp1 = vec_ld(0, dst);
216  mask = vec_lvsl(0, dst);
217  tmp2 = vec_ld(15, dst);
218 
219  d = vec_avg(a, b);
220 
221  edges = vec_perm(tmp2, tmp1, mask);
222 
223  align = vec_lvsr(0, dst);
224 
225  tmp2 = vec_perm(d, edges, align);
226  tmp1 = vec_perm(edges, d, align);
227 
228  vec_st(tmp2, 15, dst);
229  vec_st(tmp1, 0 , dst);
230 
231  dst += dst_stride;
232  }
233 }
234 
235 static inline void avg_pixels16_l2_altivec( uint8_t * dst, const uint8_t * src1,
236  const uint8_t * src2, int dst_stride,
237  int src_stride1, int h)
238 {
239  int i;
240  vec_u8 a, b, d, tmp1, tmp2, mask, mask_, edges, align;
241 
242  mask_ = vec_lvsl(0, src2);
243 
244  for (i = 0; i < h; i++) {
245 
246  tmp1 = vec_ld(i * src_stride1, src1);
247  mask = vec_lvsl(i * src_stride1, src1);
248  tmp2 = vec_ld(i * src_stride1 + 15, src1);
249 
250  a = vec_perm(tmp1, tmp2, mask);
251 
252  tmp1 = vec_ld(i * 16, src2);
253  tmp2 = vec_ld(i * 16 + 15, src2);
254 
255  b = vec_perm(tmp1, tmp2, mask_);
256 
257  tmp1 = vec_ld(0, dst);
258  mask = vec_lvsl(0, dst);
259  tmp2 = vec_ld(15, dst);
260 
261  d = vec_avg(vec_perm(tmp1, tmp2, mask), vec_avg(a, b));
262 
263  edges = vec_perm(tmp2, tmp1, mask);
264 
265  align = vec_lvsr(0, dst);
266 
267  tmp2 = vec_perm(d, edges, align);
268  tmp1 = vec_perm(edges, d, align);
269 
270  vec_st(tmp2, 15, dst);
271  vec_st(tmp1, 0 , dst);
272 
273  dst += dst_stride;
274  }
275 }
276 
277 /* Implemented but could be faster
278 #define put_pixels16_l2_altivec(d,s1,s2,ds,s1s,h) put_pixels16_l2(d,s1,s2,ds,s1s,16,h)
279 #define avg_pixels16_l2_altivec(d,s1,s2,ds,s1s,h) avg_pixels16_l2(d,s1,s2,ds,s1s,16,h)
280  */
281 
282 H264_MC(put_, 16, altivec)
283 H264_MC(avg_, 16, altivec)
284 #endif /* HAVE_ALTIVEC */
285 
287 {
288 #if HAVE_ALTIVEC
289  const int high_bit_depth = bit_depth > 8;
290 
292  if (!high_bit_depth) {
293 #define dspfunc(PFX, IDX, NUM) \
294  c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_altivec; \
295  c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_altivec; \
296  c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_altivec; \
297  c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_altivec; \
298  c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_altivec; \
299  c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_altivec; \
300  c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_altivec; \
301  c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_altivec; \
302  c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_altivec; \
303  c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_altivec; \
304  c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_altivec; \
305  c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_altivec; \
306  c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_altivec; \
307  c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_altivec; \
308  c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_altivec; \
309  c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_altivec
310 
311  dspfunc(put_h264_qpel, 0, 16);
312  dspfunc(avg_h264_qpel, 0, 16);
313 #undef dspfunc
314  }
315  }
316 #endif /* HAVE_ALTIVEC */
317 }