FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
h264_qpel.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2004-2005 Michael Niedermayer, Loren Merritt
3  * Copyright (c) 2011 Daniel Kang
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 #include "libavutil/attributes.h"
23 #include "libavutil/cpu.h"
24 #include "libavutil/x86/asm.h"
25 #include "libavutil/x86/cpu.h"
26 #include "libavcodec/h264.h"
27 #include "libavcodec/h264qpel.h"
28 #include "libavcodec/pixels.h"
29 #include "fpel.h"
30 
31 #if HAVE_YASM
32 void ff_put_pixels4_l2_mmxext(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
33  int dstStride, int src1Stride, int h);
34 void ff_avg_pixels4_l2_mmxext(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
35  int dstStride, int src1Stride, int h);
36 void ff_put_pixels8_l2_mmxext(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
37  int dstStride, int src1Stride, int h);
38 void ff_avg_pixels8_l2_mmxext(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
39  int dstStride, int src1Stride, int h);
40 void ff_put_pixels16_l2_mmxext(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
41  int dstStride, int src1Stride, int h);
42 void ff_avg_pixels16_l2_mmxext(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
43  int dstStride, int src1Stride, int h);
44 #define ff_put_pixels8_l2_sse2 ff_put_pixels8_l2_mmxext
45 #define ff_avg_pixels8_l2_sse2 ff_avg_pixels8_l2_mmxext
46 #define ff_put_pixels16_l2_sse2 ff_put_pixels16_l2_mmxext
47 #define ff_avg_pixels16_l2_sse2 ff_avg_pixels16_l2_mmxext
48 #define ff_put_pixels16_mmxext ff_put_pixels16_mmx
49 #define ff_put_pixels8_mmxext ff_put_pixels8_mmx
50 #define ff_put_pixels4_mmxext ff_put_pixels4_mmx
51 
52 #define DEF_QPEL(OPNAME)\
53 void ff_ ## OPNAME ## _h264_qpel4_h_lowpass_mmxext(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride);\
54 void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_mmxext(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride);\
55 void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_ssse3(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride);\
56 void ff_ ## OPNAME ## _h264_qpel4_h_lowpass_l2_mmxext(uint8_t *dst, const uint8_t *src, const uint8_t *src2, int dstStride, int src2Stride);\
57 void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_l2_mmxext(uint8_t *dst, const uint8_t *src, const uint8_t *src2, int dstStride, int src2Stride);\
58 void ff_ ## OPNAME ## _h264_qpel8_h_lowpass_l2_ssse3(uint8_t *dst, const uint8_t *src, const uint8_t *src2, int dstStride, int src2Stride);\
59 void ff_ ## OPNAME ## _h264_qpel4_v_lowpass_mmxext(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride);\
60 void ff_ ## OPNAME ## _h264_qpel8or16_v_lowpass_op_mmxext(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h);\
61 void ff_ ## OPNAME ## _h264_qpel8or16_v_lowpass_sse2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h);\
62 void ff_ ## OPNAME ## _h264_qpel4_hv_lowpass_v_mmxext(const uint8_t *src, int16_t *tmp, int srcStride);\
63 void ff_ ## OPNAME ## _h264_qpel4_hv_lowpass_h_mmxext(int16_t *tmp, uint8_t *dst, int dstStride);\
64 void ff_ ## OPNAME ## _h264_qpel8or16_hv1_lowpass_op_mmxext(const uint8_t *src, int16_t *tmp, int srcStride, int size);\
65 void ff_ ## OPNAME ## _h264_qpel8or16_hv1_lowpass_op_sse2(const uint8_t *src, int16_t *tmp, int srcStride, int size);\
66 void ff_ ## OPNAME ## _h264_qpel8or16_hv2_lowpass_op_mmxext(uint8_t *dst, int16_t *tmp, int dstStride, int unused, int h);\
67 void ff_ ## OPNAME ## _h264_qpel8or16_hv2_lowpass_ssse3(uint8_t *dst, int16_t *tmp, int dstStride, int tmpStride, int size);\
68 void ff_ ## OPNAME ## _pixels4_l2_shift5_mmxext(uint8_t *dst, const int16_t *src16, const uint8_t *src8, int dstStride, int src8Stride, int h);\
69 void ff_ ## OPNAME ## _pixels8_l2_shift5_mmxext(uint8_t *dst, const int16_t *src16, const uint8_t *src8, int dstStride, int src8Stride, int h);
70 
71 DEF_QPEL(avg)
72 DEF_QPEL(put)
73 
74 #define QPEL_H264(OPNAME, OP, MMX)\
75 static av_always_inline void ff_ ## OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, const uint8_t *src, int dstStride, int tmpStride, int srcStride){\
76  int w=3;\
77  src -= 2*srcStride+2;\
78  while(w--){\
79  ff_ ## OPNAME ## h264_qpel4_hv_lowpass_v_mmxext(src, tmp, srcStride);\
80  tmp += 4;\
81  src += 4;\
82  }\
83  tmp -= 3*4;\
84  ff_ ## OPNAME ## h264_qpel4_hv_lowpass_h_mmxext(tmp, dst, dstStride);\
85 }\
86 \
87 static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h){\
88  src -= 2*srcStride;\
89  ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_op_mmxext(dst, src, dstStride, srcStride, h);\
90  src += 4;\
91  dst += 4;\
92  ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_op_mmxext(dst, src, dstStride, srcStride, h);\
93 }\
94 static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv1_lowpass_ ## MMX(int16_t *tmp, const uint8_t *src, int tmpStride, int srcStride, int size){\
95  int w = (size+8)>>2;\
96  src -= 2*srcStride+2;\
97  while(w--){\
98  ff_ ## OPNAME ## h264_qpel8or16_hv1_lowpass_op_mmxext(src, tmp, srcStride, size);\
99  tmp += 4;\
100  src += 4;\
101  }\
102 }\
103 static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, int dstStride, int tmpStride, int size){\
104  int w = size>>4;\
105  do{\
106  ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_op_mmxext(dst, tmp, dstStride, 0, size);\
107  tmp += 8;\
108  dst += 8;\
109  }while(w--);\
110 }\
111 \
112 static av_always_inline void ff_ ## OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
113  ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst , src , dstStride, srcStride, 8);\
114 }\
115 static av_always_inline void ff_ ## OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
116  ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst , src , dstStride, srcStride, 16);\
117  ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
118 }\
119 \
120 static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
121  ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst , src , dstStride, srcStride);\
122  ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
123  src += 8*srcStride;\
124  dst += 8*dstStride;\
125  ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst , src , dstStride, srcStride);\
126  ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
127 }\
128 \
129 static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, const uint8_t *src, const uint8_t *src2, int dstStride, int src2Stride){\
130  ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst , src , src2 , dstStride, src2Stride);\
131  ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
132  src += 8*dstStride;\
133  dst += 8*dstStride;\
134  src2 += 8*src2Stride;\
135  ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst , src , src2 , dstStride, src2Stride);\
136  ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
137 }\
138 \
139 static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, const uint8_t *src, int dstStride, int tmpStride, int srcStride, int size){\
140  ff_put_h264_qpel8or16_hv1_lowpass_ ## MMX(tmp, src, tmpStride, srcStride, size);\
141  ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(dst, tmp, dstStride, tmpStride, size);\
142 }\
143 static av_always_inline void ff_ ## OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, const uint8_t *src, int dstStride, int tmpStride, int srcStride){\
144  ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst , tmp , src , dstStride, tmpStride, srcStride, 8);\
145 }\
146 \
147 static av_always_inline void ff_ ## OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, const uint8_t *src, int dstStride, int tmpStride, int srcStride){\
148  ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst , tmp , src , dstStride, tmpStride, srcStride, 16);\
149 }\
150 \
151 static av_always_inline void ff_ ## OPNAME ## pixels16_l2_shift5_ ## MMX(uint8_t *dst, const int16_t *src16, const uint8_t *src8, int dstStride, int src8Stride, int h)\
152 {\
153  ff_ ## OPNAME ## pixels8_l2_shift5_ ## MMX(dst , src16 , src8 , dstStride, src8Stride, h);\
154  ff_ ## OPNAME ## pixels8_l2_shift5_ ## MMX(dst+8, src16+8, src8+8, dstStride, src8Stride, h);\
155 }\
156 
157 
158 #if ARCH_X86_64
159 #define QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
160 
161 void ff_avg_h264_qpel16_h_lowpass_l2_ssse3(uint8_t *dst, const uint8_t *src, const uint8_t *src2, int dstStride, int src2Stride);
162 void ff_put_h264_qpel16_h_lowpass_l2_ssse3(uint8_t *dst, const uint8_t *src, const uint8_t *src2, int dstStride, int src2Stride);
163 
164 #else // ARCH_X86_64
165 #define QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
166 static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, const uint8_t *src, const uint8_t *src2, int dstStride, int src2Stride){\
167  ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst , src , src2 , dstStride, src2Stride);\
168  ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
169  src += 8*dstStride;\
170  dst += 8*dstStride;\
171  src2 += 8*src2Stride;\
172  ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst , src , src2 , dstStride, src2Stride);\
173  ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
174 }
175 #endif // ARCH_X86_64
176 
177 #define QPEL_H264_H_XMM(OPNAME, OP, MMX)\
178 QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
179 static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
180  ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst , src , dstStride, srcStride);\
181  ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
182  src += 8*srcStride;\
183  dst += 8*dstStride;\
184  ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst , src , dstStride, srcStride);\
185  ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
186 }\
187 
188 #define QPEL_H264_V_XMM(OPNAME, OP, MMX)\
189 static av_always_inline void ff_ ## OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
190  ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst , src , dstStride, srcStride, 8);\
191 }\
192 static av_always_inline void ff_ ## OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
193  ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst , src , dstStride, srcStride, 16);\
194  ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
195 }
196 
197 static av_always_inline void put_h264_qpel8or16_hv1_lowpass_sse2(int16_t *tmp,
198  const uint8_t *src,
199  int tmpStride,
200  int srcStride,
201  int size)
202 {
203  int w = (size+8)>>3;
204  src -= 2*srcStride+2;
205  while(w--){
206  ff_put_h264_qpel8or16_hv1_lowpass_op_sse2(src, tmp, srcStride, size);
207  tmp += 8;
208  src += 8;
209  }
210 }
211 
212 #define QPEL_H264_HV_XMM(OPNAME, OP, MMX)\
213 static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, const uint8_t *src, int dstStride, int tmpStride, int srcStride, int size){\
214  put_h264_qpel8or16_hv1_lowpass_sse2(tmp, src, tmpStride, srcStride, size);\
215  ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(dst, tmp, dstStride, tmpStride, size);\
216 }\
217 static av_always_inline void ff_ ## OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, const uint8_t *src, int dstStride, int tmpStride, int srcStride){\
218  ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst, tmp, src, dstStride, tmpStride, srcStride, 8);\
219 }\
220 static av_always_inline void ff_ ## OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, const uint8_t *src, int dstStride, int tmpStride, int srcStride){\
221  ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst, tmp, src, dstStride, tmpStride, srcStride, 16);\
222 }\
223 
224 #define ff_put_h264_qpel8_h_lowpass_l2_sse2 ff_put_h264_qpel8_h_lowpass_l2_mmxext
225 #define ff_avg_h264_qpel8_h_lowpass_l2_sse2 ff_avg_h264_qpel8_h_lowpass_l2_mmxext
226 #define ff_put_h264_qpel16_h_lowpass_l2_sse2 ff_put_h264_qpel16_h_lowpass_l2_mmxext
227 #define ff_avg_h264_qpel16_h_lowpass_l2_sse2 ff_avg_h264_qpel16_h_lowpass_l2_mmxext
228 
229 #define ff_put_h264_qpel8_v_lowpass_ssse3 ff_put_h264_qpel8_v_lowpass_sse2
230 #define ff_avg_h264_qpel8_v_lowpass_ssse3 ff_avg_h264_qpel8_v_lowpass_sse2
231 #define ff_put_h264_qpel16_v_lowpass_ssse3 ff_put_h264_qpel16_v_lowpass_sse2
232 #define ff_avg_h264_qpel16_v_lowpass_ssse3 ff_avg_h264_qpel16_v_lowpass_sse2
233 
234 #define ff_put_h264_qpel8or16_hv2_lowpass_sse2 ff_put_h264_qpel8or16_hv2_lowpass_mmxext
235 #define ff_avg_h264_qpel8or16_hv2_lowpass_sse2 ff_avg_h264_qpel8or16_hv2_lowpass_mmxext
236 
237 #define H264_MC(OPNAME, SIZE, MMX, ALIGN) \
238 H264_MC_C(OPNAME, SIZE, MMX, ALIGN)\
239 H264_MC_V(OPNAME, SIZE, MMX, ALIGN)\
240 H264_MC_H(OPNAME, SIZE, MMX, ALIGN)\
241 H264_MC_HV(OPNAME, SIZE, MMX, ALIGN)\
242 
243 static void put_h264_qpel16_mc00_sse2 (uint8_t *dst, const uint8_t *src,
244  ptrdiff_t stride)
245 {
246  ff_put_pixels16_sse2(dst, src, stride, 16);
247 }
248 static void avg_h264_qpel16_mc00_sse2 (uint8_t *dst, const uint8_t *src,
249  ptrdiff_t stride)
250 {
251  ff_avg_pixels16_sse2(dst, src, stride, 16);
252 }
253 #define put_h264_qpel8_mc00_sse2 put_h264_qpel8_mc00_mmxext
254 #define avg_h264_qpel8_mc00_sse2 avg_h264_qpel8_mc00_mmxext
255 
256 #define H264_MC_C(OPNAME, SIZE, MMX, ALIGN) \
257 static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## MMX (uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
258 {\
259  ff_ ## OPNAME ## pixels ## SIZE ## _ ## MMX(dst, src, stride, SIZE);\
260 }\
261 
262 #define H264_MC_H(OPNAME, SIZE, MMX, ALIGN) \
263 static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
264 {\
265  ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src, stride, stride);\
266 }\
267 \
268 static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
269 {\
270  ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## MMX(dst, src, stride, stride);\
271 }\
272 \
273 static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
274 {\
275  ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src+1, stride, stride);\
276 }\
277 
278 #define H264_MC_V(OPNAME, SIZE, MMX, ALIGN) \
279 static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
280 {\
281  LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
282  ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
283  ff_ ## OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src, temp, stride, stride, SIZE);\
284 }\
285 \
286 static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
287 {\
288  ff_ ## OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## MMX(dst, src, stride, stride);\
289 }\
290 \
291 static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
292 {\
293  LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
294  ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
295  ff_ ## OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src+stride, temp, stride, stride, SIZE);\
296 }\
297 
298 #define H264_MC_HV(OPNAME, SIZE, MMX, ALIGN) \
299 static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
300 {\
301  LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
302  ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
303  ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
304 }\
305 \
306 static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
307 {\
308  LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
309  ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
310  ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
311 }\
312 \
313 static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
314 {\
315  LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
316  ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
317  ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
318 }\
319 \
320 static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
321 {\
322  LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
323  ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
324  ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
325 }\
326 \
327 static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
328 {\
329  LOCAL_ALIGNED(ALIGN, uint16_t, temp, [SIZE*(SIZE<8?12:24)]);\
330  ff_ ## OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(dst, temp, src, stride, SIZE, stride);\
331 }\
332 \
333 static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
334 {\
335  LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
336  uint8_t * const halfHV= temp;\
337  int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
338  av_assert2(((int)temp & 7) == 0);\
339  ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
340  ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfHV, stride, SIZE);\
341 }\
342 \
343 static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
344 {\
345  LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
346  uint8_t * const halfHV= temp;\
347  int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
348  av_assert2(((int)temp & 7) == 0);\
349  ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
350  ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfHV, stride, SIZE);\
351 }\
352 \
353 static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
354 {\
355  LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
356  uint8_t * const halfHV= temp;\
357  int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
358  av_assert2(((int)temp & 7) == 0);\
359  ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
360  ff_ ## OPNAME ## pixels ## SIZE ## _l2_shift5_mmxext(dst, halfV+2, halfHV, stride, SIZE, SIZE);\
361 }\
362 \
363 static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
364 {\
365  LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
366  uint8_t * const halfHV= temp;\
367  int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
368  av_assert2(((int)temp & 7) == 0);\
369  ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
370  ff_ ## OPNAME ## pixels ## SIZE ## _l2_shift5_mmxext(dst, halfV+3, halfHV, stride, SIZE, SIZE);\
371 }\
372 
373 #define H264_MC_4816(MMX)\
374 H264_MC(put_, 4, MMX, 8)\
375 H264_MC(put_, 8, MMX, 8)\
376 H264_MC(put_, 16,MMX, 8)\
377 H264_MC(avg_, 4, MMX, 8)\
378 H264_MC(avg_, 8, MMX, 8)\
379 H264_MC(avg_, 16,MMX, 8)\
380 
381 #define H264_MC_816(QPEL, XMM)\
382 QPEL(put_, 8, XMM, 16)\
383 QPEL(put_, 16,XMM, 16)\
384 QPEL(avg_, 8, XMM, 16)\
385 QPEL(avg_, 16,XMM, 16)\
386 
387 QPEL_H264(put_, PUT_OP, mmxext)
388 QPEL_H264(avg_, AVG_MMXEXT_OP, mmxext)
389 QPEL_H264_V_XMM(put_, PUT_OP, sse2)
390 QPEL_H264_V_XMM(avg_,AVG_MMXEXT_OP, sse2)
391 QPEL_H264_HV_XMM(put_, PUT_OP, sse2)
392 QPEL_H264_HV_XMM(avg_,AVG_MMXEXT_OP, sse2)
393 QPEL_H264_H_XMM(put_, PUT_OP, ssse3)
394 QPEL_H264_H_XMM(avg_,AVG_MMXEXT_OP, ssse3)
395 QPEL_H264_HV_XMM(put_, PUT_OP, ssse3)
396 QPEL_H264_HV_XMM(avg_,AVG_MMXEXT_OP, ssse3)
397 
398 H264_MC_4816(mmxext)
399 H264_MC_816(H264_MC_V, sse2)
400 H264_MC_816(H264_MC_HV, sse2)
401 H264_MC_816(H264_MC_H, ssse3)
402 H264_MC_816(H264_MC_HV, ssse3)
403 
404 
405 //10bit
406 #define LUMA_MC_OP(OP, NUM, DEPTH, TYPE, OPT) \
407 void ff_ ## OP ## _h264_qpel ## NUM ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT \
408  (uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
409 
410 #define LUMA_MC_ALL(DEPTH, TYPE, OPT) \
411  LUMA_MC_OP(put, 4, DEPTH, TYPE, OPT) \
412  LUMA_MC_OP(avg, 4, DEPTH, TYPE, OPT) \
413  LUMA_MC_OP(put, 8, DEPTH, TYPE, OPT) \
414  LUMA_MC_OP(avg, 8, DEPTH, TYPE, OPT) \
415  LUMA_MC_OP(put, 16, DEPTH, TYPE, OPT) \
416  LUMA_MC_OP(avg, 16, DEPTH, TYPE, OPT)
417 
418 #define LUMA_MC_816(DEPTH, TYPE, OPT) \
419  LUMA_MC_OP(put, 8, DEPTH, TYPE, OPT) \
420  LUMA_MC_OP(avg, 8, DEPTH, TYPE, OPT) \
421  LUMA_MC_OP(put, 16, DEPTH, TYPE, OPT) \
422  LUMA_MC_OP(avg, 16, DEPTH, TYPE, OPT)
423 
424 LUMA_MC_ALL(10, mc00, mmxext)
425 LUMA_MC_ALL(10, mc10, mmxext)
426 LUMA_MC_ALL(10, mc20, mmxext)
427 LUMA_MC_ALL(10, mc30, mmxext)
428 LUMA_MC_ALL(10, mc01, mmxext)
429 LUMA_MC_ALL(10, mc11, mmxext)
430 LUMA_MC_ALL(10, mc21, mmxext)
431 LUMA_MC_ALL(10, mc31, mmxext)
432 LUMA_MC_ALL(10, mc02, mmxext)
433 LUMA_MC_ALL(10, mc12, mmxext)
434 LUMA_MC_ALL(10, mc22, mmxext)
435 LUMA_MC_ALL(10, mc32, mmxext)
436 LUMA_MC_ALL(10, mc03, mmxext)
437 LUMA_MC_ALL(10, mc13, mmxext)
438 LUMA_MC_ALL(10, mc23, mmxext)
439 LUMA_MC_ALL(10, mc33, mmxext)
440 
441 LUMA_MC_816(10, mc00, sse2)
442 LUMA_MC_816(10, mc10, sse2)
443 LUMA_MC_816(10, mc10, sse2_cache64)
444 LUMA_MC_816(10, mc10, ssse3_cache64)
445 LUMA_MC_816(10, mc20, sse2)
446 LUMA_MC_816(10, mc20, sse2_cache64)
447 LUMA_MC_816(10, mc20, ssse3_cache64)
448 LUMA_MC_816(10, mc30, sse2)
449 LUMA_MC_816(10, mc30, sse2_cache64)
450 LUMA_MC_816(10, mc30, ssse3_cache64)
451 LUMA_MC_816(10, mc01, sse2)
452 LUMA_MC_816(10, mc11, sse2)
453 LUMA_MC_816(10, mc21, sse2)
454 LUMA_MC_816(10, mc31, sse2)
455 LUMA_MC_816(10, mc02, sse2)
456 LUMA_MC_816(10, mc12, sse2)
457 LUMA_MC_816(10, mc22, sse2)
458 LUMA_MC_816(10, mc32, sse2)
459 LUMA_MC_816(10, mc03, sse2)
460 LUMA_MC_816(10, mc13, sse2)
461 LUMA_MC_816(10, mc23, sse2)
462 LUMA_MC_816(10, mc33, sse2)
463 
464 #define QPEL16_OPMC(OP, MC, MMX)\
465 void ff_ ## OP ## _h264_qpel16_ ## MC ## _10_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride){\
466  ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst , src , stride);\
467  ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst+16, src+16, stride);\
468  src += 8*stride;\
469  dst += 8*stride;\
470  ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst , src , stride);\
471  ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst+16, src+16, stride);\
472 }
473 
474 #define QPEL16_OP(MC, MMX)\
475 QPEL16_OPMC(put, MC, MMX)\
476 QPEL16_OPMC(avg, MC, MMX)
477 
478 #define QPEL16(MMX)\
479 QPEL16_OP(mc00, MMX)\
480 QPEL16_OP(mc01, MMX)\
481 QPEL16_OP(mc02, MMX)\
482 QPEL16_OP(mc03, MMX)\
483 QPEL16_OP(mc10, MMX)\
484 QPEL16_OP(mc11, MMX)\
485 QPEL16_OP(mc12, MMX)\
486 QPEL16_OP(mc13, MMX)\
487 QPEL16_OP(mc20, MMX)\
488 QPEL16_OP(mc21, MMX)\
489 QPEL16_OP(mc22, MMX)\
490 QPEL16_OP(mc23, MMX)\
491 QPEL16_OP(mc30, MMX)\
492 QPEL16_OP(mc31, MMX)\
493 QPEL16_OP(mc32, MMX)\
494 QPEL16_OP(mc33, MMX)
495 
496 #if ARCH_X86_32 // ARCH_X86_64 implies SSE2+
497 QPEL16(mmxext)
498 #endif
499 
500 #endif /* HAVE_YASM */
501 
502 #define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU, PREFIX) \
503  do { \
504  c->PFX ## _pixels_tab[IDX][ 0] = PREFIX ## PFX ## SIZE ## _mc00_ ## CPU; \
505  c->PFX ## _pixels_tab[IDX][ 1] = PREFIX ## PFX ## SIZE ## _mc10_ ## CPU; \
506  c->PFX ## _pixels_tab[IDX][ 2] = PREFIX ## PFX ## SIZE ## _mc20_ ## CPU; \
507  c->PFX ## _pixels_tab[IDX][ 3] = PREFIX ## PFX ## SIZE ## _mc30_ ## CPU; \
508  c->PFX ## _pixels_tab[IDX][ 4] = PREFIX ## PFX ## SIZE ## _mc01_ ## CPU; \
509  c->PFX ## _pixels_tab[IDX][ 5] = PREFIX ## PFX ## SIZE ## _mc11_ ## CPU; \
510  c->PFX ## _pixels_tab[IDX][ 6] = PREFIX ## PFX ## SIZE ## _mc21_ ## CPU; \
511  c->PFX ## _pixels_tab[IDX][ 7] = PREFIX ## PFX ## SIZE ## _mc31_ ## CPU; \
512  c->PFX ## _pixels_tab[IDX][ 8] = PREFIX ## PFX ## SIZE ## _mc02_ ## CPU; \
513  c->PFX ## _pixels_tab[IDX][ 9] = PREFIX ## PFX ## SIZE ## _mc12_ ## CPU; \
514  c->PFX ## _pixels_tab[IDX][10] = PREFIX ## PFX ## SIZE ## _mc22_ ## CPU; \
515  c->PFX ## _pixels_tab[IDX][11] = PREFIX ## PFX ## SIZE ## _mc32_ ## CPU; \
516  c->PFX ## _pixels_tab[IDX][12] = PREFIX ## PFX ## SIZE ## _mc03_ ## CPU; \
517  c->PFX ## _pixels_tab[IDX][13] = PREFIX ## PFX ## SIZE ## _mc13_ ## CPU; \
518  c->PFX ## _pixels_tab[IDX][14] = PREFIX ## PFX ## SIZE ## _mc23_ ## CPU; \
519  c->PFX ## _pixels_tab[IDX][15] = PREFIX ## PFX ## SIZE ## _mc33_ ## CPU; \
520  } while (0)
521 
522 #define H264_QPEL_FUNCS(x, y, CPU) \
523  do { \
524  c->put_h264_qpel_pixels_tab[0][x + y * 4] = put_h264_qpel16_mc ## x ## y ## _ ## CPU; \
525  c->put_h264_qpel_pixels_tab[1][x + y * 4] = put_h264_qpel8_mc ## x ## y ## _ ## CPU; \
526  c->avg_h264_qpel_pixels_tab[0][x + y * 4] = avg_h264_qpel16_mc ## x ## y ## _ ## CPU; \
527  c->avg_h264_qpel_pixels_tab[1][x + y * 4] = avg_h264_qpel8_mc ## x ## y ## _ ## CPU; \
528  } while (0)
529 
530 #define H264_QPEL_FUNCS_10(x, y, CPU) \
531  do { \
532  c->put_h264_qpel_pixels_tab[0][x + y * 4] = ff_put_h264_qpel16_mc ## x ## y ## _10_ ## CPU; \
533  c->put_h264_qpel_pixels_tab[1][x + y * 4] = ff_put_h264_qpel8_mc ## x ## y ## _10_ ## CPU; \
534  c->avg_h264_qpel_pixels_tab[0][x + y * 4] = ff_avg_h264_qpel16_mc ## x ## y ## _10_ ## CPU; \
535  c->avg_h264_qpel_pixels_tab[1][x + y * 4] = ff_avg_h264_qpel8_mc ## x ## y ## _10_ ## CPU; \
536  } while (0)
537 
539 {
540 #if HAVE_YASM
541  int high_bit_depth = bit_depth > 8;
542  int cpu_flags = av_get_cpu_flags();
543 
544  if (EXTERNAL_MMXEXT(cpu_flags)) {
545  if (!high_bit_depth) {
546  SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmxext, );
547  SET_QPEL_FUNCS(put_h264_qpel, 1, 8, mmxext, );
548  SET_QPEL_FUNCS(put_h264_qpel, 2, 4, mmxext, );
549  SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmxext, );
550  SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, mmxext, );
551  SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, mmxext, );
552  } else if (bit_depth == 10) {
553 #if ARCH_X86_32
554  SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_mmxext, ff_);
555  SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_mmxext, ff_);
556  SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 10_mmxext, ff_);
557  SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 10_mmxext, ff_);
558 #endif
559  SET_QPEL_FUNCS(put_h264_qpel, 2, 4, 10_mmxext, ff_);
560  SET_QPEL_FUNCS(avg_h264_qpel, 2, 4, 10_mmxext, ff_);
561  }
562  }
563 
564  if (EXTERNAL_SSE2(cpu_flags)) {
565  if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW) && !high_bit_depth) {
566  // these functions are slower than mmx on AMD, but faster on Intel
567  H264_QPEL_FUNCS(0, 0, sse2);
568  }
569 
570  if (!high_bit_depth) {
571  H264_QPEL_FUNCS(0, 1, sse2);
572  H264_QPEL_FUNCS(0, 2, sse2);
573  H264_QPEL_FUNCS(0, 3, sse2);
574  H264_QPEL_FUNCS(1, 1, sse2);
575  H264_QPEL_FUNCS(1, 2, sse2);
576  H264_QPEL_FUNCS(1, 3, sse2);
577  H264_QPEL_FUNCS(2, 1, sse2);
578  H264_QPEL_FUNCS(2, 2, sse2);
579  H264_QPEL_FUNCS(2, 3, sse2);
580  H264_QPEL_FUNCS(3, 1, sse2);
581  H264_QPEL_FUNCS(3, 2, sse2);
582  H264_QPEL_FUNCS(3, 3, sse2);
583  }
584 
585  if (bit_depth == 10) {
586  SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_sse2, ff_);
587  SET_QPEL_FUNCS(put_h264_qpel, 1, 8, 10_sse2, ff_);
588  SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_sse2, ff_);
589  SET_QPEL_FUNCS(avg_h264_qpel, 1, 8, 10_sse2, ff_);
590  H264_QPEL_FUNCS_10(1, 0, sse2_cache64);
591  H264_QPEL_FUNCS_10(2, 0, sse2_cache64);
592  H264_QPEL_FUNCS_10(3, 0, sse2_cache64);
593  }
594  }
595 
596  if (EXTERNAL_SSSE3(cpu_flags)) {
597  if (!high_bit_depth) {
598  H264_QPEL_FUNCS(1, 0, ssse3);
599  H264_QPEL_FUNCS(1, 1, ssse3);
600  H264_QPEL_FUNCS(1, 2, ssse3);
601  H264_QPEL_FUNCS(1, 3, ssse3);
602  H264_QPEL_FUNCS(2, 0, ssse3);
603  H264_QPEL_FUNCS(2, 1, ssse3);
604  H264_QPEL_FUNCS(2, 2, ssse3);
605  H264_QPEL_FUNCS(2, 3, ssse3);
606  H264_QPEL_FUNCS(3, 0, ssse3);
607  H264_QPEL_FUNCS(3, 1, ssse3);
608  H264_QPEL_FUNCS(3, 2, ssse3);
609  H264_QPEL_FUNCS(3, 3, ssse3);
610  }
611 
612  if (bit_depth == 10) {
613  H264_QPEL_FUNCS_10(1, 0, ssse3_cache64);
614  H264_QPEL_FUNCS_10(2, 0, ssse3_cache64);
615  H264_QPEL_FUNCS_10(3, 0, ssse3_cache64);
616  }
617  }
618 
619  if (EXTERNAL_AVX(cpu_flags)) {
620  /* AVX implies 64 byte cache lines without the need to avoid unaligned
621  * memory accesses that cross the boundary between two cache lines.
622  * TODO: Port X264_CPU_CACHELINE_32/64 detection from x264 to avoid
623  * having to treat SSE2 functions with such properties as AVX. */
624  if (bit_depth == 10) {
625  H264_QPEL_FUNCS_10(1, 0, sse2);
626  H264_QPEL_FUNCS_10(2, 0, sse2);
627  H264_QPEL_FUNCS_10(3, 0, sse2);
628  }
629  }
630 #endif
631 }
void ff_avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h)
Macro definitions for various function/variable attributes.
#define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU, PREFIX)
Definition: h264_qpel.c:502
uint8_t
#define av_cold
Definition: attributes.h:82
#define AV_CPU_FLAG_SSE2SLOW
SSE2 supported, but usually not faster.
Definition: cpu.h:35
ptrdiff_t size
Definition: opengl_enc.c:101
#define EXTERNAL_SSE2(flags)
Definition: cpu.h:57
H.264 / AVC / MPEG-4 part10 codec.
void ff_put_pixels8_l2_mmxext(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dstStride, int src1Stride, int h)
#define src
Definition: vp9dsp.c:530
void ff_avg_pixels8_l2_mmxext(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dstStride, int src1Stride, int h)
#define src1
Definition: h264pred.c:139
av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth)
Definition: h264_qpel.c:538
void ff_put_pixels16_l2_mmxext(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dstStride, int src1Stride, int h)
#define EXTERNAL_SSSE3(flags)
Definition: cpu.h:63
#define H264_QPEL_FUNCS_10(x, y, CPU)
Definition: h264_qpel.c:530
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
Definition: cpu.c:76
#define EXTERNAL_MMXEXT(flags)
Definition: cpu.h:55
#define avg(a, b, c, d)
GLint GLenum GLboolean GLsizei stride
Definition: opengl_enc.c:105
static double c[64]
static unsigned bit_depth(uint64_t mask)
Definition: af_astats.c:142
void ff_put_pixels16_sse2(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h)
static uint8_t tmp[8]
Definition: des.c:38
#define av_always_inline
Definition: attributes.h:39
#define H264_QPEL_FUNCS(x, y, CPU)
Definition: h264_qpel.c:522
#define EXTERNAL_AVX(flags)
Definition: cpu.h:66
void ff_avg_pixels16_l2_mmxext(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dstStride, int src1Stride, int h)