GCC Code Coverage Report
Directory: ../../../ffmpeg/ Exec Total Coverage
File: src/libavcodec/x86/h264_qpel.c Lines: 100 100 100.0 %
Date: 2020-09-25 23:16:12 Branches: 35 36 97.2 %

Line Branch Exec Source
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/h264dec.h"
27
#include "libavcodec/h264qpel.h"
28
#include "libavcodec/pixels.h"
29
#include "fpel.h"
30
31
#if HAVE_X86ASM
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
40
static av_always_inline void ff_put_h264_qpel8or16_hv1_lowpass_mmxext(int16_t *tmp, const uint8_t *src, int tmpStride, int srcStride, int size)
75
{
76
40
    int w = (size + 8) >> 2;
77
40
    src -= 2 * srcStride + 2;
78
240
    while (w--) {
79
200
        ff_put_h264_qpel8or16_hv1_lowpass_op_mmxext(src, tmp, srcStride, size);
80
200
        tmp += 4;
81
200
        src += 4;
82
    }
83
40
}
84
85
#define QPEL_H264(OPNAME, OP, MMX)\
86
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){\
87
    int w=3;\
88
    src -= 2*srcStride+2;\
89
    while(w--){\
90
        ff_ ## OPNAME ## h264_qpel4_hv_lowpass_v_mmxext(src, tmp, srcStride);\
91
        tmp += 4;\
92
        src += 4;\
93
    }\
94
    tmp -= 3*4;\
95
    ff_ ## OPNAME ## h264_qpel4_hv_lowpass_h_mmxext(tmp, dst, dstStride);\
96
}\
97
\
98
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){\
99
    src -= 2*srcStride;\
100
    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_op_mmxext(dst, src, dstStride, srcStride, h);\
101
    src += 4;\
102
    dst += 4;\
103
    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_op_mmxext(dst, src, dstStride, srcStride, h);\
104
}\
105
static av_always_inline void ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, int dstStride, int tmpStride, int size){\
106
    int w = size>>4;\
107
    do{\
108
    ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_op_mmxext(dst, tmp, dstStride, 0, size);\
109
    tmp += 8;\
110
    dst += 8;\
111
    }while(w--);\
112
}\
113
\
114
static av_always_inline void ff_ ## OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
115
    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 8);\
116
}\
117
static av_always_inline void ff_ ## OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
118
    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 16);\
119
    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
120
}\
121
\
122
static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
123
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
124
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
125
    src += 8*srcStride;\
126
    dst += 8*dstStride;\
127
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
128
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
129
}\
130
\
131
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){\
132
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
133
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
134
    src += 8*dstStride;\
135
    dst += 8*dstStride;\
136
    src2 += 8*src2Stride;\
137
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
138
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
139
}\
140
\
141
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){\
142
    ff_put_h264_qpel8or16_hv1_lowpass_ ## MMX(tmp, src, tmpStride, srcStride, size);\
143
    ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(dst, tmp, dstStride, tmpStride, size);\
144
}\
145
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){\
146
    ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride, 8);\
147
}\
148
\
149
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){\
150
    ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride, 16);\
151
}\
152
\
153
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)\
154
{\
155
    ff_ ## OPNAME ## pixels8_l2_shift5_ ## MMX(dst  , src16  , src8  , dstStride, src8Stride, h);\
156
    ff_ ## OPNAME ## pixels8_l2_shift5_ ## MMX(dst+8, src16+8, src8+8, dstStride, src8Stride, h);\
157
}\
158
159
160
#if ARCH_X86_64
161
#define QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
162
163
void ff_avg_h264_qpel16_h_lowpass_l2_ssse3(uint8_t *dst, const uint8_t *src, const uint8_t *src2, int dstStride, int src2Stride);
164
void ff_put_h264_qpel16_h_lowpass_l2_ssse3(uint8_t *dst, const uint8_t *src, const uint8_t *src2, int dstStride, int src2Stride);
165
166
#else // ARCH_X86_64
167
#define QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
168
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){\
169
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
170
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
171
    src += 8*dstStride;\
172
    dst += 8*dstStride;\
173
    src2 += 8*src2Stride;\
174
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
175
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
176
}
177
#endif // ARCH_X86_64
178
179
#define QPEL_H264_H_XMM(OPNAME, OP, MMX)\
180
QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
181
static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
182
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
183
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
184
    src += 8*srcStride;\
185
    dst += 8*dstStride;\
186
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
187
    ff_ ## OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
188
}\
189
190
#define QPEL_H264_V_XMM(OPNAME, OP, MMX)\
191
static av_always_inline void ff_ ## OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
192
    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 8);\
193
}\
194
static av_always_inline void ff_ ## OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride){\
195
    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 16);\
196
    ff_ ## OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
197
}
198
199
74002
static av_always_inline void put_h264_qpel8or16_hv1_lowpass_sse2(int16_t *tmp,
200
                                                                 const uint8_t *src,
201
                                                                 int tmpStride,
202
                                                                 int srcStride,
203
                                                                 int size)
204
{
205
74002
    int w = (size+8)>>3;
206
74002
    src -= 2*srcStride+2;
207
240088
    while(w--){
208
166086
        ff_put_h264_qpel8or16_hv1_lowpass_op_sse2(src, tmp, srcStride, size);
209
166086
        tmp += 8;
210
166086
        src += 8;
211
    }
212
74002
}
213
214
#define QPEL_H264_HV_XMM(OPNAME, OP, MMX)\
215
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){\
216
    put_h264_qpel8or16_hv1_lowpass_sse2(tmp, src, tmpStride, srcStride, size);\
217
    ff_ ## OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(dst, tmp, dstStride, tmpStride, size);\
218
}\
219
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){\
220
    ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst, tmp, src, dstStride, tmpStride, srcStride, 8);\
221
}\
222
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){\
223
    ff_ ## OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(dst, tmp, src, dstStride, tmpStride, srcStride, 16);\
224
}\
225
226
#define ff_put_h264_qpel8_h_lowpass_l2_sse2  ff_put_h264_qpel8_h_lowpass_l2_mmxext
227
#define ff_avg_h264_qpel8_h_lowpass_l2_sse2  ff_avg_h264_qpel8_h_lowpass_l2_mmxext
228
#define ff_put_h264_qpel16_h_lowpass_l2_sse2 ff_put_h264_qpel16_h_lowpass_l2_mmxext
229
#define ff_avg_h264_qpel16_h_lowpass_l2_sse2 ff_avg_h264_qpel16_h_lowpass_l2_mmxext
230
231
#define ff_put_h264_qpel8_v_lowpass_ssse3  ff_put_h264_qpel8_v_lowpass_sse2
232
#define ff_avg_h264_qpel8_v_lowpass_ssse3  ff_avg_h264_qpel8_v_lowpass_sse2
233
#define ff_put_h264_qpel16_v_lowpass_ssse3 ff_put_h264_qpel16_v_lowpass_sse2
234
#define ff_avg_h264_qpel16_v_lowpass_ssse3 ff_avg_h264_qpel16_v_lowpass_sse2
235
236
#define ff_put_h264_qpel8or16_hv2_lowpass_sse2 ff_put_h264_qpel8or16_hv2_lowpass_mmxext
237
#define ff_avg_h264_qpel8or16_hv2_lowpass_sse2 ff_avg_h264_qpel8or16_hv2_lowpass_mmxext
238
239
#define H264_MC(OPNAME, SIZE, MMX, ALIGN) \
240
H264_MC_C(OPNAME, SIZE, MMX, ALIGN)\
241
H264_MC_V(OPNAME, SIZE, MMX, ALIGN)\
242
H264_MC_H(OPNAME, SIZE, MMX, ALIGN)\
243
H264_MC_HV(OPNAME, SIZE, MMX, ALIGN)\
244
245
37721
static void put_h264_qpel16_mc00_sse2 (uint8_t *dst, const uint8_t *src,
246
                                       ptrdiff_t stride)
247
{
248
37721
    ff_put_pixels16_sse2(dst, src, stride, 16);
249
37721
}
250
13426
static void avg_h264_qpel16_mc00_sse2 (uint8_t *dst, const uint8_t *src,
251
                                       ptrdiff_t stride)
252
{
253
13426
    ff_avg_pixels16_sse2(dst, src, stride, 16);
254
13426
}
255
#define put_h264_qpel8_mc00_sse2 put_h264_qpel8_mc00_mmxext
256
#define avg_h264_qpel8_mc00_sse2 avg_h264_qpel8_mc00_mmxext
257
258
#define H264_MC_C(OPNAME, SIZE, MMX, ALIGN) \
259
static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## MMX (uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
260
{\
261
    ff_ ## OPNAME ## pixels ## SIZE ## _ ## MMX(dst, src, stride, SIZE);\
262
}\
263
264
#define H264_MC_H(OPNAME, SIZE, MMX, ALIGN) \
265
static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
266
{\
267
    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src, stride, stride);\
268
}\
269
\
270
static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
271
{\
272
    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## MMX(dst, src, stride, stride);\
273
}\
274
\
275
static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
276
{\
277
    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, src+1, stride, stride);\
278
}\
279
280
#define H264_MC_V(OPNAME, SIZE, MMX, ALIGN) \
281
static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
282
{\
283
    LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
284
    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
285
    ff_ ## OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src, temp, stride, stride, SIZE);\
286
}\
287
\
288
static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
289
{\
290
    ff_ ## OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## MMX(dst, src, stride, stride);\
291
}\
292
\
293
static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
294
{\
295
    LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
296
    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
297
    ff_ ## OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src+stride, temp, stride, stride, SIZE);\
298
}\
299
300
#define H264_MC_HV(OPNAME, SIZE, MMX, ALIGN) \
301
static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
302
{\
303
    LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
304
    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
305
    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
306
}\
307
\
308
static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
309
{\
310
    LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
311
    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
312
    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, temp, stride, SIZE);\
313
}\
314
\
315
static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
316
{\
317
    LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
318
    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src, SIZE, stride);\
319
    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
320
}\
321
\
322
static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
323
{\
324
    LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*SIZE]);\
325
    ff_put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(temp, src+1, SIZE, stride);\
326
    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, temp, stride, SIZE);\
327
}\
328
\
329
static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
330
{\
331
    LOCAL_ALIGNED(ALIGN, uint16_t, temp, [SIZE*(SIZE<8?12:24)]);\
332
    ff_ ## OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(dst, temp, src, stride, SIZE, stride);\
333
}\
334
\
335
static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
336
{\
337
    LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
338
    uint8_t * const halfHV= temp;\
339
    int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
340
    av_assert2(((int)temp & 7) == 0);\
341
    ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
342
    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src, halfHV, stride, SIZE);\
343
}\
344
\
345
static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
346
{\
347
    LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
348
    uint8_t * const halfHV= temp;\
349
    int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
350
    av_assert2(((int)temp & 7) == 0);\
351
    ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
352
    ff_ ## OPNAME ## h264_qpel ## SIZE ## _h_lowpass_l2_ ## MMX(dst, src+stride, halfHV, stride, SIZE);\
353
}\
354
\
355
static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
356
{\
357
    LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
358
    uint8_t * const halfHV= temp;\
359
    int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
360
    av_assert2(((int)temp & 7) == 0);\
361
    ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
362
    ff_ ## OPNAME ## pixels ## SIZE ## _l2_shift5_mmxext(dst, halfV+2, halfHV, stride, SIZE, SIZE);\
363
}\
364
\
365
static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)\
366
{\
367
    LOCAL_ALIGNED(ALIGN, uint8_t, temp, [SIZE*(SIZE<8?12:24)*2 + SIZE*SIZE]);\
368
    uint8_t * const halfHV= temp;\
369
    int16_t * const halfV= (int16_t*)(temp + SIZE*SIZE);\
370
    av_assert2(((int)temp & 7) == 0);\
371
    ff_put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, halfV, src, SIZE, SIZE, stride);\
372
    ff_ ## OPNAME ## pixels ## SIZE ## _l2_shift5_mmxext(dst, halfV+3, halfHV, stride, SIZE, SIZE);\
373
}\
374
375
#define H264_MC_4816(MMX)\
376
H264_MC(put_, 4, MMX, 8)\
377
H264_MC(put_, 8, MMX, 8)\
378
H264_MC(put_, 16,MMX, 8)\
379
H264_MC(avg_, 4, MMX, 8)\
380
H264_MC(avg_, 8, MMX, 8)\
381
H264_MC(avg_, 16,MMX, 8)\
382
383
#define H264_MC_816(QPEL, XMM)\
384
QPEL(put_, 8, XMM, 16)\
385
QPEL(put_, 16,XMM, 16)\
386
QPEL(avg_, 8, XMM, 16)\
387
QPEL(avg_, 16,XMM, 16)\
388
389
16326
QPEL_H264(put_,        PUT_OP, mmxext)
390
207
QPEL_H264(avg_, AVG_MMXEXT_OP, mmxext)
391
150940
QPEL_H264_V_XMM(put_,       PUT_OP, sse2)
392
208
QPEL_H264_V_XMM(avg_,AVG_MMXEXT_OP, sse2)
393
72
QPEL_H264_HV_XMM(put_,       PUT_OP, sse2)
394
8
QPEL_H264_HV_XMM(avg_,AVG_MMXEXT_OP, sse2)
395
5711
QPEL_H264_H_XMM(put_,       PUT_OP, ssse3)
396
624
QPEL_H264_H_XMM(avg_,AVG_MMXEXT_OP, ssse3)
397
295488
QPEL_H264_HV_XMM(put_,       PUT_OP, ssse3)
398
180
QPEL_H264_HV_XMM(avg_,AVG_MMXEXT_OP, ssse3)
399
400
51826
H264_MC_4816(mmxext)
401
49298
H264_MC_816(H264_MC_V, sse2)
402
72
H264_MC_816(H264_MC_HV, sse2)
403
60354
H264_MC_816(H264_MC_H, ssse3)
404
175780
H264_MC_816(H264_MC_HV, ssse3)
405
406
407
//10bit
408
#define LUMA_MC_OP(OP, NUM, DEPTH, TYPE, OPT) \
409
void ff_ ## OP ## _h264_qpel ## NUM ## _ ## TYPE ## _ ## DEPTH ## _ ## OPT \
410
    (uint8_t *dst, const uint8_t *src, ptrdiff_t stride);
411
412
#define LUMA_MC_ALL(DEPTH, TYPE, OPT) \
413
    LUMA_MC_OP(put,  4, DEPTH, TYPE, OPT) \
414
    LUMA_MC_OP(avg,  4, DEPTH, TYPE, OPT) \
415
    LUMA_MC_OP(put,  8, DEPTH, TYPE, OPT) \
416
    LUMA_MC_OP(avg,  8, DEPTH, TYPE, OPT) \
417
    LUMA_MC_OP(put, 16, DEPTH, TYPE, OPT) \
418
    LUMA_MC_OP(avg, 16, DEPTH, TYPE, OPT)
419
420
#define LUMA_MC_816(DEPTH, TYPE, OPT) \
421
    LUMA_MC_OP(put,  8, DEPTH, TYPE, OPT) \
422
    LUMA_MC_OP(avg,  8, DEPTH, TYPE, OPT) \
423
    LUMA_MC_OP(put, 16, DEPTH, TYPE, OPT) \
424
    LUMA_MC_OP(avg, 16, DEPTH, TYPE, OPT)
425
426
LUMA_MC_ALL(10, mc00, mmxext)
427
LUMA_MC_ALL(10, mc10, mmxext)
428
LUMA_MC_ALL(10, mc20, mmxext)
429
LUMA_MC_ALL(10, mc30, mmxext)
430
LUMA_MC_ALL(10, mc01, mmxext)
431
LUMA_MC_ALL(10, mc11, mmxext)
432
LUMA_MC_ALL(10, mc21, mmxext)
433
LUMA_MC_ALL(10, mc31, mmxext)
434
LUMA_MC_ALL(10, mc02, mmxext)
435
LUMA_MC_ALL(10, mc12, mmxext)
436
LUMA_MC_ALL(10, mc22, mmxext)
437
LUMA_MC_ALL(10, mc32, mmxext)
438
LUMA_MC_ALL(10, mc03, mmxext)
439
LUMA_MC_ALL(10, mc13, mmxext)
440
LUMA_MC_ALL(10, mc23, mmxext)
441
LUMA_MC_ALL(10, mc33, mmxext)
442
443
LUMA_MC_816(10, mc00, sse2)
444
LUMA_MC_816(10, mc10, sse2)
445
LUMA_MC_816(10, mc10, sse2_cache64)
446
LUMA_MC_816(10, mc10, ssse3_cache64)
447
LUMA_MC_816(10, mc20, sse2)
448
LUMA_MC_816(10, mc20, sse2_cache64)
449
LUMA_MC_816(10, mc20, ssse3_cache64)
450
LUMA_MC_816(10, mc30, sse2)
451
LUMA_MC_816(10, mc30, sse2_cache64)
452
LUMA_MC_816(10, mc30, ssse3_cache64)
453
LUMA_MC_816(10, mc01, sse2)
454
LUMA_MC_816(10, mc11, sse2)
455
LUMA_MC_816(10, mc21, sse2)
456
LUMA_MC_816(10, mc31, sse2)
457
LUMA_MC_816(10, mc02, sse2)
458
LUMA_MC_816(10, mc12, sse2)
459
LUMA_MC_816(10, mc22, sse2)
460
LUMA_MC_816(10, mc32, sse2)
461
LUMA_MC_816(10, mc03, sse2)
462
LUMA_MC_816(10, mc13, sse2)
463
LUMA_MC_816(10, mc23, sse2)
464
LUMA_MC_816(10, mc33, sse2)
465
466
#define QPEL16_OPMC(OP, MC, MMX)\
467
void ff_ ## OP ## _h264_qpel16_ ## MC ## _10_ ## MMX(uint8_t *dst, const uint8_t *src, ptrdiff_t stride){\
468
    ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst   , src   , stride);\
469
    ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst+16, src+16, stride);\
470
    src += 8*stride;\
471
    dst += 8*stride;\
472
    ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst   , src   , stride);\
473
    ff_ ## OP ## _h264_qpel8_ ## MC ## _10_ ## MMX(dst+16, src+16, stride);\
474
}
475
476
#define QPEL16_OP(MC, MMX)\
477
QPEL16_OPMC(put, MC, MMX)\
478
QPEL16_OPMC(avg, MC, MMX)
479
480
#define QPEL16(MMX)\
481
QPEL16_OP(mc00, MMX)\
482
QPEL16_OP(mc01, MMX)\
483
QPEL16_OP(mc02, MMX)\
484
QPEL16_OP(mc03, MMX)\
485
QPEL16_OP(mc10, MMX)\
486
QPEL16_OP(mc11, MMX)\
487
QPEL16_OP(mc12, MMX)\
488
QPEL16_OP(mc13, MMX)\
489
QPEL16_OP(mc20, MMX)\
490
QPEL16_OP(mc21, MMX)\
491
QPEL16_OP(mc22, MMX)\
492
QPEL16_OP(mc23, MMX)\
493
QPEL16_OP(mc30, MMX)\
494
QPEL16_OP(mc31, MMX)\
495
QPEL16_OP(mc32, MMX)\
496
QPEL16_OP(mc33, MMX)
497
498
#if ARCH_X86_32 // ARCH_X86_64 implies SSE2+
499
QPEL16(mmxext)
500
#endif
501
502
#endif /* HAVE_X86ASM */
503
504
#define SET_QPEL_FUNCS(PFX, IDX, SIZE, CPU, PREFIX)                          \
505
    do {                                                                     \
506
    c->PFX ## _pixels_tab[IDX][ 0] = PREFIX ## PFX ## SIZE ## _mc00_ ## CPU; \
507
    c->PFX ## _pixels_tab[IDX][ 1] = PREFIX ## PFX ## SIZE ## _mc10_ ## CPU; \
508
    c->PFX ## _pixels_tab[IDX][ 2] = PREFIX ## PFX ## SIZE ## _mc20_ ## CPU; \
509
    c->PFX ## _pixels_tab[IDX][ 3] = PREFIX ## PFX ## SIZE ## _mc30_ ## CPU; \
510
    c->PFX ## _pixels_tab[IDX][ 4] = PREFIX ## PFX ## SIZE ## _mc01_ ## CPU; \
511
    c->PFX ## _pixels_tab[IDX][ 5] = PREFIX ## PFX ## SIZE ## _mc11_ ## CPU; \
512
    c->PFX ## _pixels_tab[IDX][ 6] = PREFIX ## PFX ## SIZE ## _mc21_ ## CPU; \
513
    c->PFX ## _pixels_tab[IDX][ 7] = PREFIX ## PFX ## SIZE ## _mc31_ ## CPU; \
514
    c->PFX ## _pixels_tab[IDX][ 8] = PREFIX ## PFX ## SIZE ## _mc02_ ## CPU; \
515
    c->PFX ## _pixels_tab[IDX][ 9] = PREFIX ## PFX ## SIZE ## _mc12_ ## CPU; \
516
    c->PFX ## _pixels_tab[IDX][10] = PREFIX ## PFX ## SIZE ## _mc22_ ## CPU; \
517
    c->PFX ## _pixels_tab[IDX][11] = PREFIX ## PFX ## SIZE ## _mc32_ ## CPU; \
518
    c->PFX ## _pixels_tab[IDX][12] = PREFIX ## PFX ## SIZE ## _mc03_ ## CPU; \
519
    c->PFX ## _pixels_tab[IDX][13] = PREFIX ## PFX ## SIZE ## _mc13_ ## CPU; \
520
    c->PFX ## _pixels_tab[IDX][14] = PREFIX ## PFX ## SIZE ## _mc23_ ## CPU; \
521
    c->PFX ## _pixels_tab[IDX][15] = PREFIX ## PFX ## SIZE ## _mc33_ ## CPU; \
522
    } while (0)
523
524
#define H264_QPEL_FUNCS(x, y, CPU)                                                            \
525
    do {                                                                                      \
526
        c->put_h264_qpel_pixels_tab[0][x + y * 4] = put_h264_qpel16_mc ## x ## y ## _ ## CPU; \
527
        c->put_h264_qpel_pixels_tab[1][x + y * 4] = put_h264_qpel8_mc  ## x ## y ## _ ## CPU; \
528
        c->avg_h264_qpel_pixels_tab[0][x + y * 4] = avg_h264_qpel16_mc ## x ## y ## _ ## CPU; \
529
        c->avg_h264_qpel_pixels_tab[1][x + y * 4] = avg_h264_qpel8_mc  ## x ## y ## _ ## CPU; \
530
    } while (0)
531
532
#define H264_QPEL_FUNCS_10(x, y, CPU)                                                               \
533
    do {                                                                                            \
534
        c->put_h264_qpel_pixels_tab[0][x + y * 4] = ff_put_h264_qpel16_mc ## x ## y ## _10_ ## CPU; \
535
        c->put_h264_qpel_pixels_tab[1][x + y * 4] = ff_put_h264_qpel8_mc  ## x ## y ## _10_ ## CPU; \
536
        c->avg_h264_qpel_pixels_tab[0][x + y * 4] = ff_avg_h264_qpel16_mc ## x ## y ## _10_ ## CPU; \
537
        c->avg_h264_qpel_pixels_tab[1][x + y * 4] = ff_avg_h264_qpel8_mc  ## x ## y ## _10_ ## CPU; \
538
    } while (0)
539
540
652
av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth)
541
{
542
#if HAVE_X86ASM
543
652
    int high_bit_depth = bit_depth > 8;
544
652
    int cpu_flags = av_get_cpu_flags();
545
546
652
    if (EXTERNAL_MMXEXT(cpu_flags)) {
547
100
        if (!high_bit_depth) {
548
56
            SET_QPEL_FUNCS(put_h264_qpel, 0, 16, mmxext, );
549
56
            SET_QPEL_FUNCS(put_h264_qpel, 1,  8, mmxext, );
550
56
            SET_QPEL_FUNCS(put_h264_qpel, 2,  4, mmxext, );
551
56
            SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, mmxext, );
552
56
            SET_QPEL_FUNCS(avg_h264_qpel, 1,  8, mmxext, );
553
56
            SET_QPEL_FUNCS(avg_h264_qpel, 2,  4, mmxext, );
554
44
        } else if (bit_depth == 10) {
555
#if ARCH_X86_32
556
            SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_mmxext, ff_);
557
            SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_mmxext, ff_);
558
            SET_QPEL_FUNCS(put_h264_qpel, 1,  8, 10_mmxext, ff_);
559
            SET_QPEL_FUNCS(avg_h264_qpel, 1,  8, 10_mmxext, ff_);
560
#endif
561
22
            SET_QPEL_FUNCS(put_h264_qpel, 2, 4,  10_mmxext, ff_);
562
22
            SET_QPEL_FUNCS(avg_h264_qpel, 2, 4,  10_mmxext, ff_);
563
        }
564
    }
565
566
652
    if (EXTERNAL_SSE2(cpu_flags)) {
567
88
        if (!high_bit_depth) {
568
52
            H264_QPEL_FUNCS(0, 1, sse2);
569
52
            H264_QPEL_FUNCS(0, 2, sse2);
570
52
            H264_QPEL_FUNCS(0, 3, sse2);
571
52
            H264_QPEL_FUNCS(1, 1, sse2);
572
52
            H264_QPEL_FUNCS(1, 2, sse2);
573
52
            H264_QPEL_FUNCS(1, 3, sse2);
574
52
            H264_QPEL_FUNCS(2, 1, sse2);
575
52
            H264_QPEL_FUNCS(2, 2, sse2);
576
52
            H264_QPEL_FUNCS(2, 3, sse2);
577
52
            H264_QPEL_FUNCS(3, 1, sse2);
578
52
            H264_QPEL_FUNCS(3, 2, sse2);
579
52
            H264_QPEL_FUNCS(3, 3, sse2);
580
        }
581
582
88
        if (bit_depth == 10) {
583
18
            SET_QPEL_FUNCS(put_h264_qpel, 0, 16, 10_sse2, ff_);
584
18
            SET_QPEL_FUNCS(put_h264_qpel, 1,  8, 10_sse2, ff_);
585
18
            SET_QPEL_FUNCS(avg_h264_qpel, 0, 16, 10_sse2, ff_);
586
18
            SET_QPEL_FUNCS(avg_h264_qpel, 1,  8, 10_sse2, ff_);
587
18
            H264_QPEL_FUNCS_10(1, 0, sse2_cache64);
588
18
            H264_QPEL_FUNCS_10(2, 0, sse2_cache64);
589
18
            H264_QPEL_FUNCS_10(3, 0, sse2_cache64);
590
        }
591
    }
592
593

652
    if (EXTERNAL_SSE2_FAST(cpu_flags)) {
594
88
        if (!high_bit_depth) {
595
52
            H264_QPEL_FUNCS(0, 0, sse2);
596
        }
597
    }
598
599
652
    if (EXTERNAL_SSSE3(cpu_flags)) {
600
76
        if (!high_bit_depth) {
601
48
            H264_QPEL_FUNCS(1, 0, ssse3);
602
48
            H264_QPEL_FUNCS(1, 1, ssse3);
603
48
            H264_QPEL_FUNCS(1, 2, ssse3);
604
48
            H264_QPEL_FUNCS(1, 3, ssse3);
605
48
            H264_QPEL_FUNCS(2, 0, ssse3);
606
48
            H264_QPEL_FUNCS(2, 1, ssse3);
607
48
            H264_QPEL_FUNCS(2, 2, ssse3);
608
48
            H264_QPEL_FUNCS(2, 3, ssse3);
609
48
            H264_QPEL_FUNCS(3, 0, ssse3);
610
48
            H264_QPEL_FUNCS(3, 1, ssse3);
611
48
            H264_QPEL_FUNCS(3, 2, ssse3);
612
48
            H264_QPEL_FUNCS(3, 3, ssse3);
613
        }
614
615
76
        if (bit_depth == 10) {
616
14
            H264_QPEL_FUNCS_10(1, 0, ssse3_cache64);
617
14
            H264_QPEL_FUNCS_10(2, 0, ssse3_cache64);
618
14
            H264_QPEL_FUNCS_10(3, 0, ssse3_cache64);
619
        }
620
    }
621
622
652
    if (EXTERNAL_AVX(cpu_flags)) {
623
        /* AVX implies 64 byte cache lines without the need to avoid unaligned
624
         * memory accesses that cross the boundary between two cache lines.
625
         * TODO: Port X264_CPU_CACHELINE_32/64 detection from x264 to avoid
626
         * having to treat SSE2 functions with such properties as AVX. */
627
52
        if (bit_depth == 10) {
628
6
            H264_QPEL_FUNCS_10(1, 0, sse2);
629
6
            H264_QPEL_FUNCS_10(2, 0, sse2);
630
6
            H264_QPEL_FUNCS_10(3, 0, sse2);
631
        }
632
    }
633
#endif
634
652
}