1 /*
2 * This file is part of FFmpeg.
3 *
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
8 *
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
13 *
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19 #include "config.h"
20
21 #include "pixelutils.h"
22 #include "cpu.h"
23
24 int ff_pixelutils_sad_8x8_mmx(const uint8_t *src1, ptrdiff_t stride1,
25 const uint8_t *src2, ptrdiff_t stride2);
26 int ff_pixelutils_sad_8x8_mmxext(const uint8_t *src1, ptrdiff_t stride1,
27 const uint8_t *src2, ptrdiff_t stride2);
28
29 int ff_pixelutils_sad_16x16_mmxext(const uint8_t *src1, ptrdiff_t stride1,
30 const uint8_t *src2, ptrdiff_t stride2);
31 int ff_pixelutils_sad_16x16_sse2(const uint8_t *src1, ptrdiff_t stride1,
32 const uint8_t *src2, ptrdiff_t stride2);
33 int ff_pixelutils_sad_a_16x16_sse2(const uint8_t *src1, ptrdiff_t stride1,
34 const uint8_t *src2, ptrdiff_t stride2);
35 int ff_pixelutils_sad_u_16x16_sse2(const uint8_t *src1, ptrdiff_t stride1,
36 const uint8_t *src2, ptrdiff_t stride2);
37
38 int ff_pixelutils_sad_32x32_sse2(const uint8_t *src1, ptrdiff_t stride1,
39 const uint8_t *src2, ptrdiff_t stride2);
40 int ff_pixelutils_sad_a_32x32_sse2(const uint8_t *src1, ptrdiff_t stride1,
41 const uint8_t *src2, ptrdiff_t stride2);
42 int ff_pixelutils_sad_u_32x32_sse2(const uint8_t *src1, ptrdiff_t stride1,
43 const uint8_t *src2, ptrdiff_t stride2);
44
45 int ff_pixelutils_sad_32x32_avx2(const uint8_t *src1, ptrdiff_t stride1,
46 const uint8_t *src2, ptrdiff_t stride2);
47 int ff_pixelutils_sad_a_32x32_avx2(const uint8_t *src1, ptrdiff_t stride1,
48 const uint8_t *src2, ptrdiff_t stride2);
49 int ff_pixelutils_sad_u_32x32_avx2(const uint8_t *src1, ptrdiff_t stride1,
50 const uint8_t *src2, ptrdiff_t stride2);
51
ff_pixelutils_sad_init_x86(av_pixelutils_sad_fn * sad,int aligned)52 void ff_pixelutils_sad_init_x86(av_pixelutils_sad_fn *sad, int aligned)
53 {
54 int cpu_flags = av_get_cpu_flags();
55
56 if (EXTERNAL_MMX(cpu_flags)) {
57 sad[2] = ff_pixelutils_sad_8x8_mmx;
58 }
59
60 // The best way to use SSE2 would be to do 2 SADs in parallel,
61 // but we'd have to modify the pixelutils API to return SIMD functions.
62
63 // It's probably not faster to shuffle data around
64 // to get two lines of 8 pixels into a single 16byte register,
65 // so just use the MMX 8x8 version even when SSE2 is available.
66 if (EXTERNAL_MMXEXT(cpu_flags)) {
67 sad[2] = ff_pixelutils_sad_8x8_mmxext;
68 sad[3] = ff_pixelutils_sad_16x16_mmxext;
69 }
70
71 if (EXTERNAL_SSE2(cpu_flags)) {
72 switch (aligned) {
73 case 0: sad[3] = ff_pixelutils_sad_16x16_sse2; break; // src1 unaligned, src2 unaligned
74 case 1: sad[3] = ff_pixelutils_sad_u_16x16_sse2; break; // src1 aligned, src2 unaligned
75 case 2: sad[3] = ff_pixelutils_sad_a_16x16_sse2; break; // src1 aligned, src2 aligned
76 }
77 }
78
79 if (EXTERNAL_SSE2(cpu_flags)) {
80 switch (aligned) {
81 case 0: sad[4] = ff_pixelutils_sad_32x32_sse2; break; // src1 unaligned, src2 unaligned
82 case 1: sad[4] = ff_pixelutils_sad_u_32x32_sse2; break; // src1 aligned, src2 unaligned
83 case 2: sad[4] = ff_pixelutils_sad_a_32x32_sse2; break; // src1 aligned, src2 aligned
84 }
85 }
86
87 if (EXTERNAL_AVX2_FAST(cpu_flags)) {
88 switch (aligned) {
89 case 0: sad[4] = ff_pixelutils_sad_32x32_avx2; break; // src1 unaligned, src2 unaligned
90 case 1: sad[4] = ff_pixelutils_sad_u_32x32_avx2; break; // src1 aligned, src2 unaligned
91 case 2: sad[4] = ff_pixelutils_sad_a_32x32_avx2; break; // src1 aligned, src2 aligned
92 }
93 }
94 }
95