mpegvideoenc.c 8.66 KB
Newer Older
1 2 3 4
/*
 * The simplest mpeg encoder (well, it was the simplest!)
 * Copyright (c) 2000,2001 Fabrice Bellard
 *
5
 * This file is part of FFmpeg.
6
 *
7
 * FFmpeg is free software; you can redistribute it and/or
8 9 10 11
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * 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 14 15 16 17
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * 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 20 21
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

22
#include "libavutil/attributes.h"
23 24
#include "libavutil/cpu.h"
#include "libavutil/x86/asm.h"
25
#include "libavutil/x86/cpu.h"
26
#include "libavcodec/avcodec.h"
27
#include "libavcodec/dct.h"
28 29
#include "libavcodec/mpegvideo.h"

30 31
/* not permutated inverse zigzag_direct + 1 for MMX quantizer */
DECLARE_ALIGNED(16, static uint16_t, inv_zigzag_direct16)[64];
32

33 34
#if HAVE_6REGS

35
#if HAVE_MMX_INLINE
36 37 38
#define COMPILE_TEMPLATE_MMXEXT 0
#define COMPILE_TEMPLATE_SSE2   0
#define COMPILE_TEMPLATE_SSSE3  0
39 40
#define RENAME(a)      a ## _mmx
#define RENAME_FDCT(a) a ## _mmx
41
#include "mpegvideoenc_template.c"
42
#endif /* HAVE_MMX_INLINE */
43

44
#if HAVE_MMXEXT_INLINE
45 46 47 48 49 50
#undef COMPILE_TEMPLATE_SSSE3
#undef COMPILE_TEMPLATE_SSE2
#undef COMPILE_TEMPLATE_MMXEXT
#define COMPILE_TEMPLATE_MMXEXT 1
#define COMPILE_TEMPLATE_SSE2   0
#define COMPILE_TEMPLATE_SSSE3  0
51
#undef RENAME
52 53 54
#undef RENAME_FDCT
#define RENAME(a)      a ## _mmxext
#define RENAME_FDCT(a) a ## _mmxext
55
#include "mpegvideoenc_template.c"
56
#endif /* HAVE_MMXEXT_INLINE */
57

58
#if HAVE_SSE2_INLINE
59 60 61 62 63 64
#undef COMPILE_TEMPLATE_MMXEXT
#undef COMPILE_TEMPLATE_SSE2
#undef COMPILE_TEMPLATE_SSSE3
#define COMPILE_TEMPLATE_MMXEXT 0
#define COMPILE_TEMPLATE_SSE2   1
#define COMPILE_TEMPLATE_SSSE3  0
65
#undef RENAME
66 67 68
#undef RENAME_FDCT
#define RENAME(a)      a ## _sse2
#define RENAME_FDCT(a) a ## _sse2
69
#include "mpegvideoenc_template.c"
70
#endif /* HAVE_SSE2_INLINE */
71

72
#if HAVE_SSSE3_INLINE
73 74 75 76 77 78
#undef COMPILE_TEMPLATE_MMXEXT
#undef COMPILE_TEMPLATE_SSE2
#undef COMPILE_TEMPLATE_SSSE3
#define COMPILE_TEMPLATE_MMXEXT 0
#define COMPILE_TEMPLATE_SSE2   1
#define COMPILE_TEMPLATE_SSSE3  1
79
#undef RENAME
80 81 82
#undef RENAME_FDCT
#define RENAME(a)      a ## _ssse3
#define RENAME_FDCT(a) a ## _sse2
83
#include "mpegvideoenc_template.c"
84
#endif /* HAVE_SSSE3_INLINE */
85

86 87
#endif /* HAVE_6REGS */

88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199
#if HAVE_INLINE_ASM
static void  denoise_dct_mmx(MpegEncContext *s, int16_t *block){
    const int intra= s->mb_intra;
    int *sum= s->dct_error_sum[intra];
    uint16_t *offset= s->dct_offset[intra];

    s->dct_count[intra]++;

    __asm__ volatile(
        "pxor %%mm7, %%mm7                      \n\t"
        "1:                                     \n\t"
        "pxor %%mm0, %%mm0                      \n\t"
        "pxor %%mm1, %%mm1                      \n\t"
        "movq (%0), %%mm2                       \n\t"
        "movq 8(%0), %%mm3                      \n\t"
        "pcmpgtw %%mm2, %%mm0                   \n\t"
        "pcmpgtw %%mm3, %%mm1                   \n\t"
        "pxor %%mm0, %%mm2                      \n\t"
        "pxor %%mm1, %%mm3                      \n\t"
        "psubw %%mm0, %%mm2                     \n\t"
        "psubw %%mm1, %%mm3                     \n\t"
        "movq %%mm2, %%mm4                      \n\t"
        "movq %%mm3, %%mm5                      \n\t"
        "psubusw (%2), %%mm2                    \n\t"
        "psubusw 8(%2), %%mm3                   \n\t"
        "pxor %%mm0, %%mm2                      \n\t"
        "pxor %%mm1, %%mm3                      \n\t"
        "psubw %%mm0, %%mm2                     \n\t"
        "psubw %%mm1, %%mm3                     \n\t"
        "movq %%mm2, (%0)                       \n\t"
        "movq %%mm3, 8(%0)                      \n\t"
        "movq %%mm4, %%mm2                      \n\t"
        "movq %%mm5, %%mm3                      \n\t"
        "punpcklwd %%mm7, %%mm4                 \n\t"
        "punpckhwd %%mm7, %%mm2                 \n\t"
        "punpcklwd %%mm7, %%mm5                 \n\t"
        "punpckhwd %%mm7, %%mm3                 \n\t"
        "paddd (%1), %%mm4                      \n\t"
        "paddd 8(%1), %%mm2                     \n\t"
        "paddd 16(%1), %%mm5                    \n\t"
        "paddd 24(%1), %%mm3                    \n\t"
        "movq %%mm4, (%1)                       \n\t"
        "movq %%mm2, 8(%1)                      \n\t"
        "movq %%mm5, 16(%1)                     \n\t"
        "movq %%mm3, 24(%1)                     \n\t"
        "add $16, %0                            \n\t"
        "add $32, %1                            \n\t"
        "add $16, %2                            \n\t"
        "cmp %3, %0                             \n\t"
            " jb 1b                             \n\t"
        : "+r" (block), "+r" (sum), "+r" (offset)
        : "r"(block+64)
    );
}

static void  denoise_dct_sse2(MpegEncContext *s, int16_t *block){
    const int intra= s->mb_intra;
    int *sum= s->dct_error_sum[intra];
    uint16_t *offset= s->dct_offset[intra];

    s->dct_count[intra]++;

    __asm__ volatile(
        "pxor %%xmm7, %%xmm7                    \n\t"
        "1:                                     \n\t"
        "pxor %%xmm0, %%xmm0                    \n\t"
        "pxor %%xmm1, %%xmm1                    \n\t"
        "movdqa (%0), %%xmm2                    \n\t"
        "movdqa 16(%0), %%xmm3                  \n\t"
        "pcmpgtw %%xmm2, %%xmm0                 \n\t"
        "pcmpgtw %%xmm3, %%xmm1                 \n\t"
        "pxor %%xmm0, %%xmm2                    \n\t"
        "pxor %%xmm1, %%xmm3                    \n\t"
        "psubw %%xmm0, %%xmm2                   \n\t"
        "psubw %%xmm1, %%xmm3                   \n\t"
        "movdqa %%xmm2, %%xmm4                  \n\t"
        "movdqa %%xmm3, %%xmm5                  \n\t"
        "psubusw (%2), %%xmm2                   \n\t"
        "psubusw 16(%2), %%xmm3                 \n\t"
        "pxor %%xmm0, %%xmm2                    \n\t"
        "pxor %%xmm1, %%xmm3                    \n\t"
        "psubw %%xmm0, %%xmm2                   \n\t"
        "psubw %%xmm1, %%xmm3                   \n\t"
        "movdqa %%xmm2, (%0)                    \n\t"
        "movdqa %%xmm3, 16(%0)                  \n\t"
        "movdqa %%xmm4, %%xmm6                  \n\t"
        "movdqa %%xmm5, %%xmm0                  \n\t"
        "punpcklwd %%xmm7, %%xmm4               \n\t"
        "punpckhwd %%xmm7, %%xmm6               \n\t"
        "punpcklwd %%xmm7, %%xmm5               \n\t"
        "punpckhwd %%xmm7, %%xmm0               \n\t"
        "paddd (%1), %%xmm4                     \n\t"
        "paddd 16(%1), %%xmm6                   \n\t"
        "paddd 32(%1), %%xmm5                   \n\t"
        "paddd 48(%1), %%xmm0                   \n\t"
        "movdqa %%xmm4, (%1)                    \n\t"
        "movdqa %%xmm6, 16(%1)                  \n\t"
        "movdqa %%xmm5, 32(%1)                  \n\t"
        "movdqa %%xmm0, 48(%1)                  \n\t"
        "add $32, %0                            \n\t"
        "add $64, %1                            \n\t"
        "add $32, %2                            \n\t"
        "cmp %3, %0                             \n\t"
            " jb 1b                             \n\t"
        : "+r" (block), "+r" (sum), "+r" (offset)
        : "r"(block+64)
          XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm2", "%xmm3",
                            "%xmm4", "%xmm5", "%xmm6", "%xmm7")
    );
}
#endif /* HAVE_INLINE_ASM */

200
av_cold void ff_dct_encode_init_x86(MpegEncContext *s)
201 202
{
    const int dct_algo = s->avctx->dct_algo;
203 204 205 206
    int i;

    for (i = 0; i < 64; i++)
        inv_zigzag_direct16[ff_zigzag_direct[i]] = i + 1;
207 208

    if (dct_algo == FF_DCT_AUTO || dct_algo == FF_DCT_MMX) {
209
#if HAVE_MMX_INLINE
210
        int cpu_flags = av_get_cpu_flags();
211
        if (INLINE_MMX(cpu_flags)) {
212
#if HAVE_6REGS
213
            s->dct_quantize = dct_quantize_mmx;
214
#endif
215 216
            s->denoise_dct  = denoise_dct_mmx;
        }
217
#endif
218
#if HAVE_6REGS && HAVE_MMXEXT_INLINE
219
        if (INLINE_MMXEXT(cpu_flags))
220
            s->dct_quantize = dct_quantize_mmxext;
221
#endif
222
#if HAVE_SSE2_INLINE
223
        if (INLINE_SSE2(cpu_flags)) {
224
#if HAVE_6REGS
225
            s->dct_quantize = dct_quantize_sse2;
226
#endif
227 228
            s->denoise_dct  = denoise_dct_sse2;
        }
229
#endif
230
#if HAVE_6REGS && HAVE_SSSE3_INLINE
231
        if (INLINE_SSSE3(cpu_flags))
232
            s->dct_quantize = dct_quantize_ssse3;
233 234 235
#endif
    }
}