libavcodec/x86/mpegvideoenc.c

   1 /*
   2  * The simplest mpeg encoder (well, it was the simplest!)
   3  * Copyright (c) 2000,2001 Fabrice Bellard
   4  *
   5  * This file is part of Libav.
   6  *
   7  * Libav 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  * Libav 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 Libav; 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/avcodec.h"
  27 #include "libavcodec/dct.h"
  28 #include "libavcodec/mpegvideo.h"
  29
  30 /* not permutated inverse zigzag_direct + 1 for MMX quantizer */
  31 DECLARE_ALIGNED(16, static const uint16_t, inv_zigzag_direct16)[64] = {
  32     1,  2,  6,  7,  15, 16, 28, 29, 3,  5,  8,  14, 17, 27, 30, 43, 4,  9,  13,
  33     18, 26, 31, 42, 44, 10, 12, 19, 25, 32, 41, 45, 54, 11, 20, 24, 33, 40, 46,
  34     53, 55, 21, 23, 34, 39, 47, 52, 56, 61, 22, 35, 38, 48, 51, 57, 60, 62, 36,
  35     37, 49, 50, 58, 59, 63, 64,
  36 };
  37
  38 #if HAVE_MMX_INLINE
  39 #define COMPILE_TEMPLATE_MMXEXT 0
  40 #define COMPILE_TEMPLATE_SSE2   0
  41 #define COMPILE_TEMPLATE_SSSE3  0
  42 #define RENAME(a)      a ## _mmx
  43 #define RENAME_FDCT(a) a ## _mmx
  44 #include "mpegvideoenc_template.c"
  45 #endif /* HAVE_MMX_INLINE */
  46
  47 #if HAVE_MMXEXT_INLINE
  48 #undef COMPILE_TEMPLATE_SSSE3
  49 #undef COMPILE_TEMPLATE_SSE2
  50 #undef COMPILE_TEMPLATE_MMXEXT
  51 #define COMPILE_TEMPLATE_MMXEXT 1
  52 #define COMPILE_TEMPLATE_SSE2   0
  53 #define COMPILE_TEMPLATE_SSSE3  0
  54 #undef RENAME
  55 #undef RENAME_FDCT
  56 #define RENAME(a)      a ## _mmxext
  57 #define RENAME_FDCT(a) a ## _mmxext
  58 #include "mpegvideoenc_template.c"
  59 #endif /* HAVE_MMXEXT_INLINE */
  60
  61 #if HAVE_SSE2_INLINE
  62 #undef COMPILE_TEMPLATE_MMXEXT
  63 #undef COMPILE_TEMPLATE_SSE2
  64 #undef COMPILE_TEMPLATE_SSSE3
  65 #define COMPILE_TEMPLATE_MMXEXT 0
  66 #define COMPILE_TEMPLATE_SSE2   1
  67 #define COMPILE_TEMPLATE_SSSE3  0
  68 #undef RENAME
  69 #undef RENAME_FDCT
  70 #define RENAME(a)      a ## _sse2
  71 #define RENAME_FDCT(a) a ## _sse2
  72 #include "mpegvideoenc_template.c"
  73 #endif /* HAVE_SSE2_INLINE */
  74
  75 #if HAVE_SSSE3_INLINE
  76 #undef COMPILE_TEMPLATE_MMXEXT
  77 #undef COMPILE_TEMPLATE_SSE2
  78 #undef COMPILE_TEMPLATE_SSSE3
  79 #define COMPILE_TEMPLATE_MMXEXT 0
  80 #define COMPILE_TEMPLATE_SSE2   1
  81 #define COMPILE_TEMPLATE_SSSE3  1
  82 #undef RENAME
  83 #undef RENAME_FDCT
  84 #define RENAME(a)      a ## _ssse3
  85 #define RENAME_FDCT(a) a ## _sse2
  86 #include "mpegvideoenc_template.c"
  87 #endif /* HAVE_SSSE3_INLINE */
  88
  89 #if HAVE_INLINE_ASM
  90 static void  denoise_dct_mmx(MpegEncContext *s, int16_t *block){
  91     const int intra= s->mb_intra;
  92     int *sum= s->dct_error_sum[intra];
  93     uint16_t *offset= s->dct_offset[intra];
  94
  95     s->dct_count[intra]++;
  96
  97     __asm__ volatile(
  98         "pxor %%mm7, %%mm7                      \n\t"
  99         "1:                                     \n\t"
 100         "pxor %%mm0, %%mm0                      \n\t"
 101         "pxor %%mm1, %%mm1                      \n\t"
 102         "movq (%0), %%mm2                       \n\t"
 103         "movq 8(%0), %%mm3                      \n\t"
 104         "pcmpgtw %%mm2, %%mm0                   \n\t"
 105         "pcmpgtw %%mm3, %%mm1                   \n\t"
 106         "pxor %%mm0, %%mm2                      \n\t"
 107         "pxor %%mm1, %%mm3                      \n\t"
 108         "psubw %%mm0, %%mm2                     \n\t"
 109         "psubw %%mm1, %%mm3                     \n\t"
 110         "movq %%mm2, %%mm4                      \n\t"
 111         "movq %%mm3, %%mm5                      \n\t"
 112         "psubusw (%2), %%mm2                    \n\t"
 113         "psubusw 8(%2), %%mm3                   \n\t"
 114         "pxor %%mm0, %%mm2                      \n\t"
 115         "pxor %%mm1, %%mm3                      \n\t"
 116         "psubw %%mm0, %%mm2                     \n\t"
 117         "psubw %%mm1, %%mm3                     \n\t"
 118         "movq %%mm2, (%0)                       \n\t"
 119         "movq %%mm3, 8(%0)                      \n\t"
 120         "movq %%mm4, %%mm2                      \n\t"
 121         "movq %%mm5, %%mm3                      \n\t"
 122         "punpcklwd %%mm7, %%mm4                 \n\t"
 123         "punpckhwd %%mm7, %%mm2                 \n\t"
 124         "punpcklwd %%mm7, %%mm5                 \n\t"
 125         "punpckhwd %%mm7, %%mm3                 \n\t"
 126         "paddd (%1), %%mm4                      \n\t"
 127         "paddd 8(%1), %%mm2                     \n\t"
 128         "paddd 16(%1), %%mm5                    \n\t"
 129         "paddd 24(%1), %%mm3                    \n\t"
 130         "movq %%mm4, (%1)                       \n\t"
 131         "movq %%mm2, 8(%1)                      \n\t"
 132         "movq %%mm5, 16(%1)                     \n\t"
 133         "movq %%mm3, 24(%1)                     \n\t"
 134         "add $16, %0                            \n\t"
 135         "add $32, %1                            \n\t"
 136         "add $16, %2                            \n\t"
 137         "cmp %3, %0                             \n\t"
 138             " jb 1b                             \n\t"
 139         : "+r" (block), "+r" (sum), "+r" (offset)
 140         : "r"(block+64)
 141     );
 142 }
 143
 144 static void  denoise_dct_sse2(MpegEncContext *s, int16_t *block){
 145     const int intra= s->mb_intra;
 146     int *sum= s->dct_error_sum[intra];
 147     uint16_t *offset= s->dct_offset[intra];
 148
 149     s->dct_count[intra]++;
 150
 151     __asm__ volatile(
 152         "pxor %%xmm7, %%xmm7                    \n\t"
 153         "1:                                     \n\t"
 154         "pxor %%xmm0, %%xmm0                    \n\t"
 155         "pxor %%xmm1, %%xmm1                    \n\t"
 156         "movdqa (%0), %%xmm2                    \n\t"
 157         "movdqa 16(%0), %%xmm3                  \n\t"
 158         "pcmpgtw %%xmm2, %%xmm0                 \n\t"
 159         "pcmpgtw %%xmm3, %%xmm1                 \n\t"
 160         "pxor %%xmm0, %%xmm2                    \n\t"
 161         "pxor %%xmm1, %%xmm3                    \n\t"
 162         "psubw %%xmm0, %%xmm2                   \n\t"
 163         "psubw %%xmm1, %%xmm3                   \n\t"
 164         "movdqa %%xmm2, %%xmm4                  \n\t"
 165         "movdqa %%xmm3, %%xmm5                  \n\t"
 166         "psubusw (%2), %%xmm2                   \n\t"
 167         "psubusw 16(%2), %%xmm3                 \n\t"
 168         "pxor %%xmm0, %%xmm2                    \n\t"
 169         "pxor %%xmm1, %%xmm3                    \n\t"
 170         "psubw %%xmm0, %%xmm2                   \n\t"
 171         "psubw %%xmm1, %%xmm3                   \n\t"
 172         "movdqa %%xmm2, (%0)                    \n\t"
 173         "movdqa %%xmm3, 16(%0)                  \n\t"
 174         "movdqa %%xmm4, %%xmm6                  \n\t"
 175         "movdqa %%xmm5, %%xmm0                  \n\t"
 176         "punpcklwd %%xmm7, %%xmm4               \n\t"
 177         "punpckhwd %%xmm7, %%xmm6               \n\t"
 178         "punpcklwd %%xmm7, %%xmm5               \n\t"
 179         "punpckhwd %%xmm7, %%xmm0               \n\t"
 180         "paddd (%1), %%xmm4                     \n\t"
 181         "paddd 16(%1), %%xmm6                   \n\t"
 182         "paddd 32(%1), %%xmm5                   \n\t"
 183         "paddd 48(%1), %%xmm0                   \n\t"
 184         "movdqa %%xmm4, (%1)                    \n\t"
 185         "movdqa %%xmm6, 16(%1)                  \n\t"
 186         "movdqa %%xmm5, 32(%1)                  \n\t"
 187         "movdqa %%xmm0, 48(%1)                  \n\t"
 188         "add $32, %0                            \n\t"
 189         "add $64, %1                            \n\t"
 190         "add $32, %2                            \n\t"
 191         "cmp %3, %0                             \n\t"
 192             " jb 1b                             \n\t"
 193         : "+r" (block), "+r" (sum), "+r" (offset)
 194         : "r"(block+64)
 195           XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm2", "%xmm3",
 196                             "%xmm4", "%xmm5", "%xmm6", "%xmm7")
 197     );
 198 }
 199 #endif /* HAVE_INLINE_ASM */
 200
 201 av_cold void ff_mpv_encode_init_x86(MpegEncContext *s)
 202 {
 203     const int dct_algo = s->avctx->dct_algo;
 204
 205     if (dct_algo == FF_DCT_AUTO || dct_algo == FF_DCT_MMX) {
 206 #if HAVE_MMX_INLINE
 207         int cpu_flags = av_get_cpu_flags();
 208         if (INLINE_MMX(cpu_flags)) {
 209             s->dct_quantize = dct_quantize_mmx;
 210             s->denoise_dct  = denoise_dct_mmx;
 211         }
 212 #endif
 213 #if HAVE_MMXEXT_INLINE
 214         if (INLINE_MMXEXT(cpu_flags))
 215             s->dct_quantize = dct_quantize_mmxext;
 216 #endif
 217 #if HAVE_SSE2_INLINE
 218         if (INLINE_SSE2(cpu_flags)) {
 219             s->dct_quantize = dct_quantize_sse2;
 220             s->denoise_dct  = denoise_dct_sse2;
 221         }
 222 #endif
 223 #if HAVE_SSSE3_INLINE
 224         if (INLINE_SSSE3(cpu_flags))
 225             s->dct_quantize = dct_quantize_ssse3;
 226 #endif
 227     }
 228 }