| 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 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/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 uint16_t, inv_zigzag_direct16)[64]; |
| 32 | |
| 33 | #if HAVE_6REGS |
| 34 | |
| 35 | #if HAVE_MMX_INLINE |
| 36 | #define COMPILE_TEMPLATE_MMXEXT 0 |
| 37 | #define COMPILE_TEMPLATE_SSE2 0 |
| 38 | #define COMPILE_TEMPLATE_SSSE3 0 |
| 39 | #define RENAME(a) a ## _mmx |
| 40 | #define RENAME_FDCT(a) a ## _mmx |
| 41 | #include "mpegvideoenc_template.c" |
| 42 | #endif /* HAVE_MMX_INLINE */ |
| 43 | |
| 44 | #if HAVE_MMXEXT_INLINE |
| 45 | #undef COMPILE_TEMPLATE_SSSE3 |
| 46 | #undef COMPILE_TEMPLATE_SSE2 |
| 47 | #undef COMPILE_TEMPLATE_MMXEXT |
| 48 | #define COMPILE_TEMPLATE_MMXEXT 1 |
| 49 | #define COMPILE_TEMPLATE_SSE2 0 |
| 50 | #define COMPILE_TEMPLATE_SSSE3 0 |
| 51 | #undef RENAME |
| 52 | #undef RENAME_FDCT |
| 53 | #define RENAME(a) a ## _mmxext |
| 54 | #define RENAME_FDCT(a) a ## _mmxext |
| 55 | #include "mpegvideoenc_template.c" |
| 56 | #endif /* HAVE_MMXEXT_INLINE */ |
| 57 | |
| 58 | #if HAVE_SSE2_INLINE |
| 59 | #undef COMPILE_TEMPLATE_MMXEXT |
| 60 | #undef COMPILE_TEMPLATE_SSE2 |
| 61 | #undef COMPILE_TEMPLATE_SSSE3 |
| 62 | #define COMPILE_TEMPLATE_MMXEXT 0 |
| 63 | #define COMPILE_TEMPLATE_SSE2 1 |
| 64 | #define COMPILE_TEMPLATE_SSSE3 0 |
| 65 | #undef RENAME |
| 66 | #undef RENAME_FDCT |
| 67 | #define RENAME(a) a ## _sse2 |
| 68 | #define RENAME_FDCT(a) a ## _sse2 |
| 69 | #include "mpegvideoenc_template.c" |
| 70 | #endif /* HAVE_SSE2_INLINE */ |
| 71 | |
| 72 | #if HAVE_SSSE3_INLINE |
| 73 | #undef COMPILE_TEMPLATE_MMXEXT |
| 74 | #undef COMPILE_TEMPLATE_SSE2 |
| 75 | #undef COMPILE_TEMPLATE_SSSE3 |
| 76 | #define COMPILE_TEMPLATE_MMXEXT 0 |
| 77 | #define COMPILE_TEMPLATE_SSE2 1 |
| 78 | #define COMPILE_TEMPLATE_SSSE3 1 |
| 79 | #undef RENAME |
| 80 | #undef RENAME_FDCT |
| 81 | #define RENAME(a) a ## _ssse3 |
| 82 | #define RENAME_FDCT(a) a ## _sse2 |
| 83 | #include "mpegvideoenc_template.c" |
| 84 | #endif /* HAVE_SSSE3_INLINE */ |
| 85 | |
| 86 | #endif /* HAVE_6REGS */ |
| 87 | |
| 88 | #if HAVE_INLINE_ASM |
| 89 | static void denoise_dct_mmx(MpegEncContext *s, int16_t *block){ |
| 90 | const int intra= s->mb_intra; |
| 91 | int *sum= s->dct_error_sum[intra]; |
| 92 | uint16_t *offset= s->dct_offset[intra]; |
| 93 | |
| 94 | s->dct_count[intra]++; |
| 95 | |
| 96 | __asm__ volatile( |
| 97 | "pxor %%mm7, %%mm7 \n\t" |
| 98 | "1: \n\t" |
| 99 | "pxor %%mm0, %%mm0 \n\t" |
| 100 | "pxor %%mm1, %%mm1 \n\t" |
| 101 | "movq (%0), %%mm2 \n\t" |
| 102 | "movq 8(%0), %%mm3 \n\t" |
| 103 | "pcmpgtw %%mm2, %%mm0 \n\t" |
| 104 | "pcmpgtw %%mm3, %%mm1 \n\t" |
| 105 | "pxor %%mm0, %%mm2 \n\t" |
| 106 | "pxor %%mm1, %%mm3 \n\t" |
| 107 | "psubw %%mm0, %%mm2 \n\t" |
| 108 | "psubw %%mm1, %%mm3 \n\t" |
| 109 | "movq %%mm2, %%mm4 \n\t" |
| 110 | "movq %%mm3, %%mm5 \n\t" |
| 111 | "psubusw (%2), %%mm2 \n\t" |
| 112 | "psubusw 8(%2), %%mm3 \n\t" |
| 113 | "pxor %%mm0, %%mm2 \n\t" |
| 114 | "pxor %%mm1, %%mm3 \n\t" |
| 115 | "psubw %%mm0, %%mm2 \n\t" |
| 116 | "psubw %%mm1, %%mm3 \n\t" |
| 117 | "movq %%mm2, (%0) \n\t" |
| 118 | "movq %%mm3, 8(%0) \n\t" |
| 119 | "movq %%mm4, %%mm2 \n\t" |
| 120 | "movq %%mm5, %%mm3 \n\t" |
| 121 | "punpcklwd %%mm7, %%mm4 \n\t" |
| 122 | "punpckhwd %%mm7, %%mm2 \n\t" |
| 123 | "punpcklwd %%mm7, %%mm5 \n\t" |
| 124 | "punpckhwd %%mm7, %%mm3 \n\t" |
| 125 | "paddd (%1), %%mm4 \n\t" |
| 126 | "paddd 8(%1), %%mm2 \n\t" |
| 127 | "paddd 16(%1), %%mm5 \n\t" |
| 128 | "paddd 24(%1), %%mm3 \n\t" |
| 129 | "movq %%mm4, (%1) \n\t" |
| 130 | "movq %%mm2, 8(%1) \n\t" |
| 131 | "movq %%mm5, 16(%1) \n\t" |
| 132 | "movq %%mm3, 24(%1) \n\t" |
| 133 | "add $16, %0 \n\t" |
| 134 | "add $32, %1 \n\t" |
| 135 | "add $16, %2 \n\t" |
| 136 | "cmp %3, %0 \n\t" |
| 137 | " jb 1b \n\t" |
| 138 | : "+r" (block), "+r" (sum), "+r" (offset) |
| 139 | : "r"(block+64) |
| 140 | ); |
| 141 | } |
| 142 | |
| 143 | static void denoise_dct_sse2(MpegEncContext *s, int16_t *block){ |
| 144 | const int intra= s->mb_intra; |
| 145 | int *sum= s->dct_error_sum[intra]; |
| 146 | uint16_t *offset= s->dct_offset[intra]; |
| 147 | |
| 148 | s->dct_count[intra]++; |
| 149 | |
| 150 | __asm__ volatile( |
| 151 | "pxor %%xmm7, %%xmm7 \n\t" |
| 152 | "1: \n\t" |
| 153 | "pxor %%xmm0, %%xmm0 \n\t" |
| 154 | "pxor %%xmm1, %%xmm1 \n\t" |
| 155 | "movdqa (%0), %%xmm2 \n\t" |
| 156 | "movdqa 16(%0), %%xmm3 \n\t" |
| 157 | "pcmpgtw %%xmm2, %%xmm0 \n\t" |
| 158 | "pcmpgtw %%xmm3, %%xmm1 \n\t" |
| 159 | "pxor %%xmm0, %%xmm2 \n\t" |
| 160 | "pxor %%xmm1, %%xmm3 \n\t" |
| 161 | "psubw %%xmm0, %%xmm2 \n\t" |
| 162 | "psubw %%xmm1, %%xmm3 \n\t" |
| 163 | "movdqa %%xmm2, %%xmm4 \n\t" |
| 164 | "movdqa %%xmm3, %%xmm5 \n\t" |
| 165 | "psubusw (%2), %%xmm2 \n\t" |
| 166 | "psubusw 16(%2), %%xmm3 \n\t" |
| 167 | "pxor %%xmm0, %%xmm2 \n\t" |
| 168 | "pxor %%xmm1, %%xmm3 \n\t" |
| 169 | "psubw %%xmm0, %%xmm2 \n\t" |
| 170 | "psubw %%xmm1, %%xmm3 \n\t" |
| 171 | "movdqa %%xmm2, (%0) \n\t" |
| 172 | "movdqa %%xmm3, 16(%0) \n\t" |
| 173 | "movdqa %%xmm4, %%xmm6 \n\t" |
| 174 | "movdqa %%xmm5, %%xmm0 \n\t" |
| 175 | "punpcklwd %%xmm7, %%xmm4 \n\t" |
| 176 | "punpckhwd %%xmm7, %%xmm6 \n\t" |
| 177 | "punpcklwd %%xmm7, %%xmm5 \n\t" |
| 178 | "punpckhwd %%xmm7, %%xmm0 \n\t" |
| 179 | "paddd (%1), %%xmm4 \n\t" |
| 180 | "paddd 16(%1), %%xmm6 \n\t" |
| 181 | "paddd 32(%1), %%xmm5 \n\t" |
| 182 | "paddd 48(%1), %%xmm0 \n\t" |
| 183 | "movdqa %%xmm4, (%1) \n\t" |
| 184 | "movdqa %%xmm6, 16(%1) \n\t" |
| 185 | "movdqa %%xmm5, 32(%1) \n\t" |
| 186 | "movdqa %%xmm0, 48(%1) \n\t" |
| 187 | "add $32, %0 \n\t" |
| 188 | "add $64, %1 \n\t" |
| 189 | "add $32, %2 \n\t" |
| 190 | "cmp %3, %0 \n\t" |
| 191 | " jb 1b \n\t" |
| 192 | : "+r" (block), "+r" (sum), "+r" (offset) |
| 193 | : "r"(block+64) |
| 194 | XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm2", "%xmm3", |
| 195 | "%xmm4", "%xmm5", "%xmm6", "%xmm7") |
| 196 | ); |
| 197 | } |
| 198 | #endif /* HAVE_INLINE_ASM */ |
| 199 | |
| 200 | av_cold void ff_dct_encode_init_x86(MpegEncContext *s) |
| 201 | { |
| 202 | const int dct_algo = s->avctx->dct_algo; |
| 203 | int i; |
| 204 | |
| 205 | for (i = 0; i < 64; i++) |
| 206 | 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 | s->denoise_dct = denoise_dct_mmx; |
| 216 | } |
| 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 | s->denoise_dct = denoise_dct_sse2; |
| 228 | } |
| 229 | #endif |
| 230 | #if HAVE_6REGS && HAVE_SSSE3_INLINE |
| 231 | if (INLINE_SSSE3(cpu_flags)) |
| 232 | s->dct_quantize = dct_quantize_ssse3; |
| 233 | #endif |
| 234 | } |
| 235 | } |