| 1 | /* |
| 2 | * SIMD-optimized HuffYUV encoding functions |
| 3 | * Copyright (c) 2000, 2001 Fabrice Bellard |
| 4 | * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at> |
| 5 | * |
| 6 | * MMX optimization by Nick Kurshev <nickols_k@mail.ru> |
| 7 | * |
| 8 | * This file is part of FFmpeg. |
| 9 | * |
| 10 | * FFmpeg is free software; you can redistribute it and/or |
| 11 | * modify it under the terms of the GNU Lesser General Public |
| 12 | * License as published by the Free Software Foundation; either |
| 13 | * version 2.1 of the License, or (at your option) any later version. |
| 14 | * |
| 15 | * FFmpeg is distributed in the hope that it will be useful, |
| 16 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 17 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 18 | * Lesser General Public License for more details. |
| 19 | * |
| 20 | * You should have received a copy of the GNU Lesser General Public |
| 21 | * License along with FFmpeg; if not, write to the Free Software |
| 22 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 23 | */ |
| 24 | |
| 25 | #include "libavutil/attributes.h" |
| 26 | #include "libavutil/cpu.h" |
| 27 | #include "libavutil/x86/asm.h" |
| 28 | #include "libavutil/x86/cpu.h" |
| 29 | #include "libavcodec/huffyuvencdsp.h" |
| 30 | #include "libavcodec/mathops.h" |
| 31 | |
| 32 | #if HAVE_INLINE_ASM |
| 33 | |
| 34 | static void diff_bytes_mmx(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w) |
| 35 | { |
| 36 | x86_reg i = 0; |
| 37 | |
| 38 | if (w >= 16) |
| 39 | __asm__ volatile ( |
| 40 | "1: \n\t" |
| 41 | "movq (%2, %0), %%mm0 \n\t" |
| 42 | "movq (%1, %0), %%mm1 \n\t" |
| 43 | "psubb %%mm0, %%mm1 \n\t" |
| 44 | "movq %%mm1, (%3, %0) \n\t" |
| 45 | "movq 8(%2, %0), %%mm0 \n\t" |
| 46 | "movq 8(%1, %0), %%mm1 \n\t" |
| 47 | "psubb %%mm0, %%mm1 \n\t" |
| 48 | "movq %%mm1, 8(%3, %0) \n\t" |
| 49 | "add $16, %0 \n\t" |
| 50 | "cmp %4, %0 \n\t" |
| 51 | " jb 1b \n\t" |
| 52 | : "+r" (i) |
| 53 | : "r" (src1), "r" (src2), "r" (dst), "r" ((x86_reg) w - 15)); |
| 54 | |
| 55 | for (; i < w; i++) |
| 56 | dst[i + 0] = src1[i + 0] - src2[i + 0]; |
| 57 | } |
| 58 | |
| 59 | static void sub_hfyu_median_pred_mmxext(uint8_t *dst, const uint8_t *src1, |
| 60 | const uint8_t *src2, int w, |
| 61 | int *left, int *left_top) |
| 62 | { |
| 63 | x86_reg i = 0; |
| 64 | uint8_t l, lt; |
| 65 | |
| 66 | __asm__ volatile ( |
| 67 | "movq (%1, %0), %%mm0 \n\t" // LT |
| 68 | "psllq $8, %%mm0 \n\t" |
| 69 | "1: \n\t" |
| 70 | "movq (%1, %0), %%mm1 \n\t" // T |
| 71 | "movq -1(%2, %0), %%mm2 \n\t" // L |
| 72 | "movq (%2, %0), %%mm3 \n\t" // X |
| 73 | "movq %%mm2, %%mm4 \n\t" // L |
| 74 | "psubb %%mm0, %%mm2 \n\t" |
| 75 | "paddb %%mm1, %%mm2 \n\t" // L + T - LT |
| 76 | "movq %%mm4, %%mm5 \n\t" // L |
| 77 | "pmaxub %%mm1, %%mm4 \n\t" // max(T, L) |
| 78 | "pminub %%mm5, %%mm1 \n\t" // min(T, L) |
| 79 | "pminub %%mm2, %%mm4 \n\t" |
| 80 | "pmaxub %%mm1, %%mm4 \n\t" |
| 81 | "psubb %%mm4, %%mm3 \n\t" // dst - pred |
| 82 | "movq %%mm3, (%3, %0) \n\t" |
| 83 | "add $8, %0 \n\t" |
| 84 | "movq -1(%1, %0), %%mm0 \n\t" // LT |
| 85 | "cmp %4, %0 \n\t" |
| 86 | " jb 1b \n\t" |
| 87 | : "+r" (i) |
| 88 | : "r" (src1), "r" (src2), "r" (dst), "r" ((x86_reg) w)); |
| 89 | |
| 90 | l = *left; |
| 91 | lt = *left_top; |
| 92 | |
| 93 | dst[0] = src2[0] - mid_pred(l, src1[0], (l + src1[0] - lt) & 0xFF); |
| 94 | |
| 95 | *left_top = src1[w - 1]; |
| 96 | *left = src2[w - 1]; |
| 97 | } |
| 98 | |
| 99 | #endif /* HAVE_INLINE_ASM */ |
| 100 | |
| 101 | av_cold void ff_huffyuvencdsp_init_x86(HuffYUVEncDSPContext *c) |
| 102 | { |
| 103 | #if HAVE_INLINE_ASM |
| 104 | int cpu_flags = av_get_cpu_flags(); |
| 105 | |
| 106 | if (INLINE_MMX(cpu_flags)) { |
| 107 | c->diff_bytes = diff_bytes_mmx; |
| 108 | } |
| 109 | |
| 110 | if (INLINE_MMXEXT(cpu_flags)) { |
| 111 | c->sub_hfyu_median_pred = sub_hfyu_median_pred_mmxext; |
| 112 | } |
| 113 | #endif /* HAVE_INLINE_ASM */ |
| 114 | } |