| 1 | /* |
| 2 | * Copyright (c) 2002 Michael Niedermayer <michaelni@gmx.at> |
| 3 | * Copyright (c) 2013 Paul B Mahol |
| 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/x86/cpu.h" |
| 24 | #include "libavutil/x86/asm.h" |
| 25 | #include "libavfilter/vf_noise.h" |
| 26 | |
| 27 | #if HAVE_INLINE_ASM |
| 28 | static void line_noise_mmx(uint8_t *dst, const uint8_t *src, |
| 29 | const int8_t *noise, int len, int shift) |
| 30 | { |
| 31 | x86_reg mmx_len= len & (~7); |
| 32 | noise += shift; |
| 33 | |
| 34 | __asm__ volatile( |
| 35 | "mov %3, %%"REG_a" \n\t" |
| 36 | "pcmpeqb %%mm7, %%mm7 \n\t" |
| 37 | "psllw $15, %%mm7 \n\t" |
| 38 | "packsswb %%mm7, %%mm7 \n\t" |
| 39 | ".p2align 4 \n\t" |
| 40 | "1: \n\t" |
| 41 | "movq (%0, %%"REG_a"), %%mm0 \n\t" |
| 42 | "movq (%1, %%"REG_a"), %%mm1 \n\t" |
| 43 | "pxor %%mm7, %%mm0 \n\t" |
| 44 | "paddsb %%mm1, %%mm0 \n\t" |
| 45 | "pxor %%mm7, %%mm0 \n\t" |
| 46 | "movq %%mm0, (%2, %%"REG_a") \n\t" |
| 47 | "add $8, %%"REG_a" \n\t" |
| 48 | " js 1b \n\t" |
| 49 | :: "r" (src+mmx_len), "r" (noise+mmx_len), "r" (dst+mmx_len), "g" (-mmx_len) |
| 50 | : "%"REG_a |
| 51 | ); |
| 52 | if (mmx_len != len) |
| 53 | ff_line_noise_c(dst+mmx_len, src+mmx_len, noise+mmx_len, len-mmx_len, 0); |
| 54 | } |
| 55 | |
| 56 | #if HAVE_6REGS |
| 57 | static void line_noise_avg_mmx(uint8_t *dst, const uint8_t *src, |
| 58 | int len, const int8_t * const *shift) |
| 59 | { |
| 60 | x86_reg mmx_len = len & (~7); |
| 61 | |
| 62 | __asm__ volatile( |
| 63 | "mov %5, %%"REG_a" \n\t" |
| 64 | ".p2align 4 \n\t" |
| 65 | "1: \n\t" |
| 66 | "movq (%1, %%"REG_a"), %%mm1 \n\t" |
| 67 | "movq (%0, %%"REG_a"), %%mm0 \n\t" |
| 68 | "paddb (%2, %%"REG_a"), %%mm1 \n\t" |
| 69 | "paddb (%3, %%"REG_a"), %%mm1 \n\t" |
| 70 | "movq %%mm0, %%mm2 \n\t" |
| 71 | "movq %%mm1, %%mm3 \n\t" |
| 72 | "punpcklbw %%mm0, %%mm0 \n\t" |
| 73 | "punpckhbw %%mm2, %%mm2 \n\t" |
| 74 | "punpcklbw %%mm1, %%mm1 \n\t" |
| 75 | "punpckhbw %%mm3, %%mm3 \n\t" |
| 76 | "pmulhw %%mm0, %%mm1 \n\t" |
| 77 | "pmulhw %%mm2, %%mm3 \n\t" |
| 78 | "paddw %%mm1, %%mm1 \n\t" |
| 79 | "paddw %%mm3, %%mm3 \n\t" |
| 80 | "paddw %%mm0, %%mm1 \n\t" |
| 81 | "paddw %%mm2, %%mm3 \n\t" |
| 82 | "psrlw $8, %%mm1 \n\t" |
| 83 | "psrlw $8, %%mm3 \n\t" |
| 84 | "packuswb %%mm3, %%mm1 \n\t" |
| 85 | "movq %%mm1, (%4, %%"REG_a") \n\t" |
| 86 | "add $8, %%"REG_a" \n\t" |
| 87 | " js 1b \n\t" |
| 88 | :: "r" (src+mmx_len), "r" (shift[0]+mmx_len), "r" (shift[1]+mmx_len), "r" (shift[2]+mmx_len), |
| 89 | "r" (dst+mmx_len), "g" (-mmx_len) |
| 90 | : "%"REG_a |
| 91 | ); |
| 92 | |
| 93 | if (mmx_len != len){ |
| 94 | const int8_t *shift2[3] = { shift[0]+mmx_len, shift[1]+mmx_len, shift[2]+mmx_len }; |
| 95 | ff_line_noise_avg_c(dst+mmx_len, src+mmx_len, len-mmx_len, shift2); |
| 96 | } |
| 97 | } |
| 98 | #endif /* HAVE_6REGS */ |
| 99 | |
| 100 | static void line_noise_mmxext(uint8_t *dst, const uint8_t *src, |
| 101 | const int8_t *noise, int len, int shift) |
| 102 | { |
| 103 | x86_reg mmx_len = len & (~7); |
| 104 | noise += shift; |
| 105 | |
| 106 | __asm__ volatile( |
| 107 | "mov %3, %%"REG_a" \n\t" |
| 108 | "pcmpeqb %%mm7, %%mm7 \n\t" |
| 109 | "psllw $15, %%mm7 \n\t" |
| 110 | "packsswb %%mm7, %%mm7 \n\t" |
| 111 | ".p2align 4 \n\t" |
| 112 | "1: \n\t" |
| 113 | "movq (%0, %%"REG_a"), %%mm0 \n\t" |
| 114 | "movq (%1, %%"REG_a"), %%mm1 \n\t" |
| 115 | "pxor %%mm7, %%mm0 \n\t" |
| 116 | "paddsb %%mm1, %%mm0 \n\t" |
| 117 | "pxor %%mm7, %%mm0 \n\t" |
| 118 | "movntq %%mm0, (%2, %%"REG_a") \n\t" |
| 119 | "add $8, %%"REG_a" \n\t" |
| 120 | " js 1b \n\t" |
| 121 | :: "r" (src+mmx_len), "r" (noise+mmx_len), "r" (dst+mmx_len), "g" (-mmx_len) |
| 122 | : "%"REG_a |
| 123 | ); |
| 124 | if (mmx_len != len) |
| 125 | ff_line_noise_c(dst+mmx_len, src+mmx_len, noise+mmx_len, len-mmx_len, 0); |
| 126 | } |
| 127 | #endif /* HAVE_INLINE_ASM */ |
| 128 | |
| 129 | av_cold void ff_noise_init_x86(NoiseContext *n) |
| 130 | { |
| 131 | #if HAVE_INLINE_ASM |
| 132 | int cpu_flags = av_get_cpu_flags(); |
| 133 | |
| 134 | if (INLINE_MMX(cpu_flags)) { |
| 135 | n->line_noise = line_noise_mmx; |
| 136 | #if HAVE_6REGS |
| 137 | n->line_noise_avg = line_noise_avg_mmx; |
| 138 | #endif |
| 139 | } |
| 140 | if (INLINE_MMXEXT(cpu_flags)) { |
| 141 | n->line_noise = line_noise_mmxext; |
| 142 | } |
| 143 | #endif |
| 144 | } |