| Line | Branch | Exec | Source |
|---|---|---|---|
| 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 | #if HAVE_6REGS | ||
| 29 | ✗ | static void line_noise_avg_sse2(uint8_t *dst, const uint8_t *src, | |
| 30 | int len, const int8_t * const *shift) | ||
| 31 | { | ||
| 32 | ✗ | x86_reg xmm_len = len & (~15); | |
| 33 | |||
| 34 | ✗ | __asm__ volatile( | |
| 35 | "mov %5, %%"FF_REG_a" \n\t" | ||
| 36 | "pxor %%xmm4, %%xmm4 \n\t" | ||
| 37 | ".p2align 4 \n\t" | ||
| 38 | "1: \n\t" | ||
| 39 | "movdqu (%1, %%"FF_REG_a"), %%xmm1 \n\t" | ||
| 40 | "movdqu (%2, %%"FF_REG_a"), %%xmm2 \n\t" | ||
| 41 | "movdqu (%3, %%"FF_REG_a"), %%xmm3 \n\t" | ||
| 42 | "movdqu (%0, %%"FF_REG_a"), %%xmm0 \n\t" | ||
| 43 | "paddb %%xmm2, %%xmm1 \n\t" | ||
| 44 | "paddb %%xmm3, %%xmm1 \n\t" | ||
| 45 | "movdqa %%xmm4, %%xmm5 \n\t" | ||
| 46 | "pcmpgtb %%xmm0, %%xmm5 \n\t" | ||
| 47 | "movdqa %%xmm0, %%xmm6 \n\t" | ||
| 48 | "movdqa %%xmm0, %%xmm2 \n\t" | ||
| 49 | "punpcklbw %%xmm5, %%xmm0 \n\t" | ||
| 50 | "punpckhbw %%xmm5, %%xmm2 \n\t" | ||
| 51 | "movdqa %%xmm4, %%xmm5 \n\t" | ||
| 52 | "pcmpgtb %%xmm1, %%xmm5 \n\t" | ||
| 53 | "movdqa %%xmm1, %%xmm3 \n\t" | ||
| 54 | "punpcklbw %%xmm5, %%xmm1 \n\t" | ||
| 55 | "punpckhbw %%xmm5, %%xmm3 \n\t" | ||
| 56 | "pmullw %%xmm0, %%xmm1 \n\t" | ||
| 57 | "pmullw %%xmm2, %%xmm3 \n\t" | ||
| 58 | "psraw $7, %%xmm1 \n\t" | ||
| 59 | "psraw $7, %%xmm3 \n\t" | ||
| 60 | "packsswb %%xmm3, %%xmm1 \n\t" | ||
| 61 | "paddb %%xmm6, %%xmm1 \n\t" | ||
| 62 | "movdqu %%xmm1, (%4, %%"FF_REG_a") \n\t" | ||
| 63 | "add $16, %%"FF_REG_a" \n\t" | ||
| 64 | " js 1b \n\t" | ||
| 65 | ✗ | :: "r" (src+xmm_len), "r" (shift[0]+xmm_len), "r" (shift[1]+xmm_len), "r" (shift[2]+xmm_len), | |
| 66 | ✗ | "r" (dst+xmm_len), "g" (-xmm_len) | |
| 67 | : XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2", "%xmm3", | ||
| 68 | "%xmm4", "%xmm5", "%xmm6",) "%"FF_REG_a | ||
| 69 | ); | ||
| 70 | |||
| 71 | ✗ | if (xmm_len != len){ | |
| 72 | ✗ | const int8_t *shift2[3] = { shift[0]+xmm_len, shift[1]+xmm_len, shift[2]+xmm_len }; | |
| 73 | ✗ | ff_line_noise_avg_c(dst + xmm_len, src + xmm_len, len - xmm_len, shift2); | |
| 74 | } | ||
| 75 | ✗ | } | |
| 76 | #endif /* HAVE_6REGS */ | ||
| 77 | |||
| 78 | ✗ | static void line_noise_sse2(uint8_t *dst, const uint8_t *src, | |
| 79 | const int8_t *noise, int len, int shift) | ||
| 80 | { | ||
| 81 | ✗ | x86_reg xmm_len = len & (~15); | |
| 82 | ✗ | noise += shift; | |
| 83 | |||
| 84 | ✗ | __asm__ volatile( | |
| 85 | "mov %3, %%"FF_REG_a" \n\t" | ||
| 86 | "pcmpeqb %%xmm2, %%xmm2 \n\t" | ||
| 87 | "psllw $15, %%xmm2 \n\t" | ||
| 88 | "packsswb %%xmm2, %%xmm2 \n\t" | ||
| 89 | ".p2align 4 \n\t" | ||
| 90 | "1: \n\t" | ||
| 91 | "movdqu (%0, %%"FF_REG_a"), %%xmm0 \n\t" | ||
| 92 | "movdqu (%1, %%"FF_REG_a"), %%xmm1 \n\t" | ||
| 93 | "pxor %%xmm2, %%xmm0 \n\t" | ||
| 94 | "paddsb %%xmm1, %%xmm0 \n\t" | ||
| 95 | "pxor %%xmm2, %%xmm0 \n\t" | ||
| 96 | "movdqu %%xmm0, (%2, %%"FF_REG_a") \n\t" | ||
| 97 | "add $16, %%"FF_REG_a" \n\t" | ||
| 98 | " js 1b \n\t" | ||
| 99 | ✗ | :: "r" (src+xmm_len), "r" (noise+xmm_len), "r" (dst+xmm_len), "g" (-xmm_len) | |
| 100 | : XMM_CLOBBERS("%xmm0", "%xmm1", "%xmm2",) "%"FF_REG_a | ||
| 101 | ); | ||
| 102 | ✗ | if (xmm_len != len) | |
| 103 | ✗ | ff_line_noise_c(dst+xmm_len, src + xmm_len, noise + xmm_len, len - xmm_len, 0); | |
| 104 | ✗ | } | |
| 105 | #endif /* HAVE_INLINE_ASM */ | ||
| 106 | |||
| 107 | ✗ | av_cold void ff_noise_init_x86(NoiseContext *n) | |
| 108 | { | ||
| 109 | #if HAVE_INLINE_ASM | ||
| 110 | ✗ | int cpu_flags = av_get_cpu_flags(); | |
| 111 | |||
| 112 | ✗ | if (INLINE_SSE2(cpu_flags)) { | |
| 113 | #if HAVE_6REGS | ||
| 114 | ✗ | n->line_noise_avg = line_noise_avg_sse2; | |
| 115 | #endif | ||
| 116 | ✗ | n->line_noise = line_noise_sse2; | |
| 117 | } | ||
| 118 | #endif | ||
| 119 | ✗ | } | |
| 120 |