| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | /* | ||
| 2 | * Copyright (C) 2009 Loren Merritt <lorenm@u.washington.edu> | ||
| 3 | * | ||
| 4 | * This file is part of FFmpeg. | ||
| 5 | * | ||
| 6 | * FFmpeg is free software; you can redistribute it and/or | ||
| 7 | * modify it under the terms of the GNU Lesser General Public | ||
| 8 | * License as published by the Free Software Foundation; either | ||
| 9 | * version 2.1 of the License, or (at your option) any later version. | ||
| 10 | * | ||
| 11 | * FFmpeg is distributed in the hope that it will be useful, | ||
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
| 14 | * Lesser General Public License for more details. | ||
| 15 | * | ||
| 16 | * You should have received a copy of the GNU Lesser General Public | ||
| 17 | * License along with FFmpeg; if not, write to the Free Software | ||
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
| 19 | */ | ||
| 20 | |||
| 21 | #include "config.h" | ||
| 22 | #include "libavutil/attributes.h" | ||
| 23 | #include "libavutil/cpu.h" | ||
| 24 | #include "libavutil/x86/cpu.h" | ||
| 25 | #include "libavfilter/gradfun.h" | ||
| 26 | |||
| 27 | void ff_gradfun_filter_line_ssse3(intptr_t x, uint8_t *dst, const uint8_t *src, | ||
| 28 | const uint16_t *dc, int thresh, | ||
| 29 | const uint16_t *dithers); | ||
| 30 | |||
| 31 | void ff_gradfun_blur_line_movdqa_sse2(intptr_t x, uint16_t *buf, | ||
| 32 | const uint16_t *buf1, uint16_t *dc, | ||
| 33 | const uint8_t *src1, const uint8_t *src2); | ||
| 34 | void ff_gradfun_blur_line_movdqu_sse2(intptr_t x, uint16_t *buf, | ||
| 35 | const uint16_t *buf1, uint16_t *dc, | ||
| 36 | const uint8_t *src1, const uint8_t *src2); | ||
| 37 | |||
| 38 | #if HAVE_X86ASM | ||
| 39 | ✗ | static void gradfun_filter_line_ssse3(uint8_t *dst, const uint8_t *src, const uint16_t *dc, | |
| 40 | int width, int thresh, | ||
| 41 | const uint16_t *dithers) | ||
| 42 | { | ||
| 43 | intptr_t x; | ||
| 44 | ✗ | if (width & 7) { | |
| 45 | // could be 10% faster if I somehow eliminated this | ||
| 46 | ✗ | x = width & ~7; | |
| 47 | ✗ | ff_gradfun_filter_line_c(dst + x, src + x, dc + x / 2, | |
| 48 | ✗ | width - x, thresh, dithers); | |
| 49 | ✗ | width = x; | |
| 50 | } | ||
| 51 | ✗ | x = -width; | |
| 52 | ✗ | ff_gradfun_filter_line_ssse3(x, dst + width, src + width, dc + width / 2, | |
| 53 | thresh, dithers); | ||
| 54 | ✗ | } | |
| 55 | |||
| 56 | ✗ | static void gradfun_blur_line_sse2(uint16_t *dc, uint16_t *buf, const uint16_t *buf1, | |
| 57 | const uint8_t *src, int src_linesize, int width) | ||
| 58 | { | ||
| 59 | ✗ | intptr_t x = -2 * width; | |
| 60 | ✗ | if (((intptr_t) src | src_linesize) & 15) | |
| 61 | ✗ | ff_gradfun_blur_line_movdqu_sse2(x, buf + width, buf1 + width, | |
| 62 | ✗ | dc + width, src + width * 2, | |
| 63 | ✗ | src + width * 2 + src_linesize); | |
| 64 | else | ||
| 65 | ✗ | ff_gradfun_blur_line_movdqa_sse2(x, buf + width, buf1 + width, | |
| 66 | ✗ | dc + width, src + width * 2, | |
| 67 | ✗ | src + width * 2 + src_linesize); | |
| 68 | ✗ | } | |
| 69 | #endif /* HAVE_X86ASM */ | ||
| 70 | |||
| 71 | 4 | av_cold void ff_gradfun_init_x86(GradFunContext *gf) | |
| 72 | { | ||
| 73 | #if HAVE_X86ASM | ||
| 74 | 4 | int cpu_flags = av_get_cpu_flags(); | |
| 75 | |||
| 76 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
|
4 | if (EXTERNAL_SSSE3(cpu_flags)) |
| 77 | ✗ | gf->filter_line = gradfun_filter_line_ssse3; | |
| 78 | |||
| 79 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 4 times.
|
4 | if (EXTERNAL_SSE2(cpu_flags)) |
| 80 | ✗ | gf->blur_line = gradfun_blur_line_sse2; | |
| 81 | #endif /* HAVE_X86ASM */ | ||
| 82 | 4 | } | |
| 83 |