| 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/mem.h" |
| 25 | #include "libavutil/x86/asm.h" |
| 26 | #include "libavutil/x86/cpu.h" |
| 27 | #include "libavfilter/gradfun.h" |
| 28 | |
| 29 | void ff_gradfun_filter_line_mmxext(intptr_t x, uint8_t *dst, const uint8_t *src, |
| 30 | const uint16_t *dc, int thresh, |
| 31 | const uint16_t *dithers); |
| 32 | void ff_gradfun_filter_line_ssse3(intptr_t x, uint8_t *dst, const uint8_t *src, |
| 33 | const uint16_t *dc, int thresh, |
| 34 | const uint16_t *dithers); |
| 35 | |
| 36 | void ff_gradfun_blur_line_movdqa_sse2(intptr_t x, uint16_t *buf, |
| 37 | const uint16_t *buf1, uint16_t *dc, |
| 38 | const uint8_t *src1, const uint8_t *src2); |
| 39 | void ff_gradfun_blur_line_movdqu_sse2(intptr_t x, uint16_t *buf, |
| 40 | const uint16_t *buf1, uint16_t *dc, |
| 41 | const uint8_t *src1, const uint8_t *src2); |
| 42 | |
| 43 | #if HAVE_YASM |
| 44 | static void gradfun_filter_line_mmxext(uint8_t *dst, const uint8_t *src, |
| 45 | const uint16_t *dc, |
| 46 | int width, int thresh, |
| 47 | const uint16_t *dithers) |
| 48 | { |
| 49 | intptr_t x; |
| 50 | if (width & 3) { |
| 51 | x = width & ~3; |
| 52 | ff_gradfun_filter_line_c(dst + x, src + x, dc + x / 2, |
| 53 | width - x, thresh, dithers); |
| 54 | width = x; |
| 55 | } |
| 56 | x = -width; |
| 57 | ff_gradfun_filter_line_mmxext(x, dst + width, src + width, dc + width / 2, |
| 58 | thresh, dithers); |
| 59 | } |
| 60 | |
| 61 | static void gradfun_filter_line_ssse3(uint8_t *dst, const uint8_t *src, const uint16_t *dc, |
| 62 | int width, int thresh, |
| 63 | const uint16_t *dithers) |
| 64 | { |
| 65 | intptr_t x; |
| 66 | if (width & 7) { |
| 67 | // could be 10% faster if I somehow eliminated this |
| 68 | x = width & ~7; |
| 69 | ff_gradfun_filter_line_c(dst + x, src + x, dc + x / 2, |
| 70 | width - x, thresh, dithers); |
| 71 | width = x; |
| 72 | } |
| 73 | x = -width; |
| 74 | ff_gradfun_filter_line_ssse3(x, dst + width, src + width, dc + width / 2, |
| 75 | thresh, dithers); |
| 76 | } |
| 77 | |
| 78 | static void gradfun_blur_line_sse2(uint16_t *dc, uint16_t *buf, const uint16_t *buf1, |
| 79 | const uint8_t *src, int src_linesize, int width) |
| 80 | { |
| 81 | intptr_t x = -2 * width; |
| 82 | if (((intptr_t) src | src_linesize) & 15) |
| 83 | ff_gradfun_blur_line_movdqu_sse2(x, buf + width, buf1 + width, |
| 84 | dc + width, src + width * 2, |
| 85 | src + width * 2 + src_linesize); |
| 86 | else |
| 87 | ff_gradfun_blur_line_movdqa_sse2(x, buf + width, buf1 + width, |
| 88 | dc + width, src + width * 2, |
| 89 | src + width * 2 + src_linesize); |
| 90 | } |
| 91 | #endif /* HAVE_YASM */ |
| 92 | |
| 93 | av_cold void ff_gradfun_init_x86(GradFunContext *gf) |
| 94 | { |
| 95 | #if HAVE_YASM |
| 96 | int cpu_flags = av_get_cpu_flags(); |
| 97 | |
| 98 | if (EXTERNAL_MMXEXT(cpu_flags)) |
| 99 | gf->filter_line = gradfun_filter_line_mmxext; |
| 100 | if (EXTERNAL_SSSE3(cpu_flags)) |
| 101 | gf->filter_line = gradfun_filter_line_ssse3; |
| 102 | |
| 103 | if (EXTERNAL_SSE2(cpu_flags)) |
| 104 | gf->blur_line = gradfun_blur_line_sse2; |
| 105 | #endif /* HAVE_YASM */ |
| 106 | } |