2 * Copyright (C) 2002-2012 Michael Niedermayer
3 * Copyright (C) 2012 Ronald S. Bultje
5 * This file is part of FFmpeg.
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.
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.
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
23 #include "libavutil/attributes.h"
24 #include "libavutil/avassert.h"
25 #include "libavutil/common.h"
26 #include "libavutil/cpu.h"
27 #include "libavutil/mem.h"
28 #include "libavutil/x86/asm.h"
29 #include "libavcodec/videodsp.h"
32 typedef void emu_edge_core_func(uint8_t *buf, const uint8_t *src,
33 x86_reg linesize, x86_reg start_y,
34 x86_reg end_y, x86_reg block_h,
35 x86_reg start_x, x86_reg end_x,
37 extern emu_edge_core_func ff_emu_edge_core_mmx;
38 extern emu_edge_core_func ff_emu_edge_core_sse;
40 static av_always_inline void emulated_edge_mc(uint8_t *buf, const uint8_t *src,
41 ptrdiff_t linesize_arg,
42 int block_w, int block_h,
45 emu_edge_core_func *core_fn)
47 int start_y, start_x, end_y, end_x, src_y_add = 0;
48 int linesize = linesize_arg;
54 src -= src_y*linesize;
57 } else if (src_y <= -block_h) {
58 src -= src_y*linesize;
59 src_y_add = 1 - block_h;
65 } else if (src_x <= -block_w) {
66 src += 1 - block_w - src_x;
70 start_y = FFMAX(0, -src_y);
71 start_x = FFMAX(0, -src_x);
72 end_y = FFMIN(block_h, h-src_y);
73 end_x = FFMIN(block_w, w-src_x);
74 av_assert2(start_x < end_x && block_w > 0);
75 av_assert2(start_y < end_y && block_h > 0);
77 // fill in the to-be-copied part plus all above/below
78 src += (src_y_add + start_y) * linesize + start_x;
80 core_fn(buf, src, linesize, start_y, end_y,
81 block_h, start_x, end_x, block_w);
85 static av_noinline void emulated_edge_mc_mmx(uint8_t *buf, const uint8_t *src,
87 int block_w, int block_h,
88 int src_x, int src_y, int w, int h)
90 emulated_edge_mc(buf, src, linesize, block_w, block_h, src_x, src_y,
91 w, h, &ff_emu_edge_core_mmx);
95 static av_noinline void emulated_edge_mc_sse(uint8_t *buf, const uint8_t *src,
97 int block_w, int block_h,
98 int src_x, int src_y, int w, int h)
100 emulated_edge_mc(buf, src, linesize, block_w, block_h, src_x, src_y,
101 w, h, &ff_emu_edge_core_sse);
103 #endif /* HAVE_YASM */
105 void ff_prefetch_mmxext(uint8_t *buf, ptrdiff_t stride, int h);
106 void ff_prefetch_3dnow(uint8_t *buf, ptrdiff_t stride, int h);
108 av_cold void ff_videodsp_init_x86(VideoDSPContext *ctx, int bpc)
111 int mm_flags = av_get_cpu_flags();
114 if (bpc <= 8 && mm_flags & AV_CPU_FLAG_MMX) {
115 ctx->emulated_edge_mc = emulated_edge_mc_mmx;
117 if (mm_flags & AV_CPU_FLAG_3DNOW) {
118 ctx->prefetch = ff_prefetch_3dnow;
120 #endif /* ARCH_X86_32 */
121 if (mm_flags & AV_CPU_FLAG_MMXEXT) {
122 ctx->prefetch = ff_prefetch_mmxext;
124 if (bpc <= 8 && mm_flags & AV_CPU_FLAG_SSE) {
125 ctx->emulated_edge_mc = emulated_edge_mc_sse;
127 #endif /* HAVE_YASM */