FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
dsputil_mmx.c
Go to the documentation of this file.
1 /*
2  * MMX optimized DSP utils
3  * Copyright (c) 2000, 2001 Fabrice Bellard
4  * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
5  *
6  * This file is part of FFmpeg.
7  *
8  * FFmpeg is free software; you can redistribute it and/or
9  * modify it under the terms of the GNU Lesser General Public
10  * License as published by the Free Software Foundation; either
11  * version 2.1 of the License, or (at your option) any later version.
12  *
13  * FFmpeg is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16  * Lesser General Public License for more details.
17  *
18  * You should have received a copy of the GNU Lesser General Public
19  * License along with FFmpeg; if not, write to the Free Software
20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21  *
22  * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
23  */
24 
25 #include "config.h"
26 #include "libavutil/avassert.h"
27 #include "libavutil/cpu.h"
28 #include "libavutil/x86/asm.h"
29 #include "libavcodec/pixels.h"
30 #include "libavcodec/videodsp.h"
31 #include "dsputil_x86.h"
32 #include "inline_asm.h"
33 
34 #if HAVE_INLINE_ASM
35 
36 /* Draw the edges of width 'w' of an image of size width, height
37  * this MMX version can only handle w == 8 || w == 16. */
38 void ff_draw_edges_mmx(uint8_t *buf, int wrap, int width, int height,
39  int w, int h, int sides)
40 {
41  uint8_t *ptr, *last_line;
42  int i;
43 
44  last_line = buf + (height - 1) * wrap;
45  /* left and right */
46  ptr = buf;
47  if (w == 8) {
48  __asm__ volatile (
49  "1: \n\t"
50  "movd (%0), %%mm0 \n\t"
51  "punpcklbw %%mm0, %%mm0 \n\t"
52  "punpcklwd %%mm0, %%mm0 \n\t"
53  "punpckldq %%mm0, %%mm0 \n\t"
54  "movq %%mm0, -8(%0) \n\t"
55  "movq -8(%0, %2), %%mm1 \n\t"
56  "punpckhbw %%mm1, %%mm1 \n\t"
57  "punpckhwd %%mm1, %%mm1 \n\t"
58  "punpckhdq %%mm1, %%mm1 \n\t"
59  "movq %%mm1, (%0, %2) \n\t"
60  "add %1, %0 \n\t"
61  "cmp %3, %0 \n\t"
62  "jb 1b \n\t"
63  : "+r" (ptr)
64  : "r" ((x86_reg) wrap), "r" ((x86_reg) width),
65  "r" (ptr + wrap * height));
66  } else if (w == 16) {
67  __asm__ volatile (
68  "1: \n\t"
69  "movd (%0), %%mm0 \n\t"
70  "punpcklbw %%mm0, %%mm0 \n\t"
71  "punpcklwd %%mm0, %%mm0 \n\t"
72  "punpckldq %%mm0, %%mm0 \n\t"
73  "movq %%mm0, -8(%0) \n\t"
74  "movq %%mm0, -16(%0) \n\t"
75  "movq -8(%0, %2), %%mm1 \n\t"
76  "punpckhbw %%mm1, %%mm1 \n\t"
77  "punpckhwd %%mm1, %%mm1 \n\t"
78  "punpckhdq %%mm1, %%mm1 \n\t"
79  "movq %%mm1, (%0, %2) \n\t"
80  "movq %%mm1, 8(%0, %2) \n\t"
81  "add %1, %0 \n\t"
82  "cmp %3, %0 \n\t"
83  "jb 1b \n\t"
84  : "+r"(ptr)
85  : "r"((x86_reg)wrap), "r"((x86_reg)width), "r"(ptr + wrap * height)
86  );
87  } else {
88  av_assert1(w == 4);
89  __asm__ volatile (
90  "1: \n\t"
91  "movd (%0), %%mm0 \n\t"
92  "punpcklbw %%mm0, %%mm0 \n\t"
93  "punpcklwd %%mm0, %%mm0 \n\t"
94  "movd %%mm0, -4(%0) \n\t"
95  "movd -4(%0, %2), %%mm1 \n\t"
96  "punpcklbw %%mm1, %%mm1 \n\t"
97  "punpckhwd %%mm1, %%mm1 \n\t"
98  "punpckhdq %%mm1, %%mm1 \n\t"
99  "movd %%mm1, (%0, %2) \n\t"
100  "add %1, %0 \n\t"
101  "cmp %3, %0 \n\t"
102  "jb 1b \n\t"
103  : "+r" (ptr)
104  : "r" ((x86_reg) wrap), "r" ((x86_reg) width),
105  "r" (ptr + wrap * height));
106  }
107 
108  /* top and bottom (and hopefully also the corners) */
109  if (sides & EDGE_TOP) {
110  for (i = 0; i < h; i += 4) {
111  ptr = buf - (i + 1) * wrap - w;
112  __asm__ volatile (
113  "1: \n\t"
114  "movq (%1, %0), %%mm0 \n\t"
115  "movq %%mm0, (%0) \n\t"
116  "movq %%mm0, (%0, %2) \n\t"
117  "movq %%mm0, (%0, %2, 2) \n\t"
118  "movq %%mm0, (%0, %3) \n\t"
119  "add $8, %0 \n\t"
120  "cmp %4, %0 \n\t"
121  "jb 1b \n\t"
122  : "+r" (ptr)
123  : "r" ((x86_reg) buf - (x86_reg) ptr - w),
124  "r" ((x86_reg) - wrap), "r" ((x86_reg) - wrap * 3),
125  "r" (ptr + width + 2 * w));
126  }
127  }
128 
129  if (sides & EDGE_BOTTOM) {
130  for (i = 0; i < h; i += 4) {
131  ptr = last_line + (i + 1) * wrap - w;
132  __asm__ volatile (
133  "1: \n\t"
134  "movq (%1, %0), %%mm0 \n\t"
135  "movq %%mm0, (%0) \n\t"
136  "movq %%mm0, (%0, %2) \n\t"
137  "movq %%mm0, (%0, %2, 2) \n\t"
138  "movq %%mm0, (%0, %3) \n\t"
139  "add $8, %0 \n\t"
140  "cmp %4, %0 \n\t"
141  "jb 1b \n\t"
142  : "+r" (ptr)
143  : "r" ((x86_reg) last_line - (x86_reg) ptr - w),
144  "r" ((x86_reg) wrap), "r" ((x86_reg) wrap * 3),
145  "r" (ptr + width + 2 * w));
146  }
147  }
148 }
149 
150 #endif /* HAVE_INLINE_ASM */
Generated on Sun Jul 6 2014 19:21:12 for FFmpeg by   doxygen 1.8.2