1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
|
/*
* DSP utils mmx functions are compiled twice for rnd/no_rnd
* Copyright (c) 2000, 2001 Fabrice Bellard.
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
* MMX optimization by Nick Kurshev <nickols_k@mail.ru>
* mostly rewritten by Michael Niedermayer <michaelni@gmx.at>
* and improved by Zdenek Kabelac <kabi@users.sf.net>
*/
// will have to be check if it's better to have bigger
// unrolled code also on Celerons - for now yes
#define LONG_UNROLL 1
// put_pixels
static void DEF(put, pixels_x2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)
{
__asm __volatile(
MOVQ_BFE(%%mm7)
"lea (%3, %3), %%eax \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%1, %3), %%mm2 \n\t"
"movq 1(%1), %%mm1 \n\t"
"movq 1(%1, %3), %%mm3 \n\t"
PAVGB(%%mm0, %%mm1)
"movq %%mm6, (%2) \n\t"
PAVGB(%%mm2, %%mm3)
"movq %%mm6, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
#if LONG_UNROLL
"movq (%1), %%mm0 \n\t"
"movq (%1, %3), %%mm2 \n\t"
"movq 1(%1), %%mm1 \n\t"
"movq 1(%1, %3), %%mm3 \n\t"
PAVGB(%%mm0, %%mm1)
"movq %%mm6, (%2) \n\t"
PAVGB(%%mm2, %%mm3)
"movq %%mm6, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"subl $4, %0 \n\t"
#else
"subl $2, %0 \n\t"
#endif
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r"(line_size)
:"eax", "memory");
}
static void DEF(put, pixels_y2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)
{
__asm __volatile(
MOVQ_BFE(%%mm7)
"lea (%3, %3), %%eax \n\t"
"movq (%1), %%mm0 \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax),%%mm2 \n\t"
PAVGB(%%mm1, %%mm0)
"movq %%mm6, (%2) \n\t"
PAVGB(%%mm2, %%mm1)
"movq %%mm6, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
#ifdef LONG_UNROLL
"movq (%1, %3), %%mm1 \n\t"
"movq (%1, %%eax),%%mm0 \n\t"
PAVGB(%%mm1, %%mm2)
"movq %%mm6, (%2) \n\t"
PAVGB(%%mm0, %%mm1)
"movq %%mm6, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"subl $4, %0 \n\t"
#else
"subl $2, %0 \n\t"
#endif
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r"(line_size)
:"eax", "memory");
}
// ((a + b)/2 + (c + d)/2)/2
// not sure if this is properly replacing original code
static void DEF(put, pixels_xy2)(UINT8 *block, const UINT8 *pixels, int line_size, int h)
{
__asm __volatile(
MOVQ_BFE(%%mm7)
"lea (%3, %3), %%eax \n\t"
"movq (%1), %%mm0 \n\t"
"movq (%1), %%mm0 \n\t"
"movq 1(%1), %%mm1 \n\t"
".balign 8 \n\t"
"1: \n\t"
"movq (%1, %3), %%mm2 \n\t"
"movq 1(%1, %3), %%mm3 \n\t"
PAVGBR(%%mm2, %%mm0, %%mm4)
PAVGBR(%%mm3, %%mm1, %%mm5)
PAVGB(%%mm4, %%mm5)
"movq %%mm6, (%2) \n\t"
"movq (%1, %%eax), %%mm0 \n\t"
"movq 1(%1, %%eax), %%mm1 \n\t"
PAVGBR(%%mm0, %%mm2, %%mm4)
PAVGBR(%%mm1, %%mm3, %%mm5)
PAVGB(%%mm4, %%mm5)
"movq %%mm6, (%2, %3) \n\t"
"addl %%eax, %1 \n\t"
"addl %%eax, %2 \n\t"
"subl $2, %0 \n\t"
"jnz 1b \n\t"
:"+g"(h), "+S"(pixels), "+D"(block)
:"r"(line_size)
:"eax", "memory");
}
// avg_pixels
|
