Parent Directory | Revision Log
Revision 1535 - (view) (download)
1 : | suxen_drol | 1493 | ;/***************************************************************************** |
2 : | ; * | ||
3 : | ; * XVID MPEG-4 VIDEO CODEC | ||
4 : | ; * - sse2 post processing - | ||
5 : | ; * | ||
6 : | ; * Copyright(C) 2004 Peter Ross <pross@xvid.org> | ||
7 : | ; * 2004 Dcoder <dcoder@alexandria.cc> | ||
8 : | ; * | ||
9 : | ; * XviD is free software; you can redistribute it and/or modify it | ||
10 : | ; * under the terms of the GNU General Public License as published by | ||
11 : | ; * the Free Software Foundation; either version 2 of the License, or | ||
12 : | ; * (at your option) any later version. | ||
13 : | ; * | ||
14 : | ; * This program is distributed in the hope that it will be useful, | ||
15 : | ; * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
16 : | ; * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | ||
17 : | ; * GNU General Public License for more details. | ||
18 : | ; * | ||
19 : | ; * You should have received a copy of the GNU General Public License | ||
20 : | ; * along with this program; if not, write to the Free Software | ||
21 : | ; * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | ||
22 : | ; * | ||
23 : | ; *************************************************************************/ | ||
24 : | |||
25 : | BITS 32 | ||
26 : | |||
27 : | %macro cglobal 1 | ||
28 : | %ifdef PREFIX | ||
29 : | edgomez | 1535 | %ifdef MARK_FUNCS |
30 : | global _%1:function | ||
31 : | %define %1 _%1:function | ||
32 : | %else | ||
33 : | global _%1 | ||
34 : | %define %1 _%1 | ||
35 : | %endif | ||
36 : | suxen_drol | 1493 | %else |
37 : | edgomez | 1535 | %ifdef MARK_FUNCS |
38 : | global %1:function | ||
39 : | %else | ||
40 : | global %1 | ||
41 : | %endif | ||
42 : | suxen_drol | 1493 | %endif |
43 : | %endmacro | ||
44 : | |||
45 : | %macro FILLBYTES 2 | ||
46 : | |||
47 : | mov [%1], %2 | ||
48 : | mov [%1 + 1], %2 | ||
49 : | mov [%1 + 2], %2 | ||
50 : | mov [%1 + 3], %2 | ||
51 : | mov [%1 + 4], %2 | ||
52 : | mov [%1 + 5], %2 | ||
53 : | mov [%1 + 6], %2 | ||
54 : | mov [%1 + 7], %2 | ||
55 : | mov [%1 + 8], %2 | ||
56 : | mov [%1 + 9], %2 | ||
57 : | mov [%1 + 10], %2 | ||
58 : | mov [%1 + 11], %2 | ||
59 : | mov [%1 + 12], %2 | ||
60 : | mov [%1 + 13], %2 | ||
61 : | mov [%1 + 14], %2 | ||
62 : | mov [%1 + 15], %2 | ||
63 : | |||
64 : | %endmacro | ||
65 : | |||
66 : | |||
67 : | ;=========================================================================== | ||
68 : | ; read only data | ||
69 : | ;=========================================================================== | ||
70 : | |||
71 : | %ifdef FORMAT_COFF | ||
72 : | edgomez | 1519 | SECTION .rodata |
73 : | suxen_drol | 1493 | %else |
74 : | edgomez | 1519 | SECTION .rodata align=16 |
75 : | suxen_drol | 1493 | %endif |
76 : | |||
77 : | xmm_0x80: | ||
78 : | times 16 db 0x80 | ||
79 : | |||
80 : | offset_xmm: | ||
81 : | times 16 db 0x00 | ||
82 : | |||
83 : | ;============================================================================= | ||
84 : | ; Code | ||
85 : | ;============================================================================= | ||
86 : | |||
87 : | SECTION .text | ||
88 : | |||
89 : | cglobal image_brightness_sse2 | ||
90 : | |||
91 : | |||
92 : | ;////////////////////////////////////////////////////////////////////// | ||
93 : | ;// image_brightness_sse2 | ||
94 : | ;////////////////////////////////////////////////////////////////////// | ||
95 : | |||
96 : | align 16 | ||
97 : | image_brightness_sse2: | ||
98 : | |||
99 : | push esi | ||
100 : | push edi | ||
101 : | |||
102 : | movdqa xmm6, [xmm_0x80] | ||
103 : | |||
104 : | mov eax, [esp+8+20] ; offset | ||
105 : | |||
106 : | FILLBYTES offset_xmm, al | ||
107 : | |||
108 : | movdqa xmm7, [offset_xmm] | ||
109 : | |||
110 : | mov edx, [esp+8+4] ; Dst | ||
111 : | mov ecx, [esp+8+8] ; stride | ||
112 : | mov esi, [esp+8+12] ; width | ||
113 : | mov edi, [esp+8+16] ; height | ||
114 : | |||
115 : | .yloop | ||
116 : | xor eax, eax | ||
117 : | |||
118 : | .xloop | ||
119 : | movdqa xmm0, [edx + eax] | ||
120 : | movdqa xmm1, [edx + eax + 16] ; xmm0 = [dst] | ||
121 : | |||
122 : | paddb xmm0, xmm6 ; unsigned -> signed domain | ||
123 : | paddb xmm1, xmm6 | ||
124 : | paddsb xmm0, xmm7 | ||
125 : | paddsb xmm1, xmm7 ; xmm0 += offset | ||
126 : | psubb xmm0, xmm6 | ||
127 : | psubb xmm1, xmm6 ; signed -> unsigned domain | ||
128 : | |||
129 : | movdqa [edx + eax], xmm0 | ||
130 : | movdqa [edx + eax + 16], xmm1 ; [dst] = xmm0 | ||
131 : | |||
132 : | add eax,32 | ||
133 : | cmp eax,esi | ||
134 : | jl .xloop | ||
135 : | |||
136 : | add edx, ecx ; dst += stride | ||
137 : | sub edi, 1 | ||
138 : | jg .yloop | ||
139 : | |||
140 : | pop edi | ||
141 : | pop esi | ||
142 : | |||
143 : | ret | ||
144 : | ;////////////////////////////////////////////////////////////////////// |
No admin address has been configured | ViewVC Help |
Powered by ViewVC 1.0.4 |