Parent Directory | Revision Log
Revision 651 -
(view)
(download)
Original Path: trunk/xvidcore/src/image/x86_asm/interpolate8x8_mmx.asm
1 : | chl | 434 | ;/***************************************************************************** |
2 : | Isibaar | 262 | ; * |
3 : | chl | 434 | ; * XVID MPEG-4 VIDEO CODEC |
4 : | ; * mmx 8x8 block-based halfpel interpolation | ||
5 : | Isibaar | 262 | ; * |
6 : | chl | 434 | ; * Copyright(C) 2002 Peter Ross <pross@xvid.org> |
7 : | Isibaar | 262 | ; * |
8 : | edgomez | 651 | ; * This file is part of XviD, a free MPEG-4 video encoder/decoder |
9 : | Isibaar | 262 | ; * |
10 : | edgomez | 651 | ; * XviD is free software; you can redistribute it and/or modify it |
11 : | ; * under the terms of the GNU General Public License as published by | ||
12 : | chl | 434 | ; * the Free Software Foundation; either version 2 of the License, or |
13 : | ; * (at your option) any later version. | ||
14 : | Isibaar | 262 | ; * |
15 : | chl | 434 | ; * This program is distributed in the hope that it will be useful, |
16 : | ; * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
17 : | ; * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | ||
18 : | ; * GNU General Public License for more details. | ||
19 : | Isibaar | 262 | ; * |
20 : | chl | 434 | ; * You should have received a copy of the GNU General Public License |
21 : | ; * along with this program; if not, write to the Free Software | ||
22 : | ; * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | ||
23 : | Isibaar | 262 | ; * |
24 : | edgomez | 651 | ; * Under section 8 of the GNU General Public License, the copyright |
25 : | ; * holders of XVID explicitly forbid distribution in the following | ||
26 : | ; * countries: | ||
27 : | ; * | ||
28 : | ; * - Japan | ||
29 : | ; * - United States of America | ||
30 : | ; * | ||
31 : | ; * Linking XviD statically or dynamically with other modules is making a | ||
32 : | ; * combined work based on XviD. Thus, the terms and conditions of the | ||
33 : | ; * GNU General Public License cover the whole combination. | ||
34 : | ; * | ||
35 : | ; * As a special exception, the copyright holders of XviD give you | ||
36 : | ; * permission to link XviD with independent modules that communicate with | ||
37 : | ; * XviD solely through the VFW1.1 and DShow interfaces, regardless of the | ||
38 : | ; * license terms of these independent modules, and to copy and distribute | ||
39 : | ; * the resulting combined work under terms of your choice, provided that | ||
40 : | ; * every copy of the combined work is accompanied by a complete copy of | ||
41 : | ; * the source code of XviD (the version of XviD used to produce the | ||
42 : | ; * combined work), being distributed under the terms of the GNU General | ||
43 : | ; * Public License plus this exception. An independent module is a module | ||
44 : | ; * which is not derived from or based on XviD. | ||
45 : | ; * | ||
46 : | ; * Note that people who make modified versions of XviD are not obligated | ||
47 : | ; * to grant this special exception for their modified versions; it is | ||
48 : | ; * their choice whether to do so. The GNU General Public License gives | ||
49 : | ; * permission to release a modified version without this exception; this | ||
50 : | ; * exception also makes it possible to release a modified version which | ||
51 : | ; * carries forward this exception. | ||
52 : | ; * | ||
53 : | ; * $Id: interpolate8x8_mmx.asm,v 1.11 2002-11-17 00:20:30 edgomez Exp $ | ||
54 : | ; * | ||
55 : | chl | 434 | ; ****************************************************************************/ |
56 : | Isibaar | 262 | |
57 : | bits 32 | ||
58 : | |||
59 : | %macro cglobal 1 | ||
60 : | %ifdef PREFIX | ||
61 : | global _%1 | ||
62 : | %define %1 _%1 | ||
63 : | %else | ||
64 : | global %1 | ||
65 : | %endif | ||
66 : | %endmacro | ||
67 : | |||
68 : | section .data | ||
69 : | |||
70 : | align 16 | ||
71 : | |||
72 : | ;=========================================================================== | ||
73 : | ; (1 - r) rounding table | ||
74 : | ;=========================================================================== | ||
75 : | |||
76 : | rounding1_mmx | ||
77 : | times 4 dw 1 | ||
78 : | times 4 dw 0 | ||
79 : | |||
80 : | ;=========================================================================== | ||
81 : | ; (2 - r) rounding table | ||
82 : | ;=========================================================================== | ||
83 : | |||
84 : | rounding2_mmx | ||
85 : | times 4 dw 2 | ||
86 : | times 4 dw 1 | ||
87 : | |||
88 : | mmx_one | ||
89 : | times 8 db 1 | ||
90 : | |||
91 : | section .text | ||
92 : | |||
93 : | %macro CALC_AVG 6 | ||
94 : | punpcklbw %3, %6 | ||
95 : | punpckhbw %4, %6 | ||
96 : | |||
97 : | paddusw %1, %3 ; mm01 += mm23 | ||
98 : | paddusw %2, %4 | ||
99 : | paddusw %1, %5 ; mm01 += rounding | ||
100 : | paddusw %2, %5 | ||
101 : | |||
102 : | psrlw %1, 1 ; mm01 >>= 1 | ||
103 : | psrlw %2, 1 | ||
104 : | |||
105 : | %endmacro | ||
106 : | |||
107 : | |||
108 : | ;=========================================================================== | ||
109 : | ; | ||
110 : | ; void interpolate8x8_halfpel_h_mmx(uint8_t * const dst, | ||
111 : | ; const uint8_t * const src, | ||
112 : | ; const uint32_t stride, | ||
113 : | ; const uint32_t rounding); | ||
114 : | ; | ||
115 : | ;=========================================================================== | ||
116 : | |||
117 : | %macro COPY_H_MMX 0 | ||
118 : | movq mm0, [esi] | ||
119 : | movq mm2, [esi + 1] | ||
120 : | movq mm1, mm0 | ||
121 : | movq mm3, mm2 | ||
122 : | |||
123 : | punpcklbw mm0, mm6 ; mm01 = [src] | ||
124 : | punpckhbw mm1, mm6 ; mm23 = [src + 1] | ||
125 : | |||
126 : | CALC_AVG mm0, mm1, mm2, mm3, mm7, mm6 | ||
127 : | |||
128 : | packuswb mm0, mm1 | ||
129 : | movq [edi], mm0 ; [dst] = mm01 | ||
130 : | |||
131 : | add esi, edx ; src += stride | ||
132 : | add edi, edx ; dst += stride | ||
133 : | %endmacro | ||
134 : | |||
135 : | align 16 | ||
136 : | cglobal interpolate8x8_halfpel_h_mmx | ||
137 : | interpolate8x8_halfpel_h_mmx | ||
138 : | |||
139 : | push esi | ||
140 : | push edi | ||
141 : | |||
142 : | mov eax, [esp + 8 + 16] ; rounding | ||
143 : | |||
144 : | interpolate8x8_halfpel_h_mmx.start | ||
145 : | movq mm7, [rounding1_mmx + eax * 8] | ||
146 : | |||
147 : | mov edi, [esp + 8 + 4] ; dst | ||
148 : | mov esi, [esp + 8 + 8] ; src | ||
149 : | mov edx, [esp + 8 + 12] ; stride | ||
150 : | |||
151 : | pxor mm6, mm6 ; zero | ||
152 : | |||
153 : | COPY_H_MMX | ||
154 : | COPY_H_MMX | ||
155 : | COPY_H_MMX | ||
156 : | COPY_H_MMX | ||
157 : | COPY_H_MMX | ||
158 : | COPY_H_MMX | ||
159 : | COPY_H_MMX | ||
160 : | COPY_H_MMX | ||
161 : | |||
162 : | pop edi | ||
163 : | pop esi | ||
164 : | |||
165 : | ret | ||
166 : | |||
167 : | |||
168 : | ;=========================================================================== | ||
169 : | ; | ||
170 : | ; void interpolate8x8_halfpel_v_mmx(uint8_t * const dst, | ||
171 : | ; const uint8_t * const src, | ||
172 : | ; const uint32_t stride, | ||
173 : | ; const uint32_t rounding); | ||
174 : | ; | ||
175 : | ;=========================================================================== | ||
176 : | |||
177 : | %macro COPY_V_MMX 0 | ||
178 : | movq mm0, [esi] | ||
179 : | movq mm2, [esi + edx] | ||
180 : | movq mm1, mm0 | ||
181 : | movq mm3, mm2 | ||
182 : | |||
183 : | punpcklbw mm0, mm6 ; mm01 = [src] | ||
184 : | punpckhbw mm1, mm6 ; mm23 = [src + 1] | ||
185 : | |||
186 : | CALC_AVG mm0, mm1, mm2, mm3, mm7, mm6 | ||
187 : | |||
188 : | packuswb mm0, mm1 | ||
189 : | movq [edi], mm0 ; [dst] = mm01 | ||
190 : | |||
191 : | add esi, edx ; src += stride | ||
192 : | add edi, edx ; dst += stride | ||
193 : | %endmacro | ||
194 : | |||
195 : | align 16 | ||
196 : | cglobal interpolate8x8_halfpel_v_mmx | ||
197 : | interpolate8x8_halfpel_v_mmx | ||
198 : | |||
199 : | push esi | ||
200 : | push edi | ||
201 : | |||
202 : | mov eax, [esp + 8 + 16] ; rounding | ||
203 : | |||
204 : | interpolate8x8_halfpel_v_mmx.start | ||
205 : | movq mm7, [rounding1_mmx + eax * 8] | ||
206 : | |||
207 : | mov edi, [esp + 8 + 4] ; dst | ||
208 : | mov esi, [esp + 8 + 8] ; src | ||
209 : | mov edx, [esp + 8 + 12] ; stride | ||
210 : | |||
211 : | pxor mm6, mm6 ; zero | ||
212 : | |||
213 : | |||
214 : | COPY_V_MMX | ||
215 : | COPY_V_MMX | ||
216 : | COPY_V_MMX | ||
217 : | COPY_V_MMX | ||
218 : | COPY_V_MMX | ||
219 : | COPY_V_MMX | ||
220 : | COPY_V_MMX | ||
221 : | COPY_V_MMX | ||
222 : | |||
223 : | pop edi | ||
224 : | pop esi | ||
225 : | |||
226 : | ret | ||
227 : | |||
228 : | |||
229 : | ;=========================================================================== | ||
230 : | ; | ||
231 : | ; void interpolate8x8_halfpel_hv_mmx(uint8_t * const dst, | ||
232 : | ; const uint8_t * const src, | ||
233 : | ; const uint32_t stride, | ||
234 : | ; const uint32_t rounding); | ||
235 : | ; | ||
236 : | ; | ||
237 : | ;=========================================================================== | ||
238 : | |||
239 : | %macro COPY_HV_MMX 0 | ||
240 : | ; current row | ||
241 : | |||
242 : | movq mm0, [esi] | ||
243 : | movq mm2, [esi + 1] | ||
244 : | |||
245 : | movq mm1, mm0 | ||
246 : | movq mm3, mm2 | ||
247 : | |||
248 : | punpcklbw mm0, mm6 ; mm01 = [src] | ||
249 : | punpcklbw mm2, mm6 ; mm23 = [src + 1] | ||
250 : | punpckhbw mm1, mm6 | ||
251 : | punpckhbw mm3, mm6 | ||
252 : | |||
253 : | paddusw mm0, mm2 ; mm01 += mm23 | ||
254 : | paddusw mm1, mm3 | ||
255 : | |||
256 : | ; next row | ||
257 : | |||
258 : | movq mm4, [esi + edx] | ||
259 : | movq mm2, [esi + edx + 1] | ||
260 : | |||
261 : | movq mm5, mm4 | ||
262 : | movq mm3, mm2 | ||
263 : | |||
264 : | punpcklbw mm4, mm6 ; mm45 = [src + stride] | ||
265 : | punpcklbw mm2, mm6 ; mm23 = [src + stride + 1] | ||
266 : | punpckhbw mm5, mm6 | ||
267 : | punpckhbw mm3, mm6 | ||
268 : | |||
269 : | paddusw mm4, mm2 ; mm45 += mm23 | ||
270 : | paddusw mm5, mm3 | ||
271 : | |||
272 : | ; add current + next row | ||
273 : | |||
274 : | paddusw mm0, mm4 ; mm01 += mm45 | ||
275 : | paddusw mm1, mm5 | ||
276 : | paddusw mm0, mm7 ; mm01 += rounding2 | ||
277 : | paddusw mm1, mm7 | ||
278 : | |||
279 : | psrlw mm0, 2 ; mm01 >>= 2 | ||
280 : | psrlw mm1, 2 | ||
281 : | |||
282 : | packuswb mm0, mm1 | ||
283 : | movq [edi], mm0 ; [dst] = mm01 | ||
284 : | |||
285 : | add esi, edx ; src += stride | ||
286 : | add edi, edx ; dst += stride | ||
287 : | %endmacro | ||
288 : | |||
289 : | align 16 | ||
290 : | cglobal interpolate8x8_halfpel_hv_mmx | ||
291 : | interpolate8x8_halfpel_hv_mmx | ||
292 : | |||
293 : | push esi | ||
294 : | push edi | ||
295 : | |||
296 : | mov eax, [esp + 8 + 16] ; rounding | ||
297 : | interpolate8x8_halfpel_hv_mmx.start | ||
298 : | |||
299 : | movq mm7, [rounding2_mmx + eax * 8] | ||
300 : | |||
301 : | mov edi, [esp + 8 + 4] ; dst | ||
302 : | mov esi, [esp + 8 + 8] ; src | ||
303 : | |||
304 : | mov eax, 8 | ||
305 : | |||
306 : | pxor mm6, mm6 ; zero | ||
307 : | |||
308 : | mov edx, [esp + 8 + 12] ; stride | ||
309 : | |||
310 : | COPY_HV_MMX | ||
311 : | COPY_HV_MMX | ||
312 : | COPY_HV_MMX | ||
313 : | COPY_HV_MMX | ||
314 : | COPY_HV_MMX | ||
315 : | COPY_HV_MMX | ||
316 : | COPY_HV_MMX | ||
317 : | COPY_HV_MMX | ||
318 : | |||
319 : | pop edi | ||
320 : | pop esi | ||
321 : | |||
322 : | chl | 434 | ret |
No admin address has been configured | ViewVC Help |
Powered by ViewVC 1.0.4 |