1 |
;/************************************************************************** |
;/***************************************************************************** |
2 |
; * |
; * |
3 |
; * XVID MPEG-4 VIDEO CODEC |
; * XVID MPEG-4 VIDEO CODEC |
4 |
; * xmm sum of absolute difference |
; * xmm (extended mmx) sum of absolute difference |
5 |
|
; * |
6 |
|
; * Copyright(C) 2002 Peter Ross <pross@xvid.org> |
7 |
|
; * Copyright(C) 2002 Michael Militzer <michael@xvid.org> |
8 |
|
; * Copyright(C) 2002 Pascal Massimino <skal@planet-d.net> |
9 |
|
; * |
10 |
|
; * This program is an implementation of a part of one or more MPEG-4 |
11 |
|
; * Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
12 |
|
; * to use this software module in hardware or software products are |
13 |
|
; * advised that its use may infringe existing patents or copyrights, and |
14 |
|
; * any such use would be at such party's own risk. The original |
15 |
|
; * developer of this software module and his/her company, and subsequent |
16 |
|
; * editors and their companies, will have no liability for use of this |
17 |
|
; * software or modifications or derivatives thereof. |
18 |
; * |
; * |
19 |
; * This program is free software; you can redistribute it and/or modify |
; * This program is free software; you can redistribute it and/or modify |
20 |
; * it under the terms of the GNU General Public License as published by |
; * it under the terms of the GNU General Public License as published by |
28 |
; * |
; * |
29 |
; * You should have received a copy of the GNU General Public License |
; * You should have received a copy of the GNU General Public License |
30 |
; * along with this program; if not, write to the Free Software |
; * along with this program; if not, write to the Free Software |
31 |
; * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
; * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
|
; * |
|
|
; *************************************************************************/ |
|
|
|
|
|
;/************************************************************************** |
|
|
; * |
|
|
; * History: |
|
|
; * |
|
|
; * 23.07.2002 sad8bi_xmm; <pross@xvid.org> |
|
|
; * 04.06.2002 rewrote some funcs (XMM mainly) -Skal- |
|
|
; * 17.11.2001 bugfix and small improvement for dev16_xmm, |
|
|
; * removed terminate early in sad16_xmm (Isibaar) |
|
|
; * 12.11.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
32 |
; * |
; * |
33 |
; *************************************************************************/ |
; *************************************************************************/ |
34 |
|
|
35 |
bits 32 |
bits 32 |
36 |
|
|
37 |
%macro cglobal 1 |
%macro cglobal 1 |
38 |
%ifdef PREFIX |
%ifdef PREFIX |
39 |
global _%1 |
global _%1 |
60 |
; |
; |
61 |
; uint32_t sad16_xmm(const uint8_t * const cur, |
; uint32_t sad16_xmm(const uint8_t * const cur, |
62 |
; const uint8_t * const ref, |
; const uint8_t * const ref, |
63 |
|
; const uint32_t stride, |
64 |
; const uint32_t best_sad); |
; const uint32_t best_sad); |
65 |
cglobal sad8_xmm |
; |
66 |
;=========================================================================== |
;=========================================================================== |
67 |
|
|
68 |
%macro SAD_16x16_SSE 0 |
%macro SAD_16x16_SSE 0 |
117 |
; |
; |
118 |
;=========================================================================== |
;=========================================================================== |
119 |
|
|
120 |
|
%macro SAD_8x8_SSE 0 |
121 |
|
movq mm0, [eax] |
122 |
|
movq mm1, [eax+ecx] |
123 |
|
|
124 |
|
psadbw mm0, [edx] |
125 |
|
psadbw mm1, [edx+ecx] |
126 |
|
add eax, ebx |
127 |
|
add edx, ebx |
128 |
|
|
129 |
|
paddusw mm5,mm0 |
130 |
|
paddusw mm6,mm1 |
131 |
|
%endmacro |
132 |
|
|
133 |
|
align 16 |
134 |
|
sad8_xmm: |
135 |
|
|
136 |
|
mov eax, [esp+ 4] ; Src1 |
137 |
|
mov edx, [esp+ 8] ; Src2 |
138 |
|
mov ecx, [esp+12] ; Stride |
139 |
|
push ebx |
140 |
|
lea ebx, [ecx+ecx] |
141 |
|
|
142 |
|
pxor mm5, mm5 ; accum1 |
143 |
|
pxor mm6, mm6 ; accum2 |
144 |
|
|
145 |
|
SAD_8x8_SSE |
146 |
|
SAD_8x8_SSE |
147 |
|
SAD_8x8_SSE |
148 |
|
|
149 |
|
movq mm0, [eax] |
150 |
|
movq mm1, [eax+ecx] |
151 |
|
psadbw mm0, [edx] |
152 |
|
psadbw mm1, [edx+ecx] |
153 |
|
|
154 |
|
pop ebx |
155 |
|
|
156 |
|
paddusw mm5,mm0 |
157 |
|
paddusw mm6,mm1 |
158 |
|
|
159 |
|
paddusw mm6,mm5 |
160 |
|
movd eax, mm6 |
161 |
|
|
162 |
|
ret |
163 |
|
|
164 |
|
|
165 |
|
;=========================================================================== |
166 |
|
; |
167 |
|
; uint32_t sad16bi_xmm(const uint8_t * const cur, |
168 |
|
; const uint8_t * const ref1, |
169 |
|
; const uint8_t * const ref2, |
170 |
|
; const uint32_t stride); |
171 |
|
; |
172 |
|
;=========================================================================== |
173 |
|
|
174 |
%macro SADBI_16x16_SSE 0 |
%macro SADBI_16x16_SSE 0 |
175 |
movq mm0, [eax] |
movq mm0, [eax] |
176 |
movq mm1, [eax+8] |
movq mm1, [eax+8] |
236 |
; |
; |
237 |
;=========================================================================== |
;=========================================================================== |
238 |
|
|
239 |
;=========================================================================== |
%macro SADBI_8x8_XMM 0 |
|
; |
|
|
; uint32_t sad8_xmm(const uint8_t * const cur, |
|
|
; const uint8_t * const ref, |
|
|
; const uint32_t stride); |
|
|
; |
|
|
;=========================================================================== |
|
|
|
|
|
%macro SAD_8x8_SSE 0 |
|
240 |
movq mm0, [eax] |
movq mm0, [eax] |
241 |
movq mm1, [eax+ecx] |
movq mm1, [eax+ecx] |
|
%macro MEAN_16x16_SSE 0 |
|
|
psadbw mm0, [edx] |
|
|
psadbw mm1, [edx+ecx] |
|
|
add eax, ebx |
|
|
add edx, ebx |
|
242 |
|
|
243 |
|
movq mm2, [edx] |
244 |
|
movq mm3, [edx+ecx] |
245 |
|
|
246 |
|
pavgb mm2, [ebx] |
247 |
|
lea edx, [edx+2*ecx] |
248 |
|
|
249 |
|
pavgb mm3, [ebx+ecx] |
250 |
|
lea ebx, [ebx+2*ecx] |
251 |
|
|
252 |
|
psadbw mm0, mm2 |
253 |
|
lea eax, [eax+2*ecx] |
254 |
|
|
255 |
|
psadbw mm1, mm3 |
256 |
paddusw mm5,mm0 |
paddusw mm5,mm0 |
257 |
|
|
258 |
paddusw mm6,mm1 |
paddusw mm6,mm1 |
259 |
%endmacro |
%endmacro |
260 |
|
|
261 |
align 16 |
align 16 |
262 |
sad8_xmm: |
sad8bi_xmm: |
|
|
|
|
mov eax, [esp+ 4] ; Src1 |
|
|
mov edx, [esp+ 8] ; Src2 |
|
|
mov ecx, [esp+12] ; Stride |
|
263 |
push ebx |
push ebx |
264 |
lea ebx, [ecx+ecx] |
mov eax, [esp+4+ 4] ; Src |
265 |
|
mov edx, [esp+4+ 8] ; Ref1 |
266 |
|
mov ebx, [esp+4+12] ; Ref2 |
267 |
|
mov ecx, [esp+4+16] ; Stride |
268 |
|
|
269 |
pxor mm5, mm5 ; accum1 |
pxor mm5, mm5 ; accum1 |
270 |
pxor mm6, mm6 ; accum2 |
pxor mm6, mm6 ; accum2 |
271 |
|
.Loop |
272 |
|
SADBI_8x8_XMM |
273 |
|
SADBI_8x8_XMM |
274 |
|
SADBI_8x8_XMM |
275 |
|
SADBI_8x8_XMM |
276 |
|
|
277 |
SAD_8x8_SSE |
paddusw mm6,mm5 |
278 |
SAD_8x8_SSE |
movd eax, mm6 |
|
SAD_8x8_SSE |
|
|
|
|
|
movq mm0, [eax] |
|
|
movq mm1, [eax+ecx] |
|
|
psadbw mm0, [edx] |
|
|
psadbw mm1, [edx+ecx] |
|
|
|
|
279 |
pop ebx |
pop ebx |
280 |
|
ret |
281 |
|
|
|
paddusw mm5,mm0 |
|
|
paddusw mm6,mm1 |
|
282 |
|
|
283 |
paddusw mm6,mm5 |
;=========================================================================== |
284 |
movd eax, mm6 |
; |
285 |
|
; uint32_t dev16_xmm(const uint8_t * const cur, |
286 |
|
; const uint32_t stride); |
287 |
|
; |
288 |
|
;=========================================================================== |
289 |
|
|
290 |
ret |
%macro MEAN_16x16_SSE 0 |
291 |
movq mm0, [eax] |
movq mm0, [eax] |
292 |
movq mm1, [eax+8] |
movq mm1, [eax+8] |
293 |
psadbw mm0, mm7 |
psadbw mm0, mm7 |