21 |
; * along with this program ; if not, write to the Free Software |
; * along with this program ; if not, write to the Free Software |
22 |
; * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
; * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
23 |
; * |
; * |
24 |
; * $Id: quantize_h263_mmx.asm,v 1.6 2004-08-22 11:46:10 edgomez Exp $ |
; * $Id: quantize_h263_mmx.asm,v 1.11 2008-11-26 23:35:50 Isibaar Exp $ |
25 |
; * |
; * |
26 |
; ****************************************************************************/ |
; ****************************************************************************/ |
27 |
|
|
28 |
; enable dequant saturate [-2048,2047], test purposes only. |
; enable dequant saturate [-2048,2047], test purposes only. |
29 |
%define SATURATE |
%define SATURATE |
30 |
|
|
31 |
BITS 32 |
%include "nasm.inc" |
|
|
|
|
%macro cglobal 1 |
|
|
%ifdef PREFIX |
|
|
%ifdef MARK_FUNCS |
|
|
global _%1:function |
|
|
%define %1 _%1:function |
|
|
%else |
|
|
global _%1 |
|
|
%define %1 _%1 |
|
|
%endif |
|
|
%else |
|
|
%ifdef MARK_FUNCS |
|
|
global %1:function |
|
|
%else |
|
|
global %1 |
|
|
%endif |
|
|
%endif |
|
|
%endmacro |
|
32 |
|
|
33 |
;============================================================================= |
;============================================================================= |
34 |
; Read only Local data |
; Read only Local data |
35 |
;============================================================================= |
;============================================================================= |
36 |
|
|
37 |
%ifdef FORMAT_COFF |
DATA |
|
SECTION .rodata |
|
|
%else |
|
|
SECTION .rodata align=16 |
|
|
%endif |
|
38 |
|
|
39 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
40 |
plus_one: |
plus_one: |
41 |
times 8 dw 1 |
times 8 dw 1 |
42 |
|
|
46 |
; |
; |
47 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
48 |
|
|
49 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
50 |
mmx_quant: |
mmx_quant: |
51 |
%assign quant 0 |
%assign quant 0 |
52 |
%rep 32 |
%rep 32 |
60 |
; |
; |
61 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
62 |
|
|
63 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
64 |
mmx_sub: |
mmx_sub: |
65 |
%assign quant 1 |
%assign quant 1 |
66 |
%rep 31 |
%rep 31 |
78 |
; |
; |
79 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
80 |
|
|
81 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
82 |
mmx_div: |
mmx_div: |
83 |
%assign quant 1 |
%assign quant 1 |
84 |
%rep 31 |
%rep 31 |
90 |
; Code |
; Code |
91 |
;============================================================================= |
;============================================================================= |
92 |
|
|
93 |
SECTION .text |
SECTION .rotext align=SECTION_ALIGN |
94 |
|
|
95 |
cglobal quant_h263_intra_mmx |
cglobal quant_h263_intra_mmx |
96 |
cglobal quant_h263_intra_sse2 |
cglobal quant_h263_intra_sse2 |
113 |
; |
; |
114 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
115 |
|
|
116 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
117 |
quant_h263_intra_mmx: |
quant_h263_intra_mmx: |
118 |
|
|
119 |
push esi |
mov _EAX, prm2 ; data |
120 |
|
mov TMP0, prm4 ; dcscalar |
121 |
mov esi, [esp + 4 + 8] ; data |
movsx _EAX, word [_EAX] ; data[0] |
122 |
mov ecx,[esp + 4 + 16] ; dcscalar |
|
123 |
movsx eax, word [esi] ; data[0] |
sar TMP0, 1 ; dcscalar /2 |
124 |
|
mov TMP1, _EAX |
125 |
sar ecx,1 ; dcscalar /2 |
sar TMP1, 31 ; sgn(data[0]) |
126 |
mov edx,eax |
xor TMP0,TMP1 ; *sgn(data[0]) |
127 |
sar edx,31 ; sgn(data[0]) |
sub _EAX,TMP1 |
128 |
xor ecx,edx ; *sgn(data[0]) |
add _EAX,TMP0 ; + (dcscalar/2)*sgn(data[0]) |
129 |
sub eax,edx |
|
130 |
add eax,ecx ; + (dcscalar/2)*sgn(data[0]) |
mov TMP0, prm3 ; quant |
131 |
|
lea TMP1, [mmx_div] |
132 |
mov ecx, [esp + 4 + 12] ; quant |
movq mm7, [TMP1+TMP0 * 8 - 8] |
133 |
|
%ifdef ARCH_IS_X86_64 |
134 |
|
%ifdef WINDOWS |
135 |
|
mov TMP1, prm2 |
136 |
|
%endif |
137 |
|
%endif |
138 |
cdq |
cdq |
139 |
idiv dword [esp + 4 + 16] ; dcscalar |
idiv prm4d ; dcscalar |
140 |
cmp ecx, 1 |
%ifdef ARCH_IS_X86_64 |
141 |
mov edx, [esp + 4 + 4] ; coeff |
%ifdef WINDOWS |
142 |
|
mov prm2, TMP1 |
143 |
|
%endif |
144 |
|
%endif |
145 |
|
cmp TMP0, 1 |
146 |
|
mov TMP1, prm1 ; coeff |
147 |
je .low |
je .low |
148 |
|
|
149 |
movq mm7, [mmx_div+ecx * 8 - 8] |
mov TMP0, prm2 ; data |
150 |
mov ecx,4 |
push _EAX ; DC |
151 |
|
mov _EAX, TMP0 |
152 |
|
|
153 |
|
mov TMP0,4 |
154 |
|
|
155 |
.loop |
.loop: |
156 |
movq mm0, [esi] ; data |
movq mm0, [_EAX] ; data |
157 |
pxor mm4,mm4 |
pxor mm4,mm4 |
158 |
movq mm1, [esi + 8] |
movq mm1, [_EAX + 8] |
159 |
pcmpgtw mm4,mm0 ; (data<0) |
pcmpgtw mm4,mm0 ; (data<0) |
160 |
pxor mm5,mm5 |
pxor mm5,mm5 |
161 |
pmulhw mm0,mm7 ; /(2*quant) |
pmulhw mm0,mm7 ; /(2*quant) |
162 |
pcmpgtw mm5,mm1 |
pcmpgtw mm5,mm1 |
163 |
movq mm2, [esi+16] |
movq mm2, [_EAX+16] |
164 |
psubw mm0,mm4 ; +(data<0) |
psubw mm0,mm4 ; +(data<0) |
165 |
pmulhw mm1,mm7 |
pmulhw mm1,mm7 |
166 |
pxor mm4,mm4 |
pxor mm4,mm4 |
167 |
movq mm3,[esi+24] |
movq mm3,[_EAX+24] |
168 |
pcmpgtw mm4,mm2 |
pcmpgtw mm4,mm2 |
169 |
psubw mm1,mm5 |
psubw mm1,mm5 |
170 |
pmulhw mm2,mm7 |
pmulhw mm2,mm7 |
173 |
pmulhw mm3,mm7 |
pmulhw mm3,mm7 |
174 |
psubw mm2,mm4 |
psubw mm2,mm4 |
175 |
psubw mm3,mm5 |
psubw mm3,mm5 |
176 |
movq [edx], mm0 |
movq [TMP1], mm0 |
177 |
lea esi, [esi+32] |
lea _EAX, [_EAX+32] |
178 |
movq [edx + 8], mm1 |
movq [TMP1 + 8], mm1 |
179 |
movq [edx + 16], mm2 |
movq [TMP1 + 16], mm2 |
180 |
movq [edx + 24], mm3 |
movq [TMP1 + 24], mm3 |
181 |
|
|
182 |
dec ecx |
dec TMP0 |
183 |
lea edx, [edx+32] |
lea TMP1, [TMP1+32] |
184 |
jne .loop |
jne .loop |
185 |
jmp .end |
jmp .end |
186 |
|
|
187 |
.low |
.low: |
188 |
movd mm7,ecx |
movd mm7,TMP0d |
189 |
mov ecx,4 |
|
190 |
.loop_low |
mov TMP0, prm2 |
191 |
movq mm0, [esi] |
push _EAX |
192 |
|
mov _EAX, TMP0 |
193 |
|
|
194 |
|
mov TMP0,4 |
195 |
|
.loop_low: |
196 |
|
movq mm0, [_EAX] |
197 |
pxor mm4,mm4 |
pxor mm4,mm4 |
198 |
movq mm1, [esi + 8] |
movq mm1, [_EAX + 8] |
199 |
pcmpgtw mm4,mm0 |
pcmpgtw mm4,mm0 |
200 |
pxor mm5,mm5 |
pxor mm5,mm5 |
201 |
psubw mm0,mm4 |
psubw mm0,mm4 |
202 |
pcmpgtw mm5,mm1 |
pcmpgtw mm5,mm1 |
203 |
psraw mm0,mm7 |
psraw mm0,mm7 |
204 |
psubw mm1,mm5 |
psubw mm1,mm5 |
205 |
movq mm2,[esi+16] |
movq mm2,[_EAX+16] |
206 |
pxor mm4,mm4 |
pxor mm4,mm4 |
207 |
psraw mm1,mm7 |
psraw mm1,mm7 |
208 |
pcmpgtw mm4,mm2 |
pcmpgtw mm4,mm2 |
209 |
pxor mm5,mm5 |
pxor mm5,mm5 |
210 |
psubw mm2,mm4 |
psubw mm2,mm4 |
211 |
movq mm3,[esi+24] |
movq mm3,[_EAX+24] |
212 |
pcmpgtw mm5,mm3 |
pcmpgtw mm5,mm3 |
213 |
psraw mm2,mm7 |
psraw mm2,mm7 |
214 |
psubw mm3,mm5 |
psubw mm3,mm5 |
215 |
movq [edx], mm0 |
movq [TMP1], mm0 |
216 |
psraw mm3,mm7 |
psraw mm3,mm7 |
217 |
movq [edx + 8], mm1 |
movq [TMP1 + 8], mm1 |
218 |
movq [edx+16],mm2 |
movq [TMP1+16],mm2 |
219 |
lea esi, [esi+32] |
lea _EAX, [_EAX+32] |
220 |
movq [edx+24],mm3 |
movq [TMP1+24],mm3 |
221 |
|
|
222 |
dec ecx |
dec TMP0 |
223 |
lea edx, [edx+32] |
lea TMP1, [TMP1+32] |
224 |
jne .loop_low |
jne .loop_low |
225 |
|
|
226 |
.end |
.end: |
227 |
mov edx, [esp + 4 + 4] ; coeff |
|
228 |
mov [edx],ax |
pop _EAX |
229 |
xor eax,eax ; return 0 |
|
230 |
|
mov TMP1, prm1 ; coeff |
231 |
|
mov [TMP1],ax |
232 |
|
xor _EAX,_EAX ; return 0 |
233 |
|
|
|
pop esi |
|
234 |
ret |
ret |
235 |
|
ENDFUNC |
236 |
|
|
237 |
|
|
238 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
245 |
; |
; |
246 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
247 |
|
|
248 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
249 |
quant_h263_intra_sse2: |
quant_h263_intra_sse2: |
250 |
|
|
251 |
push esi |
mov _EAX, prm2 ; data |
|
|
|
|
mov esi, [esp + 4 + 8] ; data |
|
252 |
|
|
253 |
movsx eax, word [esi] ; data[0] |
movsx _EAX, word [_EAX] ; data[0] |
254 |
|
|
255 |
mov ecx,[esp + 4 + 16] ; dcscalar |
mov TMP0,prm4 ; dcscalar |
256 |
mov edx,eax |
mov TMP1,_EAX |
257 |
sar ecx,1 |
sar TMP0,1 |
258 |
add eax,ecx |
add _EAX,TMP0 |
259 |
sub edx,ecx |
sub TMP1,TMP0 |
260 |
cmovl eax,edx ; +/- dcscalar/2 |
cmovl _EAX,TMP1 ; +/- dcscalar/2 |
261 |
mov ecx, [esp + 4 + 12] ; quant |
mov TMP0, prm3 ; quant |
262 |
|
lea TMP1, [mmx_div] |
263 |
|
movq xmm7, [TMP1+TMP0 * 8 - 8] |
264 |
|
|
265 |
|
%ifdef ARCH_IS_X86_64 |
266 |
|
%ifdef WINDOWS |
267 |
|
mov TMP1, prm2 |
268 |
|
%endif |
269 |
|
%endif |
270 |
cdq |
cdq |
271 |
idiv dword [esp + 4 + 16] ; dcscalar |
idiv prm4d ; dcscalar |
272 |
cmp ecx, 1 |
%ifdef ARCH_IS_X86_64 |
273 |
mov edx, [esp + 4 + 4] ; coeff |
%ifdef WINDOWS |
274 |
movq xmm7, [mmx_div+ecx * 8 - 8] |
mov prm2, TMP1 |
275 |
je .low |
%endif |
276 |
|
%endif |
277 |
|
cmp TMP0, 1 |
278 |
|
mov TMP1, prm1 ; coeff |
279 |
|
je near .low |
280 |
|
|
281 |
|
mov TMP0, prm2 |
282 |
|
push _EAX ; DC |
283 |
|
mov _EAX, TMP0 |
284 |
|
|
285 |
mov ecx,2 |
mov TMP0,2 |
286 |
movlhps xmm7,xmm7 |
movlhps xmm7,xmm7 |
287 |
|
|
288 |
.loop |
.loop: |
289 |
movdqa xmm0, [esi] |
movdqa xmm0, [_EAX] |
290 |
pxor xmm4,xmm4 |
pxor xmm4,xmm4 |
291 |
movdqa xmm1, [esi + 16] |
movdqa xmm1, [_EAX + 16] |
292 |
pcmpgtw xmm4,xmm0 |
pcmpgtw xmm4,xmm0 |
293 |
pxor xmm5,xmm5 |
pxor xmm5,xmm5 |
294 |
pmulhw xmm0,xmm7 |
pmulhw xmm0,xmm7 |
295 |
pcmpgtw xmm5,xmm1 |
pcmpgtw xmm5,xmm1 |
296 |
movdqa xmm2, [esi+32] |
movdqa xmm2, [_EAX+32] |
297 |
psubw xmm0,xmm4 |
psubw xmm0,xmm4 |
298 |
pmulhw xmm1,xmm7 |
pmulhw xmm1,xmm7 |
299 |
pxor xmm4,xmm4 |
pxor xmm4,xmm4 |
300 |
movdqa xmm3,[esi+48] |
movdqa xmm3,[_EAX+48] |
301 |
pcmpgtw xmm4,xmm2 |
pcmpgtw xmm4,xmm2 |
302 |
psubw xmm1,xmm5 |
psubw xmm1,xmm5 |
303 |
pmulhw xmm2,xmm7 |
pmulhw xmm2,xmm7 |
306 |
pmulhw xmm3,xmm7 |
pmulhw xmm3,xmm7 |
307 |
psubw xmm2,xmm4 |
psubw xmm2,xmm4 |
308 |
psubw xmm3,xmm5 |
psubw xmm3,xmm5 |
309 |
movdqa [edx], xmm0 |
movdqa [TMP1], xmm0 |
310 |
lea esi, [esi+64] |
lea _EAX, [_EAX+64] |
311 |
movdqa [edx + 16], xmm1 |
movdqa [TMP1 + 16], xmm1 |
312 |
movdqa [edx + 32], xmm2 |
movdqa [TMP1 + 32], xmm2 |
313 |
movdqa [edx + 48], xmm3 |
movdqa [TMP1 + 48], xmm3 |
314 |
|
|
315 |
dec ecx |
dec TMP0 |
316 |
lea edx, [edx+64] |
lea TMP1, [TMP1+64] |
317 |
jne .loop |
jne .loop |
318 |
jmp .end |
jmp .end |
319 |
|
|
320 |
.low |
.low: |
321 |
movd xmm7,ecx |
movd xmm7,TMP0d |
322 |
mov ecx,2 |
|
323 |
.loop_low |
mov TMP0, prm2 |
324 |
movdqa xmm0, [esi] |
push _EAX ; DC |
325 |
|
mov _EAX, TMP0 |
326 |
|
|
327 |
|
mov TMP0,2 |
328 |
|
.loop_low: |
329 |
|
movdqa xmm0, [_EAX] |
330 |
pxor xmm4,xmm4 |
pxor xmm4,xmm4 |
331 |
movdqa xmm1, [esi + 16] |
movdqa xmm1, [_EAX + 16] |
332 |
pcmpgtw xmm4,xmm0 |
pcmpgtw xmm4,xmm0 |
333 |
pxor xmm5,xmm5 |
pxor xmm5,xmm5 |
334 |
psubw xmm0,xmm4 |
psubw xmm0,xmm4 |
335 |
pcmpgtw xmm5,xmm1 |
pcmpgtw xmm5,xmm1 |
336 |
psraw xmm0,xmm7 |
psraw xmm0,xmm7 |
337 |
psubw xmm1,xmm5 |
psubw xmm1,xmm5 |
338 |
movdqa xmm2,[esi+32] |
movdqa xmm2,[_EAX+32] |
339 |
pxor xmm4,xmm4 |
pxor xmm4,xmm4 |
340 |
psraw xmm1,xmm7 |
psraw xmm1,xmm7 |
341 |
pcmpgtw xmm4,xmm2 |
pcmpgtw xmm4,xmm2 |
342 |
pxor xmm5,xmm5 |
pxor xmm5,xmm5 |
343 |
psubw xmm2,xmm4 |
psubw xmm2,xmm4 |
344 |
movdqa xmm3,[esi+48] |
movdqa xmm3,[_EAX+48] |
345 |
pcmpgtw xmm5,xmm3 |
pcmpgtw xmm5,xmm3 |
346 |
psraw xmm2,xmm7 |
psraw xmm2,xmm7 |
347 |
psubw xmm3,xmm5 |
psubw xmm3,xmm5 |
348 |
movdqa [edx], xmm0 |
movdqa [TMP1], xmm0 |
349 |
psraw xmm3,xmm7 |
psraw xmm3,xmm7 |
350 |
movdqa [edx+16], xmm1 |
movdqa [TMP1+16], xmm1 |
351 |
movdqa [edx+32],xmm2 |
movdqa [TMP1+32],xmm2 |
352 |
lea esi, [esi+64] |
lea _EAX, [_EAX+64] |
353 |
movdqa [edx+48],xmm3 |
movdqa [TMP1+48],xmm3 |
354 |
|
|
355 |
dec ecx |
dec TMP0 |
356 |
lea edx, [edx+64] |
lea TMP1, [TMP1+64] |
357 |
jne .loop_low |
jne .loop_low |
358 |
|
|
359 |
.end |
.end: |
360 |
mov edx, [esp + 4 + 4] ; coeff |
|
361 |
mov [edx],ax |
pop _EAX |
362 |
xor eax,eax ; return 0 |
|
363 |
|
mov TMP1, prm1 ; coeff |
364 |
|
mov [TMP1],ax |
365 |
|
xor _EAX,_EAX ; return 0 |
366 |
|
|
|
pop esi |
|
367 |
ret |
ret |
368 |
|
ENDFUNC |
369 |
|
|
370 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
371 |
; |
; |
376 |
; |
; |
377 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
378 |
|
|
379 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
380 |
quant_h263_inter_mmx: |
quant_h263_inter_mmx: |
381 |
|
|
382 |
push ecx |
mov TMP1, prm1 ; coeff |
383 |
push esi |
mov _EAX, prm3 ; quant |
|
push edi |
|
|
|
|
|
mov edi, [esp + 12 + 4] ; coeff |
|
|
mov esi, [esp + 12 + 8] ; data |
|
|
mov eax, [esp + 12 + 12] ; quant |
|
|
|
|
|
xor ecx, ecx |
|
384 |
|
|
385 |
pxor mm5, mm5 ; sum |
pxor mm5, mm5 ; sum |
386 |
movq mm6, [mmx_sub + eax * 8 - 8] ; sub |
lea TMP0, [mmx_sub] |
387 |
|
movq mm6, [TMP0 + _EAX * 8 - 8] ; sub |
388 |
|
|
389 |
cmp al, 1 |
cmp al, 1 |
390 |
jz .q1loop |
jz near .q1routine |
391 |
|
|
392 |
|
lea TMP0, [mmx_div] |
393 |
|
movq mm7, [TMP0 + _EAX * 8 - 8] ; divider |
394 |
|
|
395 |
movq mm7, [mmx_div + eax * 8 - 8] ; divider |
xor TMP0, TMP0 |
396 |
|
mov _EAX, prm2 ; data |
397 |
|
|
398 |
ALIGN 8 |
ALIGN SECTION_ALIGN |
399 |
.loop |
.loop: |
400 |
movq mm0, [esi + 8*ecx] ; mm0 = [1st] |
movq mm0, [_EAX + 8*TMP0] ; mm0 = [1st] |
401 |
movq mm3, [esi + 8*ecx + 8] |
movq mm3, [_EAX + 8*TMP0 + 8] |
402 |
pxor mm1, mm1 ; mm1 = 0 |
pxor mm1, mm1 ; mm1 = 0 |
403 |
pxor mm4, mm4 ; |
pxor mm4, mm4 ; |
404 |
pcmpgtw mm1, mm0 ; mm1 = (0 > mm0) |
pcmpgtw mm1, mm0 ; mm1 = (0 > mm0) |
417 |
pxor mm3, mm4 ; |
pxor mm3, mm4 ; |
418 |
psubw mm0, mm1 ; undisplace |
psubw mm0, mm1 ; undisplace |
419 |
psubw mm3, mm4 |
psubw mm3, mm4 |
420 |
movq [edi + 8*ecx], mm0 |
movq [TMP1 + 8*TMP0], mm0 |
421 |
movq [edi + 8*ecx + 8], mm3 |
movq [TMP1 + 8*TMP0 + 8], mm3 |
422 |
|
|
423 |
add ecx, 2 |
add TMP0, 2 |
424 |
cmp ecx, 16 |
cmp TMP0, 16 |
425 |
jnz .loop |
jnz .loop |
426 |
|
|
427 |
.done |
.done: |
428 |
pmaddwd mm5, [plus_one] |
pmaddwd mm5, [plus_one] |
429 |
movq mm0, mm5 |
movq mm0, mm5 |
430 |
psrlq mm5, 32 |
psrlq mm5, 32 |
431 |
paddd mm0, mm5 |
paddd mm0, mm5 |
432 |
|
|
433 |
movd eax, mm0 ; return sum |
movd eax, mm0 ; return sum |
|
pop edi |
|
|
pop esi |
|
|
pop ecx |
|
434 |
|
|
435 |
ret |
ret |
436 |
|
|
437 |
ALIGN 8 |
.q1routine: |
438 |
.q1loop |
xor TMP0, TMP0 |
439 |
movq mm0, [esi + 8*ecx] ; mm0 = [1st] |
mov _EAX, prm2 ; data |
440 |
movq mm3, [esi + 8*ecx+ 8] ; |
|
441 |
|
ALIGN SECTION_ALIGN |
442 |
|
.q1loop: |
443 |
|
movq mm0, [_EAX + 8*TMP0] ; mm0 = [1st] |
444 |
|
movq mm3, [_EAX + 8*TMP0+ 8] ; |
445 |
pxor mm1, mm1 ; mm1 = 0 |
pxor mm1, mm1 ; mm1 = 0 |
446 |
pxor mm4, mm4 ; |
pxor mm4, mm4 ; |
447 |
pcmpgtw mm1, mm0 ; mm1 = (0 > mm0) |
pcmpgtw mm1, mm0 ; mm1 = (0 > mm0) |
460 |
pxor mm3, mm4 ; |
pxor mm3, mm4 ; |
461 |
psubw mm0, mm1 ; undisplace |
psubw mm0, mm1 ; undisplace |
462 |
psubw mm3, mm4 |
psubw mm3, mm4 |
463 |
movq [edi + 8*ecx], mm0 |
movq [TMP1 + 8*TMP0], mm0 |
464 |
movq [edi + 8*ecx + 8], mm3 |
movq [TMP1 + 8*TMP0 + 8], mm3 |
465 |
|
|
466 |
add ecx, 2 |
add TMP0, 2 |
467 |
cmp ecx, 16 |
cmp TMP0, 16 |
468 |
jnz .q1loop |
jnz .q1loop |
469 |
|
|
470 |
jmp .done |
jmp .done |
471 |
|
ENDFUNC |
472 |
|
|
473 |
|
|
474 |
|
|
481 |
; |
; |
482 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
483 |
|
|
484 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
485 |
quant_h263_inter_sse2: |
quant_h263_inter_sse2: |
486 |
|
|
487 |
push esi |
mov TMP1, prm1 ; coeff |
488 |
push edi |
mov _EAX, prm3 ; quant |
|
|
|
|
mov edi, [esp + 8 + 4] ; coeff |
|
|
mov esi, [esp + 8 + 8] ; data |
|
|
mov eax, [esp + 8 + 12] ; quant |
|
|
|
|
|
xor ecx, ecx |
|
489 |
|
|
490 |
pxor xmm5, xmm5 ; sum |
pxor xmm5, xmm5 ; sum |
491 |
|
|
492 |
movq mm0, [mmx_sub + eax*8 - 8] ; sub |
lea TMP0, [mmx_sub] |
493 |
|
movq mm0, [TMP0 + _EAX*8 - 8] ; sub |
494 |
movq2dq xmm6, mm0 ; load into low 8 bytes |
movq2dq xmm6, mm0 ; load into low 8 bytes |
495 |
movlhps xmm6, xmm6 ; duplicate into high 8 bytes |
movlhps xmm6, xmm6 ; duplicate into high 8 bytes |
496 |
|
|
497 |
cmp al, 1 |
cmp al, 1 |
498 |
jz near .qes2_q1loop |
jz near .qes2_q1_routine |
499 |
|
|
500 |
|
.qes2_not1: |
501 |
|
lea TMP0, [mmx_div] |
502 |
|
movq mm0, [TMP0 + _EAX*8 - 8] ; divider |
503 |
|
|
504 |
|
xor TMP0, TMP0 |
505 |
|
mov _EAX, prm2 ; data |
506 |
|
|
|
.qes2_not1 |
|
|
movq mm0, [mmx_div + eax*8 - 8] ; divider |
|
507 |
movq2dq xmm7, mm0 |
movq2dq xmm7, mm0 |
508 |
movlhps xmm7, xmm7 |
movlhps xmm7, xmm7 |
509 |
|
|
510 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
511 |
.qes2_loop |
.qes2_loop: |
512 |
movdqa xmm0, [esi + ecx*8] ; xmm0 = [1st] |
movdqa xmm0, [_EAX + TMP0*8] ; xmm0 = [1st] |
513 |
movdqa xmm3, [esi + ecx*8 + 16] ; xmm3 = [2nd] |
movdqa xmm3, [_EAX + TMP0*8 + 16] ; xmm3 = [2nd] |
514 |
pxor xmm1, xmm1 |
pxor xmm1, xmm1 |
515 |
pxor xmm4, xmm4 |
pxor xmm4, xmm4 |
516 |
pcmpgtw xmm1, xmm0 |
pcmpgtw xmm1, xmm0 |
529 |
pxor xmm3, xmm4 |
pxor xmm3, xmm4 |
530 |
psubw xmm0, xmm1 |
psubw xmm0, xmm1 |
531 |
psubw xmm3, xmm4 |
psubw xmm3, xmm4 |
532 |
movdqa [edi + ecx*8], xmm0 |
movdqa [TMP1 + TMP0*8], xmm0 |
533 |
movdqa [edi + ecx*8 + 16], xmm3 |
movdqa [TMP1 + TMP0*8 + 16], xmm3 |
534 |
|
|
535 |
add ecx, 4 |
add TMP0, 4 |
536 |
cmp ecx, 16 |
cmp TMP0, 16 |
537 |
jnz .qes2_loop |
jnz .qes2_loop |
538 |
|
|
539 |
.qes2_done |
.qes2_done: |
540 |
movdqu xmm6, [plus_one] |
movdqu xmm6, [plus_one] |
541 |
pmaddwd xmm5, xmm6 |
pmaddwd xmm5, xmm6 |
542 |
movhlps xmm6, xmm5 |
movhlps xmm6, xmm5 |
549 |
|
|
550 |
movd eax, mm0 ; return sum |
movd eax, mm0 ; return sum |
551 |
|
|
|
pop edi |
|
|
pop esi |
|
|
|
|
552 |
ret |
ret |
553 |
|
|
554 |
ALIGN 16 |
.qes2_q1_routine: |
555 |
.qes2_q1loop |
xor TMP0, TMP0 |
556 |
movdqa xmm0, [esi + ecx*8] ; xmm0 = [1st] |
mov _EAX, prm2 ; data |
557 |
movdqa xmm3, [esi + ecx*8 + 16] ; xmm3 = [2nd] |
|
558 |
|
ALIGN SECTION_ALIGN |
559 |
|
.qes2_q1loop: |
560 |
|
movdqa xmm0, [_EAX + TMP0*8] ; xmm0 = [1st] |
561 |
|
movdqa xmm3, [_EAX + TMP0*8 + 16] ; xmm3 = [2nd] |
562 |
pxor xmm1, xmm1 |
pxor xmm1, xmm1 |
563 |
pxor xmm4, xmm4 |
pxor xmm4, xmm4 |
564 |
pcmpgtw xmm1, xmm0 |
pcmpgtw xmm1, xmm0 |
577 |
pxor xmm3, xmm4 |
pxor xmm3, xmm4 |
578 |
psubw xmm0, xmm1 |
psubw xmm0, xmm1 |
579 |
psubw xmm3, xmm4 |
psubw xmm3, xmm4 |
580 |
movdqa [edi + ecx*8], xmm0 |
movdqa [TMP1 + TMP0*8], xmm0 |
581 |
movdqa [edi + ecx*8 + 16], xmm3 |
movdqa [TMP1 + TMP0*8 + 16], xmm3 |
582 |
|
|
583 |
add ecx, 4 |
add TMP0, 4 |
584 |
cmp ecx, 16 |
cmp TMP0, 16 |
585 |
jnz .qes2_q1loop |
jnz .qes2_q1loop |
586 |
jmp .qes2_done |
jmp .qes2_done |
587 |
|
ENDFUNC |
588 |
|
|
589 |
|
|
590 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
597 |
; |
; |
598 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
599 |
|
|
600 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
601 |
dequant_h263_intra_mmx: |
dequant_h263_intra_mmx: |
602 |
|
|
603 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
604 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
605 |
pcmpeqw mm0,mm0 |
pcmpeqw mm0,mm0 |
606 |
movq mm6, [mmx_quant + ecx*8] ; quant |
lea TMP1, [mmx_quant] |
607 |
shl ecx,31 ; quant & 1 ? 0 : - 1 |
movq mm6, [TMP1 + TMP0*8] ; quant |
608 |
|
shl TMP0,31 ; quant & 1 ? 0 : - 1 |
609 |
movq mm7,mm6 |
movq mm7,mm6 |
610 |
movq mm5,mm0 |
movq mm5,mm0 |
611 |
movd mm1,ecx |
movd mm1,TMP0d |
612 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
613 |
psllw mm0,mm1 |
psllw mm0,mm1 |
614 |
paddw mm7,mm7 ; 2*quant |
paddw mm7,mm7 ; 2*quant |
615 |
paddw mm6,mm0 ; quant-1 |
paddw mm6,mm0 ; quant-1 |
616 |
psllw mm5,12 |
psllw mm5,12 |
617 |
mov ecx,8 |
mov TMP0,8 |
618 |
psrlw mm5,1 |
psrlw mm5,1 |
619 |
|
|
620 |
.loop: |
.loop: |
621 |
movq mm0,[eax] |
movq mm0,[_EAX] |
622 |
pxor mm2,mm2 |
pxor mm2,mm2 |
623 |
pxor mm4,mm4 |
pxor mm4,mm4 |
624 |
pcmpgtw mm2,mm0 |
pcmpgtw mm2,mm0 |
625 |
pcmpeqw mm4,mm0 |
pcmpeqw mm4,mm0 |
626 |
pmullw mm0,mm7 ; * 2 * quant |
pmullw mm0,mm7 ; * 2 * quant |
627 |
movq mm1,[eax+8] |
movq mm1,[_EAX+8] |
628 |
psubw mm0,mm2 |
psubw mm0,mm2 |
629 |
pxor mm2,mm6 |
pxor mm2,mm6 |
630 |
pxor mm3,mm3 |
pxor mm3,mm3 |
648 |
paddsw mm0, mm5 |
paddsw mm0, mm5 |
649 |
paddsw mm1, mm5 |
paddsw mm1, mm5 |
650 |
|
|
651 |
movq [edx],mm0 |
movq [TMP1],mm0 |
652 |
lea eax,[eax+16] |
lea _EAX,[_EAX+16] |
653 |
movq [edx+8],mm1 |
movq [TMP1+8],mm1 |
654 |
|
|
655 |
dec ecx |
dec TMP0 |
656 |
lea edx,[edx+16] |
lea TMP1,[TMP1+16] |
657 |
jne .loop |
jne .loop |
658 |
|
|
659 |
; deal with DC |
; deal with DC |
660 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
661 |
movd mm1,[esp+16] ; dcscalar |
movd mm1,prm4d ; dcscalar |
662 |
movd mm0,[eax] ; coeff[0] |
movd mm0,[_EAX] ; coeff[0] |
663 |
pmullw mm0,mm1 ; * dcscalar |
pmullw mm0,mm1 ; * dcscalar |
664 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
665 |
paddsw mm0, mm5 ; saturate + |
paddsw mm0, mm5 ; saturate + |
666 |
psubsw mm0, mm5 |
psubsw mm0, mm5 |
667 |
psubsw mm0, mm5 ; saturate - |
psubsw mm0, mm5 ; saturate - |
668 |
paddsw mm0, mm5 |
paddsw mm0, mm5 |
669 |
movd eax,mm0 |
movd eax,mm0 |
670 |
mov [edx], ax |
mov [TMP1], ax |
671 |
|
|
672 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
673 |
ret |
ret |
674 |
|
ENDFUNC |
675 |
|
|
676 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
677 |
; |
; |
684 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
685 |
|
|
686 |
|
|
687 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
688 |
dequant_h263_intra_xmm: |
dequant_h263_intra_xmm: |
689 |
|
|
690 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
691 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
692 |
|
|
693 |
movd mm6,ecx ; quant |
movd mm6,TMP0d ; quant |
694 |
pcmpeqw mm0,mm0 |
pcmpeqw mm0,mm0 |
695 |
pshufw mm6,mm6,0 ; all quant |
pshufw mm6,mm6,0 ; all quant |
696 |
shl ecx,31 |
shl TMP0,31 |
697 |
movq mm5,mm0 |
movq mm5,mm0 |
698 |
movq mm7,mm6 |
movq mm7,mm6 |
699 |
movd mm1,ecx |
movd mm1,TMP0d |
700 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
701 |
psllw mm0,mm1 ; quant & 1 ? 0 : - 1 |
psllw mm0,mm1 ; quant & 1 ? 0 : - 1 |
702 |
movq mm4,mm5 |
movq mm4,mm5 |
703 |
paddw mm7,mm7 ; quant*2 |
paddw mm7,mm7 ; quant*2 |
704 |
paddw mm6,mm0 ; quant-1 |
paddw mm6,mm0 ; quant-1 |
705 |
psrlw mm4,5 ; mm4=2047 |
psrlw mm4,5 ; mm4=2047 |
706 |
mov ecx,8 |
mov TMP0,8 |
707 |
pxor mm5,mm4 ; mm5=-2048 |
pxor mm5,mm4 ; mm5=-2048 |
708 |
|
|
709 |
.loop: |
.loop: |
710 |
movq mm0,[eax] |
movq mm0,[_EAX] |
711 |
pxor mm2,mm2 |
pxor mm2,mm2 |
712 |
pxor mm3,mm3 |
pxor mm3,mm3 |
713 |
|
|
714 |
pcmpgtw mm2,mm0 |
pcmpgtw mm2,mm0 |
715 |
pcmpeqw mm3,mm0 ; if coeff==0... |
pcmpeqw mm3,mm0 ; if coeff==0... |
716 |
pmullw mm0,mm7 ; * 2 * quant |
pmullw mm0,mm7 ; * 2 * quant |
717 |
movq mm1,[eax+8] |
movq mm1,[_EAX+8] |
718 |
|
|
719 |
psubw mm0,mm2 |
psubw mm0,mm2 |
720 |
pxor mm2,mm6 |
pxor mm2,mm6 |
734 |
|
|
735 |
pmaxsw mm0,mm5 |
pmaxsw mm0,mm5 |
736 |
pminsw mm1,mm4 |
pminsw mm1,mm4 |
737 |
movq [edx],mm0 |
movq [TMP1],mm0 |
738 |
pmaxsw mm1,mm5 |
pmaxsw mm1,mm5 |
739 |
lea eax,[eax+16] |
lea _EAX,[_EAX+16] |
740 |
movq [edx+8],mm1 |
movq [TMP1+8],mm1 |
741 |
|
|
742 |
dec ecx |
dec TMP0 |
743 |
lea edx,[edx+16] |
lea TMP1,[TMP1+16] |
744 |
jne .loop |
jne .loop |
745 |
|
|
746 |
; deal with DC |
; deal with DC |
747 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
748 |
movd mm1,[esp+16] ; dcscalar |
movd mm1,prm4d ; dcscalar |
749 |
movd mm0, [eax] |
movd mm0, [_EAX] |
750 |
pmullw mm0, mm1 |
pmullw mm0, mm1 |
751 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
752 |
pminsw mm0,mm4 |
pminsw mm0,mm4 |
753 |
pmaxsw mm0,mm5 |
pmaxsw mm0,mm5 |
754 |
movd eax, mm0 |
movd eax, mm0 |
755 |
mov [edx], ax |
mov [TMP1], ax |
756 |
|
|
757 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
758 |
ret |
ret |
759 |
|
ENDFUNC |
760 |
|
|
761 |
|
|
762 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
769 |
; |
; |
770 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
771 |
|
|
772 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
773 |
dequant_h263_intra_sse2: |
dequant_h263_intra_sse2: |
774 |
|
|
775 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
776 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
777 |
|
|
778 |
movd xmm6,ecx ; quant |
movd xmm6,TMP0d ; quant |
779 |
|
|
780 |
shl ecx,31 |
shl TMP0,31 |
781 |
pshuflw xmm6,xmm6,0 |
pshuflw xmm6,xmm6,0 |
782 |
pcmpeqw xmm0,xmm0 |
pcmpeqw xmm0,xmm0 |
783 |
movlhps xmm6,xmm6 ; all quant |
movlhps xmm6,xmm6 ; all quant |
784 |
movd xmm1,ecx |
movd xmm1,TMP0d |
785 |
movdqa xmm5,xmm0 |
movdqa xmm5,xmm0 |
786 |
movdqa xmm7,xmm6 |
movdqa xmm7,xmm6 |
787 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
788 |
paddw xmm7,xmm7 ; quant *2 |
paddw xmm7,xmm7 ; quant *2 |
789 |
psllw xmm0,xmm1 ; quant & 1 ? 0 : - 1 |
psllw xmm0,xmm1 ; quant & 1 ? 0 : - 1 |
790 |
movdqa xmm4,xmm5 |
movdqa xmm4,xmm5 |
791 |
paddw xmm6,xmm0 ; quant-1 |
paddw xmm6,xmm0 ; quant-1 |
792 |
psrlw xmm4,5 ; 2047 |
psrlw xmm4,5 ; 2047 |
793 |
mov ecx,4 |
mov TMP0,4 |
794 |
pxor xmm5,xmm4 ; mm5=-2048 |
pxor xmm5,xmm4 ; mm5=-2048 |
795 |
|
|
796 |
.loop: |
.loop: |
797 |
movdqa xmm0,[eax] |
movdqa xmm0,[_EAX] |
798 |
pxor xmm2,xmm2 |
pxor xmm2,xmm2 |
799 |
pxor xmm3,xmm3 |
pxor xmm3,xmm3 |
800 |
|
|
801 |
pcmpgtw xmm2,xmm0 |
pcmpgtw xmm2,xmm0 |
802 |
pcmpeqw xmm3,xmm0 |
pcmpeqw xmm3,xmm0 |
803 |
pmullw xmm0,xmm7 ; * 2 * quant |
pmullw xmm0,xmm7 ; * 2 * quant |
804 |
movdqa xmm1,[eax+16] |
movdqa xmm1,[_EAX+16] |
805 |
|
|
806 |
psubw xmm0,xmm2 |
psubw xmm0,xmm2 |
807 |
pxor xmm2,xmm6 |
pxor xmm2,xmm6 |
821 |
|
|
822 |
pmaxsw xmm0,xmm5 |
pmaxsw xmm0,xmm5 |
823 |
pminsw xmm1,xmm4 |
pminsw xmm1,xmm4 |
824 |
movdqa [edx],xmm0 |
movdqa [TMP1],xmm0 |
825 |
pmaxsw xmm1,xmm5 |
pmaxsw xmm1,xmm5 |
826 |
lea eax,[eax+32] |
lea _EAX,[_EAX+32] |
827 |
movdqa [edx+16],xmm1 |
movdqa [TMP1+16],xmm1 |
828 |
|
|
829 |
dec ecx |
dec TMP0 |
830 |
lea edx,[edx+32] |
lea TMP1,[TMP1+32] |
831 |
jne .loop |
jne .loop |
832 |
|
|
833 |
; deal with DC |
; deal with DC |
834 |
|
|
835 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
836 |
movsx eax,word [eax] |
movsx _EAX,word [_EAX] |
837 |
imul dword [esp+16] ; dcscalar |
imul prm4d ; dcscalar |
838 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
839 |
movd xmm0,eax |
movd xmm0,eax |
840 |
pminsw xmm0,xmm4 |
pminsw xmm0,xmm4 |
841 |
pmaxsw xmm0,xmm5 |
pmaxsw xmm0,xmm5 |
842 |
movd eax,xmm0 |
movd eax,xmm0 |
843 |
|
|
844 |
mov [edx], ax |
mov [TMP1], ax |
845 |
|
|
846 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
847 |
ret |
ret |
848 |
|
ENDFUNC |
849 |
|
|
850 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
851 |
; |
; |
856 |
; |
; |
857 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
858 |
|
|
859 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
860 |
dequant_h263_inter_mmx: |
dequant_h263_inter_mmx: |
861 |
|
|
862 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
863 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
864 |
pcmpeqw mm0,mm0 |
pcmpeqw mm0,mm0 |
865 |
movq mm6, [mmx_quant + ecx*8] ; quant |
lea TMP1, [mmx_quant] |
866 |
shl ecx,31 ; odd/even |
movq mm6, [TMP1 + TMP0*8] ; quant |
867 |
|
shl TMP0,31 ; odd/even |
868 |
movq mm7,mm6 |
movq mm7,mm6 |
869 |
movd mm1,ecx |
movd mm1,TMP0d |
870 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
871 |
movq mm5,mm0 |
movq mm5,mm0 |
872 |
psllw mm0,mm1 ; quant & 1 ? 0 : - 1 |
psllw mm0,mm1 ; quant & 1 ? 0 : - 1 |
873 |
paddw mm7,mm7 ; quant*2 |
paddw mm7,mm7 ; quant*2 |
874 |
paddw mm6,mm0 ; quant & 1 ? quant : quant - 1 |
paddw mm6,mm0 ; quant & 1 ? quant : quant - 1 |
875 |
psllw mm5,12 |
psllw mm5,12 |
876 |
mov ecx,8 |
mov TMP0,8 |
877 |
psrlw mm5,1 ; 32767-2047 (32768-2048) |
psrlw mm5,1 ; 32767-2047 (32768-2048) |
878 |
|
|
879 |
.loop: |
.loop: |
880 |
movq mm0,[eax] |
movq mm0,[_EAX] |
881 |
pxor mm4,mm4 |
pxor mm4,mm4 |
882 |
pxor mm2,mm2 |
pxor mm2,mm2 |
883 |
pcmpeqw mm4,mm0 ; if coeff==0... |
pcmpeqw mm4,mm0 ; if coeff==0... |
885 |
pmullw mm0,mm7 ; * 2 * quant |
pmullw mm0,mm7 ; * 2 * quant |
886 |
pxor mm3,mm3 |
pxor mm3,mm3 |
887 |
psubw mm0,mm2 |
psubw mm0,mm2 |
888 |
movq mm1,[eax+8] |
movq mm1,[_EAX+8] |
889 |
pxor mm2,mm6 |
pxor mm2,mm6 |
890 |
pcmpgtw mm3,mm1 |
pcmpgtw mm3,mm1 |
891 |
pandn mm4,mm2 ; ... then data==0 |
pandn mm4,mm2 ; ... then data==0 |
907 |
paddsw mm0, mm5 |
paddsw mm0, mm5 |
908 |
paddsw mm1, mm5 |
paddsw mm1, mm5 |
909 |
|
|
910 |
movq [edx],mm0 |
movq [TMP1],mm0 |
911 |
lea eax,[eax+16] |
lea _EAX,[_EAX+16] |
912 |
movq [edx+8],mm1 |
movq [TMP1+8],mm1 |
913 |
|
|
914 |
dec ecx |
dec TMP0 |
915 |
lea edx,[edx+16] |
lea TMP1,[TMP1+16] |
916 |
jne .loop |
jne .loop |
917 |
|
|
918 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
919 |
ret |
ret |
920 |
|
ENDFUNC |
921 |
|
|
922 |
|
|
923 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
928 |
; const uint16_t *mpeg_matrices); |
; const uint16_t *mpeg_matrices); |
929 |
; |
; |
930 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
931 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
932 |
dequant_h263_inter_xmm: |
dequant_h263_inter_xmm: |
933 |
|
|
934 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
935 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
936 |
pcmpeqw mm0,mm0 |
pcmpeqw mm0,mm0 |
937 |
movq mm6, [mmx_quant + ecx*8] ; quant |
lea TMP1, [mmx_quant] |
938 |
shl ecx,31 |
movq mm6, [TMP1 + TMP0*8] ; quant |
939 |
|
shl TMP0,31 |
940 |
movq mm5,mm0 |
movq mm5,mm0 |
941 |
movd mm1,ecx |
movd mm1,TMP0d |
942 |
movq mm7,mm6 |
movq mm7,mm6 |
943 |
psllw mm0,mm1 |
psllw mm0,mm1 |
944 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
945 |
movq mm4,mm5 |
movq mm4,mm5 |
946 |
paddw mm7,mm7 |
paddw mm7,mm7 |
947 |
paddw mm6,mm0 ; quant-1 |
paddw mm6,mm0 ; quant-1 |
948 |
|
|
949 |
psrlw mm4,5 |
psrlw mm4,5 |
950 |
mov ecx,8 |
mov TMP0,8 |
951 |
pxor mm5,mm4 ; mm5=-2048 |
pxor mm5,mm4 ; mm5=-2048 |
952 |
|
|
953 |
.loop: |
.loop: |
954 |
movq mm0,[eax] |
movq mm0,[_EAX] |
955 |
pxor mm3,mm3 |
pxor mm3,mm3 |
956 |
pxor mm2,mm2 |
pxor mm2,mm2 |
957 |
pcmpeqw mm3,mm0 |
pcmpeqw mm3,mm0 |
958 |
pcmpgtw mm2,mm0 |
pcmpgtw mm2,mm0 |
959 |
pmullw mm0,mm7 ; * 2 * quant |
pmullw mm0,mm7 ; * 2 * quant |
960 |
pandn mm3,mm6 |
pandn mm3,mm6 |
961 |
movq mm1,[eax+8] |
movq mm1,[_EAX+8] |
962 |
psubw mm0,mm2 |
psubw mm0,mm2 |
963 |
pxor mm2,mm3 |
pxor mm2,mm3 |
964 |
pxor mm3,mm3 |
pxor mm3,mm3 |
977 |
pmaxsw mm0,mm5 |
pmaxsw mm0,mm5 |
978 |
pmaxsw mm1,mm5 |
pmaxsw mm1,mm5 |
979 |
|
|
980 |
movq [edx],mm0 |
movq [TMP1],mm0 |
981 |
lea eax,[eax+16] |
lea _EAX,[_EAX+16] |
982 |
movq [edx+8],mm1 |
movq [TMP1+8],mm1 |
983 |
|
|
984 |
dec ecx |
dec TMP0 |
985 |
lea edx,[edx+16] |
lea TMP1,[TMP1+16] |
986 |
jne .loop |
jne .loop |
987 |
|
|
988 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
989 |
ret |
ret |
990 |
|
ENDFUNC |
991 |
|
|
992 |
|
|
993 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
999 |
; |
; |
1000 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
1001 |
|
|
1002 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
1003 |
dequant_h263_inter_sse2: |
dequant_h263_inter_sse2: |
1004 |
|
|
1005 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
1006 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
1007 |
|
|
1008 |
movq xmm6, [mmx_quant + ecx*8] ; quant |
lea TMP1, [mmx_quant] |
1009 |
inc ecx |
movq xmm6, [TMP1 + TMP0*8] ; quant |
1010 |
|
inc TMP0 |
1011 |
pcmpeqw xmm5,xmm5 |
pcmpeqw xmm5,xmm5 |
1012 |
and ecx,1 |
and TMP0,1 |
1013 |
movlhps xmm6,xmm6 |
movlhps xmm6,xmm6 |
1014 |
movd xmm0,ecx |
movd xmm0,TMP0d |
1015 |
movdqa xmm7,xmm6 |
movdqa xmm7,xmm6 |
1016 |
pshuflw xmm0,xmm0,0 |
pshuflw xmm0,xmm0,0 |
1017 |
movdqa xmm4,xmm5 |
movdqa xmm4,xmm5 |
1018 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
1019 |
movlhps xmm0,xmm0 |
movlhps xmm0,xmm0 |
1020 |
paddw xmm7,xmm7 |
paddw xmm7,xmm7 |
1021 |
psubw xmm6,xmm0 |
psubw xmm6,xmm0 |
1022 |
psrlw xmm4,5 ; 2047 |
psrlw xmm4,5 ; 2047 |
1023 |
mov ecx,4 |
mov TMP0,4 |
1024 |
pxor xmm5,xmm4 ; mm5=-2048 |
pxor xmm5,xmm4 ; mm5=-2048 |
1025 |
|
|
1026 |
.loop: |
.loop: |
1027 |
movdqa xmm0,[eax] |
movdqa xmm0,[_EAX] |
1028 |
pxor xmm3,xmm3 |
pxor xmm3,xmm3 |
1029 |
pxor xmm2,xmm2 |
pxor xmm2,xmm2 |
1030 |
pcmpeqw xmm3,xmm0 |
pcmpeqw xmm3,xmm0 |
1031 |
pcmpgtw xmm2,xmm0 |
pcmpgtw xmm2,xmm0 |
1032 |
pmullw xmm0,xmm7 ; * 2 * quant |
pmullw xmm0,xmm7 ; * 2 * quant |
1033 |
pandn xmm3,xmm6 |
pandn xmm3,xmm6 |
1034 |
movdqa xmm1,[eax+16] |
movdqa xmm1,[_EAX+16] |
1035 |
psubw xmm0,xmm2 |
psubw xmm0,xmm2 |
1036 |
pxor xmm2,xmm3 |
pxor xmm2,xmm3 |
1037 |
pxor xmm3,xmm3 |
pxor xmm3,xmm3 |
1050 |
pmaxsw xmm0,xmm5 |
pmaxsw xmm0,xmm5 |
1051 |
pmaxsw xmm1,xmm5 |
pmaxsw xmm1,xmm5 |
1052 |
|
|
1053 |
movdqa [edx],xmm0 |
movdqa [TMP1],xmm0 |
1054 |
lea eax,[eax+32] |
lea _EAX,[_EAX+32] |
1055 |
movdqa [edx+16],xmm1 |
movdqa [TMP1+16],xmm1 |
1056 |
|
|
1057 |
dec ecx |
dec TMP0 |
1058 |
lea edx,[edx+32] |
lea TMP1,[TMP1+32] |
1059 |
jne .loop |
jne .loop |
1060 |
|
|
1061 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
1062 |
ret |
ret |
1063 |
|
ENDFUNC |
1064 |
|
|
1065 |
|
|
1066 |
|
%ifidn __OUTPUT_FORMAT__,elf |
1067 |
|
section ".note.GNU-stack" noalloc noexec nowrite progbits |
1068 |
|
%endif |
1069 |
|
|