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 %1.endfunc-%1 |
|
|
%define %1 _%1:function %1.endfunc-%1 |
|
|
%define ENDFUNC .endfunc |
|
|
%else |
|
|
global _%1 |
|
|
%define %1 _%1 |
|
|
%define ENDFUNC |
|
|
%endif |
|
|
%else |
|
|
%ifdef MARK_FUNCS |
|
|
global %1:function %1.endfunc-%1 |
|
|
%define ENDFUNC .endfunc |
|
|
%else |
|
|
global %1 |
|
|
%define ENDFUNC |
|
|
%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]) |
|
sub eax,edx |
|
|
add eax,ecx ; + (dcscalar/2)*sgn(data[0]) |
|
129 |
|
|
130 |
mov ecx, [esp + 4 + 12] ; quant |
mov TMP0, prm3 ; quant |
131 |
cdq |
cdq |
132 |
idiv dword [esp + 4 + 16] ; dcscalar |
idiv prm4d ; dcscalar |
133 |
cmp ecx, 1 |
lea TMP1, [mmx_div] |
134 |
mov edx, [esp + 4 + 4] ; coeff |
movq mm7, [TMP1+TMP0 * 8 - 8] |
135 |
|
cmp TMP0, 1 |
136 |
|
mov TMP1, prm1 ; coeff |
137 |
je .low |
je .low |
138 |
|
|
139 |
movq mm7, [mmx_div+ecx * 8 - 8] |
mov TMP0, prm2 ; data |
140 |
mov ecx,4 |
push _EAX ; DC |
141 |
|
mov _EAX, TMP0 |
142 |
|
|
143 |
|
mov TMP0,4 |
144 |
|
|
145 |
.loop: |
.loop: |
146 |
movq mm0, [esi] ; data |
movq mm0, [_EAX] ; data |
147 |
pxor mm4,mm4 |
pxor mm4,mm4 |
148 |
movq mm1, [esi + 8] |
movq mm1, [_EAX + 8] |
149 |
pcmpgtw mm4,mm0 ; (data<0) |
pcmpgtw mm4,mm0 ; (data<0) |
150 |
pxor mm5,mm5 |
pxor mm5,mm5 |
151 |
pmulhw mm0,mm7 ; /(2*quant) |
pmulhw mm0,mm7 ; /(2*quant) |
152 |
pcmpgtw mm5,mm1 |
pcmpgtw mm5,mm1 |
153 |
movq mm2, [esi+16] |
movq mm2, [_EAX+16] |
154 |
psubw mm0,mm4 ; +(data<0) |
psubw mm0,mm4 ; +(data<0) |
155 |
pmulhw mm1,mm7 |
pmulhw mm1,mm7 |
156 |
pxor mm4,mm4 |
pxor mm4,mm4 |
157 |
movq mm3,[esi+24] |
movq mm3,[_EAX+24] |
158 |
pcmpgtw mm4,mm2 |
pcmpgtw mm4,mm2 |
159 |
psubw mm1,mm5 |
psubw mm1,mm5 |
160 |
pmulhw mm2,mm7 |
pmulhw mm2,mm7 |
163 |
pmulhw mm3,mm7 |
pmulhw mm3,mm7 |
164 |
psubw mm2,mm4 |
psubw mm2,mm4 |
165 |
psubw mm3,mm5 |
psubw mm3,mm5 |
166 |
movq [edx], mm0 |
movq [TMP1], mm0 |
167 |
lea esi, [esi+32] |
lea _EAX, [_EAX+32] |
168 |
movq [edx + 8], mm1 |
movq [TMP1 + 8], mm1 |
169 |
movq [edx + 16], mm2 |
movq [TMP1 + 16], mm2 |
170 |
movq [edx + 24], mm3 |
movq [TMP1 + 24], mm3 |
171 |
|
|
172 |
dec ecx |
dec TMP0 |
173 |
lea edx, [edx+32] |
lea TMP1, [TMP1+32] |
174 |
jne .loop |
jne .loop |
175 |
jmp .end |
jmp .end |
176 |
|
|
177 |
.low: |
.low: |
178 |
movd mm7,ecx |
movd mm7,TMP0d |
179 |
mov ecx,4 |
|
180 |
|
mov TMP0, prm2 |
181 |
|
push _EAX |
182 |
|
mov _EAX, TMP0 |
183 |
|
|
184 |
|
mov TMP0,4 |
185 |
.loop_low: |
.loop_low: |
186 |
movq mm0, [esi] |
movq mm0, [_EAX] |
187 |
pxor mm4,mm4 |
pxor mm4,mm4 |
188 |
movq mm1, [esi + 8] |
movq mm1, [_EAX + 8] |
189 |
pcmpgtw mm4,mm0 |
pcmpgtw mm4,mm0 |
190 |
pxor mm5,mm5 |
pxor mm5,mm5 |
191 |
psubw mm0,mm4 |
psubw mm0,mm4 |
192 |
pcmpgtw mm5,mm1 |
pcmpgtw mm5,mm1 |
193 |
psraw mm0,mm7 |
psraw mm0,mm7 |
194 |
psubw mm1,mm5 |
psubw mm1,mm5 |
195 |
movq mm2,[esi+16] |
movq mm2,[_EAX+16] |
196 |
pxor mm4,mm4 |
pxor mm4,mm4 |
197 |
psraw mm1,mm7 |
psraw mm1,mm7 |
198 |
pcmpgtw mm4,mm2 |
pcmpgtw mm4,mm2 |
199 |
pxor mm5,mm5 |
pxor mm5,mm5 |
200 |
psubw mm2,mm4 |
psubw mm2,mm4 |
201 |
movq mm3,[esi+24] |
movq mm3,[_EAX+24] |
202 |
pcmpgtw mm5,mm3 |
pcmpgtw mm5,mm3 |
203 |
psraw mm2,mm7 |
psraw mm2,mm7 |
204 |
psubw mm3,mm5 |
psubw mm3,mm5 |
205 |
movq [edx], mm0 |
movq [TMP1], mm0 |
206 |
psraw mm3,mm7 |
psraw mm3,mm7 |
207 |
movq [edx + 8], mm1 |
movq [TMP1 + 8], mm1 |
208 |
movq [edx+16],mm2 |
movq [TMP1+16],mm2 |
209 |
lea esi, [esi+32] |
lea _EAX, [_EAX+32] |
210 |
movq [edx+24],mm3 |
movq [TMP1+24],mm3 |
211 |
|
|
212 |
dec ecx |
dec TMP0 |
213 |
lea edx, [edx+32] |
lea TMP1, [TMP1+32] |
214 |
jne .loop_low |
jne .loop_low |
215 |
|
|
216 |
.end: |
.end: |
|
mov edx, [esp + 4 + 4] ; coeff |
|
|
mov [edx],ax |
|
|
xor eax,eax ; return 0 |
|
217 |
|
|
218 |
pop esi |
pop _EAX |
219 |
|
|
220 |
|
mov TMP1, prm1 ; coeff |
221 |
|
mov [TMP1],ax |
222 |
|
xor _EAX,_EAX ; return 0 |
223 |
|
|
224 |
ret |
ret |
225 |
ENDFUNC |
ENDFUNC |
226 |
|
|
235 |
; |
; |
236 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
237 |
|
|
238 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
239 |
quant_h263_intra_sse2: |
quant_h263_intra_sse2: |
240 |
|
|
241 |
push esi |
mov _EAX, prm2 ; data |
|
|
|
|
mov esi, [esp + 4 + 8] ; data |
|
242 |
|
|
243 |
movsx eax, word [esi] ; data[0] |
movsx _EAX, word [_EAX] ; data[0] |
244 |
|
|
245 |
mov ecx,[esp + 4 + 16] ; dcscalar |
mov TMP0,prm4 ; dcscalar |
246 |
mov edx,eax |
mov TMP1,_EAX |
247 |
sar ecx,1 |
sar TMP0,1 |
248 |
add eax,ecx |
add _EAX,TMP0 |
249 |
sub edx,ecx |
sub TMP1,TMP0 |
250 |
cmovl eax,edx ; +/- dcscalar/2 |
cmovl _EAX,TMP1 ; +/- dcscalar/2 |
251 |
mov ecx, [esp + 4 + 12] ; quant |
mov TMP0, prm3 ; quant |
252 |
cdq |
cdq |
253 |
idiv dword [esp + 4 + 16] ; dcscalar |
idiv prm4d ; dcscalar |
254 |
cmp ecx, 1 |
cmp TMP0, 1 |
255 |
mov edx, [esp + 4 + 4] ; coeff |
lea TMP1, [mmx_div] |
256 |
movq xmm7, [mmx_div+ecx * 8 - 8] |
movq xmm7, [TMP1+TMP0 * 8 - 8] |
257 |
je .low |
mov TMP1, prm1 ; coeff |
258 |
|
je near .low |
259 |
|
|
260 |
|
mov TMP0, prm2 |
261 |
|
push _EAX ; DC |
262 |
|
mov _EAX, TMP0 |
263 |
|
|
264 |
mov ecx,2 |
mov TMP0,2 |
265 |
movlhps xmm7,xmm7 |
movlhps xmm7,xmm7 |
266 |
|
|
267 |
.loop: |
.loop: |
268 |
movdqa xmm0, [esi] |
movdqa xmm0, [_EAX] |
269 |
pxor xmm4,xmm4 |
pxor xmm4,xmm4 |
270 |
movdqa xmm1, [esi + 16] |
movdqa xmm1, [_EAX + 16] |
271 |
pcmpgtw xmm4,xmm0 |
pcmpgtw xmm4,xmm0 |
272 |
pxor xmm5,xmm5 |
pxor xmm5,xmm5 |
273 |
pmulhw xmm0,xmm7 |
pmulhw xmm0,xmm7 |
274 |
pcmpgtw xmm5,xmm1 |
pcmpgtw xmm5,xmm1 |
275 |
movdqa xmm2, [esi+32] |
movdqa xmm2, [_EAX+32] |
276 |
psubw xmm0,xmm4 |
psubw xmm0,xmm4 |
277 |
pmulhw xmm1,xmm7 |
pmulhw xmm1,xmm7 |
278 |
pxor xmm4,xmm4 |
pxor xmm4,xmm4 |
279 |
movdqa xmm3,[esi+48] |
movdqa xmm3,[_EAX+48] |
280 |
pcmpgtw xmm4,xmm2 |
pcmpgtw xmm4,xmm2 |
281 |
psubw xmm1,xmm5 |
psubw xmm1,xmm5 |
282 |
pmulhw xmm2,xmm7 |
pmulhw xmm2,xmm7 |
285 |
pmulhw xmm3,xmm7 |
pmulhw xmm3,xmm7 |
286 |
psubw xmm2,xmm4 |
psubw xmm2,xmm4 |
287 |
psubw xmm3,xmm5 |
psubw xmm3,xmm5 |
288 |
movdqa [edx], xmm0 |
movdqa [TMP1], xmm0 |
289 |
lea esi, [esi+64] |
lea _EAX, [_EAX+64] |
290 |
movdqa [edx + 16], xmm1 |
movdqa [TMP1 + 16], xmm1 |
291 |
movdqa [edx + 32], xmm2 |
movdqa [TMP1 + 32], xmm2 |
292 |
movdqa [edx + 48], xmm3 |
movdqa [TMP1 + 48], xmm3 |
293 |
|
|
294 |
dec ecx |
dec TMP0 |
295 |
lea edx, [edx+64] |
lea TMP1, [TMP1+64] |
296 |
jne .loop |
jne .loop |
297 |
jmp .end |
jmp .end |
298 |
|
|
299 |
.low: |
.low: |
300 |
movd xmm7,ecx |
movd xmm7,TMP0d |
301 |
mov ecx,2 |
|
302 |
|
mov TMP0, prm2 |
303 |
|
push _EAX ; DC |
304 |
|
mov _EAX, TMP0 |
305 |
|
|
306 |
|
mov TMP0,2 |
307 |
.loop_low: |
.loop_low: |
308 |
movdqa xmm0, [esi] |
movdqa xmm0, [_EAX] |
309 |
pxor xmm4,xmm4 |
pxor xmm4,xmm4 |
310 |
movdqa xmm1, [esi + 16] |
movdqa xmm1, [_EAX + 16] |
311 |
pcmpgtw xmm4,xmm0 |
pcmpgtw xmm4,xmm0 |
312 |
pxor xmm5,xmm5 |
pxor xmm5,xmm5 |
313 |
psubw xmm0,xmm4 |
psubw xmm0,xmm4 |
314 |
pcmpgtw xmm5,xmm1 |
pcmpgtw xmm5,xmm1 |
315 |
psraw xmm0,xmm7 |
psraw xmm0,xmm7 |
316 |
psubw xmm1,xmm5 |
psubw xmm1,xmm5 |
317 |
movdqa xmm2,[esi+32] |
movdqa xmm2,[_EAX+32] |
318 |
pxor xmm4,xmm4 |
pxor xmm4,xmm4 |
319 |
psraw xmm1,xmm7 |
psraw xmm1,xmm7 |
320 |
pcmpgtw xmm4,xmm2 |
pcmpgtw xmm4,xmm2 |
321 |
pxor xmm5,xmm5 |
pxor xmm5,xmm5 |
322 |
psubw xmm2,xmm4 |
psubw xmm2,xmm4 |
323 |
movdqa xmm3,[esi+48] |
movdqa xmm3,[_EAX+48] |
324 |
pcmpgtw xmm5,xmm3 |
pcmpgtw xmm5,xmm3 |
325 |
psraw xmm2,xmm7 |
psraw xmm2,xmm7 |
326 |
psubw xmm3,xmm5 |
psubw xmm3,xmm5 |
327 |
movdqa [edx], xmm0 |
movdqa [TMP1], xmm0 |
328 |
psraw xmm3,xmm7 |
psraw xmm3,xmm7 |
329 |
movdqa [edx+16], xmm1 |
movdqa [TMP1+16], xmm1 |
330 |
movdqa [edx+32],xmm2 |
movdqa [TMP1+32],xmm2 |
331 |
lea esi, [esi+64] |
lea _EAX, [_EAX+64] |
332 |
movdqa [edx+48],xmm3 |
movdqa [TMP1+48],xmm3 |
333 |
|
|
334 |
dec ecx |
dec TMP0 |
335 |
lea edx, [edx+64] |
lea TMP1, [TMP1+64] |
336 |
jne .loop_low |
jne .loop_low |
337 |
|
|
338 |
.end: |
.end: |
|
mov edx, [esp + 4 + 4] ; coeff |
|
|
mov [edx],ax |
|
|
xor eax,eax ; return 0 |
|
339 |
|
|
340 |
pop esi |
pop _EAX |
341 |
|
|
342 |
|
mov TMP1, prm1 ; coeff |
343 |
|
mov [TMP1],ax |
344 |
|
xor _EAX,_EAX ; return 0 |
345 |
|
|
346 |
ret |
ret |
347 |
ENDFUNC |
ENDFUNC |
348 |
|
|
355 |
; |
; |
356 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
357 |
|
|
358 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
359 |
quant_h263_inter_mmx: |
quant_h263_inter_mmx: |
360 |
|
|
361 |
push ecx |
mov TMP1, prm1 ; coeff |
362 |
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 |
|
363 |
|
|
364 |
pxor mm5, mm5 ; sum |
pxor mm5, mm5 ; sum |
365 |
movq mm6, [mmx_sub + eax * 8 - 8] ; sub |
lea TMP0, [mmx_sub] |
366 |
|
movq mm6, [TMP0 + _EAX * 8 - 8] ; sub |
367 |
|
|
368 |
cmp al, 1 |
cmp al, 1 |
369 |
jz .q1loop |
jz near .q1routine |
370 |
|
|
371 |
|
lea TMP0, [mmx_div] |
372 |
|
movq mm7, [TMP0 + _EAX * 8 - 8] ; divider |
373 |
|
|
374 |
movq mm7, [mmx_div + eax * 8 - 8] ; divider |
xor TMP0, TMP0 |
375 |
|
mov _EAX, prm2 ; data |
376 |
|
|
377 |
ALIGN 8 |
ALIGN SECTION_ALIGN |
378 |
.loop: |
.loop: |
379 |
movq mm0, [esi + 8*ecx] ; mm0 = [1st] |
movq mm0, [_EAX + 8*TMP0] ; mm0 = [1st] |
380 |
movq mm3, [esi + 8*ecx + 8] |
movq mm3, [_EAX + 8*TMP0 + 8] |
381 |
pxor mm1, mm1 ; mm1 = 0 |
pxor mm1, mm1 ; mm1 = 0 |
382 |
pxor mm4, mm4 ; |
pxor mm4, mm4 ; |
383 |
pcmpgtw mm1, mm0 ; mm1 = (0 > mm0) |
pcmpgtw mm1, mm0 ; mm1 = (0 > mm0) |
396 |
pxor mm3, mm4 ; |
pxor mm3, mm4 ; |
397 |
psubw mm0, mm1 ; undisplace |
psubw mm0, mm1 ; undisplace |
398 |
psubw mm3, mm4 |
psubw mm3, mm4 |
399 |
movq [edi + 8*ecx], mm0 |
movq [TMP1 + 8*TMP0], mm0 |
400 |
movq [edi + 8*ecx + 8], mm3 |
movq [TMP1 + 8*TMP0 + 8], mm3 |
401 |
|
|
402 |
add ecx, 2 |
add TMP0, 2 |
403 |
cmp ecx, 16 |
cmp TMP0, 16 |
404 |
jnz .loop |
jnz .loop |
405 |
|
|
406 |
.done: |
.done: |
410 |
paddd mm0, mm5 |
paddd mm0, mm5 |
411 |
|
|
412 |
movd eax, mm0 ; return sum |
movd eax, mm0 ; return sum |
|
pop edi |
|
|
pop esi |
|
|
pop ecx |
|
413 |
|
|
414 |
ret |
ret |
415 |
|
|
416 |
ALIGN 8 |
.q1routine: |
417 |
|
xor TMP0, TMP0 |
418 |
|
mov _EAX, prm2 ; data |
419 |
|
|
420 |
|
ALIGN SECTION_ALIGN |
421 |
.q1loop: |
.q1loop: |
422 |
movq mm0, [esi + 8*ecx] ; mm0 = [1st] |
movq mm0, [_EAX + 8*TMP0] ; mm0 = [1st] |
423 |
movq mm3, [esi + 8*ecx+ 8] ; |
movq mm3, [_EAX + 8*TMP0+ 8] ; |
424 |
pxor mm1, mm1 ; mm1 = 0 |
pxor mm1, mm1 ; mm1 = 0 |
425 |
pxor mm4, mm4 ; |
pxor mm4, mm4 ; |
426 |
pcmpgtw mm1, mm0 ; mm1 = (0 > mm0) |
pcmpgtw mm1, mm0 ; mm1 = (0 > mm0) |
439 |
pxor mm3, mm4 ; |
pxor mm3, mm4 ; |
440 |
psubw mm0, mm1 ; undisplace |
psubw mm0, mm1 ; undisplace |
441 |
psubw mm3, mm4 |
psubw mm3, mm4 |
442 |
movq [edi + 8*ecx], mm0 |
movq [TMP1 + 8*TMP0], mm0 |
443 |
movq [edi + 8*ecx + 8], mm3 |
movq [TMP1 + 8*TMP0 + 8], mm3 |
444 |
|
|
445 |
add ecx, 2 |
add TMP0, 2 |
446 |
cmp ecx, 16 |
cmp TMP0, 16 |
447 |
jnz .q1loop |
jnz .q1loop |
448 |
|
|
449 |
jmp .done |
jmp .done |
460 |
; |
; |
461 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
462 |
|
|
463 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
464 |
quant_h263_inter_sse2: |
quant_h263_inter_sse2: |
465 |
|
|
466 |
push esi |
mov TMP1, prm1 ; coeff |
467 |
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 |
|
468 |
|
|
469 |
pxor xmm5, xmm5 ; sum |
pxor xmm5, xmm5 ; sum |
470 |
|
|
471 |
movq mm0, [mmx_sub + eax*8 - 8] ; sub |
lea TMP0, [mmx_sub] |
472 |
|
movq mm0, [TMP0 + _EAX*8 - 8] ; sub |
473 |
movq2dq xmm6, mm0 ; load into low 8 bytes |
movq2dq xmm6, mm0 ; load into low 8 bytes |
474 |
movlhps xmm6, xmm6 ; duplicate into high 8 bytes |
movlhps xmm6, xmm6 ; duplicate into high 8 bytes |
475 |
|
|
476 |
cmp al, 1 |
cmp al, 1 |
477 |
jz near .qes2_q1loop |
jz near .qes2_q1_routine |
478 |
|
|
479 |
.qes2_not1: |
.qes2_not1: |
480 |
movq mm0, [mmx_div + eax*8 - 8] ; divider |
lea TMP0, [mmx_div] |
481 |
|
movq mm0, [TMP0 + _EAX*8 - 8] ; divider |
482 |
|
|
483 |
|
xor TMP0, TMP0 |
484 |
|
mov _EAX, prm2 ; data |
485 |
|
|
486 |
movq2dq xmm7, mm0 |
movq2dq xmm7, mm0 |
487 |
movlhps xmm7, xmm7 |
movlhps xmm7, xmm7 |
488 |
|
|
489 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
490 |
.qes2_loop: |
.qes2_loop: |
491 |
movdqa xmm0, [esi + ecx*8] ; xmm0 = [1st] |
movdqa xmm0, [_EAX + TMP0*8] ; xmm0 = [1st] |
492 |
movdqa xmm3, [esi + ecx*8 + 16] ; xmm3 = [2nd] |
movdqa xmm3, [_EAX + TMP0*8 + 16] ; xmm3 = [2nd] |
493 |
pxor xmm1, xmm1 |
pxor xmm1, xmm1 |
494 |
pxor xmm4, xmm4 |
pxor xmm4, xmm4 |
495 |
pcmpgtw xmm1, xmm0 |
pcmpgtw xmm1, xmm0 |
508 |
pxor xmm3, xmm4 |
pxor xmm3, xmm4 |
509 |
psubw xmm0, xmm1 |
psubw xmm0, xmm1 |
510 |
psubw xmm3, xmm4 |
psubw xmm3, xmm4 |
511 |
movdqa [edi + ecx*8], xmm0 |
movdqa [TMP1 + TMP0*8], xmm0 |
512 |
movdqa [edi + ecx*8 + 16], xmm3 |
movdqa [TMP1 + TMP0*8 + 16], xmm3 |
513 |
|
|
514 |
add ecx, 4 |
add TMP0, 4 |
515 |
cmp ecx, 16 |
cmp TMP0, 16 |
516 |
jnz .qes2_loop |
jnz .qes2_loop |
517 |
|
|
518 |
.qes2_done: |
.qes2_done: |
528 |
|
|
529 |
movd eax, mm0 ; return sum |
movd eax, mm0 ; return sum |
530 |
|
|
|
pop edi |
|
|
pop esi |
|
|
|
|
531 |
ret |
ret |
532 |
|
|
533 |
ALIGN 16 |
.qes2_q1_routine: |
534 |
|
xor TMP0, TMP0 |
535 |
|
mov _EAX, prm2 ; data |
536 |
|
|
537 |
|
ALIGN SECTION_ALIGN |
538 |
.qes2_q1loop: |
.qes2_q1loop: |
539 |
movdqa xmm0, [esi + ecx*8] ; xmm0 = [1st] |
movdqa xmm0, [_EAX + TMP0*8] ; xmm0 = [1st] |
540 |
movdqa xmm3, [esi + ecx*8 + 16] ; xmm3 = [2nd] |
movdqa xmm3, [_EAX + TMP0*8 + 16] ; xmm3 = [2nd] |
541 |
pxor xmm1, xmm1 |
pxor xmm1, xmm1 |
542 |
pxor xmm4, xmm4 |
pxor xmm4, xmm4 |
543 |
pcmpgtw xmm1, xmm0 |
pcmpgtw xmm1, xmm0 |
556 |
pxor xmm3, xmm4 |
pxor xmm3, xmm4 |
557 |
psubw xmm0, xmm1 |
psubw xmm0, xmm1 |
558 |
psubw xmm3, xmm4 |
psubw xmm3, xmm4 |
559 |
movdqa [edi + ecx*8], xmm0 |
movdqa [TMP1 + TMP0*8], xmm0 |
560 |
movdqa [edi + ecx*8 + 16], xmm3 |
movdqa [TMP1 + TMP0*8 + 16], xmm3 |
561 |
|
|
562 |
add ecx, 4 |
add TMP0, 4 |
563 |
cmp ecx, 16 |
cmp TMP0, 16 |
564 |
jnz .qes2_q1loop |
jnz .qes2_q1loop |
565 |
jmp .qes2_done |
jmp .qes2_done |
566 |
ENDFUNC |
ENDFUNC |
576 |
; |
; |
577 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
578 |
|
|
579 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
580 |
dequant_h263_intra_mmx: |
dequant_h263_intra_mmx: |
581 |
|
|
582 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
583 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
584 |
pcmpeqw mm0,mm0 |
pcmpeqw mm0,mm0 |
585 |
movq mm6, [mmx_quant + ecx*8] ; quant |
lea TMP1, [mmx_quant] |
586 |
shl ecx,31 ; quant & 1 ? 0 : - 1 |
movq mm6, [TMP1 + TMP0*8] ; quant |
587 |
|
shl TMP0,31 ; quant & 1 ? 0 : - 1 |
588 |
movq mm7,mm6 |
movq mm7,mm6 |
589 |
movq mm5,mm0 |
movq mm5,mm0 |
590 |
movd mm1,ecx |
movd mm1,TMP0d |
591 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
592 |
psllw mm0,mm1 |
psllw mm0,mm1 |
593 |
paddw mm7,mm7 ; 2*quant |
paddw mm7,mm7 ; 2*quant |
594 |
paddw mm6,mm0 ; quant-1 |
paddw mm6,mm0 ; quant-1 |
595 |
psllw mm5,12 |
psllw mm5,12 |
596 |
mov ecx,8 |
mov TMP0,8 |
597 |
psrlw mm5,1 |
psrlw mm5,1 |
598 |
|
|
599 |
.loop: |
.loop: |
600 |
movq mm0,[eax] |
movq mm0,[_EAX] |
601 |
pxor mm2,mm2 |
pxor mm2,mm2 |
602 |
pxor mm4,mm4 |
pxor mm4,mm4 |
603 |
pcmpgtw mm2,mm0 |
pcmpgtw mm2,mm0 |
604 |
pcmpeqw mm4,mm0 |
pcmpeqw mm4,mm0 |
605 |
pmullw mm0,mm7 ; * 2 * quant |
pmullw mm0,mm7 ; * 2 * quant |
606 |
movq mm1,[eax+8] |
movq mm1,[_EAX+8] |
607 |
psubw mm0,mm2 |
psubw mm0,mm2 |
608 |
pxor mm2,mm6 |
pxor mm2,mm6 |
609 |
pxor mm3,mm3 |
pxor mm3,mm3 |
627 |
paddsw mm0, mm5 |
paddsw mm0, mm5 |
628 |
paddsw mm1, mm5 |
paddsw mm1, mm5 |
629 |
|
|
630 |
movq [edx],mm0 |
movq [TMP1],mm0 |
631 |
lea eax,[eax+16] |
lea _EAX,[_EAX+16] |
632 |
movq [edx+8],mm1 |
movq [TMP1+8],mm1 |
633 |
|
|
634 |
dec ecx |
dec TMP0 |
635 |
lea edx,[edx+16] |
lea TMP1,[TMP1+16] |
636 |
jne .loop |
jne .loop |
637 |
|
|
638 |
; deal with DC |
; deal with DC |
639 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
640 |
movd mm1,[esp+16] ; dcscalar |
movd mm1,prm4d ; dcscalar |
641 |
movd mm0,[eax] ; coeff[0] |
movd mm0,[_EAX] ; coeff[0] |
642 |
pmullw mm0,mm1 ; * dcscalar |
pmullw mm0,mm1 ; * dcscalar |
643 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
644 |
paddsw mm0, mm5 ; saturate + |
paddsw mm0, mm5 ; saturate + |
645 |
psubsw mm0, mm5 |
psubsw mm0, mm5 |
646 |
psubsw mm0, mm5 ; saturate - |
psubsw mm0, mm5 ; saturate - |
647 |
paddsw mm0, mm5 |
paddsw mm0, mm5 |
648 |
movd eax,mm0 |
movd eax,mm0 |
649 |
mov [edx], ax |
mov [TMP1], ax |
650 |
|
|
651 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
652 |
ret |
ret |
653 |
ENDFUNC |
ENDFUNC |
654 |
|
|
663 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
664 |
|
|
665 |
|
|
666 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
667 |
dequant_h263_intra_xmm: |
dequant_h263_intra_xmm: |
668 |
|
|
669 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
670 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
671 |
|
|
672 |
movd mm6,ecx ; quant |
movd mm6,TMP0d ; quant |
673 |
pcmpeqw mm0,mm0 |
pcmpeqw mm0,mm0 |
674 |
pshufw mm6,mm6,0 ; all quant |
pshufw mm6,mm6,0 ; all quant |
675 |
shl ecx,31 |
shl TMP0,31 |
676 |
movq mm5,mm0 |
movq mm5,mm0 |
677 |
movq mm7,mm6 |
movq mm7,mm6 |
678 |
movd mm1,ecx |
movd mm1,TMP0d |
679 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
680 |
psllw mm0,mm1 ; quant & 1 ? 0 : - 1 |
psllw mm0,mm1 ; quant & 1 ? 0 : - 1 |
681 |
movq mm4,mm5 |
movq mm4,mm5 |
682 |
paddw mm7,mm7 ; quant*2 |
paddw mm7,mm7 ; quant*2 |
683 |
paddw mm6,mm0 ; quant-1 |
paddw mm6,mm0 ; quant-1 |
684 |
psrlw mm4,5 ; mm4=2047 |
psrlw mm4,5 ; mm4=2047 |
685 |
mov ecx,8 |
mov TMP0,8 |
686 |
pxor mm5,mm4 ; mm5=-2048 |
pxor mm5,mm4 ; mm5=-2048 |
687 |
|
|
688 |
.loop: |
.loop: |
689 |
movq mm0,[eax] |
movq mm0,[_EAX] |
690 |
pxor mm2,mm2 |
pxor mm2,mm2 |
691 |
pxor mm3,mm3 |
pxor mm3,mm3 |
692 |
|
|
693 |
pcmpgtw mm2,mm0 |
pcmpgtw mm2,mm0 |
694 |
pcmpeqw mm3,mm0 ; if coeff==0... |
pcmpeqw mm3,mm0 ; if coeff==0... |
695 |
pmullw mm0,mm7 ; * 2 * quant |
pmullw mm0,mm7 ; * 2 * quant |
696 |
movq mm1,[eax+8] |
movq mm1,[_EAX+8] |
697 |
|
|
698 |
psubw mm0,mm2 |
psubw mm0,mm2 |
699 |
pxor mm2,mm6 |
pxor mm2,mm6 |
713 |
|
|
714 |
pmaxsw mm0,mm5 |
pmaxsw mm0,mm5 |
715 |
pminsw mm1,mm4 |
pminsw mm1,mm4 |
716 |
movq [edx],mm0 |
movq [TMP1],mm0 |
717 |
pmaxsw mm1,mm5 |
pmaxsw mm1,mm5 |
718 |
lea eax,[eax+16] |
lea _EAX,[_EAX+16] |
719 |
movq [edx+8],mm1 |
movq [TMP1+8],mm1 |
720 |
|
|
721 |
dec ecx |
dec TMP0 |
722 |
lea edx,[edx+16] |
lea TMP1,[TMP1+16] |
723 |
jne .loop |
jne .loop |
724 |
|
|
725 |
; deal with DC |
; deal with DC |
726 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
727 |
movd mm1,[esp+16] ; dcscalar |
movd mm1,prm4d ; dcscalar |
728 |
movd mm0, [eax] |
movd mm0, [_EAX] |
729 |
pmullw mm0, mm1 |
pmullw mm0, mm1 |
730 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
731 |
pminsw mm0,mm4 |
pminsw mm0,mm4 |
732 |
pmaxsw mm0,mm5 |
pmaxsw mm0,mm5 |
733 |
movd eax, mm0 |
movd eax, mm0 |
734 |
mov [edx], ax |
mov [TMP1], ax |
735 |
|
|
736 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
737 |
ret |
ret |
738 |
ENDFUNC |
ENDFUNC |
739 |
|
|
748 |
; |
; |
749 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
750 |
|
|
751 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
752 |
dequant_h263_intra_sse2: |
dequant_h263_intra_sse2: |
753 |
|
|
754 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
755 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
756 |
|
|
757 |
movd xmm6,ecx ; quant |
movd xmm6,TMP0d ; quant |
758 |
|
|
759 |
shl ecx,31 |
shl TMP0,31 |
760 |
pshuflw xmm6,xmm6,0 |
pshuflw xmm6,xmm6,0 |
761 |
pcmpeqw xmm0,xmm0 |
pcmpeqw xmm0,xmm0 |
762 |
movlhps xmm6,xmm6 ; all quant |
movlhps xmm6,xmm6 ; all quant |
763 |
movd xmm1,ecx |
movd xmm1,TMP0d |
764 |
movdqa xmm5,xmm0 |
movdqa xmm5,xmm0 |
765 |
movdqa xmm7,xmm6 |
movdqa xmm7,xmm6 |
766 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
767 |
paddw xmm7,xmm7 ; quant *2 |
paddw xmm7,xmm7 ; quant *2 |
768 |
psllw xmm0,xmm1 ; quant & 1 ? 0 : - 1 |
psllw xmm0,xmm1 ; quant & 1 ? 0 : - 1 |
769 |
movdqa xmm4,xmm5 |
movdqa xmm4,xmm5 |
770 |
paddw xmm6,xmm0 ; quant-1 |
paddw xmm6,xmm0 ; quant-1 |
771 |
psrlw xmm4,5 ; 2047 |
psrlw xmm4,5 ; 2047 |
772 |
mov ecx,4 |
mov TMP0,4 |
773 |
pxor xmm5,xmm4 ; mm5=-2048 |
pxor xmm5,xmm4 ; mm5=-2048 |
774 |
|
|
775 |
.loop: |
.loop: |
776 |
movdqa xmm0,[eax] |
movdqa xmm0,[_EAX] |
777 |
pxor xmm2,xmm2 |
pxor xmm2,xmm2 |
778 |
pxor xmm3,xmm3 |
pxor xmm3,xmm3 |
779 |
|
|
780 |
pcmpgtw xmm2,xmm0 |
pcmpgtw xmm2,xmm0 |
781 |
pcmpeqw xmm3,xmm0 |
pcmpeqw xmm3,xmm0 |
782 |
pmullw xmm0,xmm7 ; * 2 * quant |
pmullw xmm0,xmm7 ; * 2 * quant |
783 |
movdqa xmm1,[eax+16] |
movdqa xmm1,[_EAX+16] |
784 |
|
|
785 |
psubw xmm0,xmm2 |
psubw xmm0,xmm2 |
786 |
pxor xmm2,xmm6 |
pxor xmm2,xmm6 |
800 |
|
|
801 |
pmaxsw xmm0,xmm5 |
pmaxsw xmm0,xmm5 |
802 |
pminsw xmm1,xmm4 |
pminsw xmm1,xmm4 |
803 |
movdqa [edx],xmm0 |
movdqa [TMP1],xmm0 |
804 |
pmaxsw xmm1,xmm5 |
pmaxsw xmm1,xmm5 |
805 |
lea eax,[eax+32] |
lea _EAX,[_EAX+32] |
806 |
movdqa [edx+16],xmm1 |
movdqa [TMP1+16],xmm1 |
807 |
|
|
808 |
dec ecx |
dec TMP0 |
809 |
lea edx,[edx+32] |
lea TMP1,[TMP1+32] |
810 |
jne .loop |
jne .loop |
811 |
|
|
812 |
; deal with DC |
; deal with DC |
813 |
|
|
814 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
815 |
movsx eax,word [eax] |
movsx _EAX,word [_EAX] |
816 |
imul dword [esp+16] ; dcscalar |
imul prm4d ; dcscalar |
817 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
818 |
movd xmm0,eax |
movd xmm0,eax |
819 |
pminsw xmm0,xmm4 |
pminsw xmm0,xmm4 |
820 |
pmaxsw xmm0,xmm5 |
pmaxsw xmm0,xmm5 |
821 |
movd eax,xmm0 |
movd eax,xmm0 |
822 |
|
|
823 |
mov [edx], ax |
mov [TMP1], ax |
824 |
|
|
825 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
826 |
ret |
ret |
827 |
ENDFUNC |
ENDFUNC |
828 |
|
|
835 |
; |
; |
836 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
837 |
|
|
838 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
839 |
dequant_h263_inter_mmx: |
dequant_h263_inter_mmx: |
840 |
|
|
841 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
842 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
843 |
pcmpeqw mm0,mm0 |
pcmpeqw mm0,mm0 |
844 |
movq mm6, [mmx_quant + ecx*8] ; quant |
lea TMP1, [mmx_quant] |
845 |
shl ecx,31 ; odd/even |
movq mm6, [TMP1 + TMP0*8] ; quant |
846 |
|
shl TMP0,31 ; odd/even |
847 |
movq mm7,mm6 |
movq mm7,mm6 |
848 |
movd mm1,ecx |
movd mm1,TMP0d |
849 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
850 |
movq mm5,mm0 |
movq mm5,mm0 |
851 |
psllw mm0,mm1 ; quant & 1 ? 0 : - 1 |
psllw mm0,mm1 ; quant & 1 ? 0 : - 1 |
852 |
paddw mm7,mm7 ; quant*2 |
paddw mm7,mm7 ; quant*2 |
853 |
paddw mm6,mm0 ; quant & 1 ? quant : quant - 1 |
paddw mm6,mm0 ; quant & 1 ? quant : quant - 1 |
854 |
psllw mm5,12 |
psllw mm5,12 |
855 |
mov ecx,8 |
mov TMP0,8 |
856 |
psrlw mm5,1 ; 32767-2047 (32768-2048) |
psrlw mm5,1 ; 32767-2047 (32768-2048) |
857 |
|
|
858 |
.loop: |
.loop: |
859 |
movq mm0,[eax] |
movq mm0,[_EAX] |
860 |
pxor mm4,mm4 |
pxor mm4,mm4 |
861 |
pxor mm2,mm2 |
pxor mm2,mm2 |
862 |
pcmpeqw mm4,mm0 ; if coeff==0... |
pcmpeqw mm4,mm0 ; if coeff==0... |
864 |
pmullw mm0,mm7 ; * 2 * quant |
pmullw mm0,mm7 ; * 2 * quant |
865 |
pxor mm3,mm3 |
pxor mm3,mm3 |
866 |
psubw mm0,mm2 |
psubw mm0,mm2 |
867 |
movq mm1,[eax+8] |
movq mm1,[_EAX+8] |
868 |
pxor mm2,mm6 |
pxor mm2,mm6 |
869 |
pcmpgtw mm3,mm1 |
pcmpgtw mm3,mm1 |
870 |
pandn mm4,mm2 ; ... then data==0 |
pandn mm4,mm2 ; ... then data==0 |
886 |
paddsw mm0, mm5 |
paddsw mm0, mm5 |
887 |
paddsw mm1, mm5 |
paddsw mm1, mm5 |
888 |
|
|
889 |
movq [edx],mm0 |
movq [TMP1],mm0 |
890 |
lea eax,[eax+16] |
lea _EAX,[_EAX+16] |
891 |
movq [edx+8],mm1 |
movq [TMP1+8],mm1 |
892 |
|
|
893 |
dec ecx |
dec TMP0 |
894 |
lea edx,[edx+16] |
lea TMP1,[TMP1+16] |
895 |
jne .loop |
jne .loop |
896 |
|
|
897 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
898 |
ret |
ret |
899 |
ENDFUNC |
ENDFUNC |
900 |
|
|
907 |
; const uint16_t *mpeg_matrices); |
; const uint16_t *mpeg_matrices); |
908 |
; |
; |
909 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
910 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
911 |
dequant_h263_inter_xmm: |
dequant_h263_inter_xmm: |
912 |
|
|
913 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
914 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
915 |
pcmpeqw mm0,mm0 |
pcmpeqw mm0,mm0 |
916 |
movq mm6, [mmx_quant + ecx*8] ; quant |
lea TMP1, [mmx_quant] |
917 |
shl ecx,31 |
movq mm6, [TMP1 + TMP0*8] ; quant |
918 |
|
shl TMP0,31 |
919 |
movq mm5,mm0 |
movq mm5,mm0 |
920 |
movd mm1,ecx |
movd mm1,TMP0d |
921 |
movq mm7,mm6 |
movq mm7,mm6 |
922 |
psllw mm0,mm1 |
psllw mm0,mm1 |
923 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
924 |
movq mm4,mm5 |
movq mm4,mm5 |
925 |
paddw mm7,mm7 |
paddw mm7,mm7 |
926 |
paddw mm6,mm0 ; quant-1 |
paddw mm6,mm0 ; quant-1 |
927 |
|
|
928 |
psrlw mm4,5 |
psrlw mm4,5 |
929 |
mov ecx,8 |
mov TMP0,8 |
930 |
pxor mm5,mm4 ; mm5=-2048 |
pxor mm5,mm4 ; mm5=-2048 |
931 |
|
|
932 |
.loop: |
.loop: |
933 |
movq mm0,[eax] |
movq mm0,[_EAX] |
934 |
pxor mm3,mm3 |
pxor mm3,mm3 |
935 |
pxor mm2,mm2 |
pxor mm2,mm2 |
936 |
pcmpeqw mm3,mm0 |
pcmpeqw mm3,mm0 |
937 |
pcmpgtw mm2,mm0 |
pcmpgtw mm2,mm0 |
938 |
pmullw mm0,mm7 ; * 2 * quant |
pmullw mm0,mm7 ; * 2 * quant |
939 |
pandn mm3,mm6 |
pandn mm3,mm6 |
940 |
movq mm1,[eax+8] |
movq mm1,[_EAX+8] |
941 |
psubw mm0,mm2 |
psubw mm0,mm2 |
942 |
pxor mm2,mm3 |
pxor mm2,mm3 |
943 |
pxor mm3,mm3 |
pxor mm3,mm3 |
956 |
pmaxsw mm0,mm5 |
pmaxsw mm0,mm5 |
957 |
pmaxsw mm1,mm5 |
pmaxsw mm1,mm5 |
958 |
|
|
959 |
movq [edx],mm0 |
movq [TMP1],mm0 |
960 |
lea eax,[eax+16] |
lea _EAX,[_EAX+16] |
961 |
movq [edx+8],mm1 |
movq [TMP1+8],mm1 |
962 |
|
|
963 |
dec ecx |
dec TMP0 |
964 |
lea edx,[edx+16] |
lea TMP1,[TMP1+16] |
965 |
jne .loop |
jne .loop |
966 |
|
|
967 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
968 |
ret |
ret |
969 |
ENDFUNC |
ENDFUNC |
970 |
|
|
978 |
; |
; |
979 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
980 |
|
|
981 |
ALIGN 16 |
ALIGN SECTION_ALIGN |
982 |
dequant_h263_inter_sse2: |
dequant_h263_inter_sse2: |
983 |
|
|
984 |
mov ecx, [esp+12] ; quant |
mov TMP0, prm3 ; quant |
985 |
mov eax, [esp+ 8] ; coeff |
mov _EAX, prm2 ; coeff |
986 |
|
|
987 |
movq xmm6, [mmx_quant + ecx*8] ; quant |
lea TMP1, [mmx_quant] |
988 |
inc ecx |
movq xmm6, [TMP1 + TMP0*8] ; quant |
989 |
|
inc TMP0 |
990 |
pcmpeqw xmm5,xmm5 |
pcmpeqw xmm5,xmm5 |
991 |
and ecx,1 |
and TMP0,1 |
992 |
movlhps xmm6,xmm6 |
movlhps xmm6,xmm6 |
993 |
movd xmm0,ecx |
movd xmm0,TMP0d |
994 |
movdqa xmm7,xmm6 |
movdqa xmm7,xmm6 |
995 |
pshuflw xmm0,xmm0,0 |
pshuflw xmm0,xmm0,0 |
996 |
movdqa xmm4,xmm5 |
movdqa xmm4,xmm5 |
997 |
mov edx, [esp+ 4] ; data |
mov TMP1, prm1 ; data |
998 |
movlhps xmm0,xmm0 |
movlhps xmm0,xmm0 |
999 |
paddw xmm7,xmm7 |
paddw xmm7,xmm7 |
1000 |
psubw xmm6,xmm0 |
psubw xmm6,xmm0 |
1001 |
psrlw xmm4,5 ; 2047 |
psrlw xmm4,5 ; 2047 |
1002 |
mov ecx,4 |
mov TMP0,4 |
1003 |
pxor xmm5,xmm4 ; mm5=-2048 |
pxor xmm5,xmm4 ; mm5=-2048 |
1004 |
|
|
1005 |
.loop: |
.loop: |
1006 |
movdqa xmm0,[eax] |
movdqa xmm0,[_EAX] |
1007 |
pxor xmm3,xmm3 |
pxor xmm3,xmm3 |
1008 |
pxor xmm2,xmm2 |
pxor xmm2,xmm2 |
1009 |
pcmpeqw xmm3,xmm0 |
pcmpeqw xmm3,xmm0 |
1010 |
pcmpgtw xmm2,xmm0 |
pcmpgtw xmm2,xmm0 |
1011 |
pmullw xmm0,xmm7 ; * 2 * quant |
pmullw xmm0,xmm7 ; * 2 * quant |
1012 |
pandn xmm3,xmm6 |
pandn xmm3,xmm6 |
1013 |
movdqa xmm1,[eax+16] |
movdqa xmm1,[_EAX+16] |
1014 |
psubw xmm0,xmm2 |
psubw xmm0,xmm2 |
1015 |
pxor xmm2,xmm3 |
pxor xmm2,xmm3 |
1016 |
pxor xmm3,xmm3 |
pxor xmm3,xmm3 |
1029 |
pmaxsw xmm0,xmm5 |
pmaxsw xmm0,xmm5 |
1030 |
pmaxsw xmm1,xmm5 |
pmaxsw xmm1,xmm5 |
1031 |
|
|
1032 |
movdqa [edx],xmm0 |
movdqa [TMP1],xmm0 |
1033 |
lea eax,[eax+32] |
lea _EAX,[_EAX+32] |
1034 |
movdqa [edx+16],xmm1 |
movdqa [TMP1+16],xmm1 |
1035 |
|
|
1036 |
dec ecx |
dec TMP0 |
1037 |
lea edx,[edx+32] |
lea TMP1,[TMP1+32] |
1038 |
jne .loop |
jne .loop |
1039 |
|
|
1040 |
xor eax, eax ; return 0 |
xor _EAX, _EAX ; return 0 |
1041 |
ret |
ret |
1042 |
ENDFUNC |
ENDFUNC |
1043 |
|
|