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_mpeg_mmx.asm,v 1.7 2006-07-11 10:01:27 chl Exp $ |
; * $Id: quantize_mpeg_mmx.asm,v 1.10 2008-11-11 20:46:24 Isibaar Exp $ |
25 |
; * |
; * |
26 |
; *************************************************************************/ |
; *************************************************************************/ |
27 |
|
|
34 |
%ifdef MARK_FUNCS |
%ifdef MARK_FUNCS |
35 |
global _%1:function %1.endfunc-%1 |
global _%1:function %1.endfunc-%1 |
36 |
%define %1 _%1:function %1.endfunc-%1 |
%define %1 _%1:function %1.endfunc-%1 |
37 |
|
%define ENDFUNC .endfunc |
38 |
%else |
%else |
39 |
global _%1 |
global _%1 |
40 |
%define %1 _%1 |
%define %1 _%1 |
41 |
|
%define ENDFUNC |
42 |
%endif |
%endif |
43 |
%else |
%else |
44 |
%ifdef MARK_FUNCS |
%ifdef MARK_FUNCS |
45 |
global %1:function %1.endfunc-%1 |
global %1:function %1.endfunc-%1 |
46 |
|
%define ENDFUNC .endfunc |
47 |
%else |
%else |
48 |
global %1 |
global %1 |
49 |
|
%define ENDFUNC |
50 |
%endif |
%endif |
51 |
%endif |
%endif |
52 |
%endmacro |
%endmacro |
225 |
; calculate DC |
; calculate DC |
226 |
movsx eax, word [eax] ; data[0] |
movsx eax, word [eax] ; data[0] |
227 |
mov ecx, [esp + 4 + 16] ; dcscalar |
mov ecx, [esp + 4 + 16] ; dcscalar |
228 |
mov edx, ecx |
mov edx, eax |
229 |
shr edx, 1 ; edx = dcscalar /2 |
mov edi, ecx |
230 |
mov edi, edx |
shr ecx, 1 ; ecx = dcscalar/2 |
231 |
neg edi |
sar edx, 31 ; edx = sign extend of eax (ready for division too) |
232 |
|
xor ecx, edx ; adjust ecx according to the sign of data[0] |
233 |
cmp eax, 0 |
sub ecx, edx |
234 |
cmovg edx, edi |
add eax, ecx |
235 |
sub eax, edx |
|
236 |
|
mov ecx, [esp + 4 + 4] ; coeff again |
237 |
mov edi, [esp + 4 + 4] ; coeff again |
idiv edi ; eax = edx:eax / dcscalar |
238 |
|
mov [ecx], ax ; coeff[0] = ax |
|
cdq ; expand eax -> edx:eax |
|
|
idiv ecx ; eax = edx:eax / dcscalar |
|
|
|
|
|
mov [edi], ax ; coeff[0] = ax |
|
239 |
|
|
240 |
pop edi |
pop edi |
241 |
|
|
242 |
xor eax, eax ; return(0); |
xor eax, eax ; return(0); |
243 |
ret |
ret |
244 |
.endfunc |
ENDFUNC |
245 |
|
|
246 |
|
|
247 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
279 |
movq mm7, [mmx_div + eax * 8 - 8] ; divider |
movq mm7, [mmx_div + eax * 8 - 8] ; divider |
280 |
|
|
281 |
ALIGN 16 |
ALIGN 16 |
282 |
.loop |
.loop: |
283 |
movq mm0, [esi + 8*ecx] ; mm0 = [1st] |
movq mm0, [esi + 8*ecx] ; mm0 = [1st] |
284 |
movq mm3, [esi + 8*ecx + 8] ; |
movq mm3, [esi + 8*ecx + 8] ; |
285 |
pxor mm1, mm1 ; mm1 = 0 |
pxor mm1, mm1 ; mm1 = 0 |
319 |
cmp ecx, 16 |
cmp ecx, 16 |
320 |
jnz near .loop |
jnz near .loop |
321 |
|
|
322 |
.done |
.done: |
323 |
pmaddwd mm5, [mmx_one] |
pmaddwd mm5, [mmx_one] |
324 |
movq mm0, mm5 |
movq mm0, mm5 |
325 |
psrlq mm5, 32 |
psrlq mm5, 32 |
334 |
ret |
ret |
335 |
|
|
336 |
ALIGN 16 |
ALIGN 16 |
337 |
.q1loop |
.q1loop: |
338 |
movq mm0, [esi + 8*ecx] ; mm0 = [1st] |
movq mm0, [esi + 8*ecx] ; mm0 = [1st] |
339 |
movq mm3, [esi + 8*ecx+ 8] |
movq mm3, [esi + 8*ecx+ 8] |
340 |
pxor mm1, mm1 ; mm1 = 0 |
pxor mm1, mm1 ; mm1 = 0 |
375 |
jmp .done |
jmp .done |
376 |
|
|
377 |
ALIGN 16 |
ALIGN 16 |
378 |
.q2loop |
.q2loop: |
379 |
movq mm0, [esi + 8*ecx] ; mm0 = [1st] |
movq mm0, [esi + 8*ecx] ; mm0 = [1st] |
380 |
movq mm3, [esi + 8*ecx+ 8] |
movq mm3, [esi + 8*ecx+ 8] |
381 |
pxor mm1, mm1 ; mm1 = 0 |
pxor mm1, mm1 ; mm1 = 0 |
414 |
jnz near .q2loop |
jnz near .q2loop |
415 |
|
|
416 |
jmp .done |
jmp .done |
417 |
.endfunc |
ENDFUNC |
418 |
|
|
419 |
|
|
420 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
482 |
pxor mm6, mm6 ; this is a NOP |
pxor mm6, mm6 ; this is a NOP |
483 |
|
|
484 |
ALIGN 16 |
ALIGN 16 |
485 |
.loop |
.loop: |
486 |
movq mm0, [ecx+8*eax + 8*16] ; mm0 = c = coeff[i] |
movq mm0, [ecx+8*eax + 8*16] ; mm0 = c = coeff[i] |
487 |
movq mm3, [ecx+8*eax + 8*16 +8]; mm3 = c' = coeff[i+1] |
movq mm3, [ecx+8*eax + 8*16 +8]; mm3 = c' = coeff[i+1] |
488 |
pxor mm1, mm1 |
pxor mm1, mm1 |
554 |
pop ebx |
pop ebx |
555 |
|
|
556 |
ret |
ret |
557 |
.endfunc |
ENDFUNC |
558 |
|
|
559 |
;----------------------------------------------------------------------------- |
;----------------------------------------------------------------------------- |
560 |
; |
; |
586 |
pxor mm6, mm6 ; mismatch sum |
pxor mm6, mm6 ; mismatch sum |
587 |
|
|
588 |
ALIGN 16 |
ALIGN 16 |
589 |
.loop |
.loop: |
590 |
movq mm0, [ecx+8*eax + 8*16 ] ; mm0 = coeff[i] |
movq mm0, [ecx+8*eax + 8*16 ] ; mm0 = coeff[i] |
591 |
movq mm2, [ecx+8*eax + 8*16 +8] ; mm2 = coeff[i+1] |
movq mm2, [ecx+8*eax + 8*16 +8] ; mm2 = coeff[i+1] |
592 |
add eax, 2 |
add eax, 2 |
668 |
pop ebx |
pop ebx |
669 |
|
|
670 |
ret |
ret |
671 |
.endfunc |
ENDFUNC |
672 |
|
|
673 |
|
|
674 |
|
%ifidn __OUTPUT_FORMAT__,elf |
675 |
|
section ".note.GNU-stack" noalloc noexec nowrite progbits |
676 |
|
%endif |
677 |
|
|