idct_3dne.asm

wince下的xvidcore开发库,可用于MP4等视频播放开发

   pmaddwd mm5, [tab_i_26_xmm+56]; x7*w31+x5*w30 x7*w27+x5*w26
   paddd mm3, mm4                ; 4 free    ; a1=sum(even1) a0=sum(even0)
   paddd mm0, mm1                ; 1 free    ; a3=sum(even3) a2=sum(even2)
   pshufw mm1, [eax+48+8],10001000b      ; x6 x4 x6 x4
   movq mm4, mm3                 ; 4     ; a1 a0
   paddd mm6, mm7                ; 7 free    ; b1=sum(odd1) b0=sum(odd0)
   paddd mm2, mm5                ; 5 free    ; b3=sum(odd3) b2=sum(odd2)
   pshufw mm5, [eax+48],10001000b; x2 x0 x2 x0   mm5 & mm0 exchanged for next cycle
   movq mm7, mm0                 ; 7     ; a3 a2
   psubd mm4, mm6                ; 6 free    ; a1-b1 a0-b0
   paddd mm6, mm3                ; mm6 = mm3+mm6+mm5+mm4; a1+b1 a0+b0
   movq mm3, [tab_i_35_xmm]      ; 3     ; w05 w04 w01 w00
   psubd mm7, mm2                ; ; a3-b3 a2-b2
   paddd mm0, mm2                ; 0 free a3+b3 a2+b2
   pshufw mm2, [eax+48],11011101b; x3 x1 x3 x1
   pmaddwd mm3, mm5              ; x2*w05+x0*w04 x2*w01+x0*w00
   pmaddwd mm5, [tab_i_35_xmm+16]; x2*w13+x0*w12 x2*w09+x0*w08
   psrad mm4, SHIFT_INV_ROW      ; y6=a1-b1 y7=a0-b0
   psrad mm7, SHIFT_INV_ROW      ; y4=a3-b3 y5=a2-b2
   psrad mm6, SHIFT_INV_ROW      ; y1=a1+b1 y0=a0+b0
   psrad mm0, SHIFT_INV_ROW      ; y3=a3+b3 y2=a2+b2
   packssdw mm7, mm4             ; 4     ; y6 y7 y4 y5
   packssdw mm6, mm0             ; 0 free    ; y3 y2 y1 y0
   pshufw mm0, [eax+48+8],11011101b  ; x7 x5 x7 x5
   movq [eax+32], mm6            ; 3     ; save y3 y2 y1 y0 stall2
 
 ;   DCT_8_INV_ROW_1_s [eax+48], [eax+48], tab_i_35_xmm, rounder_3
   movq mm4, [tab_i_35_xmm+8]    ; 4     ; w07 w06 w03 w02
   movq mm6, [tab_i_35_xmm+32]   ; 6     ; w21 w20 w17 w16
   pshufw mm7, mm7, 10110001b    ; y7 y6 y5 y4
   paddd mm3, [rounder_3]        ; +rounder stall 6
   paddd mm5, [rounder_3]        ; +rounder
   movq [eax+32+8], mm7          ; 7     ; save y7 y6 y5 y4
   movq mm7, [tab_i_35_xmm+40]   ; 7     ; w23 w22 w19 w18
   pmaddwd mm4, mm1              ; x6*w07+x4*w06 x6*w03+x4*w02
   pmaddwd mm1, [tab_i_35_xmm+24]; x6*w15+x4*w14 x6*w11+x4*w10
   pmaddwd mm6, mm2              ; x3*w21+x1*w20 x3*w17+x1*w16
   pmaddwd mm2, [tab_i_35_xmm+48]; x3*w29+x1*w28 x3*w25+x1*w24
   pmaddwd mm7, mm0              ; 7     ; x7*w23+x5*w22 x7*w19+x5*w18 ; w23 w22 w19 w18
   pmaddwd mm0, [tab_i_35_xmm+56]; x7*w31+x5*w30 x7*w27+x5*w26
   paddd mm3, mm4                ; 4 free    ; a1=sum(even1) a0=sum(even0)
   paddd mm5, mm1                ; mm1 free  ; a3=sum(even3) a2=sum(even2)
   movq mm1, [tg_3_16]
   movq mm4, mm3                 ; 4     ; a1 a0
   paddd mm6, mm7                ; 7 free    ; b1=sum(odd1) b0=sum(odd0)
   paddd mm2, mm0                ; 5 free    ; b3=sum(odd3) b2=sum(odd2)
   movq mm0, [tg_3_16]
   movq mm7, mm5                 ; 7     ; a3 a2
   psubd mm4, mm6                ; 6 free    ; a1-b1 a0-b0
   paddd mm3, mm6                ; mm3 = mm3+mm6+mm5+mm4; a1+b1 a0+b0
   psubd mm7, mm2                ; ; a3-b3 a2-b2
   paddd mm2, mm5                ; 0 free a3+b3 a2+b2
   movq mm5, [eax+16*5]
   psrad mm4, SHIFT_INV_ROW      ; y6=a1-b1 y7=a0-b0
   psrad mm7, SHIFT_INV_ROW      ; y4=a3-b3 y5=a2-b2
   psrad mm3, SHIFT_INV_ROW      ; y1=a1+b1 y0=a0+b0
   psrad mm2, SHIFT_INV_ROW      ; y3=a3+b3 y2=a2+b2
   movq mm6, [eax+16*1]
   packssdw mm7, mm4             ; 4     ; y6 y7 y4 y5
   movq mm4, [tg_1_16]
   packssdw mm3, mm2             ; 0 free    ; y3 y2 y1 y0
   pshufw mm2, mm7, 10110001b    ; y7 y6 y5 y4
 
 ;   DCT_8_INV_COL_4 [eax+0],[eax+0]
 ;   movq    mm3,mmword ptr [eax+16*3]
   movq mm7, [eax+16*7]
   pmulhw mm0, mm3           ; x3*(tg_3_16-1)
   pmulhw mm1, mm5           ; x5*(tg_3_16-1)
   movq [eax+48+8], mm2      ; 7     ; save y7 y6 y5 y4
   movq mm2, mm4             ; tg_1_16
   pmulhw mm4, mm7           ; x7*tg_1_16
   paddsw mm0, mm3           ; x3*tg_3_16
   pmulhw mm2, mm6           ; x1*tg_1_16
   paddsw mm1, mm3           ; x3+x5*(tg_3_16-1)
   psubsw mm0, mm5           ; x3*tg_3_16-x5 = tm35
   movq [eax+48], mm3        ; 3     ; save y3 y2 y1 y0
   movq mm3, [ocos_4_16]
   paddsw mm1, mm5           ; x3+x5*tg_3_16 = tp35
   paddsw mm4, mm6           ; x1+tg_1_16*x7 = tp17
   psubsw mm2, mm7           ; x1*tg_1_16-x7 = tm17
   movq mm5, mm4             ; tp17
   movq mm6, mm2             ; tm17
   paddsw mm5, mm1           ; tp17+tp35 = b0
   psubsw mm6, mm0           ; tm17-tm35 = b3
   psubsw mm4, mm1           ; tp17-tp35 = t1
   paddsw mm2, mm0           ; tm17+tm35 = t2
   movq mm7, [tg_2_16]
   movq mm1, mm4             ; t1
   movq [eax+3*16], mm5      ; save b0
   paddsw mm1, mm2           ; t1+t2
   movq [eax+5*16], mm6      ; save b3
   psubsw mm4, mm2           ; t1-t2
   movq mm5, [eax+2*16]
   movq mm0, mm7             ; tg_2_16
   movq mm6, [eax+6*16]
   pmulhw mm0, mm5           ; x2*tg_2_16
   pmulhw mm7, mm6           ; x6*tg_2_16
 ; slot
   pmulhw mm1, mm3           ; ocos_4_16*(t1+t2) = b1/2
 ; slot
   movq mm2, [eax+0*16]
   pmulhw mm4, mm3           ; ocos_4_16*(t1-t2) = b2/2
   psubsw mm0, mm6           ; t2*tg_2_16-x6 = tm26
   movq mm3, [eax+0*16]      ; x0
   movq mm6, [eax+4*16]
   paddsw mm7, mm5           ; x2+x6*tg_2_16 = tp26
   paddsw mm2, mm6           ; x0+x4 = tp04
   psubsw mm3, mm6           ; x0-x4 = tm04
   movq mm5, mm2             ; tp04
   movq mm6, mm3             ; tm04
   psubsw mm2, mm7           ; tp04-tp26 = a3
   paddsw mm3, mm0           ; tm04+tm26 = a1
   paddsw mm1, mm1           ; b1
   paddsw mm4, mm4           ; b2
   paddsw mm5, mm7           ; tp04+tp26 = a0
   psubsw mm6, mm0           ; tm04-tm26 = a2
   movq mm7, mm3             ; a1
   movq mm0, mm6             ; a2
   paddsw mm3, mm1           ; a1+b1
   paddsw mm6, mm4           ; a2+b2
   psraw mm3, SHIFT_INV_COL  ; dst1
   psubsw mm7, mm1           ; a1-b1
   psraw mm6, SHIFT_INV_COL  ; dst2
   psubsw mm0, mm4           ; a2-b2
   movq mm1, [eax+3*16]      ; load b0
   psraw mm7, SHIFT_INV_COL  ; dst6
   movq mm4, mm5             ; a0
   psraw mm0, SHIFT_INV_COL  ; dst5
   movq [eax+1*16], mm3
   paddsw mm5, mm1           ; a0+b0
   movq [eax+2*16], mm6
   psubsw mm4, mm1           ; a0-b0
   movq mm3, [eax+5*16]      ; load b3
   psraw mm5, SHIFT_INV_COL  ; dst0
   movq mm6, mm2             ; a3
   psraw mm4, SHIFT_INV_COL  ; dst7
   movq [eax+5*16], mm0
   movq mm0, [tg_3_16]
   paddsw mm2, mm3           ; a3+b3
   movq [eax+6*16], mm7
   psubsw mm6, mm3           ; a3-b3
   movq mm3, [eax+8+16*3]
   movq [eax+0*16], mm5
   psraw mm2, SHIFT_INV_COL  ; dst3
   movq [eax+7*16], mm4
 
  ;  DCT_8_INV_COL_4 [eax+8],[eax+8]
   movq mm1, mm0             ; tg_3_16
   movq mm5, [eax+8+16*5]
   psraw mm6, SHIFT_INV_COL  ; dst4
   pmulhw mm0, mm3           ; x3*(tg_3_16-1)
   movq mm4, [tg_1_16]
   pmulhw mm1, mm5           ; x5*(tg_3_16-1)
   movq mm7, [eax+8+16*7]
   movq [eax+3*16], mm2
   movq mm2, mm4             ; tg_1_16
   movq [eax+4*16], mm6
   movq mm6, [eax+8+16*1]
   pmulhw mm4, mm7           ; x7*tg_1_16
   paddsw mm0, mm3           ; x3*tg_3_16
   pmulhw mm2, mm6           ; x1*tg_1_16
   paddsw mm1, mm3           ; x3+x5*(tg_3_16-1)
   psubsw mm0, mm5           ; x3*tg_3_16-x5 = tm35
   movq mm3, [ocos_4_16]
   paddsw mm1, mm5           ; x3+x5*tg_3_16 = tp35
   paddsw mm4, mm6           ; x1+tg_1_16*x7 = tp17
   psubsw mm2, mm7           ; x1*tg_1_16-x7 = tm17
   movq mm5, mm4             ; tp17
   movq mm6, mm2             ; tm17
   paddsw mm5, mm1           ; tp17+tp35 = b0
   psubsw mm4, mm1           ; tp17-tp35 = t1
   paddsw mm2, mm0           ; tm17+tm35 = t2
   movq mm7, [tg_2_16]
   movq mm1, mm4             ; t1
   psubsw mm6, mm0           ; tm17-tm35 = b3
   movq [eax+8+3*16], mm5    ; save b0
   movq [eax+8+5*16], mm6    ; save b3
   psubsw mm4, mm2           ; t1-t2
   movq mm5, [eax+8+2*16]
   movq mm0, mm7             ; tg_2_16
   movq mm6, [eax+8+6*16]
   paddsw mm1, mm2           ; t1+t2
   pmulhw mm0, mm5           ; x2*tg_2_16
   pmulhw mm7, mm6           ; x6*tg_2_16
   movq mm2, [eax+8+0*16]
   pmulhw mm4, mm3           ; ocos_4_16*(t1-t2) = b2/2
   psubsw mm0, mm6           ; t2*tg_2_16-x6 = tm26
  ; slot
   pmulhw mm1, mm3           ; ocos_4_16*(t1+t2) = b1/2
  ; slot
   movq mm3, [eax+8+0*16]    ; x0
   movq mm6, [eax+8+4*16]
   paddsw mm7, mm5           ; x2+x6*tg_2_16 = tp26
   paddsw mm2, mm6           ; x0+x4 = tp04
   psubsw mm3, mm6           ; x0-x4 = tm04
   movq mm5, mm2             ; tp04
   movq mm6, mm3             ; tm04
   psubsw mm2, mm7           ; tp04-tp26 = a3
   paddsw mm3, mm0           ; tm04+tm26 = a1
   paddsw mm1, mm1           ; b1
   paddsw mm4, mm4           ; b2
   paddsw mm5, mm7           ; tp04+tp26 = a0
   psubsw mm6, mm0           ; tm04-tm26 = a2
   movq mm7, mm3             ; a1
   movq mm0, mm6             ; a2
   paddsw mm3, mm1           ; a1+b1
   paddsw mm6, mm4           ; a2+b2
   psraw mm3, SHIFT_INV_COL  ; dst1
   psubsw mm7, mm1           ; a1-b1
   psraw mm6, SHIFT_INV_COL  ; dst2
   psubsw mm0, mm4           ; a2-b2
   movq mm1, [eax+8+3*16]    ; load b0
   psraw mm7, SHIFT_INV_COL  ; dst6
   movq mm4, mm5             ; a0
   psraw mm0, SHIFT_INV_COL  ; dst5
   movq [eax+8+1*16], mm3
   paddsw mm5, mm1           ; a0+b0
   movq [eax+8+2*16], mm6
   psubsw mm4, mm1           ; a0-b0
   movq mm3, [eax+8+5*16]    ; load b3
   psraw mm5, SHIFT_INV_COL  ; dst0
   movq mm6, mm2         ; a3
   psraw mm4, SHIFT_INV_COL  ; dst7
   movq [eax+8+5*16], mm0
   paddsw mm2, mm3           ; a3+b3
   movq [eax+8+6*16], mm7
   psubsw mm6, mm3           ; a3-b3
   movq [eax+8+0*16], mm5
   psraw mm2, SHIFT_INV_COL  ; dst3
   movq [eax+8+7*16], mm4
   psraw mm6, SHIFT_INV_COL  ; dst4
   movq [eax+8+3*16], mm2
   movq [eax+8+4*16], mm6
 
   ret