idct_mmx.asm

quicktime linux播放器v1

	packssdw mm4, mm3 		; 3 	; y6 y7 y4 y5

	movq mm7, mm4 			; 7 	; y6 y7 y4 y5
	psrld mm4, 16 				; 0 y6 0 y4

	pslld mm7, 16 				; y7 0 y5 0
	movq [OUT_6], mm1 		; 1 	; save y3 y2 y1 y0
                             	
	por mm7, mm4 			; 4 	; y7 y6 y5 y4
	movq [OUT_6+8], mm7 		; 7 	; save y7 y6 y5 y4

	movq mm0, [INP_7] 		; 0	; x3 x2 x1 x0

	movq mm1, [INP_7+8]		; 1	; x7 x6 x5 x4
	movq mm2, mm0 			; 2	; x3 x2 x1 x0

	movq mm3, [TABLE_7]		; 3	; w06 w04 w02 w00
	punpcklwd mm0, mm1 			; x5 x1 x4 x0

	movq mm5, mm0 			; 5	; x5 x1 x4 x0
	punpckldq mm0, mm0 			; x4 x0 x4 x0

	movq mm4, [TABLE_7+8] 	; 4	; w07 w05 w03 w01
	punpckhwd mm2, mm1		; 1	; x7 x3 x6 x2

	pmaddwd mm3, mm0 			; x4*w06+x0*w04 x4*w02+x0*w00
	movq mm6, mm2 			; 6 	; x7 x3 x6 x2

	movq mm1, [TABLE_7+32] 	; 1 	; w22 w20 w18 w16
	punpckldq mm2, mm2 			; x6 x2 x6 x2

	pmaddwd mm4, mm2 			; x6*w07+x2*w05 x6*w03+x2*w01
	punpckhdq mm5, mm5 			; x5 x1 x5 x1

	pmaddwd mm0, [TABLE_7+16] 		; x4*w14+x0*w12 x4*w10+x0*w08
	punpckhdq mm6, mm6 			; x7 x3 x7 x3

	movq mm7, [TABLE_7+40] 	; 7 	; w23 w21 w19 w17
	pmaddwd mm1, mm5 			; x5*w22+x1*w20 x5*w18+x1*w16

	paddd mm3, [ROUNDER_7] 		; +rounder
	pmaddwd mm7, mm6 			; x7*w23+x3*w21 x7*w19+x3*w17

	pmaddwd mm2, [TABLE_7+24] 		; x6*w15+x2*w13 x6*w11+x2*w09
	paddd mm3, mm4 			; 4 	; a1=sum(even1) a0=sum(even0)

	pmaddwd mm5, [TABLE_7+48] 		; x5*w30+x1*w28 x5*w26+x1*w24
	movq mm4, mm3 			; 4 	; a1 a0

	pmaddwd mm6, [TABLE_7+56] 		; x7*w31+x3*w29 x7*w27+x3*w25
	paddd mm1, mm7 			; 7 	; b1=sum(odd1) b0=sum(odd0)

	paddd mm0, [ROUNDER_7]		; +rounder
	psubd mm3, mm1 				; a1-b1 a0-b0

	psrad mm3, SHIFT_INV_ROW 		; y6=a1-b1 y7=a0-b0
	paddd mm1, mm4 			; 4 	; a1+b1 a0+b0

	paddd mm0, mm2 			; 2 	; a3=sum(even3) a2=sum(even2)
	psrad mm1, SHIFT_INV_ROW 		; y1=a1+b1 y0=a0+b0

	paddd mm5, mm6 			; 6 	; b3=sum(odd3) b2=sum(odd2)
	movq mm4, mm0 			; 4 	; a3 a2

	paddd mm0, mm5 				; a3+b3 a2+b2
	psubd mm4, mm5 			; 5 	; a3-b3 a2-b2

	psrad mm0, SHIFT_INV_ROW 		; y3=a3+b3 y2=a2+b2
	psrad mm4, SHIFT_INV_ROW 		; y4=a3-b3 y5=a2-b2

	packssdw mm1, mm0 		; 0 	; y3 y2 y1 y0
	packssdw mm4, mm3 		; 3 	; y6 y7 y4 y5

	movq mm7, mm4 			; 7 	; y6 y7 y4 y5
	psrld mm4, 16 				; 0 y6 0 y4

	pslld mm7, 16 				; y7 0 y5 0
	movq [OUT_7], mm1 		; 1 	; save y3 y2 y1 y0
                             	
	por mm7, mm4 			; 4 	; y7 y6 y5 y4
	movq [OUT_7+8], mm7 		; 7 	; save y7 y6 y5 y4

	movq	mm0, [tg_3_16]

	movq	mm3, [INP_0+16*3]
	movq	mm1, mm0			; tg_3_16

	movq	mm5, [INP_0+16*5]
	pmulhw	mm0, mm3			; x3*(tg_3_16-1)

	movq	mm4, [tg_1_16]
	pmulhw	mm1, mm5			; x5*(tg_3_16-1)

	movq	mm7, [INP_0+16*7]
	movq	mm2, mm4			; tg_1_16

	movq	mm6, [INP_0+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	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	[SCRATCH+0], mm5	; save b0
	movq	[OUT_0+3*16], mm5	; save b0
	paddsw	mm1, mm2			; t1+t2

;	movq	[SCRATCH+8], mm6	; save b3
	movq	[OUT_0+5*16], mm6	; save b3
	psubsw	mm4, mm2			; t1-t2

	movq	mm5, [INP_0+2*16]
	movq	mm0, mm7			; tg_2_16

	movq	mm6, [INP_0+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, [INP_0+0*16]
	pmulhw	mm4, mm3			; ocos_4_16*(t1-t2) = b2/2

	psubsw	mm0, mm6			; t2*tg_2_16-x6 = tm26
	movq	mm3, mm2			; x0

	movq	mm6, [INP_0+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, [SCRATCH+0]	; load b0
	movq	mm1, [OUT_0+3*16]	; load b0
	psraw	mm7, SHIFT_INV_COL		; dst6

	movq	mm4, mm5			; a0
	psraw	mm0, SHIFT_INV_COL		; dst5

	movq	[OUT_0+1*16], mm3
	paddsw	mm5, mm1			; a0+b0

	movq	[OUT_0+2*16], mm6
	psubsw	mm4, mm1			; a0-b0

;	movq	mm3, [SCRATCH+8]	; load b3
	movq	mm3, [OUT_0+5*16]	; load b3
	psraw	mm5, SHIFT_INV_COL		; dst0

	movq	mm6, mm2			; a3
	psraw	mm4, SHIFT_INV_COL		; dst7

	movq	[OUT_0+5*16], mm0
	paddsw	mm2, mm3			; a3+b3

	movq	[OUT_0+6*16], mm7
	psubsw	mm6, mm3			; a3-b3

	movq	[OUT_0+0*16], mm5
	psraw	mm2, SHIFT_INV_COL		; dst3

	movq	[OUT_0+7*16], mm4
	psraw	mm6, SHIFT_INV_COL		; dst4

	movq	[OUT_0+3*16], mm2

	movq	[OUT_0+4*16], mm6

	movq	mm0, [tg_3_16]

	movq	mm3, [INP_8+16*3]
	movq	mm1, mm0			; tg_3_16

	movq	mm5, [INP_8+16*5]
	pmulhw	mm0, mm3			; x3*(tg_3_16-1)

	movq	mm4, [tg_1_16]
	pmulhw	mm1, mm5			; x5*(tg_3_16-1)

	movq	mm7, [INP_8+16*7]
	movq	mm2, mm4			; tg_1_16

	movq	mm6, [INP_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	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	[SCRATCH+0], mm5	; save b0
	movq	[OUT_8+3*16], mm5	; save b0
	paddsw	mm1, mm2			; t1+t2

;	movq	[SCRATCH+8], mm6	; save b3
	movq	[OUT_8+5*16], mm6	; save b3
	psubsw	mm4, mm2			; t1-t2

	movq	mm5, [INP_8+2*16]
	movq	mm0, mm7			; tg_2_16

	movq	mm6, [INP_8+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, [INP_8+0*16]
	pmulhw	mm4, mm3			; ocos_4_16*(t1-t2) = b2/2

	psubsw	mm0, mm6			; t2*tg_2_16-x6 = tm26
	movq	mm3, mm2			; x0

	movq	mm6, [INP_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, [SCRATCH+0]	; load b0
	movq	mm1, [OUT_8+3*16]	; load b0
	psraw	mm7, SHIFT_INV_COL		; dst6

	movq	mm4, mm5			; a0
	psraw	mm0, SHIFT_INV_COL		; dst5

	movq	[OUT_8+1*16], mm3
	paddsw	mm5, mm1			; a0+b0

	movq	[OUT_8+2*16], mm6
	psubsw	mm4, mm1			; a0-b0

;	movq	mm3, [SCRATCH+8]	; load b3
	movq	mm3, [OUT_8+5*16]	; load b3
	psraw	mm5, SHIFT_INV_COL		; dst0

	movq	mm6, mm2			; a3
	psraw	mm4, SHIFT_INV_COL		; dst7

	movq	[OUT_8+5*16], mm0
	paddsw	mm2, mm3			; a3+b3

	movq	[OUT_8+6*16], mm7
	psubsw	mm6, mm3			; a3-b3

	movq	[OUT_8+0*16], mm5
	psraw	mm2, SHIFT_INV_COL		; dst3

	movq	[OUT_8+7*16], mm4
	psraw	mm6, SHIFT_INV_COL		; dst4

	movq	[OUT_8+3*16], mm2

	movq	[OUT_8+4*16], mm6

	pop edi
	pop ebx
	
	emms

	ret