[svn] / branches / dev-api-4 / xvidcore / src / image / x86_asm / yv12_to_rgb24_mmx.asm Repository:
ViewVC logotype

View of /branches/dev-api-4/xvidcore/src/image/x86_asm/yv12_to_rgb24_mmx.asm

Parent Directory Parent Directory | Revision Log Revision Log


Revision 330 - (download) (annotate)
Tue Jul 23 15:36:59 2002 UTC (21 years, 8 months ago) by edgomez
Original Path: trunk/xvidcore/src/image/x86_asm/yv12_to_rgb24_mmx.asm
File size: 10782 byte(s)
Thread safe version.
;------------------------------------------------------------------------------
;
;  This file is part of XviD, a free MPEG-4 video encoder/decoder
;
;  XviD is free software; you can redistribute it and/or modify it
;  under the terms of the GNU General Public License as published by
;  the Free Software Foundation; either version 2 of the License, or
;  (at your option) any later version.
;
;  XviD is distributed in the hope that it will be useful, but
;  WITHOUT ANY WARRANTY; without even the implied warranty of
;  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
;  GNU General Public License for more details.
;
;  You should have received a copy of the GNU General Public License
;  along with this program; if not, write to the Free Software
;  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
;
;------------------------------------------------------------------------------
;------------------------------------------------------------------------------
;
;  yuv_to_rgb24.asm, MMX optimized color conversion
;
;  Copyright (C) 2001 - Michael Militzer <isibaar@xvid.org>
;
;  For more information visit the XviD homepage: http://www.xvid.org
;
;------------------------------------------------------------------------------
;------------------------------------------------------------------------------
;
;  Revision history :
;
;  13.12.2001 initial version  (Isibaar)
;  23.07.2002 Thread safe version (edgomez)
;
; $Id: yv12_to_rgb24_mmx.asm,v 1.2 2002-07-23 15:36:59 edgomez Exp $
; 
;------------------------------------------------------------------------------
;------------------------------------------------------------------------------
; NB:	n contrary to the c implementation this code does the conversion
;      using direct calculations. Input data width must be a multiple of 8
;      and height must be even.
;      This implementation is less precise than the c version but is
;      more than twice as fast :-)
;------------------------------------------------------------------------------
	
BITS 32


%macro cglobal 1 
	%ifdef PREFIX
		global _%1 
		%define %1 _%1
	%else
		global %1
	%endif
%endmacro


%define SCALEBITS 6


ALIGN 16

SECTION .data

Y_SUB		dw  16,  16,  16,  16
U_SUB		dw 128, 128, 128, 128
V_SUB		dw 128, 128, 128, 128

Y_MUL		dw  74,  74,  74,  74

UG_MUL		dw  25,  25,  25,  25
VG_MUL		dw  52,  52,  52,  52

UB_MUL		dw 129, 129, 129, 129
VR_MUL		dw 102, 102, 102, 102


ALIGN 16
	
SECTION .text

;------------------------------------------------------------------------------
; 
; void yv12_to_rgb24_mmx(uint8_t *dst,
;                        int dst_stride,
;                        uint8_t *y_src,
;                        uint8_t *u_src,
;                        uint8_t *v_src, 
;                        int y_stride, int uv_stride,
;                        int width, int height);
;
;------------------------------------------------------------------------------
	
cglobal yv12_to_rgb24_mmx
yv12_to_rgb24_mmx:

	push ebx
	push esi
	push edi
	push ebp

	;  local vars alloc
%define localsize 72
%define TEMP_Y1  esp
%define TEMP_Y2  esp + 8
%define TEMP_G1  esp + 16
%define TEMP_G2  esp + 24
%define TEMP_B1  esp + 32
%define TEMP_B2  esp + 40
%define y_dif    esp + 48
%define dst_dif	 esp + 52
%define uv_dif   esp + 56
%define height   esp + 60
%define width_8  esp + 64
%define height_2 esp + 68
	sub esp, localsize


	mov eax, [esp + 52 + localsize]	; height -> eax
	cmp eax, 0x00
	jge near dont_flip		; flip?
	
	neg eax
	mov [height], eax

	mov esi, [esp + 48 + localsize]	; width -> esi

	mov ebp, [esp + 40 + localsize]	; y_stride -> ebp
	mov ebx, ebp
	shl ebx, 1			; 2 * y_stride -> ebx
	neg ebx
	sub ebx, esi			; y_dif -> eax

	mov [y_dif], ebx

	sub eax, 1			; height - 1 -> eax
	mul ebp				; (height - 1) * y_stride -> ebp
	mov ecx, eax
	mov eax, [esp + 28 + localsize]	; y_src -> eax
	add eax, ecx			; y_src -> eax
	mov ebx, eax
	sub ebx, ebp			; y_src2 -> ebx

	mov ecx, [esp + 24 + localsize]	; dst_stride -> ecx
	mov edx, ecx
	add ecx, edx
	shl edx, 2
	add ecx, edx			; 6 * dst_stride -> ecx
	mov edx, ecx
	sub ecx, esi
	shl esi, 1
	sub ecx, esi			; 6 * dst_stride - 3 * width -> ecx

	mov [dst_dif], ecx

	mov esi, [esp + 20 + localsize]	; dst -> esi
	mov edi, esi
	shr edx, 1
	add edi, edx			; dst2 -> edi

	mov ebp, [esp + 48 + localsize]	; width -> ebp
	mov ecx, ebp			; width -> ecx
	shr ecx, 1
	shr ebp, 3			; width / 8 -> ebp
	mov [width_8], ebp

	mov ebp, [esp + 44 + localsize]	; uv_stride -> ebp
	mov edx, ebp
	neg edx
	sub edx, ecx
	mov [uv_dif], edx

	mov edx, ebp
	mov ebp, eax
	mov eax, [height]		; height -> eax
	shr eax, 1			; height / 2 -> eax

	mov ecx, [esp + 32 + localsize]	; u_src -> ecx
	sub eax, 1
	mul edx
	add ecx, eax

	mov edx, [esp + 36 + localsize]	; v_src -> edx
	add edx, eax

	mov eax, ebp

	mov ebp, [height]		; height -> ebp
	shr ebp, 1			; height / 2 -> ebp

	pxor mm7, mm7			; clear mm7
	jmp y_loop


dont_flip:
	mov esi, [esp + 48 + localsize]	; width -> esi

	mov ebp, [esp + 40 + localsize]	; y_stride -> ebp
	mov ebx, ebp
	shl ebx, 1			; 2 * y_stride -> ebx
	sub ebx, esi			; y_dif -> ebx

	mov [y_dif], ebx

	mov eax, [esp + 28 + localsize]	; y_src -> eax
	mov ebx, eax
	add ebx, ebp			; y_src2 -> ebp

	mov ecx, [esp + 24 + localsize]	; dst_stride -> ecx
	mov edx, ecx
	add ecx, edx
	shl edx, 2
	add ecx, edx			; 6 * dst_stride -> ecx
	mov edx, ecx
	sub ecx, esi
	shl esi, 1
	sub ecx, esi			; 6 * dst_stride - 3 * width -> ecx

	mov [dst_dif], ecx

	mov esi, [esp + 20 + localsize]	; dst -> esi
	mov edi, esi
	shr edx, 1
	add edi, edx			; dst2 -> edi

	mov ebp, [esp + 48 + localsize]	; width -> ebp
	mov ecx, ebp			; width -> ecx
	shr ecx, 1
	shr ebp, 3			; width / 8 -> ebp
	mov [width_8], ebp

	mov ebp, [esp + 44 + localsize]	; uv_stride -> ebp
	sub ebp, ecx
	mov [uv_dif], ebp

	mov ecx, [esp + 32 + localsize]	; u_src -> ecx
	mov edx, [esp + 36 + localsize]	; v_src -> edx

	mov ebp, [esp + 52 + localsize]	; height -> ebp
	shr ebp, 1			; height / 2 -> ebp

	pxor mm7, mm7

y_loop:
	mov [height_2], ebp
	mov ebp, [width_8]

x_loop:
	movd mm2, [ecx]
	movd mm3, [edx]

	punpcklbw mm2, mm7		; u3u2u1u0 -> mm2
	punpcklbw mm3, mm7		; v3v2v1v0 -> mm3

	psubsw mm2, [U_SUB]		; U - 128
	psubsw mm3, [V_SUB]		; V - 128

	movq mm4, mm2
	movq mm5, mm3

	pmullw mm2, [UG_MUL]
	pmullw mm3, [VG_MUL]

	movq mm6, mm2			; u3u2u1u0 -> mm6
	punpckhwd mm2, mm2		; u3u3u2u2 -> mm2
	punpcklwd mm6, mm6		; u1u1u0u0 -> mm6

	pmullw mm4, [UB_MUL]		; B_ADD -> mm4

	movq mm0, mm3
	punpckhwd mm3, mm3		; v3v3v2v2 -> mm2
	punpcklwd mm0, mm0		; v1v1v0v0 -> mm6

	paddsw mm2, mm3
	paddsw mm6, mm0

	pmullw mm5, [VR_MUL]		; R_ADD -> mm5

	movq mm0, [eax]			; y7y6y5y4y3y2y1y0 -> mm0

	movq mm1, mm0
	punpckhbw mm1, mm7		; y7y6y5y4 -> mm1
	punpcklbw mm0, mm7		; y3y2y1y0 -> mm0

	psubsw mm0, [Y_SUB]		; Y - Y_SUB
	psubsw mm1, [Y_SUB]		; Y - Y_SUB

	pmullw mm1, [Y_MUL] 
	pmullw mm0, [Y_MUL]

	movq [TEMP_Y2], mm1		; y7y6y5y4 -> mm3
	movq [TEMP_Y1], mm0		; y3y2y1y0 -> mm7

	psubsw mm1, mm2			; g7g6g5g4 -> mm1
	psubsw mm0, mm6			; g3g2g1g0 -> mm0

	psraw mm1, SCALEBITS
	psraw mm0, SCALEBITS

	packuswb mm0, mm1		;g7g6g5g4g3g2g1g0 -> mm0

	movq [TEMP_G1], mm0

	movq mm0, [ebx]			; y7y6y5y4y3y2y1y0 -> mm0

	movq mm1, mm0

	punpckhbw mm1, mm7		; y7y6y5y4 -> mm1
	punpcklbw mm0, mm7		; y3y2y1y0 -> mm0

	psubsw mm0, [Y_SUB]		; Y - Y_SUB
	psubsw mm1, [Y_SUB]		; Y - Y_SUB

	pmullw mm1, [Y_MUL] 
	pmullw mm0, [Y_MUL]

	movq mm3, mm1
	psubsw mm1, mm2			; g7g6g5g4 -> mm1

	movq mm2, mm0
	psubsw mm0, mm6			; g3g2g1g0 -> mm0

	psraw mm1, SCALEBITS
	psraw mm0, SCALEBITS

	packuswb mm0, mm1		; g7g6g5g4g3g2g1g0 -> mm0

	movq [TEMP_G2], mm0

	movq mm0, mm4
	punpckhwd mm4, mm4		; u3u3u2u2 -> mm2
	punpcklwd mm0, mm0		; u1u1u0u0 -> mm6

	movq mm1, mm3			; y7y6y5y4 -> mm1
	paddsw mm3, mm4			; b7b6b5b4 -> mm3

	movq mm7, mm2			; y3y2y1y0 -> mm7

	paddsw mm2, mm0			; b3b2b1b0 -> mm2

	psraw mm3, SCALEBITS
	psraw mm2, SCALEBITS

	packuswb mm2, mm3		; b7b6b5b4b3b2b1b0 -> mm2

	movq [TEMP_B2], mm2

	movq mm3, [TEMP_Y2]
	movq mm2, [TEMP_Y1]

	movq mm6, mm3			; TEMP_Y2 -> mm6
	paddsw mm3, mm4			; b7b6b5b4 -> mm3

	movq mm4, mm2			; TEMP_Y1 -> mm4
	paddsw mm2, mm0			; b3b2b1b0 -> mm2

	psraw mm3, SCALEBITS
	psraw mm2, SCALEBITS

	packuswb mm2, mm3		; b7b6b5b4b3b2b1b0 -> mm2

	movq [TEMP_B1], mm2

	movq mm0, mm5
	punpckhwd mm5, mm5		; v3v3v2v2 -> mm5
	punpcklwd mm0, mm0		; v1v1v0v0 -> mm0

	paddsw mm1, mm5			; r7r6r5r4 -> mm1
	paddsw mm7, mm0			; r3r2r1r0 -> mm7

	psraw mm1, SCALEBITS
	psraw mm7, SCALEBITS

	packuswb mm7, mm1		; r7r6r5r4r3r2r1r0 -> mm7 (TEMP_R2)

	paddsw mm6, mm5			; r7r6r5r4 -> mm6
	paddsw mm4, mm0			; r3r2r1r0 -> mm4

	psraw mm6, SCALEBITS
	psraw mm4, SCALEBITS

	packuswb mm4, mm6		; r7r6r5r4r3r2r1r0 -> mm4 (TEMP_R1)
	
	movq mm0, [TEMP_B1]
	movq mm1, [TEMP_G1]

	movq mm6, mm7

	movq mm2, mm0
	punpcklbw mm2, mm4		; r3b3r2b2r1b1r0b0 -> mm2
	punpckhbw mm0, mm4		; r7b7r6b6r5b5r4b4 -> mm0

	pxor mm7, mm7

	movq mm3, mm1
	punpcklbw mm1, mm7		; 0g30g20g10g0 -> mm1
	punpckhbw mm3, mm7		; 0g70g60g50g4 -> mm3

	movq mm4, mm2
	punpcklbw mm2, mm1		; 0r1g1b10r0g0b0 -> mm2
	punpckhbw mm4, mm1		; 0r3g3b30r2g2b2 -> mm4

	movq mm5, mm0
	punpcklbw mm0, mm3		; 0r5g5b50r4g4b4 -> mm0
	punpckhbw mm5, mm3		; 0r7g7b70r6g6b6 -> mm5

	movd [esi], mm2			
	psrlq mm2, 32

	movd [esi + 3], mm2
	movd [esi + 6], mm4		

	psrlq mm4, 32

	movd [esi + 9], mm4
	movd [esi + 12], mm0	

	psrlq mm0, 32

	movd [esi + 15], mm0
	movd [esi + 18], mm5	

	psrlq mm5, 32

	movd [esi + 21], mm5	

	movq mm0, [TEMP_B2]
	movq mm1, [TEMP_G2]

	movq mm2, mm0
	punpcklbw mm2, mm6		; r3b3r2b2r1b1r0b0 -> mm2
	punpckhbw mm0, mm6		; r7b7r6b6r5b5r4b4 -> mm0

	movq mm3, mm1 
	punpcklbw mm1, mm7		; 0g30g20g10g0 -> mm1
	punpckhbw mm3, mm7		; 0g70g60g50g4 -> mm3

	movq mm4, mm2
	punpcklbw mm2, mm1		; 0r1g1b10r0g0b0 -> mm2
	punpckhbw mm4, mm1		; 0r3g3b30r2g2b2 -> mm4

	movq mm5, mm0
	punpcklbw mm0, mm3		; 0r5g5b50r4g4b4 -> mm0
	punpckhbw mm5, mm3		; 0r7g7b70r6g6b6 -> mm5

	movd [edi], mm2
	psrlq mm2, 32

	movd [edi + 3], mm2
	movd [edi + 6], mm4

	psrlq mm4, 32

	movd [edi + 9], mm4
	movd [edi + 12], mm0

	psrlq mm0, 32

	movd [edi + 15], mm0
	movd [edi + 18], mm5

	psrlq mm5, 32

	movd [edi + 21], mm5

	add esi, 24
	add edi, 24

	add eax, 8
	add ebx, 8
	add ecx, 4
	add edx, 4

	dec ebp

	jnz near x_loop

	add esi, [dst_dif]
	add edi, [dst_dif]

	add eax, [y_dif]
	add ebx, [y_dif]

	add ecx, [uv_dif]
	add edx, [uv_dif]

	mov ebp, [height_2]
	dec ebp
	jnz near y_loop

	emms

	;; Local vars deallocation
	add esp, localsize
%undef TEMP_Y1
%undef TEMP_Y2
%undef TEMP_G1
%undef TEMP_G2
%undef TEMP_B1
%undef TEMP_B2
%undef y_dif
%undef dst_dif
%undef uv_dif
%undef height
%undef width_8
%undef height_2
%undef localsize

	pop ebp
	pop edi
	pop esi
	pop ebx

	ret

No admin address has been configured
ViewVC Help
Powered by ViewVC 1.0.4