shithub: openh264

Info  •  Files  •  Log  •  Branches

Merge pull request #252 from mstorsjo/share-processing-asm

Merge identical assembly code between the processing and encoder libs

--- a/codec/build/win32/enc/WelsEncCore.vcproj

+++ b/codec/build/win32/enc/WelsEncCore.vcproj

@@ -2106,7 +2106,7 @@

 				</FileConfiguration>

 			</File>

 			<File

-				RelativePath="..\..\..\encoder\core\asm\satd_sad.asm"

+				RelativePath="..\..\..\common\satd_sad.asm"

 				>

 				<FileConfiguration

 					Name="Debug|Win32"

--- /dev/null

+++ b/codec/common/satd_sad.asm

@@ -1,0 +1,2344 @@

+;*!

+;* \copy

+;*     Copyright (c)  2009-2013, Cisco Systems

+;*     All rights reserved.

+;*

+;*     Redistribution and use in source and binary forms, with or without

+;*     modification, are permitted provided that the following conditions

+;*     are met:

+;*

+;*        * Redistributions of source code must retain the above copyright

+;*          notice, this list of conditions and the following disclaimer.

+;*

+;*        * Redistributions in binary form must reproduce the above copyright

+;*          notice, this list of conditions and the following disclaimer in

+;*          the documentation and/or other materials provided with the

+;*          distribution.

+;*

+;*     THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS

+;*     "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT

+;*     LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS

+;*     FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE

+;*     COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,

+;*     INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,

+;*     BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;

+;*     LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER

+;*     CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT

+;*     LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN

+;*     ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE

+;*     POSSIBILITY OF SUCH DAMAGE.

+;*

+;*

+;*  satd_sad.asm

+;*

+;*  Abstract

+;*      WelsSampleSatd4x4_sse2

+;*      WelsSampleSatd8x8_sse2

+;*      WelsSampleSatd16x8_sse2

+;*      WelsSampleSatd8x16_sse2

+;*      WelsSampleSatd16x16_sse2

+;*

+;*      WelsSampleSad16x8_sse2

+;*      WelsSampleSad16x16_sse2

+;*

+;*  History

+;*      8/5/2009 Created

+;*     24/9/2009 modified

+;*

+;*

+;*************************************************************************/

+

+%include "asm_inc.asm"

+

+;***********************************************************************

+; Data

+;***********************************************************************

+SECTION .rodata align=16

+

+align 16

+HSumSubDB1:   db 1,1,1,1,1,1,1,1,1,-1,1,-1,1,-1,1,-1

+align 16

+HSumSubDW1:   dw 1,-1,1,-1,1,-1,1,-1

+align 16

+PDW1:  dw 1,1,1,1,1,1,1,1

+align 16

+PDQ2:  dw 2,0,0,0,2,0,0,0

+align 16

+HSwapSumSubDB1:   times 2 db 1, 1, 1, 1, 1, -1, 1, -1

+

+;***********************************************************************

+; Code

+;***********************************************************************

+SECTION .text

+

+;***********************************************************************

+;

+;Pixel_satd_wxh_sse2 BEGIN

+;

+;***********************************************************************

+%macro MMX_DW_1_2REG 2

+      pxor %1, %1

+      pcmpeqw %2, %2

+      psubw %1, %2

+%endmacro

+

+%macro  SSE2_SumWHorizon1 2

+	movdqa      %2, %1

+	psrldq      %2, 8

+	paddusw     %1, %2

+	movdqa      %2, %1

+	psrldq      %2, 4

+	paddusw     %1, %2

+	movdqa      %2, %1

+	psrldq      %2, 2

+	paddusw     %1, %2

+%endmacro

+

+%macro SSE2_HDMTwo4x4 5 ;in: xmm1,xmm2,xmm3,xmm4  pOut: xmm4,xmm2,xmm1,xmm3

+   SSE2_SumSub %1, %2, %5

+   SSE2_SumSub %3, %4, %5

+   SSE2_SumSub %2, %4, %5

+   SSE2_SumSub %1, %3, %5

+%endmacro

+

+%macro SSE2_SumAbs4 7

+	WELS_AbsW %1, %3

+	WELS_AbsW %2, %3

+	WELS_AbsW %4, %6

+	WELS_AbsW %5, %6

+	paddusw       %1, %2

+	paddusw       %4, %5

+	paddusw       %7, %1

+	paddusw       %7, %4

+%endmacro

+

+%macro  SSE2_SumWHorizon 3

+	movhlps		%2, %1			; x2 = xx xx xx xx d7 d6 d5 d4

+	paddw		%1, %2			; x1 = xx xx xx xx d37 d26 d15 d04

+	punpcklwd	%1, %3			; x1 =  d37  d26 d15 d04

+	movhlps		%2, %1			; x2 = xxxx xxxx d37 d26

+	paddd		%1, %2			; x1 = xxxx xxxx d1357 d0246

+	pshuflw		%2, %1, 0x4e	; x2 = xxxx xxxx d0246 d1357

+	paddd		%1, %2			; x1 = xxxx xxxx xxxx  d01234567

+%endmacro

+

+%macro SSE2_GetSatd8x8 0

+	SSE2_LoadDiff8P    xmm0,xmm4,xmm7,[r0],[r2]

+	SSE2_LoadDiff8P    xmm1,xmm5,xmm7,[r0+r1],[r2+r3]

+	lea                 r0, [r0+2*r1]

+	lea                 r2, [r2+2*r3]

+	SSE2_LoadDiff8P    xmm2,xmm4,xmm7,[r0],[r2]

+	SSE2_LoadDiff8P    xmm3,xmm5,xmm7,[r0+r1],[r2+r3]

+

+	SSE2_HDMTwo4x4       xmm0,xmm1,xmm2,xmm3,xmm4

+	SSE2_TransTwo4x4W     xmm3,xmm1,xmm0,xmm2,xmm4

+	SSE2_HDMTwo4x4       xmm3,xmm1,xmm2,xmm4,xmm5

+	SSE2_SumAbs4         xmm4,xmm1,xmm0,xmm2,xmm3,xmm5,xmm6

+

+	lea					r0,    [r0+2*r1]

+    lea					r2,    [r2+2*r3]

+	SSE2_LoadDiff8P    xmm0,xmm4,xmm7,[r0],[r2]

+	SSE2_LoadDiff8P    xmm1,xmm5,xmm7,[r0+r1],[r2+r3]

+	lea                 r0, [r0+2*r1]

+	lea                 r2, [r2+2*r3]

+	SSE2_LoadDiff8P    xmm2,xmm4,xmm7,[r0],[r2]

+	SSE2_LoadDiff8P    xmm3,xmm5,xmm7,[r0+r1],[r2+r3]

+

+	SSE2_HDMTwo4x4       xmm0,xmm1,xmm2,xmm3,xmm4

+	SSE2_TransTwo4x4W     xmm3,xmm1,xmm0,xmm2,xmm4

+	SSE2_HDMTwo4x4       xmm3,xmm1,xmm2,xmm4,xmm5

+	SSE2_SumAbs4         xmm4,xmm1,xmm0,xmm2,xmm3,xmm5,xmm6

+%endmacro

+

+;***********************************************************************

+;

+;int32_t WelsSampleSatd4x4_sse2( uint8_t *, int32_t, uint8_t *, int32_t );

+;

+;***********************************************************************

+WELS_EXTERN WelsSampleSatd4x4_sse2

+align 16

+WelsSampleSatd4x4_sse2:

+	;push      ebx

+	;mov       eax,  [esp+8]

+	;mov       ebx,  [esp+12]

+	;mov       ecx,  [esp+16]

+	;mov       edx,  [esp+20]

+

+	%assign  push_num 0

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+    movd      xmm0, [r0]

+    movd      xmm1, [r0+r1]

+    lea       r0 , [r0+2*r1]

+    movd      xmm2, [r0]

+    movd      xmm3, [r0+r1]

+    punpckldq xmm0, xmm2

+    punpckldq xmm1, xmm3

+

+    movd      xmm4, [r2]

+    movd      xmm5, [r2+r3]

+    lea       r2 , [r2+2*r3]

+    movd      xmm6, [r2]

+    movd      xmm7, [r2+r3]

+    punpckldq xmm4, xmm6

+    punpckldq xmm5, xmm7

+

+    pxor      xmm6, xmm6

+    punpcklbw xmm0, xmm6

+    punpcklbw xmm1, xmm6

+    punpcklbw xmm4, xmm6

+    punpcklbw xmm5, xmm6

+

+    psubw     xmm0, xmm4

+    psubw     xmm1, xmm5

+

+    movdqa    xmm2, xmm0

+    paddw     xmm0, xmm1

+    psubw     xmm2, xmm1

+    SSE2_XSawp qdq, xmm0, xmm2, xmm3

+

+    movdqa     xmm4, xmm0

+    paddw      xmm0, xmm3

+    psubw      xmm4, xmm3

+

+    movdqa         xmm2, xmm0

+    punpcklwd      xmm0, xmm4

+    punpckhwd      xmm4, xmm2

+

+	SSE2_XSawp     dq,  xmm0, xmm4, xmm3

+	SSE2_XSawp     qdq, xmm0, xmm3, xmm5

+

+    movdqa         xmm7, xmm0

+    paddw          xmm0, xmm5

+    psubw          xmm7, xmm5

+

+	SSE2_XSawp     qdq,  xmm0, xmm7, xmm1

+

+    movdqa         xmm2, xmm0

+    paddw          xmm0, xmm1

+    psubw          xmm2, xmm1

+

+    WELS_AbsW  xmm0, xmm3

+    paddusw        xmm6, xmm0

+	WELS_AbsW  xmm2, xmm4

+    paddusw        xmm6, xmm2

+    SSE2_SumWHorizon1  xmm6, xmm4

+	movd           retrd,  xmm6

+    and            retrd,  0xffff

+    shr            retrd,  1

+	LOAD_4_PARA_POP

+	ret

+

+ ;***********************************************************************

+ ;

+ ;int32_t WelsSampleSatd8x8_sse2( uint8_t *, int32_t, uint8_t *, int32_t, );

+ ;

+ ;***********************************************************************

+ WELS_EXTERN WelsSampleSatd8x8_sse2

+align 16

+ WelsSampleSatd8x8_sse2:

+	 ;push   ebx

+	 ;mov    eax,    [esp+8]

+	 ;mov    ebx,    [esp+12]

+	 ;mov    ecx,    [esp+16]

+	 ;mov    edx,    [esp+20]

+

+	%assign  push_num 0

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	pxor   xmm6,   xmm6

+    pxor   xmm7,   xmm7

+    SSE2_GetSatd8x8

+    psrlw   xmm6,  1

+	SSE2_SumWHorizon   xmm6,xmm4,xmm7

+	movd    retrd,   xmm6

+	LOAD_4_PARA_POP

+	ret

+

+ ;***********************************************************************

+ ;

+ ;int32_t WelsSampleSatd8x16_sse2( uint8_t *, int32_t, uint8_t *, int32_t, );

+ ;

+ ;***********************************************************************

+ WELS_EXTERN WelsSampleSatd8x16_sse2

+align 16

+ WelsSampleSatd8x16_sse2:

+	 ;push   ebx

+	 ;mov    eax,    [esp+8]

+	 ;mov    ebx,    [esp+12]

+	 ;mov    ecx,    [esp+16]

+	 ;mov    edx,    [esp+20]

+

+	 %assign  push_num 0

+	 LOAD_4_PARA

+	 SIGN_EXTENTION r1, r1d

+	 SIGN_EXTENTION r3, r3d

+	 pxor   xmm6,   xmm6

+     pxor   xmm7,   xmm7

+

+	 SSE2_GetSatd8x8

+     lea    r0,    [r0+2*r1]

+     lea    r2,    [r2+2*r3]

+	 SSE2_GetSatd8x8

+

+	 psrlw   xmm6,  1

+	 SSE2_SumWHorizon   xmm6,xmm4,xmm7

+	 movd    retrd,   xmm6

+	 LOAD_4_PARA_POP

+	 ret

+

+;***********************************************************************

+;

+;int32_t WelsSampleSatd16x8_sse2( uint8_t *, int32_t, uint8_t *, int32_t, );

+;

+;***********************************************************************

+WELS_EXTERN WelsSampleSatd16x8_sse2

+align 16

+WelsSampleSatd16x8_sse2:

+	;push   ebx

+	;mov    eax,    [esp+8]

+	;mov    ebx,    [esp+12]

+	;mov    ecx,    [esp+16]

+	;mov    edx,    [esp+20]

+

+	%assign  push_num 0

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	push r0

+	push r2

+	pxor   xmm6,   xmm6

+    pxor   xmm7,   xmm7

+

+	SSE2_GetSatd8x8

+

+	pop r2

+	pop r0

+	;mov    eax,    [esp+8]

+    ;mov    ecx,    [esp+16]

+    add    r0,    8

+    add    r2,    8

+	SSE2_GetSatd8x8

+

+	psrlw   xmm6,  1

+	SSE2_SumWHorizon   xmm6,xmm4,xmm7

+	movd    retrd,   xmm6

+	LOAD_4_PARA_POP

+	ret

+

+;***********************************************************************

+;

+;int32_t WelsSampleSatd16x16_sse2( uint8_t *, int32_t, uint8_t *, int32_t, );

+;

+;***********************************************************************

+WELS_EXTERN WelsSampleSatd16x16_sse2

+align 16

+WelsSampleSatd16x16_sse2:

+	;push   ebx

+	;mov    eax,    [esp+8]

+	;mov    ebx,    [esp+12]

+	;mov    ecx,    [esp+16]

+	;mov    edx,    [esp+20]

+

+	%assign  push_num 0

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	push r0

+	push r2

+	pxor   xmm6,   xmm6

+    pxor   xmm7,   xmm7

+

+	SSE2_GetSatd8x8

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	SSE2_GetSatd8x8

+

+	pop r2

+	pop r0

+	;mov    eax,    [esp+8]

+	;mov    ecx,    [esp+16]

+	add    r0,    8

+	add    r2,    8

+

+	SSE2_GetSatd8x8

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	SSE2_GetSatd8x8

+

+ ; each column sum of SATD is necessarily even, so we don't lose any precision by shifting first.

+    psrlw   xmm6,  1

+	SSE2_SumWHorizon   xmm6,xmm4,xmm7

+	movd    retrd,   xmm6

+	LOAD_4_PARA_POP

+	ret

+

+;***********************************************************************

+;

+;Pixel_satd_wxh_sse2 END

+;

+;***********************************************************************

+

+;***********************************************************************

+;

+;Pixel_satd_intra_sse2 BEGIN

+;

+;***********************************************************************

+

+%macro SSE41_I16x16Get8WSumSub 3 ;xmm5 HSumSubDB1, xmm6 HSumSubDW1, xmm7 PDW1 : in %1, pOut %1, %3

+	pmaddubsw    %1, xmm5

+	movdqa       %2, %1

+	pmaddwd      %1, xmm7

+	pmaddwd      %2, xmm6

+	movdqa       %3, %1

+	punpckldq    %1, %2

+	punpckhdq    %2, %3

+	movdqa       %3, %1

+	punpcklqdq   %1, %2

+	punpckhqdq   %3, %2

+	paddd        xmm4, %1 ;for dc

+	paddd        xmm4, %3 ;for dc

+	packssdw     %1, %3

+	psllw        %1, 2

+%endmacro

+%macro SSE41_ChromaGet8WSumSub 4 ;xmm5 HSumSubDB1, xmm6 HSumSubDW1, xmm7 PDW1 : in %1, pOut %1, %3 : %4 tempsse2

+	pmaddubsw    %1, xmm5

+	movdqa       %2, %1

+	pmaddwd      %1, xmm7

+	pmaddwd      %2, xmm6

+	movdqa       %3, %1

+	punpckldq    %1, %2

+	punpckhdq    %2, %3

+	movdqa       %3, %1

+	punpcklqdq   %1, %2

+	punpckhqdq   %3, %2

+;    paddd        xmm4, %1 ;for dc

+;	 paddd        xmm4, %3 ;for dc

+	movdqa       %4, %1

+	punpcklqdq   %4, %3

+	packssdw     %1, %3

+	psllw        %1, 2

+%endmacro

+

+%macro SSE41_GetX38x4SatdDec 0

+	pxor        xmm7,   xmm7

+	movq        xmm0,   [eax]

+	movq        xmm1,   [eax+ebx]

+	lea         eax,    [eax+2*ebx]

+	movq        xmm2,   [eax]

+	movq        xmm3,   [eax+ebx]

+	lea         eax,    [eax+2*ebx]

+	punpcklbw   xmm0,   xmm7

+	punpcklbw   xmm1,   xmm7

+	punpcklbw   xmm2,   xmm7

+	punpcklbw   xmm3,   xmm7

+	SSE2_HDMTwo4x4       xmm0,xmm1,xmm2,xmm3,xmm7

+	SSE2_TransTwo4x4W     xmm3,xmm1,xmm0,xmm2,xmm7

+	SSE2_HDMTwo4x4       xmm3,xmm1,xmm2,xmm7,xmm0 ;pOut xmm7,xmm1,xmm3,xmm2

+	;doesn't need another transpose

+%endmacro

+%macro SSE41_GetX38x4SatdV 2

+	pxor        xmm0,   xmm0

+	pinsrw      xmm0,   word[esi+%2],   0

+	pinsrw      xmm0,   word[esi+%2+8], 4

+	psubsw      xmm0,   xmm7

+	pabsw       xmm0,   xmm0

+	paddw       xmm4,   xmm0

+	pxor        xmm0,   xmm0

+	pinsrw      xmm0,   word[esi+%2+2],  0

+	pinsrw      xmm0,   word[esi+%2+10], 4

+	psubsw      xmm0,   xmm1

+	pabsw       xmm0,   xmm0

+	paddw       xmm4,   xmm0

+	pxor        xmm0,   xmm0

+	pinsrw      xmm0,   word[esi+%2+4],  0

+	pinsrw      xmm0,   word[esi+%2+12], 4

+	psubsw      xmm0,   xmm3

+	pabsw       xmm0,   xmm0

+	paddw       xmm4,   xmm0

+	pxor        xmm0,   xmm0

+	pinsrw      xmm0,   word[esi+%2+6],  0

+	pinsrw      xmm0,   word[esi+%2+14], 4

+	psubsw      xmm0,   xmm2

+	pabsw       xmm0,   xmm0

+	paddw       xmm4,   xmm0

+%endmacro

+%macro SSE41_GetX38x4SatdH  3

+	movq        xmm0,   [esi+%3+8*%1]

+	punpcklqdq  xmm0,   xmm0

+	psubsw      xmm0,   xmm7

+	pabsw       xmm0,   xmm0

+	paddw       xmm5,   xmm0

+	pabsw       xmm1,   xmm1

+	pabsw       xmm2,   xmm2

+	pabsw       xmm3,   xmm3

+	paddw       xmm2,   xmm1;for DC

+	paddw       xmm2,   xmm3;for DC

+	paddw       xmm5,   xmm2

+%endmacro

+%macro SSE41_I16X16GetX38x4SatdDC 0

+	pxor        xmm0,   xmm0

+	movq2dq     xmm0,   mm4

+	punpcklqdq  xmm0,   xmm0

+	psubsw      xmm0,   xmm7

+	pabsw       xmm0,   xmm0

+	paddw       xmm6,   xmm0

+	paddw       xmm6,   xmm2

+%endmacro

+%macro SSE41_ChromaGetX38x4SatdDC 1

+	shl         %1,     4

+	movdqa      xmm0,   [esi+32+%1]

+	psubsw      xmm0,   xmm7

+	pabsw       xmm0,   xmm0

+	paddw       xmm6,   xmm0

+	paddw       xmm6,   xmm2

+%endmacro

+%macro SSE41_I16x16GetX38x4Satd 2

+	SSE41_GetX38x4SatdDec

+	SSE41_GetX38x4SatdV   %1, %2

+	SSE41_GetX38x4SatdH   %1, %2, 32

+	SSE41_I16X16GetX38x4SatdDC

+%endmacro

+%macro SSE41_ChromaGetX38x4Satd 2

+	SSE41_GetX38x4SatdDec

+	SSE41_GetX38x4SatdV   %1, %2

+	SSE41_GetX38x4SatdH   %1, %2, 16

+	SSE41_ChromaGetX38x4SatdDC %1

+%endmacro

+%macro SSE41_HSum8W 3

+	pmaddwd     %1, %2

+	movhlps     %3, %1

+	paddd       %1, %3

+	pshuflw     %3, %1,0Eh

+	paddd       %1, %3

+%endmacro

+

+

+%ifdef X86_32

+WELS_EXTERN WelsIntra16x16Combined3Satd_sse41

+WelsIntra16x16Combined3Satd_sse41:

+	push   ebx

+	push   esi

+	push   edi

+	mov    ecx,    [esp+16]

+	mov    edx,    [esp+20]

+	mov    eax,    [esp+24]

+	mov    ebx,    [esp+28]

+	mov    esi,    [esp+40] ;temp_satd

+	pxor        xmm4,   xmm4

+	movdqa      xmm5,   [HSumSubDB1]

+	movdqa      xmm6,   [HSumSubDW1]

+	movdqa      xmm7,   [PDW1]

+	sub         ecx,    edx

+	movdqu 		xmm0,   [ecx]

+	movhlps		xmm1,   xmm0

+	punpcklqdq  xmm0,   xmm0

+	punpcklqdq  xmm1,   xmm1

+	SSE41_I16x16Get8WSumSub xmm0, xmm2, xmm3

+	SSE41_I16x16Get8WSumSub xmm1, xmm2, xmm3

+	movdqa      [esi],  xmm0 ;V

+	movdqa      [esi+16], xmm1

+	add         ecx,    edx

+	pinsrb      xmm0,   byte[ecx-1], 0

+	pinsrb      xmm0,   byte[ecx+edx-1], 1

+	lea         ecx,    [ecx+2*edx]

+	pinsrb      xmm0,   byte[ecx-1],     2

+	pinsrb      xmm0,   byte[ecx+edx-1], 3

+	lea         ecx,    [ecx+2*edx]

+	pinsrb      xmm0,   byte[ecx-1],     4

+	pinsrb      xmm0,   byte[ecx+edx-1], 5

+	lea         ecx,    [ecx+2*edx]

+	pinsrb      xmm0,   byte[ecx-1],     6

+	pinsrb      xmm0,   byte[ecx+edx-1], 7

+	lea         ecx,    [ecx+2*edx]

+	pinsrb      xmm0,   byte[ecx-1],     8

+	pinsrb      xmm0,   byte[ecx+edx-1], 9

+	lea         ecx,    [ecx+2*edx]

+	pinsrb      xmm0,   byte[ecx-1],     10

+	pinsrb      xmm0,   byte[ecx+edx-1], 11

+	lea         ecx,    [ecx+2*edx]

+	pinsrb      xmm0,   byte[ecx-1],     12

+	pinsrb      xmm0,   byte[ecx+edx-1], 13

+	lea         ecx,    [ecx+2*edx]

+	pinsrb      xmm0,   byte[ecx-1],     14

+	pinsrb      xmm0,   byte[ecx+edx-1], 15

+	movhlps		xmm1,   xmm0

+	punpcklqdq  xmm0,   xmm0

+	punpcklqdq  xmm1,   xmm1

+	SSE41_I16x16Get8WSumSub xmm0, xmm2, xmm3

+	SSE41_I16x16Get8WSumSub xmm1, xmm2, xmm3

+	movdqa      [esi+32], xmm0 ;H

+	movdqa      [esi+48], xmm1

+	movd        ecx,    xmm4 ;dc

+	add         ecx,    16   ;(sum+16)

+	shr         ecx,    5    ;((sum+16)>>5)

+	shl         ecx,    4    ;

+	movd        mm4,    ecx  ; mm4 copy DC

+	pxor        xmm4,   xmm4 ;V

+	pxor        xmm5,   xmm5 ;H

+	pxor        xmm6,   xmm6 ;DC

+	mov         ecx,    0

+	mov         edi,    0

+.loop16x16_get_satd:

+.loopStart1:

+	SSE41_I16x16GetX38x4Satd ecx, edi

+	inc          ecx

+	cmp         ecx, 4

+	jl          .loopStart1

+	cmp         edi, 16

+	je          .loop16x16_get_satd_end

+	mov         eax, [esp+24]

+	add         eax, 8

+	mov         ecx, 0

+	add         edi, 16

+	jmp         .loop16x16_get_satd

+ .loop16x16_get_satd_end:

+	MMX_DW_1_2REG    xmm0, xmm1

+	psrlw       xmm4, 1 ;/2

+	psrlw       xmm5, 1 ;/2

+	psrlw       xmm6, 1 ;/2

+	SSE41_HSum8W     xmm4, xmm0, xmm1

+	SSE41_HSum8W     xmm5, xmm0, xmm1

+	SSE41_HSum8W     xmm6, xmm0, xmm1

+

+	; comparing order: DC H V

+	movd      ebx, xmm6 ;DC

+	movd      edi, xmm5 ;H

+	movd      ecx, xmm4 ;V

+	mov      edx, [esp+36]

+	shl       edx, 1

+	add       edi, edx

+	add       ebx, edx

+	mov       edx, [esp+32]

+	cmp       ebx, edi

+	jge near   not_dc_16x16

+	cmp        ebx, ecx

+	jge near   not_dc_h_16x16

+

+	; for DC mode

+	mov       dword[edx], 2;I16_PRED_DC

+	mov       eax, ebx

+	jmp near return_satd_intra_16x16_x3

+not_dc_16x16:

+	; for H mode

+	cmp       edi, ecx

+	jge near   not_dc_h_16x16

+	mov       dword[edx], 1;I16_PRED_H

+	mov       eax, edi

+	jmp near return_satd_intra_16x16_x3

+not_dc_h_16x16:

+	; for V mode

+	mov       dword[edx], 0;I16_PRED_V

+	mov       eax, ecx

+return_satd_intra_16x16_x3:

+	WELSEMMS

+	pop         edi

+	pop         esi

+	pop         ebx

+ret

+

+%macro SSE41_ChromaGetX38x8Satd 0

+	movdqa      xmm5,   [HSumSubDB1]

+	movdqa      xmm6,   [HSumSubDW1]

+	movdqa      xmm7,   [PDW1]

+	sub         ecx,    edx

+	movq 		xmm0,   [ecx]

+	punpcklqdq  xmm0,   xmm0

+	SSE41_ChromaGet8WSumSub xmm0, xmm2, xmm3, xmm4

+	movdqa      [esi],  xmm0 ;V

+	add         ecx,    edx

+	pinsrb      xmm0,   byte[ecx-1], 0

+	pinsrb      xmm0,   byte[ecx+edx-1], 1

+	lea         ecx,    [ecx+2*edx]

+	pinsrb      xmm0,   byte[ecx-1],     2

+	pinsrb      xmm0,   byte[ecx+edx-1], 3

+	lea         ecx,    [ecx+2*edx]

+	pinsrb      xmm0,   byte[ecx-1],     4

+	pinsrb      xmm0,   byte[ecx+edx-1], 5

+	lea         ecx,    [ecx+2*edx]

+	pinsrb      xmm0,   byte[ecx-1],     6

+	pinsrb      xmm0,   byte[ecx+edx-1], 7

+	punpcklqdq  xmm0,   xmm0

+	SSE41_ChromaGet8WSumSub xmm0, xmm2, xmm3, xmm1

+	movdqa      [esi+16], xmm0 ;H

+;(sum+2)>>2

+	movdqa      xmm6,   [PDQ2]

+	movdqa      xmm5,   xmm4

+	punpckhqdq  xmm5,   xmm1

+	paddd       xmm5,   xmm6

+	psrld       xmm5,   2

+;(sum1+sum2+4)>>3

+	paddd       xmm6,   xmm6

+	paddd       xmm4,   xmm1

+	paddd       xmm4,   xmm6

+	psrld       xmm4,   3

+;satd *16

+	pslld       xmm5,   4

+	pslld       xmm4,   4

+;temp satd

+	movdqa      xmm6,   xmm4

+	punpcklqdq  xmm4,   xmm5

+	psllq       xmm4,   32

+	psrlq       xmm4,   32

+	movdqa      [esi+32], xmm4

+	punpckhqdq  xmm5,   xmm6

+	psllq       xmm5,   32

+	psrlq       xmm5,   32

+	movdqa      [esi+48], xmm5

+

+	pxor        xmm4,   xmm4 ;V

+	pxor        xmm5,   xmm5 ;H

+	pxor        xmm6,   xmm6 ;DC

+	mov         ecx,    0

+loop_chroma_satdx3_cb_cr:

+	SSE41_ChromaGetX38x4Satd ecx, 0

+	inc             ecx

+	cmp             ecx, 2

+	jl              loop_chroma_satdx3_cb_cr

+%endmacro

+

+%macro SSEReg2MMX 3

+	movdq2q     %2, %1

+	movhlps     %1, %1

+	movdq2q     %3, %1

+%endmacro

+%macro MMXReg2SSE 4

+	movq2dq     %1, %3

+	movq2dq     %2, %4

+	punpcklqdq  %1, %2

+%endmacro

+;for reduce the code size of WelsIntraChroma8x8Combined3Satd_sse41

+

+WELS_EXTERN WelsIntraChroma8x8Combined3Satd_sse41

+WelsIntraChroma8x8Combined3Satd_sse41:

+	push   ebx

+	push   esi

+	push   edi

+	mov    ecx,    [esp+16]

+	mov    edx,    [esp+20]

+	mov    eax,    [esp+24]

+	mov    ebx,    [esp+28]

+	mov    esi,    [esp+40] ;temp_satd

+	xor    edi,    edi

+loop_chroma_satdx3:

+	SSE41_ChromaGetX38x8Satd

+	cmp             edi, 1

+	je              loop_chroma_satdx3end

+	inc             edi

+	SSEReg2MMX  xmm4, mm0,mm1

+	SSEReg2MMX  xmm5, mm2,mm3

+	SSEReg2MMX  xmm6, mm5,mm6

+	mov         ecx,  [esp+44]

+	mov         eax,  [esp+48]

+	jmp         loop_chroma_satdx3

+loop_chroma_satdx3end:

+	MMXReg2SSE  xmm0, xmm3, mm0, mm1

+	MMXReg2SSE  xmm1, xmm3, mm2, mm3

+	MMXReg2SSE  xmm2, xmm3, mm5, mm6

+

+	paddw       xmm4, xmm0

+	paddw       xmm5, xmm1

+	paddw       xmm6, xmm2

+

+	MMX_DW_1_2REG    xmm0, xmm1

+	psrlw       xmm4, 1 ;/2

+	psrlw       xmm5, 1 ;/2

+	psrlw       xmm6, 1 ;/2

+	SSE41_HSum8W     xmm4, xmm0, xmm1

+	SSE41_HSum8W     xmm5, xmm0, xmm1

+	SSE41_HSum8W     xmm6, xmm0, xmm1

+	; comparing order: DC H V

+	movd      ebx, xmm6 ;DC

+	movd      edi, xmm5 ;H

+	movd      ecx, xmm4 ;V

+	mov       edx, [esp+36]

+	shl       edx, 1

+	add       edi, edx

+	add       ecx, edx

+	mov       edx, [esp+32]

+	cmp       ebx, edi

+	jge near   not_dc_8x8

+	cmp        ebx, ecx

+	jge near   not_dc_h_8x8

+

+	; for DC mode

+	mov       dword[edx], 0;I8_PRED_DC

+	mov       eax, ebx

+	jmp near return_satd_intra_8x8_x3

+not_dc_8x8:

+	; for H mode

+	cmp       edi, ecx

+	jge near   not_dc_h_8x8

+	mov       dword[edx], 1;I8_PRED_H

+	mov       eax, edi

+	jmp near return_satd_intra_8x8_x3

+not_dc_h_8x8:

+	; for V mode

+	mov       dword[edx], 2;I8_PRED_V

+	mov       eax, ecx

+return_satd_intra_8x8_x3:

+	WELSEMMS

+	pop         edi

+	pop         esi

+	pop         ebx

+ret

+

+

+;***********************************************************************

+;

+;Pixel_satd_intra_sse2 END

+;

+;***********************************************************************

+%macro SSSE3_Get16BSadHVDC 2

+  movd        xmm6,%1

+  pshufb      xmm6,xmm1

+  movdqa      %1,  xmm6

+  movdqa      xmm0,%2

+  psadbw      xmm0,xmm7

+  paddw       xmm4,xmm0

+  movdqa      xmm0,%2

+  psadbw      xmm0,xmm5

+  paddw       xmm2,xmm0

+  psadbw      xmm6,%2

+  paddw       xmm3,xmm6

+%endmacro

+%macro WelsAddDCValue 4

+    movzx   %2, byte %1

+    mov    %3, %2

+    add     %4, %2

+%endmacro

+

+;***********************************************************************

+;

+;Pixel_sad_intra_ssse3 BEGIN

+;

+;***********************************************************************

+WELS_EXTERN WelsIntra16x16Combined3Sad_ssse3

+WelsIntra16x16Combined3Sad_ssse3:

+	push   ebx

+	push   esi

+	push   edi

+	mov    ecx,    [esp+16]

+	mov    edx,    [esp+20]

+	mov    edi,    [esp+40] ;temp_sad

+	sub    ecx,    edx

+    movdqa      xmm5,[ecx]

+    pxor        xmm0,xmm0

+    psadbw      xmm0,xmm5

+    movhlps     xmm1,xmm0

+    paddw       xmm0,xmm1

+    movd        eax,xmm0

+

+    add         ecx,edx

+    lea         ebx, [edx+2*edx]

+    WelsAddDCValue [ecx-1      ], esi, [edi   ], eax

+    WelsAddDCValue [ecx-1+edx  ], esi, [edi+16], eax

+    WelsAddDCValue [ecx-1+edx*2], esi, [edi+32], eax

+    WelsAddDCValue [ecx-1+ebx  ], esi, [edi+48], eax

+    lea         ecx, [ecx+4*edx]

+    add         edi, 64

+    WelsAddDCValue [ecx-1      ], esi, [edi   ], eax

+    WelsAddDCValue [ecx-1+edx  ], esi, [edi+16], eax

+    WelsAddDCValue [ecx-1+edx*2], esi, [edi+32], eax

+    WelsAddDCValue [ecx-1+ebx  ], esi, [edi+48], eax

+    lea         ecx, [ecx+4*edx]

+    add         edi, 64

+    WelsAddDCValue [ecx-1      ], esi, [edi   ], eax

+    WelsAddDCValue [ecx-1+edx  ], esi, [edi+16], eax

+    WelsAddDCValue [ecx-1+edx*2], esi, [edi+32], eax

+    WelsAddDCValue [ecx-1+ebx  ], esi, [edi+48], eax

+    lea         ecx, [ecx+4*edx]

+    add         edi, 64

+    WelsAddDCValue [ecx-1      ], esi, [edi   ], eax

+    WelsAddDCValue [ecx-1+edx  ], esi, [edi+16], eax

+    WelsAddDCValue [ecx-1+edx*2], esi, [edi+32], eax

+    WelsAddDCValue [ecx-1+ebx  ], esi, [edi+48], eax

+    sub        edi, 192

+    add         eax,10h

+    shr         eax,5

+    movd        xmm7,eax

+    pxor        xmm1,xmm1

+    pshufb      xmm7,xmm1

+    pxor        xmm4,xmm4

+    pxor        xmm3,xmm3

+    pxor        xmm2,xmm2

+;sad begin

+	mov    eax,    [esp+24]

+	mov    ebx,    [esp+28]

+    lea         esi, [ebx+2*ebx]

+    SSSE3_Get16BSadHVDC [edi], [eax]

+    SSSE3_Get16BSadHVDC [edi+16], [eax+ebx]

+    SSSE3_Get16BSadHVDC [edi+32], [eax+2*ebx]

+    SSSE3_Get16BSadHVDC [edi+48], [eax+esi]

+    add         edi, 64

+    lea         eax, [eax+4*ebx]

+    SSSE3_Get16BSadHVDC [edi], [eax]

+    SSSE3_Get16BSadHVDC [edi+16], [eax+ebx]

+    SSSE3_Get16BSadHVDC [edi+32], [eax+2*ebx]

+    SSSE3_Get16BSadHVDC [edi+48], [eax+esi]

+    add         edi, 64

+    lea         eax, [eax+4*ebx]

+    SSSE3_Get16BSadHVDC [edi], [eax]

+    SSSE3_Get16BSadHVDC [edi+16], [eax+ebx]

+    SSSE3_Get16BSadHVDC [edi+32], [eax+2*ebx]

+    SSSE3_Get16BSadHVDC [edi+48], [eax+esi]

+    add         edi, 64

+    lea         eax, [eax+4*ebx]

+    SSSE3_Get16BSadHVDC [edi], [eax]

+    SSSE3_Get16BSadHVDC [edi+16], [eax+ebx]

+    SSSE3_Get16BSadHVDC [edi+32], [eax+2*ebx]

+    SSSE3_Get16BSadHVDC [edi+48], [eax+esi]

+

+    pslldq      xmm3,4

+    por         xmm3,xmm2

+    movhlps     xmm1,xmm3

+    paddw       xmm3,xmm1

+    movhlps     xmm0,xmm4

+    paddw       xmm4,xmm0

+; comparing order: DC H V

+	movd        ebx, xmm4 ;DC

+	movd        ecx, xmm3 ;V

+	psrldq      xmm3, 4

+	movd        esi, xmm3 ;H

+	mov         eax, [esp+36] ;lamda

+	shl         eax, 1

+	add         esi, eax

+	add         ebx, eax

+	mov         edx, [esp+32]

+	cmp         ebx, esi

+	jge near   not_dc_16x16_sad

+	cmp        ebx, ecx

+	jge near   not_dc_h_16x16_sad

+	; for DC mode

+	mov       dword[edx], 2;I16_PRED_DC

+	mov       eax, ebx

+    sub        edi, 192

+%assign x 0

+%rep 16

+    movdqa    [edi+16*x], xmm7

+%assign x x+1

+%endrep

+	jmp near return_sad_intra_16x16_x3

+not_dc_16x16_sad:

+	; for H mode

+	cmp       esi, ecx

+	jge near   not_dc_h_16x16_sad

+	mov       dword[edx], 1;I16_PRED_H

+	mov       eax, esi

+	jmp near return_sad_intra_16x16_x3

+not_dc_h_16x16_sad:

+	; for V mode

+	mov       dword[edx], 0;I16_PRED_V

+	mov       eax, ecx

+    sub       edi, 192

+%assign x 0

+%rep 16

+    movdqa    [edi+16*x], xmm5

+%assign x x+1

+%endrep

+return_sad_intra_16x16_x3:

+	pop    edi

+	pop    esi

+	pop    ebx

+	ret

+%endif

+;***********************************************************************

+;

+;Pixel_sad_intra_ssse3 END

+;

+;***********************************************************************

+;***********************************************************************

+;

+;Pixel_satd_wxh_sse41 BEGIN

+;

+;***********************************************************************

+

+;SSE4.1

+%macro SSE41_GetSatd8x4 0

+	movq             xmm0, [r0]

+	punpcklqdq       xmm0, xmm0

+	pmaddubsw        xmm0, xmm7

+	movq             xmm1, [r0+r1]

+	punpcklqdq       xmm1, xmm1

+	pmaddubsw        xmm1, xmm7

+	movq             xmm2, [r2]

+	punpcklqdq       xmm2, xmm2

+	pmaddubsw        xmm2, xmm7

+	movq             xmm3, [r2+r3]

+	punpcklqdq       xmm3, xmm3

+	pmaddubsw        xmm3, xmm7

+	psubsw           xmm0, xmm2

+	psubsw           xmm1, xmm3

+	movq             xmm2, [r0+2*r1]

+	punpcklqdq       xmm2, xmm2

+	pmaddubsw        xmm2, xmm7

+	movq             xmm3, [r0+r4]

+	punpcklqdq       xmm3, xmm3

+	pmaddubsw        xmm3, xmm7

+	movq             xmm4, [r2+2*r3]

+	punpcklqdq       xmm4, xmm4

+	pmaddubsw        xmm4, xmm7

+	movq             xmm5, [r2+r5]

+	punpcklqdq       xmm5, xmm5

+	pmaddubsw        xmm5, xmm7

+	psubsw           xmm2, xmm4

+	psubsw           xmm3, xmm5

+	SSE2_HDMTwo4x4   xmm0, xmm1, xmm2, xmm3, xmm4

+	pabsw            xmm0, xmm0

+	pabsw            xmm2, xmm2

+	pabsw            xmm1, xmm1

+	pabsw            xmm3, xmm3

+	movdqa           xmm4, xmm3

+	pblendw          xmm3, xmm1, 0xAA

+	pslld            xmm1, 16

+	psrld            xmm4, 16

+	por              xmm1, xmm4

+	pmaxuw           xmm1, xmm3

+	paddw            xmm6, xmm1

+	movdqa           xmm4, xmm0

+	pblendw          xmm0, xmm2, 0xAA

+	pslld            xmm2, 16

+	psrld            xmm4, 16

+	por              xmm2, xmm4

+	pmaxuw           xmm0, xmm2

+	paddw            xmm6, xmm0

+%endmacro

+

+%macro SSSE3_SumWHorizon 4 ;eax, srcSSE, tempSSE, tempSSE

+	MMX_DW_1_2REG    %3, %4

+	pmaddwd     %2, %3

+	movhlps     %4, %2

+	paddd       %2, %4

+	pshuflw     %4, %2,0Eh

+	paddd       %2, %4

+	movd		%1, %2

+%endmacro

+;***********************************************************************

+;

+;int32_t WelsSampleSatd4x4_sse41( uint8_t *, int32_t, uint8_t *, int32_t );

+;

+;***********************************************************************

+WELS_EXTERN WelsSampleSatd4x4_sse41

+WelsSampleSatd4x4_sse41:

+	;push        ebx

+	;mov         eax,[esp+8]

+	;mov         ebx,[esp+12]

+	;mov         ecx,[esp+16]

+	;mov         edx,[esp+20]

+

+	%assign  push_num 0

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	movdqa      xmm4,[HSwapSumSubDB1]

+	movd        xmm2,[r2]

+	movd        xmm5,[r2+r3]

+	shufps      xmm2,xmm5,0

+	movd        xmm3,[r2+r3*2]

+	lea         r2, [r3*2+r2]

+	movd        xmm5,[r2+r3]

+	shufps      xmm3,xmm5,0

+	movd        xmm0,[r0]

+	movd        xmm5,[r0+r1]

+	shufps      xmm0,xmm5,0

+	movd        xmm1,[r0+r1*2]

+	lea         r0, [r1*2+r0]

+	movd        xmm5,[r0+r1]

+	shufps      xmm1,xmm5,0

+	pmaddubsw   xmm0,xmm4

+	pmaddubsw   xmm1,xmm4

+	pmaddubsw   xmm2,xmm4

+	pmaddubsw   xmm3,xmm4

+	psubw       xmm0,xmm2

+	psubw       xmm1,xmm3

+	movdqa      xmm2,xmm0

+	paddw       xmm0,xmm1

+	psubw       xmm1,xmm2

+	movdqa      xmm2,xmm0

+	punpcklqdq  xmm0,xmm1

+	punpckhqdq  xmm2,xmm1

+	movdqa      xmm1,xmm0

+	paddw       xmm0,xmm2

+	psubw       xmm2,xmm1

+	movdqa      xmm1,xmm0

+	pblendw     xmm0,xmm2,0AAh

+	pslld       xmm2,16

+	psrld       xmm1,16

+	por         xmm2,xmm1

+	pabsw       xmm0,xmm0

+	pabsw       xmm2,xmm2

+	pmaxsw      xmm0,xmm2

+	SSSE3_SumWHorizon retrd, xmm0, xmm5, xmm7

+	LOAD_4_PARA_POP

+	ret

+

+;***********************************************************************

+;

+;int32_t WelsSampleSatd8x8_sse41( uint8_t *, int32_t, uint8_t *, int32_t, );

+;

+;***********************************************************************

+WELS_EXTERN WelsSampleSatd8x8_sse41

+align 16

+WelsSampleSatd8x8_sse41:

+	;push   ebx

+	;push   esi

+	;push   edi

+	;mov    eax,    [esp+16]

+	;mov    ebx,    [esp+20]

+	;mov    ecx,    [esp+24]

+	;mov    edx,    [esp+28]

+%ifdef X86_32

+	push  r4

+	push  r5

+%endif

+	%assign  push_num 2

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	movdqa      xmm7, [HSumSubDB1]

+	lea         r4,  [r1+r1*2]

+	lea         r5,  [r3+r3*2]

+	pxor		xmm6, xmm6

+	SSE41_GetSatd8x4

+	lea			r0,	 [r0+4*r1]

+	lea			r2,  [r2+4*r3]

+	SSE41_GetSatd8x4

+	SSSE3_SumWHorizon retrd, xmm6, xmm5, xmm7

+	LOAD_4_PARA_POP

+%ifdef X86_32

+	pop  r5

+	pop  r4

+%endif

+	ret

+

+;***********************************************************************

+;

+;int32_t WelsSampleSatd8x16_sse41( uint8_t *, int32_t, uint8_t *, int32_t, );

+;

+;***********************************************************************

+WELS_EXTERN WelsSampleSatd8x16_sse41

+align 16

+WelsSampleSatd8x16_sse41:

+	;push   ebx

+	;push   esi

+	;push   edi

+	;push   ebp

+	;%define pushsize   16

+	;mov    eax,    [esp+pushsize+4]

+	;mov    ebx,    [esp+pushsize+8]

+	;mov    ecx,    [esp+pushsize+12]

+	;mov    edx,    [esp+pushsize+16]

+%ifdef X86_32

+	push  r4

+	push  r5

+	push  r6

+%endif

+	%assign  push_num 3

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	movdqa      xmm7, [HSumSubDB1]

+	lea         r4,  [r1+r1*2]

+	lea         r5,  [r3+r3*2]

+	pxor        xmm6, xmm6

+	mov         r6,    0

+loop_get_satd_8x16:

+	SSE41_GetSatd8x4

+	lea			r0,  [r0+4*r1]

+	lea			r2,  [r2+4*r3]

+	inc         r6

+	cmp         r6,  4

+	jl          loop_get_satd_8x16

+	SSSE3_SumWHorizon retrd, xmm6, xmm5, xmm7

+	LOAD_4_PARA_POP

+%ifdef X86_32

+	pop  r6

+	pop  r5

+	pop  r4

+%endif

+	ret

+

+;***********************************************************************

+;

+;int32_t WelsSampleSatd16x8_sse41( uint8_t *, int32_t, uint8_t *, int32_t, );

+;

+;***********************************************************************

+WELS_EXTERN WelsSampleSatd16x8_sse41

+align 16

+WelsSampleSatd16x8_sse41:

+	;push   ebx

+	;push   esi

+	;push   edi

+	;mov    eax,    [esp+16]

+	;mov    ebx,    [esp+20]

+	;mov    ecx,    [esp+24]

+	;mov    edx,    [esp+28]

+%ifdef X86_32

+	push  r4

+	push  r5

+%endif

+	%assign  push_num 2

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	push  r0

+	push  r2

+

+	movdqa      xmm7, [HSumSubDB1]

+	lea         r4,  [r1+r1*2]

+	lea         r5,  [r3+r3*2]

+	pxor		xmm6,   xmm6

+	SSE41_GetSatd8x4

+	lea			r0,  [r0+4*r1]

+	lea			r2,  [r2+4*r3]

+	SSE41_GetSatd8x4

+

+	pop  r2

+	pop  r0

+	;mov			eax,    [esp+16]

+	;mov			ecx,    [esp+24]

+	add			r0,    8

+	add			r2,    8

+	SSE41_GetSatd8x4

+	lea			r0,  [r0+4*r1]

+	lea			r2,  [r2+4*r3]

+	SSE41_GetSatd8x4

+	SSSE3_SumWHorizon retrd, xmm6, xmm5, xmm7

+	LOAD_4_PARA_POP

+%ifdef X86_32

+	pop  r5

+	pop  r4

+%endif

+	ret

+

+;***********************************************************************

+;

+;int32_t WelsSampleSatd16x16_sse41( uint8_t *, int32_t, uint8_t *, int32_t, );

+;

+;***********************************************************************

+

+WELS_EXTERN WelsSampleSatd16x16_sse41

+align 16

+WelsSampleSatd16x16_sse41:

+	;push   ebx

+	;push   esi

+	;push   edi

+	;push   ebp

+	;%define pushsize   16

+	;mov    eax,    [esp+pushsize+4]

+	;mov    ebx,    [esp+pushsize+8]

+	;mov    ecx,    [esp+pushsize+12]

+	;mov    edx,    [esp+pushsize+16]

+%ifdef X86_32

+	push  r4

+	push  r5

+	push  r6

+%endif

+	%assign  push_num 3

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+

+	push  r0

+	push  r2

+

+	movdqa      xmm7, [HSumSubDB1]

+	lea         r4,  [r1+r1*2]

+	lea         r5,  [r3+r3*2]

+	pxor		xmm6,   xmm6

+	mov         r6,    0

+loop_get_satd_16x16_left:

+	SSE41_GetSatd8x4

+	lea			r0,  [r0+4*r1]

+	lea			r2,  [r2+4*r3]

+	inc         r6

+	cmp         r6,  4

+	jl          loop_get_satd_16x16_left

+

+	pop  r2

+	pop  r0

+	;mov			eax,    [esp+pushsize+4]

+	;mov			ecx,    [esp+pushsize+12]

+	add			r0,    8

+	add			r2,    8

+	mov         r6,    0

+loop_get_satd_16x16_right:

+	SSE41_GetSatd8x4

+	lea			r0,  [r0+4*r1]

+	lea			r2,  [r2+4*r3]

+	inc         r6

+	cmp         r6,  4

+	jl          loop_get_satd_16x16_right

+	SSSE3_SumWHorizon retrd, xmm6, xmm5, xmm7

+	;%undef pushsize

+	LOAD_4_PARA_POP

+%ifdef X86_32

+	pop  r6

+	pop  r5

+	pop  r4

+%endif

+	ret

+

+;***********************************************************************

+;

+;Pixel_satd_wxh_sse41 END

+;

+;***********************************************************************

+

+;***********************************************************************

+;

+;Pixel_sad_wxh_sse2 BEGIN

+;

+;***********************************************************************

+

+%macro SSE2_GetSad2x16 0

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movdqu xmm1,   [r2]

+	MOVDQ  xmm2,   [r0];[eax] must aligned 16

+	psadbw xmm1,   xmm2

+	paddw  xmm0,   xmm1

+	movdqu xmm1,   [r2+r3]

+	MOVDQ  xmm2,   [r0+r1]

+	psadbw xmm1,   xmm2

+	paddw  xmm0,   xmm1

+%endmacro

+

+

+%macro SSE2_GetSad4x16 0

+	movdqu xmm0,   [r2]

+	MOVDQ  xmm2,   [r0]

+	psadbw xmm0,   xmm2

+	paddw  xmm7,   xmm0

+	movdqu xmm1,   [r2+r3]

+	MOVDQ  xmm2,   [r0+r1]

+	psadbw xmm1,   xmm2

+	paddw  xmm7,   xmm1

+	movdqu xmm1,   [r2+2*r3]

+	MOVDQ  xmm2,   [r0+2*r1];[eax] must aligned 16

+	psadbw xmm1,   xmm2

+	paddw  xmm7,   xmm1

+	movdqu xmm1,   [r2+r5]

+	MOVDQ  xmm2,   [r0+r4]

+	psadbw xmm1,   xmm2

+	paddw  xmm7,   xmm1

+%endmacro

+

+

+%macro SSE2_GetSad8x4 0

+	movq   xmm0,   [r0]

+	movq   xmm1,   [r0+r1]

+	lea    r0,     [r0+2*r1]

+	movhps xmm0,   [r0]

+	movhps xmm1,   [r0+r1]

+

+	movq   xmm2,   [r2]

+	movq   xmm3,   [r2+r3]

+	lea    r2,     [r2+2*r3]

+	movhps xmm2,   [r2]

+	movhps xmm3,   [r2+r3]

+	psadbw xmm0,   xmm2

+	psadbw xmm1,   xmm3

+	paddw  xmm6,   xmm0

+	paddw  xmm6,   xmm1

+%endmacro

+

+;***********************************************************************

+;

+;int32_t WelsSampleSad16x16_sse2( uint8_t *, int32_t, uint8_t *, int32_t, )

+;First parameter can align to 16 bytes,

+;In wels, the third parameter can't align to 16 bytes.

+;

+;***********************************************************************

+WELS_EXTERN WelsSampleSad16x16_sse2

+align 16

+WelsSampleSad16x16_sse2:

+	;push ebx

+	;push edi

+	;push esi

+	;%define _STACK_SIZE		12

+	;mov eax, [esp+_STACK_SIZE+4 ]

+	;mov	ebx, [esp+_STACK_SIZE+8 ]

+	;mov ecx, [esp+_STACK_SIZE+12]

+	;mov edx, [esp+_STACK_SIZE+16]

+%ifdef X86_32

+	push  r4

+	push  r5

+%endif

+

+	%assign  push_num 2

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	lea r4, [3*r1]

+	lea r5, [3*r3]

+

+	pxor   xmm7,   xmm7

+	SSE2_GetSad4x16

+	lea	   r0,  [r0+4*r1]

+	lea	   r2,  [r2+4*r3]

+	SSE2_GetSad4x16

+	lea	   r0,  [r0+4*r1]

+	lea	   r2,  [r2+4*r3]

+	SSE2_GetSad4x16

+	lea	   r0,  [r0+4*r1]

+	lea	   r2,  [r2+4*r3]

+	SSE2_GetSad4x16

+	movhlps xmm0, xmm7

+	paddw xmm0, xmm7

+	movd retrd, xmm0

+	LOAD_4_PARA_POP

+%ifdef X86_32

+	pop  r5

+	pop  r4

+%endif

+	ret

+

+;***********************************************************************

+;

+;int32_t WelsSampleSad16x8_sse2( uint8_t *, int32_t, uint8_t *, int32_t, )

+;First parameter can align to 16 bytes,

+;In wels, the third parameter can't align to 16 bytes.

+;

+;***********************************************************************

+WELS_EXTERN WelsSampleSad16x8_sse2

+align 16

+WelsSampleSad16x8_sse2:

+	;push   ebx

+	;mov    eax,    [esp+8]

+	;mov    ebx,    [esp+12]

+	;mov    ecx,    [esp+16]

+	;mov    edx,    [esp+20]

+

+	%assign  push_num 0

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	movdqu xmm0,   [r2]

+	MOVDQ  xmm2,   [r0]

+	psadbw xmm0,   xmm2

+	movdqu xmm1,   [r2+r3]

+	MOVDQ  xmm2,   [r0+r1]

+	psadbw xmm1,   xmm2

+	paddw  xmm0,   xmm1

+

+	SSE2_GetSad2x16

+	SSE2_GetSad2x16

+	SSE2_GetSad2x16

+

+	movhlps     xmm1, xmm0

+	paddw       xmm0, xmm1

+	movd        retrd,  xmm0

+	LOAD_4_PARA_POP

+	ret

+

+

+

+WELS_EXTERN WelsSampleSad8x16_sse2

+WelsSampleSad8x16_sse2:

+	;push   ebx

+	;mov    eax,    [esp+8]

+	;mov    ebx,    [esp+12]

+	;mov    ecx,    [esp+16]

+	;mov    edx,    [esp+20]

+

+	%assign  push_num 0

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+    pxor   xmm6,   xmm6

+

+	SSE2_GetSad8x4

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+    SSE2_GetSad8x4

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	SSE2_GetSad8x4

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+    SSE2_GetSad8x4

+

+    movhlps    xmm0, xmm6

+	paddw      xmm0, xmm6

+	movd       retrd,  xmm0

+	LOAD_4_PARA_POP

+	ret

+

+

+%macro CACHE_SPLIT_CHECK 3 ; address, width, cacheline

+and    %1,  0x1f|(%3>>1)

+cmp    %1,  (32-%2)|(%3>>1)

+%endmacro

+

+WELS_EXTERN WelsSampleSad8x8_sse21

+WelsSampleSad8x8_sse21:

+    ;mov    ecx,    [esp+12]

+	;mov    edx,    ecx

+    ;CACHE_SPLIT_CHECK edx, 8, 64

+	;jle    near   .pixel_sad_8x8_nsplit

+	;push   ebx

+	;push   edi

+	;mov    eax,    [esp+12]

+	;mov    ebx,    [esp+16]

+

+	%assign  push_num 0

+	mov		r2,  arg3

+	push	r2

+	CACHE_SPLIT_CHECK r2, 8, 64

+	jle    near   .pixel_sad_8x8_nsplit

+	pop		r2

+%ifdef X86_32

+	push	r3

+	push	r4

+	push	r5

+%endif

+	%assign  push_num 3

+	mov		r0,  arg1

+	mov		r1,  arg2

+	SIGN_EXTENTION r1, r1d

+    pxor   xmm7,   xmm7

+

+    ;ecx r2, edx r4, edi r5

+

+    mov    r5,    r2

+    and    r5,    0x07

+    sub    r2,    r5

+    mov    r4,    8

+    sub    r4,    r5

+

+    shl    r5,    3

+    shl    r4,    3

+    movd   xmm5,   r5d

+    movd   xmm6,   r4d

+	mov    r5,    8

+	add    r5,    r2

+    mov    r3,    arg4

+	SIGN_EXTENTION r3, r3d

+    movq   xmm0,   [r0]

+	movhps xmm0,   [r0+r1]

+

+	movq   xmm1,   [r2]

+	movq   xmm2,   [r5]

+	movhps xmm1,   [r2+r3]

+	movhps xmm2,   [r5+r3]

+	psrlq  xmm1,   xmm5

+	psllq  xmm2,   xmm6

+	por    xmm1,   xmm2

+

+	psadbw xmm0,   xmm1

+	paddw  xmm7,   xmm0

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	lea    r5,    [r5+2*r3]

+

+    movq   xmm0,   [r0]

+	movhps xmm0,   [r0+r1]

+

+	movq   xmm1,   [r2]

+	movq   xmm2,   [r5]

+	movhps xmm1,   [r2+r3]

+	movhps xmm2,   [r5+r3]

+	psrlq  xmm1,   xmm5

+	psllq  xmm2,   xmm6

+	por    xmm1,   xmm2

+

+	psadbw xmm0,   xmm1

+	paddw  xmm7,   xmm0

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	lea    r5,    [r5+2*r3]

+

+    movq   xmm0,   [r0]

+	movhps xmm0,   [r0+r1]

+

+	movq   xmm1,   [r2]

+	movq   xmm2,   [r5]

+	movhps xmm1,   [r2+r3]

+	movhps xmm2,   [r5+r3]

+	psrlq  xmm1,   xmm5

+	psllq  xmm2,   xmm6

+	por    xmm1,   xmm2

+

+	psadbw xmm0,   xmm1

+	paddw  xmm7,   xmm0

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	lea    r5,    [r5+2*r3]

+

+    movq   xmm0,   [r0]

+	movhps xmm0,   [r0+r1]

+

+	movq   xmm1,   [r2]

+	movq   xmm2,   [r5]

+	movhps xmm1,   [r2+r3]

+	movhps xmm2,   [r5+r3]

+	psrlq  xmm1,   xmm5

+	psllq  xmm2,   xmm6

+	por    xmm1,   xmm2

+

+	psadbw xmm0,   xmm1

+	paddw  xmm7,   xmm0

+

+    movhlps    xmm0, xmm7

+	paddw      xmm0, xmm7

+	movd       retrd,  xmm0

+%ifdef X86_32

+	pop	 r5

+	pop	 r4

+	pop	 r3

+%endif

+	jmp        .return

+

+.pixel_sad_8x8_nsplit:

+    ;push   ebx

+    ;mov    eax,    [esp+8]

+	;mov    ebx,    [esp+12]

+	;mov    edx,    [esp+20]

+

+	pop r2

+	%assign  push_num 0

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	pxor   xmm6,   xmm6

+	SSE2_GetSad8x4

+    lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+    SSE2_GetSad8x4

+    movhlps    xmm0, xmm6

+	paddw      xmm0, xmm6

+	movd       retrd,  xmm0

+	LOAD_4_PARA_POP

+.return:

+	ret

+

+

+;***********************************************************************

+;

+;Pixel_sad_wxh_sse2 END

+;

+;***********************************************************************

+

+

+;***********************************************************************

+;

+;Pixel_sad_4_wxh_sse2 BEGIN

+;

+;***********************************************************************

+

+

+%macro SSE2_Get4LW16Sad 5 ;s-1l, s, s+1l, d, address

+	psadbw %1,   %4

+	paddw  xmm5, %1

+	psadbw %4,   %3

+	paddw  xmm4, %4

+	movdqu %4,   [%5-1]

+	psadbw %4,   %2

+	paddw  xmm6, %4

+	movdqu %4,   [%5+1]

+	psadbw %4,   %2

+	paddw  xmm7, %4

+%endmacro

+WELS_EXTERN WelsSampleSadFour16x16_sse2

+WelsSampleSadFour16x16_sse2:

+	;push ebx

+	;mov    eax,    [esp+8]

+	;mov    ebx,    [esp+12]

+	;mov    ecx,    [esp+16]

+	;mov    edx,    [esp+20]

+

+	%assign  push_num 0

+	LOAD_5_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	pxor   xmm4,   xmm4    ;sad pRefMb-i_stride_ref

+	pxor   xmm5,   xmm5    ;sad pRefMb+i_stride_ref

+	pxor   xmm6,   xmm6    ;sad pRefMb-1

+	pxor   xmm7,   xmm7    ;sad pRefMb+1

+	movdqa xmm0,   [r0]

+	sub    r2,    r3

+	movdqu xmm3,   [r2]

+	psadbw xmm3,   xmm0

+	paddw  xmm4,   xmm3

+

+	movdqa xmm1,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	psadbw xmm3,   xmm1

+	paddw  xmm4,   xmm3

+

+	movdqu xmm2,   [r2+r3-1]

+	psadbw xmm2,   xmm0

+	paddw  xmm6,   xmm2

+

+	movdqu xmm3,   [r2+r3+1]

+	psadbw xmm3,   xmm0

+	paddw  xmm7,   xmm3

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movdqa xmm2,   [r0]

+	movdqu xmm3,   [r2]

+	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2

+	movdqa xmm0,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2+r3

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movdqa xmm1,   [r0]

+	movdqu xmm3,   [r2]

+	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2

+	movdqa xmm2,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2+r3

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movdqa xmm0,   [r0]

+	movdqu xmm3,   [r2]

+	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2

+	movdqa xmm1,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2+r3

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movdqa xmm2,   [r0]

+	movdqu xmm3,   [r2]

+	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2

+	movdqa xmm0,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2+r3

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movdqa xmm1,   [r0]

+	movdqu xmm3,   [r2]

+	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2

+	movdqa xmm2,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2+r3

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movdqa xmm0,   [r0]

+	movdqu xmm3,   [r2]

+	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2

+	movdqa xmm1,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2+r3

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movdqa xmm2,   [r0]

+	movdqu xmm3,   [r2]

+	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2

+	movdqa xmm0,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2+r3

+	lea    r2,    [r2+2*r3]

+	movdqu xmm3,   [r2]

+	psadbw xmm2,   xmm3

+	paddw xmm5,   xmm2

+

+	movdqu xmm2,   [r2-1]

+	psadbw xmm2,   xmm0

+	paddw xmm6,   xmm2

+

+	movdqu xmm3,   [r2+1]

+	psadbw xmm3,   xmm0

+	paddw xmm7,   xmm3

+

+	movdqu xmm3,   [r2+r3]

+	psadbw xmm0,   xmm3

+	paddw xmm5,   xmm0

+

+	;mov        ecx,  [esp+24]

+	movhlps    xmm0, xmm4

+	paddw      xmm4, xmm0

+	movhlps    xmm0, xmm5

+	paddw      xmm5, xmm0

+	movhlps    xmm0, xmm6

+	paddw      xmm6, xmm0

+	movhlps    xmm0, xmm7

+	paddw      xmm7, xmm0

+	punpckldq  xmm4, xmm5

+	punpckldq  xmm6, xmm7

+	punpcklqdq xmm4, xmm6

+	movdqa     [r4],xmm4

+	LOAD_5_PARA_POP

+	ret

+

+

+WELS_EXTERN WelsSampleSadFour16x8_sse2

+WelsSampleSadFour16x8_sse2:

+	;push ebx

+	;push edi

+	;mov    eax,    [esp+12]

+	;mov    ebx,    [esp+16]

+	;mov    edi,    [esp+20]

+	;mov    edx,    [esp+24]

+

+	%assign  push_num 0

+	LOAD_5_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	pxor   xmm4,   xmm4    ;sad pRefMb-i_stride_ref

+	pxor   xmm5,   xmm5    ;sad pRefMb+i_stride_ref

+	pxor   xmm6,   xmm6    ;sad pRefMb-1

+	pxor   xmm7,   xmm7    ;sad pRefMb+1

+	movdqa xmm0,   [r0]

+	sub    r2,    r3

+	movdqu xmm3,   [r2]

+	psadbw xmm3,   xmm0

+	paddw xmm4,   xmm3

+

+	movdqa xmm1,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	psadbw xmm3,   xmm1

+	paddw xmm4,   xmm3

+

+	movdqu xmm2,   [r2+r3-1]

+	psadbw xmm2,   xmm0

+	paddw xmm6,   xmm2

+

+	movdqu xmm3,   [r2+r3+1]

+	psadbw xmm3,   xmm0

+	paddw xmm7,   xmm3

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movdqa xmm2,   [r0]

+	movdqu xmm3,   [r2]

+	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2

+	movdqa xmm0,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2+r3

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movdqa xmm1,   [r0]

+	movdqu xmm3,   [r2]

+	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2

+	movdqa xmm2,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2+r3

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movdqa xmm0,   [r0]

+	movdqu xmm3,   [r2]

+	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2

+	movdqa xmm1,   [r0+r1]

+	movdqu xmm3,   [r2+r3]

+	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2+r3

+	lea    r2,    [r2+2*r3]

+	movdqu xmm3,   [r2]

+	psadbw xmm0,   xmm3

+	paddw xmm5,   xmm0

+

+	movdqu xmm0,   [r2-1]

+	psadbw xmm0,   xmm1

+	paddw xmm6,   xmm0

+

+	movdqu xmm3,   [r2+1]

+	psadbw xmm3,   xmm1

+	paddw xmm7,   xmm3

+

+	movdqu xmm3,   [r2+r3]

+	psadbw xmm1,   xmm3

+	paddw xmm5,   xmm1

+

+	;mov        edi,  [esp+28]

+	movhlps    xmm0, xmm4

+	paddw      xmm4, xmm0

+	movhlps    xmm0, xmm5

+	paddw      xmm5, xmm0

+	movhlps    xmm0, xmm6

+	paddw      xmm6, xmm0

+	movhlps    xmm0, xmm7

+	paddw      xmm7, xmm0

+	punpckldq  xmm4, xmm5

+	punpckldq  xmm6, xmm7

+	punpcklqdq xmm4, xmm6

+	movdqa     [r4],xmm4

+	LOAD_5_PARA_POP

+	ret

+

+WELS_EXTERN WelsSampleSadFour8x16_sse2

+WelsSampleSadFour8x16_sse2:

+	;push ebx

+	;push edi

+	;mov    eax,    [esp+12]

+	;mov    ebx,    [esp+16]

+	;mov    edi,    [esp+20]

+	;mov    edx,    [esp+24]

+

+	%assign  push_num 0

+	LOAD_5_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	pxor   xmm4,   xmm4    ;sad pRefMb-i_stride_ref

+	pxor   xmm5,   xmm5    ;sad pRefMb+i_stride_ref

+	pxor   xmm6,   xmm6    ;sad pRefMb-1

+	pxor   xmm7,   xmm7    ;sad pRefMb+1

+	movq   xmm0,   [r0]

+	movhps xmm0,   [r0+r1]

+	sub    r2,    r3

+	movq   xmm3,   [r2]

+	movhps xmm3,   [r2+r3]

+	psadbw xmm3,   xmm0

+	paddw  xmm4,   xmm3

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	movq   xmm0,  [r0]

+	movhps xmm0,  [r0+r1]

+	psadbw xmm3,  xmm0

+	paddw  xmm4,  xmm3

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	movq   xmm0,  [r0]

+	movhps xmm0,  [r0+r1]

+	psadbw xmm3,  xmm0

+	paddw  xmm4,  xmm3

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	movq   xmm0,  [r0]

+	movhps xmm0,  [r0+r1]

+	psadbw xmm3,  xmm0

+	paddw  xmm4,  xmm3

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	movq   xmm0,  [r0]

+	movhps xmm0,  [r0+r1]

+	psadbw xmm3,  xmm0

+	paddw  xmm4,  xmm3

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	movq   xmm0,  [r0]

+	movhps xmm0,  [r0+r1]

+	psadbw xmm3,  xmm0

+	paddw  xmm4,  xmm3

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	movq   xmm0,  [r0]

+	movhps xmm0,  [r0+r1]

+	psadbw xmm3,  xmm0

+	paddw  xmm4,  xmm3

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	movq   xmm0,  [r0]

+	movhps xmm0,  [r0+r1]

+	psadbw xmm3,  xmm0

+	paddw  xmm4,  xmm3

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	;mov        edi,  [esp+28]

+	movhlps    xmm0, xmm4

+	paddw      xmm4, xmm0

+	movhlps    xmm0, xmm5

+	paddw      xmm5, xmm0

+	movhlps    xmm0, xmm6

+	paddw      xmm6, xmm0

+	movhlps    xmm0, xmm7

+	paddw      xmm7, xmm0

+	punpckldq  xmm4, xmm5

+	punpckldq  xmm6, xmm7

+	punpcklqdq xmm4, xmm6

+	movdqa     [r4],xmm4

+	LOAD_5_PARA_POP

+	ret

+

+

+WELS_EXTERN WelsSampleSadFour8x8_sse2

+WelsSampleSadFour8x8_sse2:

+	;push ebx

+	;push edi

+	;mov    eax,    [esp+12]

+	;mov    ebx,    [esp+16]

+	;mov    edi,    [esp+20]

+	;mov    edx,    [esp+24]

+

+	%assign  push_num 0

+	LOAD_5_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	pxor   xmm4,   xmm4    ;sad pRefMb-i_stride_ref

+	pxor   xmm5,   xmm5    ;sad pRefMb+i_stride_ref

+	pxor   xmm6,   xmm6    ;sad pRefMb-1

+	pxor   xmm7,   xmm7    ;sad pRefMb+1

+	movq   xmm0,   [r0]

+	movhps xmm0,   [r0+r1]

+	sub    r2,    r3

+	movq   xmm3,   [r2]

+	movhps xmm3,   [r2+r3]

+	psadbw xmm3,   xmm0

+	paddw  xmm4,   xmm3

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	movq   xmm0,  [r0]

+	movhps xmm0,  [r0+r1]

+	psadbw xmm3,  xmm0

+	paddw  xmm4,  xmm3

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	movq   xmm0,  [r0]

+	movhps xmm0,  [r0+r1]

+	psadbw xmm3,  xmm0

+	paddw  xmm4,  xmm3

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	movq   xmm0,  [r0]

+	movhps xmm0,  [r0+r1]

+	psadbw xmm3,  xmm0

+	paddw  xmm4,  xmm3

+

+

+	movq   xmm1,  [r2+r3-1]

+	movq   xmm3,  [r2+r3+1]

+

+	lea    r0,    [r0+2*r1]

+	lea    r2,    [r2+2*r3]

+	movhps xmm1,  [r2-1]

+	movhps xmm3,  [r2+1]

+

+	psadbw xmm1,  xmm0

+	paddw  xmm6,  xmm1

+	psadbw xmm3,  xmm0

+	paddw  xmm7,  xmm3

+

+	movq   xmm3,  [r2]

+	movhps xmm3,  [r2+r3]

+	psadbw xmm0,  xmm3

+	paddw  xmm5,  xmm0

+

+	;mov        edi,  [esp+28]

+	movhlps    xmm0, xmm4

+	paddw      xmm4, xmm0

+	movhlps    xmm0, xmm5

+	paddw      xmm5, xmm0

+	movhlps    xmm0, xmm6

+	paddw      xmm6, xmm0

+	movhlps    xmm0, xmm7

+	paddw      xmm7, xmm0

+	punpckldq  xmm4, xmm5

+	punpckldq  xmm6, xmm7

+	punpcklqdq xmm4, xmm6

+	movdqa     [r4],xmm4

+	LOAD_5_PARA_POP

+	ret

+

+WELS_EXTERN WelsSampleSadFour4x4_sse2

+WelsSampleSadFour4x4_sse2:

+	;push ebx

+	;push edi

+	;mov    eax,    [esp+12]

+	;mov    ebx,    [esp+16]

+	;mov    edi,    [esp+20]

+	;mov    edx,    [esp+24]

+

+	%assign  push_num 0

+	LOAD_5_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	movd   xmm0,   [r0]

+	movd   xmm1,   [r0+r1]

+	lea        r0,    [r0+2*r1]

+	movd       xmm2,   [r0]

+	movd       xmm3,   [r0+r1]

+	punpckldq  xmm0, xmm1

+	punpckldq  xmm2, xmm3

+	punpcklqdq xmm0, xmm2

+	sub        r2,  r3

+	movd       xmm1, [r2]

+	movd       xmm2, [r2+r3]

+	punpckldq  xmm1, xmm2

+	movd       xmm2, [r2+r3-1]

+	movd       xmm3, [r2+r3+1]

+

+	lea        r2,  [r2+2*r3]

+

+	movd       xmm4, [r2]

+	movd       xmm5, [r2-1]

+	punpckldq  xmm2, xmm5

+	movd       xmm5, [r2+1]

+	punpckldq  xmm3, xmm5

+

+	movd       xmm5, [r2+r3]

+	punpckldq  xmm4, xmm5

+

+	punpcklqdq xmm1, xmm4 ;-L

+

+	movd       xmm5, [r2+r3-1]

+	movd       xmm6, [r2+r3+1]

+

+	lea        r2,  [r2+2*r3]

+	movd       xmm7, [r2-1]

+	punpckldq  xmm5, xmm7

+	punpcklqdq xmm2, xmm5 ;-1

+	movd       xmm7, [r2+1]

+	punpckldq  xmm6, xmm7

+	punpcklqdq xmm3, xmm6 ;+1

+	movd       xmm6, [r2]

+	movd       xmm7, [r2+r3]

+	punpckldq  xmm6, xmm7

+	punpcklqdq xmm4, xmm6 ;+L

+	psadbw     xmm1, xmm0

+	psadbw     xmm2, xmm0

+	psadbw     xmm3, xmm0

+	psadbw     xmm4, xmm0

+

+	movhlps    xmm0, xmm1

+	paddw      xmm1, xmm0

+	movhlps    xmm0, xmm2

+	paddw      xmm2, xmm0

+	movhlps    xmm0, xmm3

+	paddw      xmm3, xmm0

+	movhlps    xmm0, xmm4

+	paddw      xmm4, xmm0

+	;mov        edi,  [esp+28]

+	punpckldq  xmm1, xmm4

+	punpckldq  xmm2, xmm3

+	punpcklqdq xmm1, xmm2

+	movdqa     [r4],xmm1

+	LOAD_5_PARA_POP

+	ret

+

+;***********************************************************************

+;

+;Pixel_sad_4_wxh_sse2 END

+;

+;***********************************************************************

+

+WELS_EXTERN WelsSampleSad4x4_mmx

+

+align 16

+;***********************************************************************

+;   int32_t __cdecl WelsSampleSad4x4_mmx (uint8_t *, int32_t, uint8_t *, int32_t )

+;***********************************************************************

+WelsSampleSad4x4_mmx:

+    ;push    ebx

+	;%define pushsize     4

+	;%define pix1address	 esp+pushsize+4

+	;%define pix1stride   esp+pushsize+8

+	;%define pix2address  esp+pushsize+12

+	;%define pix2stride   esp+pushsize+16

+    ;mov		  eax, [pix1address]

+    ;mov		  ebx, [pix1stride ]

+    ;mov		  ecx, [pix2address]

+    ;mov		  edx, [pix2stride ]

+

+    %assign  push_num 0

+	LOAD_4_PARA

+	SIGN_EXTENTION r1, r1d

+	SIGN_EXTENTION r3, r3d

+	movd	  mm0, [r0]

+	movd	  mm1, [r0+r1]

+	punpckldq mm0, mm1

+

+	movd      mm3, [r2]

+	movd      mm4, [r2+r3]

+	punpckldq mm3, mm4

+	psadbw    mm0, mm3

+

+	lea       r0, [r0+2*r1]

+	lea       r2, [r2+2*r3]

+

+	movd      mm1, [r0]

+	movd      mm2, [r0+r1]

+	punpckldq mm1, mm2

+

+	movd      mm3, [r2]

+	movd      mm4, [r2+r3]

+	punpckldq mm3, mm4

+	psadbw    mm1, mm3

+	paddw     mm0, mm1

+

+    movd      retrd, mm0

+

+	WELSEMMS

+    LOAD_4_PARA_POP

+    ret

--- a/codec/common/targets.mk

+++ b/codec/common/targets.mk

@@ -16,6 +16,7 @@

 	$(COMMON_SRCDIR)/./mb_copy.asm\

 	$(COMMON_SRCDIR)/./mc_chroma.asm\

 	$(COMMON_SRCDIR)/./mc_luma.asm\

+	$(COMMON_SRCDIR)/./satd_sad.asm\

 	$(COMMON_SRCDIR)/./vaa.asm\

 COMMON_OBJS += $(COMMON_ASM_SRCS:.asm=.o)

--- a/codec/encoder/core/asm/satd_sad.asm

+++ /dev/null

@@ -1,2344 +1,0 @@

-;*!

-;* \copy

-;*     Copyright (c)  2009-2013, Cisco Systems

-;*     All rights reserved.

-;*

-;*     Redistribution and use in source and binary forms, with or without

-;*     modification, are permitted provided that the following conditions

-;*     are met:

-;*

-;*        * Redistributions of source code must retain the above copyright

-;*          notice, this list of conditions and the following disclaimer.

-;*

-;*        * Redistributions in binary form must reproduce the above copyright

-;*          notice, this list of conditions and the following disclaimer in

-;*          the documentation and/or other materials provided with the

-;*          distribution.

-;*

-;*     THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS

-;*     "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT

-;*     LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS

-;*     FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE

-;*     COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,

-;*     INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,

-;*     BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;

-;*     LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER

-;*     CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT

-;*     LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN

-;*     ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE

-;*     POSSIBILITY OF SUCH DAMAGE.

-;*

-;*

-;*  satd_sad.asm

-;*

-;*  Abstract

-;*      WelsSampleSatd4x4_sse2

-;*      WelsSampleSatd8x8_sse2

-;*      WelsSampleSatd16x8_sse2

-;*      WelsSampleSatd8x16_sse2

-;*      WelsSampleSatd16x16_sse2

-;*

-;*      WelsSampleSad16x8_sse2

-;*      WelsSampleSad16x16_sse2

-;*

-;*  History

-;*      8/5/2009 Created

-;*     24/9/2009 modified

-;*

-;*

-;*************************************************************************/

-

-%include "asm_inc.asm"

-

-;***********************************************************************

-; Data

-;***********************************************************************

-SECTION .rodata align=16

-

-align 16

-HSumSubDB1:   db 1,1,1,1,1,1,1,1,1,-1,1,-1,1,-1,1,-1

-align 16

-HSumSubDW1:   dw 1,-1,1,-1,1,-1,1,-1

-align 16

-PDW1:  dw 1,1,1,1,1,1,1,1

-align 16

-PDQ2:  dw 2,0,0,0,2,0,0,0

-align 16

-HSwapSumSubDB1:   times 2 db 1, 1, 1, 1, 1, -1, 1, -1

-

-;***********************************************************************

-; Code

-;***********************************************************************

-SECTION .text

-

-;***********************************************************************

-;

-;Pixel_satd_wxh_sse2 BEGIN

-;

-;***********************************************************************

-%macro MMX_DW_1_2REG 2

-      pxor %1, %1

-      pcmpeqw %2, %2

-      psubw %1, %2

-%endmacro

-

-%macro  SSE2_SumWHorizon1 2

-	movdqa      %2, %1

-	psrldq      %2, 8

-	paddusw     %1, %2

-	movdqa      %2, %1

-	psrldq      %2, 4

-	paddusw     %1, %2

-	movdqa      %2, %1

-	psrldq      %2, 2

-	paddusw     %1, %2

-%endmacro

-

-%macro SSE2_HDMTwo4x4 5 ;in: xmm1,xmm2,xmm3,xmm4  pOut: xmm4,xmm2,xmm1,xmm3

-   SSE2_SumSub %1, %2, %5

-   SSE2_SumSub %3, %4, %5

-   SSE2_SumSub %2, %4, %5

-   SSE2_SumSub %1, %3, %5

-%endmacro

-

-%macro SSE2_SumAbs4 7

-	WELS_AbsW %1, %3

-	WELS_AbsW %2, %3

-	WELS_AbsW %4, %6

-	WELS_AbsW %5, %6

-	paddusw       %1, %2

-	paddusw       %4, %5

-	paddusw       %7, %1

-	paddusw       %7, %4

-%endmacro

-

-%macro  SSE2_SumWHorizon 3

-	movhlps		%2, %1			; x2 = xx xx xx xx d7 d6 d5 d4

-	paddw		%1, %2			; x1 = xx xx xx xx d37 d26 d15 d04

-	punpcklwd	%1, %3			; x1 =  d37  d26 d15 d04

-	movhlps		%2, %1			; x2 = xxxx xxxx d37 d26

-	paddd		%1, %2			; x1 = xxxx xxxx d1357 d0246

-	pshuflw		%2, %1, 0x4e	; x2 = xxxx xxxx d0246 d1357

-	paddd		%1, %2			; x1 = xxxx xxxx xxxx  d01234567

-%endmacro

-

-%macro SSE2_GetSatd8x8 0

-	SSE2_LoadDiff8P    xmm0,xmm4,xmm7,[r0],[r2]

-	SSE2_LoadDiff8P    xmm1,xmm5,xmm7,[r0+r1],[r2+r3]

-	lea                 r0, [r0+2*r1]

-	lea                 r2, [r2+2*r3]

-	SSE2_LoadDiff8P    xmm2,xmm4,xmm7,[r0],[r2]

-	SSE2_LoadDiff8P    xmm3,xmm5,xmm7,[r0+r1],[r2+r3]

-

-	SSE2_HDMTwo4x4       xmm0,xmm1,xmm2,xmm3,xmm4

-	SSE2_TransTwo4x4W     xmm3,xmm1,xmm0,xmm2,xmm4

-	SSE2_HDMTwo4x4       xmm3,xmm1,xmm2,xmm4,xmm5

-	SSE2_SumAbs4         xmm4,xmm1,xmm0,xmm2,xmm3,xmm5,xmm6

-

-	lea					r0,    [r0+2*r1]

-    lea					r2,    [r2+2*r3]

-	SSE2_LoadDiff8P    xmm0,xmm4,xmm7,[r0],[r2]

-	SSE2_LoadDiff8P    xmm1,xmm5,xmm7,[r0+r1],[r2+r3]

-	lea                 r0, [r0+2*r1]

-	lea                 r2, [r2+2*r3]

-	SSE2_LoadDiff8P    xmm2,xmm4,xmm7,[r0],[r2]

-	SSE2_LoadDiff8P    xmm3,xmm5,xmm7,[r0+r1],[r2+r3]

-

-	SSE2_HDMTwo4x4       xmm0,xmm1,xmm2,xmm3,xmm4

-	SSE2_TransTwo4x4W     xmm3,xmm1,xmm0,xmm2,xmm4

-	SSE2_HDMTwo4x4       xmm3,xmm1,xmm2,xmm4,xmm5

-	SSE2_SumAbs4         xmm4,xmm1,xmm0,xmm2,xmm3,xmm5,xmm6

-%endmacro

-

-;***********************************************************************

-;

-;int32_t WelsSampleSatd4x4_sse2( uint8_t *, int32_t, uint8_t *, int32_t );

-;

-;***********************************************************************

-WELS_EXTERN WelsSampleSatd4x4_sse2

-align 16

-WelsSampleSatd4x4_sse2:

-	;push      ebx

-	;mov       eax,  [esp+8]

-	;mov       ebx,  [esp+12]

-	;mov       ecx,  [esp+16]

-	;mov       edx,  [esp+20]

-

-	%assign  push_num 0

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-    movd      xmm0, [r0]

-    movd      xmm1, [r0+r1]

-    lea       r0 , [r0+2*r1]

-    movd      xmm2, [r0]

-    movd      xmm3, [r0+r1]

-    punpckldq xmm0, xmm2

-    punpckldq xmm1, xmm3

-

-    movd      xmm4, [r2]

-    movd      xmm5, [r2+r3]

-    lea       r2 , [r2+2*r3]

-    movd      xmm6, [r2]

-    movd      xmm7, [r2+r3]

-    punpckldq xmm4, xmm6

-    punpckldq xmm5, xmm7

-

-    pxor      xmm6, xmm6

-    punpcklbw xmm0, xmm6

-    punpcklbw xmm1, xmm6

-    punpcklbw xmm4, xmm6

-    punpcklbw xmm5, xmm6

-

-    psubw     xmm0, xmm4

-    psubw     xmm1, xmm5

-

-    movdqa    xmm2, xmm0

-    paddw     xmm0, xmm1

-    psubw     xmm2, xmm1

-    SSE2_XSawp qdq, xmm0, xmm2, xmm3

-

-    movdqa     xmm4, xmm0

-    paddw      xmm0, xmm3

-    psubw      xmm4, xmm3

-

-    movdqa         xmm2, xmm0

-    punpcklwd      xmm0, xmm4

-    punpckhwd      xmm4, xmm2

-

-	SSE2_XSawp     dq,  xmm0, xmm4, xmm3

-	SSE2_XSawp     qdq, xmm0, xmm3, xmm5

-

-    movdqa         xmm7, xmm0

-    paddw          xmm0, xmm5

-    psubw          xmm7, xmm5

-

-	SSE2_XSawp     qdq,  xmm0, xmm7, xmm1

-

-    movdqa         xmm2, xmm0

-    paddw          xmm0, xmm1

-    psubw          xmm2, xmm1

-

-    WELS_AbsW  xmm0, xmm3

-    paddusw        xmm6, xmm0

-	WELS_AbsW  xmm2, xmm4

-    paddusw        xmm6, xmm2

-    SSE2_SumWHorizon1  xmm6, xmm4

-	movd           retrd,  xmm6

-    and            retrd,  0xffff

-    shr            retrd,  1

-	LOAD_4_PARA_POP

-	ret

-

- ;***********************************************************************

- ;

- ;int32_t WelsSampleSatd8x8_sse2( uint8_t *, int32_t, uint8_t *, int32_t, );

- ;

- ;***********************************************************************

- WELS_EXTERN WelsSampleSatd8x8_sse2

-align 16

- WelsSampleSatd8x8_sse2:

-	 ;push   ebx

-	 ;mov    eax,    [esp+8]

-	 ;mov    ebx,    [esp+12]

-	 ;mov    ecx,    [esp+16]

-	 ;mov    edx,    [esp+20]

-

-	%assign  push_num 0

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	pxor   xmm6,   xmm6

-    pxor   xmm7,   xmm7

-    SSE2_GetSatd8x8

-    psrlw   xmm6,  1

-	SSE2_SumWHorizon   xmm6,xmm4,xmm7

-	movd    retrd,   xmm6

-	LOAD_4_PARA_POP

-	ret

-

- ;***********************************************************************

- ;

- ;int32_t WelsSampleSatd8x16_sse2( uint8_t *, int32_t, uint8_t *, int32_t, );

- ;

- ;***********************************************************************

- WELS_EXTERN WelsSampleSatd8x16_sse2

-align 16

- WelsSampleSatd8x16_sse2:

-	 ;push   ebx

-	 ;mov    eax,    [esp+8]

-	 ;mov    ebx,    [esp+12]

-	 ;mov    ecx,    [esp+16]

-	 ;mov    edx,    [esp+20]

-

-	 %assign  push_num 0

-	 LOAD_4_PARA

-	 SIGN_EXTENTION r1, r1d

-	 SIGN_EXTENTION r3, r3d

-	 pxor   xmm6,   xmm6

-     pxor   xmm7,   xmm7

-

-	 SSE2_GetSatd8x8

-     lea    r0,    [r0+2*r1]

-     lea    r2,    [r2+2*r3]

-	 SSE2_GetSatd8x8

-

-	 psrlw   xmm6,  1

-	 SSE2_SumWHorizon   xmm6,xmm4,xmm7

-	 movd    retrd,   xmm6

-	 LOAD_4_PARA_POP

-	 ret

-

-;***********************************************************************

-;

-;int32_t WelsSampleSatd16x8_sse2( uint8_t *, int32_t, uint8_t *, int32_t, );

-;

-;***********************************************************************

-WELS_EXTERN WelsSampleSatd16x8_sse2

-align 16

-WelsSampleSatd16x8_sse2:

-	;push   ebx

-	;mov    eax,    [esp+8]

-	;mov    ebx,    [esp+12]

-	;mov    ecx,    [esp+16]

-	;mov    edx,    [esp+20]

-

-	%assign  push_num 0

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	push r0

-	push r2

-	pxor   xmm6,   xmm6

-    pxor   xmm7,   xmm7

-

-	SSE2_GetSatd8x8

-

-	pop r2

-	pop r0

-	;mov    eax,    [esp+8]

-    ;mov    ecx,    [esp+16]

-    add    r0,    8

-    add    r2,    8

-	SSE2_GetSatd8x8

-

-	psrlw   xmm6,  1

-	SSE2_SumWHorizon   xmm6,xmm4,xmm7

-	movd    retrd,   xmm6

-	LOAD_4_PARA_POP

-	ret

-

-;***********************************************************************

-;

-;int32_t WelsSampleSatd16x16_sse2( uint8_t *, int32_t, uint8_t *, int32_t, );

-;

-;***********************************************************************

-WELS_EXTERN WelsSampleSatd16x16_sse2

-align 16

-WelsSampleSatd16x16_sse2:

-	;push   ebx

-	;mov    eax,    [esp+8]

-	;mov    ebx,    [esp+12]

-	;mov    ecx,    [esp+16]

-	;mov    edx,    [esp+20]

-

-	%assign  push_num 0

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	push r0

-	push r2

-	pxor   xmm6,   xmm6

-    pxor   xmm7,   xmm7

-

-	SSE2_GetSatd8x8

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	SSE2_GetSatd8x8

-

-	pop r2

-	pop r0

-	;mov    eax,    [esp+8]

-	;mov    ecx,    [esp+16]

-	add    r0,    8

-	add    r2,    8

-

-	SSE2_GetSatd8x8

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	SSE2_GetSatd8x8

-

- ; each column sum of SATD is necessarily even, so we don't lose any precision by shifting first.

-    psrlw   xmm6,  1

-	SSE2_SumWHorizon   xmm6,xmm4,xmm7

-	movd    retrd,   xmm6

-	LOAD_4_PARA_POP

-	ret

-

-;***********************************************************************

-;

-;Pixel_satd_wxh_sse2 END

-;

-;***********************************************************************

-

-;***********************************************************************

-;

-;Pixel_satd_intra_sse2 BEGIN

-;

-;***********************************************************************

-

-%macro SSE41_I16x16Get8WSumSub 3 ;xmm5 HSumSubDB1, xmm6 HSumSubDW1, xmm7 PDW1 : in %1, pOut %1, %3

-	pmaddubsw    %1, xmm5

-	movdqa       %2, %1

-	pmaddwd      %1, xmm7

-	pmaddwd      %2, xmm6

-	movdqa       %3, %1

-	punpckldq    %1, %2

-	punpckhdq    %2, %3

-	movdqa       %3, %1

-	punpcklqdq   %1, %2

-	punpckhqdq   %3, %2

-	paddd        xmm4, %1 ;for dc

-	paddd        xmm4, %3 ;for dc

-	packssdw     %1, %3

-	psllw        %1, 2

-%endmacro

-%macro SSE41_ChromaGet8WSumSub 4 ;xmm5 HSumSubDB1, xmm6 HSumSubDW1, xmm7 PDW1 : in %1, pOut %1, %3 : %4 tempsse2

-	pmaddubsw    %1, xmm5

-	movdqa       %2, %1

-	pmaddwd      %1, xmm7

-	pmaddwd      %2, xmm6

-	movdqa       %3, %1

-	punpckldq    %1, %2

-	punpckhdq    %2, %3

-	movdqa       %3, %1

-	punpcklqdq   %1, %2

-	punpckhqdq   %3, %2

-;    paddd        xmm4, %1 ;for dc

-;	 paddd        xmm4, %3 ;for dc

-	movdqa       %4, %1

-	punpcklqdq   %4, %3

-	packssdw     %1, %3

-	psllw        %1, 2

-%endmacro

-

-%macro SSE41_GetX38x4SatdDec 0

-	pxor        xmm7,   xmm7

-	movq        xmm0,   [eax]

-	movq        xmm1,   [eax+ebx]

-	lea         eax,    [eax+2*ebx]

-	movq        xmm2,   [eax]

-	movq        xmm3,   [eax+ebx]

-	lea         eax,    [eax+2*ebx]

-	punpcklbw   xmm0,   xmm7

-	punpcklbw   xmm1,   xmm7

-	punpcklbw   xmm2,   xmm7

-	punpcklbw   xmm3,   xmm7

-	SSE2_HDMTwo4x4       xmm0,xmm1,xmm2,xmm3,xmm7

-	SSE2_TransTwo4x4W     xmm3,xmm1,xmm0,xmm2,xmm7

-	SSE2_HDMTwo4x4       xmm3,xmm1,xmm2,xmm7,xmm0 ;pOut xmm7,xmm1,xmm3,xmm2

-	;doesn't need another transpose

-%endmacro

-%macro SSE41_GetX38x4SatdV 2

-	pxor        xmm0,   xmm0

-	pinsrw      xmm0,   word[esi+%2],   0

-	pinsrw      xmm0,   word[esi+%2+8], 4

-	psubsw      xmm0,   xmm7

-	pabsw       xmm0,   xmm0

-	paddw       xmm4,   xmm0

-	pxor        xmm0,   xmm0

-	pinsrw      xmm0,   word[esi+%2+2],  0

-	pinsrw      xmm0,   word[esi+%2+10], 4

-	psubsw      xmm0,   xmm1

-	pabsw       xmm0,   xmm0

-	paddw       xmm4,   xmm0

-	pxor        xmm0,   xmm0

-	pinsrw      xmm0,   word[esi+%2+4],  0

-	pinsrw      xmm0,   word[esi+%2+12], 4

-	psubsw      xmm0,   xmm3

-	pabsw       xmm0,   xmm0

-	paddw       xmm4,   xmm0

-	pxor        xmm0,   xmm0

-	pinsrw      xmm0,   word[esi+%2+6],  0

-	pinsrw      xmm0,   word[esi+%2+14], 4

-	psubsw      xmm0,   xmm2

-	pabsw       xmm0,   xmm0

-	paddw       xmm4,   xmm0

-%endmacro

-%macro SSE41_GetX38x4SatdH  3

-	movq        xmm0,   [esi+%3+8*%1]

-	punpcklqdq  xmm0,   xmm0

-	psubsw      xmm0,   xmm7

-	pabsw       xmm0,   xmm0

-	paddw       xmm5,   xmm0

-	pabsw       xmm1,   xmm1

-	pabsw       xmm2,   xmm2

-	pabsw       xmm3,   xmm3

-	paddw       xmm2,   xmm1;for DC

-	paddw       xmm2,   xmm3;for DC

-	paddw       xmm5,   xmm2

-%endmacro

-%macro SSE41_I16X16GetX38x4SatdDC 0

-	pxor        xmm0,   xmm0

-	movq2dq     xmm0,   mm4

-	punpcklqdq  xmm0,   xmm0

-	psubsw      xmm0,   xmm7

-	pabsw       xmm0,   xmm0

-	paddw       xmm6,   xmm0

-	paddw       xmm6,   xmm2

-%endmacro

-%macro SSE41_ChromaGetX38x4SatdDC 1

-	shl         %1,     4

-	movdqa      xmm0,   [esi+32+%1]

-	psubsw      xmm0,   xmm7

-	pabsw       xmm0,   xmm0

-	paddw       xmm6,   xmm0

-	paddw       xmm6,   xmm2

-%endmacro

-%macro SSE41_I16x16GetX38x4Satd 2

-	SSE41_GetX38x4SatdDec

-	SSE41_GetX38x4SatdV   %1, %2

-	SSE41_GetX38x4SatdH   %1, %2, 32

-	SSE41_I16X16GetX38x4SatdDC

-%endmacro

-%macro SSE41_ChromaGetX38x4Satd 2

-	SSE41_GetX38x4SatdDec

-	SSE41_GetX38x4SatdV   %1, %2

-	SSE41_GetX38x4SatdH   %1, %2, 16

-	SSE41_ChromaGetX38x4SatdDC %1

-%endmacro

-%macro SSE41_HSum8W 3

-	pmaddwd     %1, %2

-	movhlps     %3, %1

-	paddd       %1, %3

-	pshuflw     %3, %1,0Eh

-	paddd       %1, %3

-%endmacro

-

-

-%ifdef X86_32

-WELS_EXTERN WelsIntra16x16Combined3Satd_sse41

-WelsIntra16x16Combined3Satd_sse41:

-	push   ebx

-	push   esi

-	push   edi

-	mov    ecx,    [esp+16]

-	mov    edx,    [esp+20]

-	mov    eax,    [esp+24]

-	mov    ebx,    [esp+28]

-	mov    esi,    [esp+40] ;temp_satd

-	pxor        xmm4,   xmm4

-	movdqa      xmm5,   [HSumSubDB1]

-	movdqa      xmm6,   [HSumSubDW1]

-	movdqa      xmm7,   [PDW1]

-	sub         ecx,    edx

-	movdqu 		xmm0,   [ecx]

-	movhlps		xmm1,   xmm0

-	punpcklqdq  xmm0,   xmm0

-	punpcklqdq  xmm1,   xmm1

-	SSE41_I16x16Get8WSumSub xmm0, xmm2, xmm3

-	SSE41_I16x16Get8WSumSub xmm1, xmm2, xmm3

-	movdqa      [esi],  xmm0 ;V

-	movdqa      [esi+16], xmm1

-	add         ecx,    edx

-	pinsrb      xmm0,   byte[ecx-1], 0

-	pinsrb      xmm0,   byte[ecx+edx-1], 1

-	lea         ecx,    [ecx+2*edx]

-	pinsrb      xmm0,   byte[ecx-1],     2

-	pinsrb      xmm0,   byte[ecx+edx-1], 3

-	lea         ecx,    [ecx+2*edx]

-	pinsrb      xmm0,   byte[ecx-1],     4

-	pinsrb      xmm0,   byte[ecx+edx-1], 5

-	lea         ecx,    [ecx+2*edx]

-	pinsrb      xmm0,   byte[ecx-1],     6

-	pinsrb      xmm0,   byte[ecx+edx-1], 7

-	lea         ecx,    [ecx+2*edx]

-	pinsrb      xmm0,   byte[ecx-1],     8

-	pinsrb      xmm0,   byte[ecx+edx-1], 9

-	lea         ecx,    [ecx+2*edx]

-	pinsrb      xmm0,   byte[ecx-1],     10

-	pinsrb      xmm0,   byte[ecx+edx-1], 11

-	lea         ecx,    [ecx+2*edx]

-	pinsrb      xmm0,   byte[ecx-1],     12

-	pinsrb      xmm0,   byte[ecx+edx-1], 13

-	lea         ecx,    [ecx+2*edx]

-	pinsrb      xmm0,   byte[ecx-1],     14

-	pinsrb      xmm0,   byte[ecx+edx-1], 15

-	movhlps		xmm1,   xmm0

-	punpcklqdq  xmm0,   xmm0

-	punpcklqdq  xmm1,   xmm1

-	SSE41_I16x16Get8WSumSub xmm0, xmm2, xmm3

-	SSE41_I16x16Get8WSumSub xmm1, xmm2, xmm3

-	movdqa      [esi+32], xmm0 ;H

-	movdqa      [esi+48], xmm1

-	movd        ecx,    xmm4 ;dc

-	add         ecx,    16   ;(sum+16)

-	shr         ecx,    5    ;((sum+16)>>5)

-	shl         ecx,    4    ;

-	movd        mm4,    ecx  ; mm4 copy DC

-	pxor        xmm4,   xmm4 ;V

-	pxor        xmm5,   xmm5 ;H

-	pxor        xmm6,   xmm6 ;DC

-	mov         ecx,    0

-	mov         edi,    0

-.loop16x16_get_satd:

-.loopStart1:

-	SSE41_I16x16GetX38x4Satd ecx, edi

-	inc          ecx

-	cmp         ecx, 4

-	jl          .loopStart1

-	cmp         edi, 16

-	je          .loop16x16_get_satd_end

-	mov         eax, [esp+24]

-	add         eax, 8

-	mov         ecx, 0

-	add         edi, 16

-	jmp         .loop16x16_get_satd

- .loop16x16_get_satd_end:

-	MMX_DW_1_2REG    xmm0, xmm1

-	psrlw       xmm4, 1 ;/2

-	psrlw       xmm5, 1 ;/2

-	psrlw       xmm6, 1 ;/2

-	SSE41_HSum8W     xmm4, xmm0, xmm1

-	SSE41_HSum8W     xmm5, xmm0, xmm1

-	SSE41_HSum8W     xmm6, xmm0, xmm1

-

-	; comparing order: DC H V

-	movd      ebx, xmm6 ;DC

-	movd      edi, xmm5 ;H

-	movd      ecx, xmm4 ;V

-	mov      edx, [esp+36]

-	shl       edx, 1

-	add       edi, edx

-	add       ebx, edx

-	mov       edx, [esp+32]

-	cmp       ebx, edi

-	jge near   not_dc_16x16

-	cmp        ebx, ecx

-	jge near   not_dc_h_16x16

-

-	; for DC mode

-	mov       dword[edx], 2;I16_PRED_DC

-	mov       eax, ebx

-	jmp near return_satd_intra_16x16_x3

-not_dc_16x16:

-	; for H mode

-	cmp       edi, ecx

-	jge near   not_dc_h_16x16

-	mov       dword[edx], 1;I16_PRED_H

-	mov       eax, edi

-	jmp near return_satd_intra_16x16_x3

-not_dc_h_16x16:

-	; for V mode

-	mov       dword[edx], 0;I16_PRED_V

-	mov       eax, ecx

-return_satd_intra_16x16_x3:

-	WELSEMMS

-	pop         edi

-	pop         esi

-	pop         ebx

-ret

-

-%macro SSE41_ChromaGetX38x8Satd 0

-	movdqa      xmm5,   [HSumSubDB1]

-	movdqa      xmm6,   [HSumSubDW1]

-	movdqa      xmm7,   [PDW1]

-	sub         ecx,    edx

-	movq 		xmm0,   [ecx]

-	punpcklqdq  xmm0,   xmm0

-	SSE41_ChromaGet8WSumSub xmm0, xmm2, xmm3, xmm4

-	movdqa      [esi],  xmm0 ;V

-	add         ecx,    edx

-	pinsrb      xmm0,   byte[ecx-1], 0

-	pinsrb      xmm0,   byte[ecx+edx-1], 1

-	lea         ecx,    [ecx+2*edx]

-	pinsrb      xmm0,   byte[ecx-1],     2

-	pinsrb      xmm0,   byte[ecx+edx-1], 3

-	lea         ecx,    [ecx+2*edx]

-	pinsrb      xmm0,   byte[ecx-1],     4

-	pinsrb      xmm0,   byte[ecx+edx-1], 5

-	lea         ecx,    [ecx+2*edx]

-	pinsrb      xmm0,   byte[ecx-1],     6

-	pinsrb      xmm0,   byte[ecx+edx-1], 7

-	punpcklqdq  xmm0,   xmm0

-	SSE41_ChromaGet8WSumSub xmm0, xmm2, xmm3, xmm1

-	movdqa      [esi+16], xmm0 ;H

-;(sum+2)>>2

-	movdqa      xmm6,   [PDQ2]

-	movdqa      xmm5,   xmm4

-	punpckhqdq  xmm5,   xmm1

-	paddd       xmm5,   xmm6

-	psrld       xmm5,   2

-;(sum1+sum2+4)>>3

-	paddd       xmm6,   xmm6

-	paddd       xmm4,   xmm1

-	paddd       xmm4,   xmm6

-	psrld       xmm4,   3

-;satd *16

-	pslld       xmm5,   4

-	pslld       xmm4,   4

-;temp satd

-	movdqa      xmm6,   xmm4

-	punpcklqdq  xmm4,   xmm5

-	psllq       xmm4,   32

-	psrlq       xmm4,   32

-	movdqa      [esi+32], xmm4

-	punpckhqdq  xmm5,   xmm6

-	psllq       xmm5,   32

-	psrlq       xmm5,   32

-	movdqa      [esi+48], xmm5

-

-	pxor        xmm4,   xmm4 ;V

-	pxor        xmm5,   xmm5 ;H

-	pxor        xmm6,   xmm6 ;DC

-	mov         ecx,    0

-loop_chroma_satdx3_cb_cr:

-	SSE41_ChromaGetX38x4Satd ecx, 0

-	inc             ecx

-	cmp             ecx, 2

-	jl              loop_chroma_satdx3_cb_cr

-%endmacro

-

-%macro SSEReg2MMX 3

-	movdq2q     %2, %1

-	movhlps     %1, %1

-	movdq2q     %3, %1

-%endmacro

-%macro MMXReg2SSE 4

-	movq2dq     %1, %3

-	movq2dq     %2, %4

-	punpcklqdq  %1, %2

-%endmacro

-;for reduce the code size of WelsIntraChroma8x8Combined3Satd_sse41

-

-WELS_EXTERN WelsIntraChroma8x8Combined3Satd_sse41

-WelsIntraChroma8x8Combined3Satd_sse41:

-	push   ebx

-	push   esi

-	push   edi

-	mov    ecx,    [esp+16]

-	mov    edx,    [esp+20]

-	mov    eax,    [esp+24]

-	mov    ebx,    [esp+28]

-	mov    esi,    [esp+40] ;temp_satd

-	xor    edi,    edi

-loop_chroma_satdx3:

-	SSE41_ChromaGetX38x8Satd

-	cmp             edi, 1

-	je              loop_chroma_satdx3end

-	inc             edi

-	SSEReg2MMX  xmm4, mm0,mm1

-	SSEReg2MMX  xmm5, mm2,mm3

-	SSEReg2MMX  xmm6, mm5,mm6

-	mov         ecx,  [esp+44]

-	mov         eax,  [esp+48]

-	jmp         loop_chroma_satdx3

-loop_chroma_satdx3end:

-	MMXReg2SSE  xmm0, xmm3, mm0, mm1

-	MMXReg2SSE  xmm1, xmm3, mm2, mm3

-	MMXReg2SSE  xmm2, xmm3, mm5, mm6

-

-	paddw       xmm4, xmm0

-	paddw       xmm5, xmm1

-	paddw       xmm6, xmm2

-

-	MMX_DW_1_2REG    xmm0, xmm1

-	psrlw       xmm4, 1 ;/2

-	psrlw       xmm5, 1 ;/2

-	psrlw       xmm6, 1 ;/2

-	SSE41_HSum8W     xmm4, xmm0, xmm1

-	SSE41_HSum8W     xmm5, xmm0, xmm1

-	SSE41_HSum8W     xmm6, xmm0, xmm1

-	; comparing order: DC H V

-	movd      ebx, xmm6 ;DC

-	movd      edi, xmm5 ;H

-	movd      ecx, xmm4 ;V

-	mov       edx, [esp+36]

-	shl       edx, 1

-	add       edi, edx

-	add       ecx, edx

-	mov       edx, [esp+32]

-	cmp       ebx, edi

-	jge near   not_dc_8x8

-	cmp        ebx, ecx

-	jge near   not_dc_h_8x8

-

-	; for DC mode

-	mov       dword[edx], 0;I8_PRED_DC

-	mov       eax, ebx

-	jmp near return_satd_intra_8x8_x3

-not_dc_8x8:

-	; for H mode

-	cmp       edi, ecx

-	jge near   not_dc_h_8x8

-	mov       dword[edx], 1;I8_PRED_H

-	mov       eax, edi

-	jmp near return_satd_intra_8x8_x3

-not_dc_h_8x8:

-	; for V mode

-	mov       dword[edx], 2;I8_PRED_V

-	mov       eax, ecx

-return_satd_intra_8x8_x3:

-	WELSEMMS

-	pop         edi

-	pop         esi

-	pop         ebx

-ret

-

-

-;***********************************************************************

-;

-;Pixel_satd_intra_sse2 END

-;

-;***********************************************************************

-%macro SSSE3_Get16BSadHVDC 2

-  movd        xmm6,%1

-  pshufb      xmm6,xmm1

-  movdqa      %1,  xmm6

-  movdqa      xmm0,%2

-  psadbw      xmm0,xmm7

-  paddw       xmm4,xmm0

-  movdqa      xmm0,%2

-  psadbw      xmm0,xmm5

-  paddw       xmm2,xmm0

-  psadbw      xmm6,%2

-  paddw       xmm3,xmm6

-%endmacro

-%macro WelsAddDCValue 4

-    movzx   %2, byte %1

-    mov    %3, %2

-    add     %4, %2

-%endmacro

-

-;***********************************************************************

-;

-;Pixel_sad_intra_ssse3 BEGIN

-;

-;***********************************************************************

-WELS_EXTERN WelsIntra16x16Combined3Sad_ssse3

-WelsIntra16x16Combined3Sad_ssse3:

-	push   ebx

-	push   esi

-	push   edi

-	mov    ecx,    [esp+16]

-	mov    edx,    [esp+20]

-	mov    edi,    [esp+40] ;temp_sad

-	sub    ecx,    edx

-    movdqa      xmm5,[ecx]

-    pxor        xmm0,xmm0

-    psadbw      xmm0,xmm5

-    movhlps     xmm1,xmm0

-    paddw       xmm0,xmm1

-    movd        eax,xmm0

-

-    add         ecx,edx

-    lea         ebx, [edx+2*edx]

-    WelsAddDCValue [ecx-1      ], esi, [edi   ], eax

-    WelsAddDCValue [ecx-1+edx  ], esi, [edi+16], eax

-    WelsAddDCValue [ecx-1+edx*2], esi, [edi+32], eax

-    WelsAddDCValue [ecx-1+ebx  ], esi, [edi+48], eax

-    lea         ecx, [ecx+4*edx]

-    add         edi, 64

-    WelsAddDCValue [ecx-1      ], esi, [edi   ], eax

-    WelsAddDCValue [ecx-1+edx  ], esi, [edi+16], eax

-    WelsAddDCValue [ecx-1+edx*2], esi, [edi+32], eax

-    WelsAddDCValue [ecx-1+ebx  ], esi, [edi+48], eax

-    lea         ecx, [ecx+4*edx]

-    add         edi, 64

-    WelsAddDCValue [ecx-1      ], esi, [edi   ], eax

-    WelsAddDCValue [ecx-1+edx  ], esi, [edi+16], eax

-    WelsAddDCValue [ecx-1+edx*2], esi, [edi+32], eax

-    WelsAddDCValue [ecx-1+ebx  ], esi, [edi+48], eax

-    lea         ecx, [ecx+4*edx]

-    add         edi, 64

-    WelsAddDCValue [ecx-1      ], esi, [edi   ], eax

-    WelsAddDCValue [ecx-1+edx  ], esi, [edi+16], eax

-    WelsAddDCValue [ecx-1+edx*2], esi, [edi+32], eax

-    WelsAddDCValue [ecx-1+ebx  ], esi, [edi+48], eax

-    sub        edi, 192

-    add         eax,10h

-    shr         eax,5

-    movd        xmm7,eax

-    pxor        xmm1,xmm1

-    pshufb      xmm7,xmm1

-    pxor        xmm4,xmm4

-    pxor        xmm3,xmm3

-    pxor        xmm2,xmm2

-;sad begin

-	mov    eax,    [esp+24]

-	mov    ebx,    [esp+28]

-    lea         esi, [ebx+2*ebx]

-    SSSE3_Get16BSadHVDC [edi], [eax]

-    SSSE3_Get16BSadHVDC [edi+16], [eax+ebx]

-    SSSE3_Get16BSadHVDC [edi+32], [eax+2*ebx]

-    SSSE3_Get16BSadHVDC [edi+48], [eax+esi]

-    add         edi, 64

-    lea         eax, [eax+4*ebx]

-    SSSE3_Get16BSadHVDC [edi], [eax]

-    SSSE3_Get16BSadHVDC [edi+16], [eax+ebx]

-    SSSE3_Get16BSadHVDC [edi+32], [eax+2*ebx]

-    SSSE3_Get16BSadHVDC [edi+48], [eax+esi]

-    add         edi, 64

-    lea         eax, [eax+4*ebx]

-    SSSE3_Get16BSadHVDC [edi], [eax]

-    SSSE3_Get16BSadHVDC [edi+16], [eax+ebx]

-    SSSE3_Get16BSadHVDC [edi+32], [eax+2*ebx]

-    SSSE3_Get16BSadHVDC [edi+48], [eax+esi]

-    add         edi, 64

-    lea         eax, [eax+4*ebx]

-    SSSE3_Get16BSadHVDC [edi], [eax]

-    SSSE3_Get16BSadHVDC [edi+16], [eax+ebx]

-    SSSE3_Get16BSadHVDC [edi+32], [eax+2*ebx]

-    SSSE3_Get16BSadHVDC [edi+48], [eax+esi]

-

-    pslldq      xmm3,4

-    por         xmm3,xmm2

-    movhlps     xmm1,xmm3

-    paddw       xmm3,xmm1

-    movhlps     xmm0,xmm4

-    paddw       xmm4,xmm0

-; comparing order: DC H V

-	movd        ebx, xmm4 ;DC

-	movd        ecx, xmm3 ;V

-	psrldq      xmm3, 4

-	movd        esi, xmm3 ;H

-	mov         eax, [esp+36] ;lamda

-	shl         eax, 1

-	add         esi, eax

-	add         ebx, eax

-	mov         edx, [esp+32]

-	cmp         ebx, esi

-	jge near   not_dc_16x16_sad

-	cmp        ebx, ecx

-	jge near   not_dc_h_16x16_sad

-	; for DC mode

-	mov       dword[edx], 2;I16_PRED_DC

-	mov       eax, ebx

-    sub        edi, 192

-%assign x 0

-%rep 16

-    movdqa    [edi+16*x], xmm7

-%assign x x+1

-%endrep

-	jmp near return_sad_intra_16x16_x3

-not_dc_16x16_sad:

-	; for H mode

-	cmp       esi, ecx

-	jge near   not_dc_h_16x16_sad

-	mov       dword[edx], 1;I16_PRED_H

-	mov       eax, esi

-	jmp near return_sad_intra_16x16_x3

-not_dc_h_16x16_sad:

-	; for V mode

-	mov       dword[edx], 0;I16_PRED_V

-	mov       eax, ecx

-    sub       edi, 192

-%assign x 0

-%rep 16

-    movdqa    [edi+16*x], xmm5

-%assign x x+1

-%endrep

-return_sad_intra_16x16_x3:

-	pop    edi

-	pop    esi

-	pop    ebx

-	ret

-%endif

-;***********************************************************************

-;

-;Pixel_sad_intra_ssse3 END

-;

-;***********************************************************************

-;***********************************************************************

-;

-;Pixel_satd_wxh_sse41 BEGIN

-;

-;***********************************************************************

-

-;SSE4.1

-%macro SSE41_GetSatd8x4 0

-	movq             xmm0, [r0]

-	punpcklqdq       xmm0, xmm0

-	pmaddubsw        xmm0, xmm7

-	movq             xmm1, [r0+r1]

-	punpcklqdq       xmm1, xmm1

-	pmaddubsw        xmm1, xmm7

-	movq             xmm2, [r2]

-	punpcklqdq       xmm2, xmm2

-	pmaddubsw        xmm2, xmm7

-	movq             xmm3, [r2+r3]

-	punpcklqdq       xmm3, xmm3

-	pmaddubsw        xmm3, xmm7

-	psubsw           xmm0, xmm2

-	psubsw           xmm1, xmm3

-	movq             xmm2, [r0+2*r1]

-	punpcklqdq       xmm2, xmm2

-	pmaddubsw        xmm2, xmm7

-	movq             xmm3, [r0+r4]

-	punpcklqdq       xmm3, xmm3

-	pmaddubsw        xmm3, xmm7

-	movq             xmm4, [r2+2*r3]

-	punpcklqdq       xmm4, xmm4

-	pmaddubsw        xmm4, xmm7

-	movq             xmm5, [r2+r5]

-	punpcklqdq       xmm5, xmm5

-	pmaddubsw        xmm5, xmm7

-	psubsw           xmm2, xmm4

-	psubsw           xmm3, xmm5

-	SSE2_HDMTwo4x4   xmm0, xmm1, xmm2, xmm3, xmm4

-	pabsw            xmm0, xmm0

-	pabsw            xmm2, xmm2

-	pabsw            xmm1, xmm1

-	pabsw            xmm3, xmm3

-	movdqa           xmm4, xmm3

-	pblendw          xmm3, xmm1, 0xAA

-	pslld            xmm1, 16

-	psrld            xmm4, 16

-	por              xmm1, xmm4

-	pmaxuw           xmm1, xmm3

-	paddw            xmm6, xmm1

-	movdqa           xmm4, xmm0

-	pblendw          xmm0, xmm2, 0xAA

-	pslld            xmm2, 16

-	psrld            xmm4, 16

-	por              xmm2, xmm4

-	pmaxuw           xmm0, xmm2

-	paddw            xmm6, xmm0

-%endmacro

-

-%macro SSSE3_SumWHorizon 4 ;eax, srcSSE, tempSSE, tempSSE

-	MMX_DW_1_2REG    %3, %4

-	pmaddwd     %2, %3

-	movhlps     %4, %2

-	paddd       %2, %4

-	pshuflw     %4, %2,0Eh

-	paddd       %2, %4

-	movd		%1, %2

-%endmacro

-;***********************************************************************

-;

-;int32_t WelsSampleSatd4x4_sse41( uint8_t *, int32_t, uint8_t *, int32_t );

-;

-;***********************************************************************

-WELS_EXTERN WelsSampleSatd4x4_sse41

-WelsSampleSatd4x4_sse41:

-	;push        ebx

-	;mov         eax,[esp+8]

-	;mov         ebx,[esp+12]

-	;mov         ecx,[esp+16]

-	;mov         edx,[esp+20]

-

-	%assign  push_num 0

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	movdqa      xmm4,[HSwapSumSubDB1]

-	movd        xmm2,[r2]

-	movd        xmm5,[r2+r3]

-	shufps      xmm2,xmm5,0

-	movd        xmm3,[r2+r3*2]

-	lea         r2, [r3*2+r2]

-	movd        xmm5,[r2+r3]

-	shufps      xmm3,xmm5,0

-	movd        xmm0,[r0]

-	movd        xmm5,[r0+r1]

-	shufps      xmm0,xmm5,0

-	movd        xmm1,[r0+r1*2]

-	lea         r0, [r1*2+r0]

-	movd        xmm5,[r0+r1]

-	shufps      xmm1,xmm5,0

-	pmaddubsw   xmm0,xmm4

-	pmaddubsw   xmm1,xmm4

-	pmaddubsw   xmm2,xmm4

-	pmaddubsw   xmm3,xmm4

-	psubw       xmm0,xmm2

-	psubw       xmm1,xmm3

-	movdqa      xmm2,xmm0

-	paddw       xmm0,xmm1

-	psubw       xmm1,xmm2

-	movdqa      xmm2,xmm0

-	punpcklqdq  xmm0,xmm1

-	punpckhqdq  xmm2,xmm1

-	movdqa      xmm1,xmm0

-	paddw       xmm0,xmm2

-	psubw       xmm2,xmm1

-	movdqa      xmm1,xmm0

-	pblendw     xmm0,xmm2,0AAh

-	pslld       xmm2,16

-	psrld       xmm1,16

-	por         xmm2,xmm1

-	pabsw       xmm0,xmm0

-	pabsw       xmm2,xmm2

-	pmaxsw      xmm0,xmm2

-	SSSE3_SumWHorizon retrd, xmm0, xmm5, xmm7

-	LOAD_4_PARA_POP

-	ret

-

-;***********************************************************************

-;

-;int32_t WelsSampleSatd8x8_sse41( uint8_t *, int32_t, uint8_t *, int32_t, );

-;

-;***********************************************************************

-WELS_EXTERN WelsSampleSatd8x8_sse41

-align 16

-WelsSampleSatd8x8_sse41:

-	;push   ebx

-	;push   esi

-	;push   edi

-	;mov    eax,    [esp+16]

-	;mov    ebx,    [esp+20]

-	;mov    ecx,    [esp+24]

-	;mov    edx,    [esp+28]

-%ifdef X86_32

-	push  r4

-	push  r5

-%endif

-	%assign  push_num 2

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	movdqa      xmm7, [HSumSubDB1]

-	lea         r4,  [r1+r1*2]

-	lea         r5,  [r3+r3*2]

-	pxor		xmm6, xmm6

-	SSE41_GetSatd8x4

-	lea			r0,	 [r0+4*r1]

-	lea			r2,  [r2+4*r3]

-	SSE41_GetSatd8x4

-	SSSE3_SumWHorizon retrd, xmm6, xmm5, xmm7

-	LOAD_4_PARA_POP

-%ifdef X86_32

-	pop  r5

-	pop  r4

-%endif

-	ret

-

-;***********************************************************************

-;

-;int32_t WelsSampleSatd8x16_sse41( uint8_t *, int32_t, uint8_t *, int32_t, );

-;

-;***********************************************************************

-WELS_EXTERN WelsSampleSatd8x16_sse41

-align 16

-WelsSampleSatd8x16_sse41:

-	;push   ebx

-	;push   esi

-	;push   edi

-	;push   ebp

-	;%define pushsize   16

-	;mov    eax,    [esp+pushsize+4]

-	;mov    ebx,    [esp+pushsize+8]

-	;mov    ecx,    [esp+pushsize+12]

-	;mov    edx,    [esp+pushsize+16]

-%ifdef X86_32

-	push  r4

-	push  r5

-	push  r6

-%endif

-	%assign  push_num 3

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	movdqa      xmm7, [HSumSubDB1]

-	lea         r4,  [r1+r1*2]

-	lea         r5,  [r3+r3*2]

-	pxor        xmm6, xmm6

-	mov         r6,    0

-loop_get_satd_8x16:

-	SSE41_GetSatd8x4

-	lea			r0,  [r0+4*r1]

-	lea			r2,  [r2+4*r3]

-	inc         r6

-	cmp         r6,  4

-	jl          loop_get_satd_8x16

-	SSSE3_SumWHorizon retrd, xmm6, xmm5, xmm7

-	LOAD_4_PARA_POP

-%ifdef X86_32

-	pop  r6

-	pop  r5

-	pop  r4

-%endif

-	ret

-

-;***********************************************************************

-;

-;int32_t WelsSampleSatd16x8_sse41( uint8_t *, int32_t, uint8_t *, int32_t, );

-;

-;***********************************************************************

-WELS_EXTERN WelsSampleSatd16x8_sse41

-align 16

-WelsSampleSatd16x8_sse41:

-	;push   ebx

-	;push   esi

-	;push   edi

-	;mov    eax,    [esp+16]

-	;mov    ebx,    [esp+20]

-	;mov    ecx,    [esp+24]

-	;mov    edx,    [esp+28]

-%ifdef X86_32

-	push  r4

-	push  r5

-%endif

-	%assign  push_num 2

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	push  r0

-	push  r2

-

-	movdqa      xmm7, [HSumSubDB1]

-	lea         r4,  [r1+r1*2]

-	lea         r5,  [r3+r3*2]

-	pxor		xmm6,   xmm6

-	SSE41_GetSatd8x4

-	lea			r0,  [r0+4*r1]

-	lea			r2,  [r2+4*r3]

-	SSE41_GetSatd8x4

-

-	pop  r2

-	pop  r0

-	;mov			eax,    [esp+16]

-	;mov			ecx,    [esp+24]

-	add			r0,    8

-	add			r2,    8

-	SSE41_GetSatd8x4

-	lea			r0,  [r0+4*r1]

-	lea			r2,  [r2+4*r3]

-	SSE41_GetSatd8x4

-	SSSE3_SumWHorizon retrd, xmm6, xmm5, xmm7

-	LOAD_4_PARA_POP

-%ifdef X86_32

-	pop  r5

-	pop  r4

-%endif

-	ret

-

-;***********************************************************************

-;

-;int32_t WelsSampleSatd16x16_sse41( uint8_t *, int32_t, uint8_t *, int32_t, );

-;

-;***********************************************************************

-

-WELS_EXTERN WelsSampleSatd16x16_sse41

-align 16

-WelsSampleSatd16x16_sse41:

-	;push   ebx

-	;push   esi

-	;push   edi

-	;push   ebp

-	;%define pushsize   16

-	;mov    eax,    [esp+pushsize+4]

-	;mov    ebx,    [esp+pushsize+8]

-	;mov    ecx,    [esp+pushsize+12]

-	;mov    edx,    [esp+pushsize+16]

-%ifdef X86_32

-	push  r4

-	push  r5

-	push  r6

-%endif

-	%assign  push_num 3

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-

-	push  r0

-	push  r2

-

-	movdqa      xmm7, [HSumSubDB1]

-	lea         r4,  [r1+r1*2]

-	lea         r5,  [r3+r3*2]

-	pxor		xmm6,   xmm6

-	mov         r6,    0

-loop_get_satd_16x16_left:

-	SSE41_GetSatd8x4

-	lea			r0,  [r0+4*r1]

-	lea			r2,  [r2+4*r3]

-	inc         r6

-	cmp         r6,  4

-	jl          loop_get_satd_16x16_left

-

-	pop  r2

-	pop  r0

-	;mov			eax,    [esp+pushsize+4]

-	;mov			ecx,    [esp+pushsize+12]

-	add			r0,    8

-	add			r2,    8

-	mov         r6,    0

-loop_get_satd_16x16_right:

-	SSE41_GetSatd8x4

-	lea			r0,  [r0+4*r1]

-	lea			r2,  [r2+4*r3]

-	inc         r6

-	cmp         r6,  4

-	jl          loop_get_satd_16x16_right

-	SSSE3_SumWHorizon retrd, xmm6, xmm5, xmm7

-	;%undef pushsize

-	LOAD_4_PARA_POP

-%ifdef X86_32

-	pop  r6

-	pop  r5

-	pop  r4

-%endif

-	ret

-

-;***********************************************************************

-;

-;Pixel_satd_wxh_sse41 END

-;

-;***********************************************************************

-

-;***********************************************************************

-;

-;Pixel_sad_wxh_sse2 BEGIN

-;

-;***********************************************************************

-

-%macro SSE2_GetSad2x16 0

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movdqu xmm1,   [r2]

-	MOVDQ  xmm2,   [r0];[eax] must aligned 16

-	psadbw xmm1,   xmm2

-	paddw  xmm0,   xmm1

-	movdqu xmm1,   [r2+r3]

-	MOVDQ  xmm2,   [r0+r1]

-	psadbw xmm1,   xmm2

-	paddw  xmm0,   xmm1

-%endmacro

-

-

-%macro SSE2_GetSad4x16 0

-	movdqu xmm0,   [r2]

-	MOVDQ  xmm2,   [r0]

-	psadbw xmm0,   xmm2

-	paddw  xmm7,   xmm0

-	movdqu xmm1,   [r2+r3]

-	MOVDQ  xmm2,   [r0+r1]

-	psadbw xmm1,   xmm2

-	paddw  xmm7,   xmm1

-	movdqu xmm1,   [r2+2*r3]

-	MOVDQ  xmm2,   [r0+2*r1];[eax] must aligned 16

-	psadbw xmm1,   xmm2

-	paddw  xmm7,   xmm1

-	movdqu xmm1,   [r2+r5]

-	MOVDQ  xmm2,   [r0+r4]

-	psadbw xmm1,   xmm2

-	paddw  xmm7,   xmm1

-%endmacro

-

-

-%macro SSE2_GetSad8x4 0

-	movq   xmm0,   [r0]

-	movq   xmm1,   [r0+r1]

-	lea    r0,     [r0+2*r1]

-	movhps xmm0,   [r0]

-	movhps xmm1,   [r0+r1]

-

-	movq   xmm2,   [r2]

-	movq   xmm3,   [r2+r3]

-	lea    r2,     [r2+2*r3]

-	movhps xmm2,   [r2]

-	movhps xmm3,   [r2+r3]

-	psadbw xmm0,   xmm2

-	psadbw xmm1,   xmm3

-	paddw  xmm6,   xmm0

-	paddw  xmm6,   xmm1

-%endmacro

-

-;***********************************************************************

-;

-;int32_t WelsSampleSad16x16_sse2( uint8_t *, int32_t, uint8_t *, int32_t, )

-;First parameter can align to 16 bytes,

-;In wels, the third parameter can't align to 16 bytes.

-;

-;***********************************************************************

-WELS_EXTERN WelsSampleSad16x16_sse2

-align 16

-WelsSampleSad16x16_sse2:

-	;push ebx

-	;push edi

-	;push esi

-	;%define _STACK_SIZE		12

-	;mov eax, [esp+_STACK_SIZE+4 ]

-	;mov	ebx, [esp+_STACK_SIZE+8 ]

-	;mov ecx, [esp+_STACK_SIZE+12]

-	;mov edx, [esp+_STACK_SIZE+16]

-%ifdef X86_32

-	push  r4

-	push  r5

-%endif

-

-	%assign  push_num 2

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	lea r4, [3*r1]

-	lea r5, [3*r3]

-

-	pxor   xmm7,   xmm7

-	SSE2_GetSad4x16

-	lea	   r0,  [r0+4*r1]

-	lea	   r2,  [r2+4*r3]

-	SSE2_GetSad4x16

-	lea	   r0,  [r0+4*r1]

-	lea	   r2,  [r2+4*r3]

-	SSE2_GetSad4x16

-	lea	   r0,  [r0+4*r1]

-	lea	   r2,  [r2+4*r3]

-	SSE2_GetSad4x16

-	movhlps xmm0, xmm7

-	paddw xmm0, xmm7

-	movd retrd, xmm0

-	LOAD_4_PARA_POP

-%ifdef X86_32

-	pop  r5

-	pop  r4

-%endif

-	ret

-

-;***********************************************************************

-;

-;int32_t WelsSampleSad16x8_sse2( uint8_t *, int32_t, uint8_t *, int32_t, )

-;First parameter can align to 16 bytes,

-;In wels, the third parameter can't align to 16 bytes.

-;

-;***********************************************************************

-WELS_EXTERN WelsSampleSad16x8_sse2

-align 16

-WelsSampleSad16x8_sse2:

-	;push   ebx

-	;mov    eax,    [esp+8]

-	;mov    ebx,    [esp+12]

-	;mov    ecx,    [esp+16]

-	;mov    edx,    [esp+20]

-

-	%assign  push_num 0

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	movdqu xmm0,   [r2]

-	MOVDQ  xmm2,   [r0]

-	psadbw xmm0,   xmm2

-	movdqu xmm1,   [r2+r3]

-	MOVDQ  xmm2,   [r0+r1]

-	psadbw xmm1,   xmm2

-	paddw  xmm0,   xmm1

-

-	SSE2_GetSad2x16

-	SSE2_GetSad2x16

-	SSE2_GetSad2x16

-

-	movhlps     xmm1, xmm0

-	paddw       xmm0, xmm1

-	movd        retrd,  xmm0

-	LOAD_4_PARA_POP

-	ret

-

-

-

-WELS_EXTERN WelsSampleSad8x16_sse2

-WelsSampleSad8x16_sse2:

-	;push   ebx

-	;mov    eax,    [esp+8]

-	;mov    ebx,    [esp+12]

-	;mov    ecx,    [esp+16]

-	;mov    edx,    [esp+20]

-

-	%assign  push_num 0

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-    pxor   xmm6,   xmm6

-

-	SSE2_GetSad8x4

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-    SSE2_GetSad8x4

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	SSE2_GetSad8x4

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-    SSE2_GetSad8x4

-

-    movhlps    xmm0, xmm6

-	paddw      xmm0, xmm6

-	movd       retrd,  xmm0

-	LOAD_4_PARA_POP

-	ret

-

-

-%macro CACHE_SPLIT_CHECK 3 ; address, width, cacheline

-and    %1,  0x1f|(%3>>1)

-cmp    %1,  (32-%2)|(%3>>1)

-%endmacro

-

-WELS_EXTERN WelsSampleSad8x8_sse21

-WelsSampleSad8x8_sse21:

-    ;mov    ecx,    [esp+12]

-	;mov    edx,    ecx

-    ;CACHE_SPLIT_CHECK edx, 8, 64

-	;jle    near   .pixel_sad_8x8_nsplit

-	;push   ebx

-	;push   edi

-	;mov    eax,    [esp+12]

-	;mov    ebx,    [esp+16]

-

-	%assign  push_num 0

-	mov		r2,  arg3

-	push	r2

-	CACHE_SPLIT_CHECK r2, 8, 64

-	jle    near   .pixel_sad_8x8_nsplit

-	pop		r2

-%ifdef X86_32

-	push	r3

-	push	r4

-	push	r5

-%endif

-	%assign  push_num 3

-	mov		r0,  arg1

-	mov		r1,  arg2

-	SIGN_EXTENTION r1, r1d

-    pxor   xmm7,   xmm7

-

-    ;ecx r2, edx r4, edi r5

-

-    mov    r5,    r2

-    and    r5,    0x07

-    sub    r2,    r5

-    mov    r4,    8

-    sub    r4,    r5

-

-    shl    r5,    3

-    shl    r4,    3

-    movd   xmm5,   r5d

-    movd   xmm6,   r4d

-	mov    r5,    8

-	add    r5,    r2

-    mov    r3,    arg4

-	SIGN_EXTENTION r3, r3d

-    movq   xmm0,   [r0]

-	movhps xmm0,   [r0+r1]

-

-	movq   xmm1,   [r2]

-	movq   xmm2,   [r5]

-	movhps xmm1,   [r2+r3]

-	movhps xmm2,   [r5+r3]

-	psrlq  xmm1,   xmm5

-	psllq  xmm2,   xmm6

-	por    xmm1,   xmm2

-

-	psadbw xmm0,   xmm1

-	paddw  xmm7,   xmm0

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	lea    r5,    [r5+2*r3]

-

-    movq   xmm0,   [r0]

-	movhps xmm0,   [r0+r1]

-

-	movq   xmm1,   [r2]

-	movq   xmm2,   [r5]

-	movhps xmm1,   [r2+r3]

-	movhps xmm2,   [r5+r3]

-	psrlq  xmm1,   xmm5

-	psllq  xmm2,   xmm6

-	por    xmm1,   xmm2

-

-	psadbw xmm0,   xmm1

-	paddw  xmm7,   xmm0

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	lea    r5,    [r5+2*r3]

-

-    movq   xmm0,   [r0]

-	movhps xmm0,   [r0+r1]

-

-	movq   xmm1,   [r2]

-	movq   xmm2,   [r5]

-	movhps xmm1,   [r2+r3]

-	movhps xmm2,   [r5+r3]

-	psrlq  xmm1,   xmm5

-	psllq  xmm2,   xmm6

-	por    xmm1,   xmm2

-

-	psadbw xmm0,   xmm1

-	paddw  xmm7,   xmm0

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	lea    r5,    [r5+2*r3]

-

-    movq   xmm0,   [r0]

-	movhps xmm0,   [r0+r1]

-

-	movq   xmm1,   [r2]

-	movq   xmm2,   [r5]

-	movhps xmm1,   [r2+r3]

-	movhps xmm2,   [r5+r3]

-	psrlq  xmm1,   xmm5

-	psllq  xmm2,   xmm6

-	por    xmm1,   xmm2

-

-	psadbw xmm0,   xmm1

-	paddw  xmm7,   xmm0

-

-    movhlps    xmm0, xmm7

-	paddw      xmm0, xmm7

-	movd       retrd,  xmm0

-%ifdef X86_32

-	pop	 r5

-	pop	 r4

-	pop	 r3

-%endif

-	jmp        .return

-

-.pixel_sad_8x8_nsplit:

-    ;push   ebx

-    ;mov    eax,    [esp+8]

-	;mov    ebx,    [esp+12]

-	;mov    edx,    [esp+20]

-

-	pop r2

-	%assign  push_num 0

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	pxor   xmm6,   xmm6

-	SSE2_GetSad8x4

-    lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-    SSE2_GetSad8x4

-    movhlps    xmm0, xmm6

-	paddw      xmm0, xmm6

-	movd       retrd,  xmm0

-	LOAD_4_PARA_POP

-.return:

-	ret

-

-

-;***********************************************************************

-;

-;Pixel_sad_wxh_sse2 END

-;

-;***********************************************************************

-

-

-;***********************************************************************

-;

-;Pixel_sad_4_wxh_sse2 BEGIN

-;

-;***********************************************************************

-

-

-%macro SSE2_Get4LW16Sad 5 ;s-1l, s, s+1l, d, address

-	psadbw %1,   %4

-	paddw  xmm5, %1

-	psadbw %4,   %3

-	paddw  xmm4, %4

-	movdqu %4,   [%5-1]

-	psadbw %4,   %2

-	paddw  xmm6, %4

-	movdqu %4,   [%5+1]

-	psadbw %4,   %2

-	paddw  xmm7, %4

-%endmacro

-WELS_EXTERN WelsSampleSadFour16x16_sse2

-WelsSampleSadFour16x16_sse2:

-	;push ebx

-	;mov    eax,    [esp+8]

-	;mov    ebx,    [esp+12]

-	;mov    ecx,    [esp+16]

-	;mov    edx,    [esp+20]

-

-	%assign  push_num 0

-	LOAD_5_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	pxor   xmm4,   xmm4    ;sad pRefMb-i_stride_ref

-	pxor   xmm5,   xmm5    ;sad pRefMb+i_stride_ref

-	pxor   xmm6,   xmm6    ;sad pRefMb-1

-	pxor   xmm7,   xmm7    ;sad pRefMb+1

-	movdqa xmm0,   [r0]

-	sub    r2,    r3

-	movdqu xmm3,   [r2]

-	psadbw xmm3,   xmm0

-	paddw  xmm4,   xmm3

-

-	movdqa xmm1,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	psadbw xmm3,   xmm1

-	paddw  xmm4,   xmm3

-

-	movdqu xmm2,   [r2+r3-1]

-	psadbw xmm2,   xmm0

-	paddw  xmm6,   xmm2

-

-	movdqu xmm3,   [r2+r3+1]

-	psadbw xmm3,   xmm0

-	paddw  xmm7,   xmm3

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movdqa xmm2,   [r0]

-	movdqu xmm3,   [r2]

-	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2

-	movdqa xmm0,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2+r3

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movdqa xmm1,   [r0]

-	movdqu xmm3,   [r2]

-	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2

-	movdqa xmm2,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2+r3

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movdqa xmm0,   [r0]

-	movdqu xmm3,   [r2]

-	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2

-	movdqa xmm1,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2+r3

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movdqa xmm2,   [r0]

-	movdqu xmm3,   [r2]

-	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2

-	movdqa xmm0,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2+r3

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movdqa xmm1,   [r0]

-	movdqu xmm3,   [r2]

-	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2

-	movdqa xmm2,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2+r3

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movdqa xmm0,   [r0]

-	movdqu xmm3,   [r2]

-	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2

-	movdqa xmm1,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2+r3

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movdqa xmm2,   [r0]

-	movdqu xmm3,   [r2]

-	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2

-	movdqa xmm0,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2+r3

-	lea    r2,    [r2+2*r3]

-	movdqu xmm3,   [r2]

-	psadbw xmm2,   xmm3

-	paddw xmm5,   xmm2

-

-	movdqu xmm2,   [r2-1]

-	psadbw xmm2,   xmm0

-	paddw xmm6,   xmm2

-

-	movdqu xmm3,   [r2+1]

-	psadbw xmm3,   xmm0

-	paddw xmm7,   xmm3

-

-	movdqu xmm3,   [r2+r3]

-	psadbw xmm0,   xmm3

-	paddw xmm5,   xmm0

-

-	;mov        ecx,  [esp+24]

-	movhlps    xmm0, xmm4

-	paddw      xmm4, xmm0

-	movhlps    xmm0, xmm5

-	paddw      xmm5, xmm0

-	movhlps    xmm0, xmm6

-	paddw      xmm6, xmm0

-	movhlps    xmm0, xmm7

-	paddw      xmm7, xmm0

-	punpckldq  xmm4, xmm5

-	punpckldq  xmm6, xmm7

-	punpcklqdq xmm4, xmm6

-	movdqa     [r4],xmm4

-	LOAD_5_PARA_POP

-	ret

-

-

-WELS_EXTERN WelsSampleSadFour16x8_sse2

-WelsSampleSadFour16x8_sse2:

-	;push ebx

-	;push edi

-	;mov    eax,    [esp+12]

-	;mov    ebx,    [esp+16]

-	;mov    edi,    [esp+20]

-	;mov    edx,    [esp+24]

-

-	%assign  push_num 0

-	LOAD_5_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	pxor   xmm4,   xmm4    ;sad pRefMb-i_stride_ref

-	pxor   xmm5,   xmm5    ;sad pRefMb+i_stride_ref

-	pxor   xmm6,   xmm6    ;sad pRefMb-1

-	pxor   xmm7,   xmm7    ;sad pRefMb+1

-	movdqa xmm0,   [r0]

-	sub    r2,    r3

-	movdqu xmm3,   [r2]

-	psadbw xmm3,   xmm0

-	paddw xmm4,   xmm3

-

-	movdqa xmm1,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	psadbw xmm3,   xmm1

-	paddw xmm4,   xmm3

-

-	movdqu xmm2,   [r2+r3-1]

-	psadbw xmm2,   xmm0

-	paddw xmm6,   xmm2

-

-	movdqu xmm3,   [r2+r3+1]

-	psadbw xmm3,   xmm0

-	paddw xmm7,   xmm3

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movdqa xmm2,   [r0]

-	movdqu xmm3,   [r2]

-	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2

-	movdqa xmm0,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2+r3

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movdqa xmm1,   [r0]

-	movdqu xmm3,   [r2]

-	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2

-	movdqa xmm2,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	SSE2_Get4LW16Sad xmm0, xmm1, xmm2, xmm3, r2+r3

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movdqa xmm0,   [r0]

-	movdqu xmm3,   [r2]

-	SSE2_Get4LW16Sad xmm1, xmm2, xmm0, xmm3, r2

-	movdqa xmm1,   [r0+r1]

-	movdqu xmm3,   [r2+r3]

-	SSE2_Get4LW16Sad xmm2, xmm0, xmm1, xmm3, r2+r3

-	lea    r2,    [r2+2*r3]

-	movdqu xmm3,   [r2]

-	psadbw xmm0,   xmm3

-	paddw xmm5,   xmm0

-

-	movdqu xmm0,   [r2-1]

-	psadbw xmm0,   xmm1

-	paddw xmm6,   xmm0

-

-	movdqu xmm3,   [r2+1]

-	psadbw xmm3,   xmm1

-	paddw xmm7,   xmm3

-

-	movdqu xmm3,   [r2+r3]

-	psadbw xmm1,   xmm3

-	paddw xmm5,   xmm1

-

-	;mov        edi,  [esp+28]

-	movhlps    xmm0, xmm4

-	paddw      xmm4, xmm0

-	movhlps    xmm0, xmm5

-	paddw      xmm5, xmm0

-	movhlps    xmm0, xmm6

-	paddw      xmm6, xmm0

-	movhlps    xmm0, xmm7

-	paddw      xmm7, xmm0

-	punpckldq  xmm4, xmm5

-	punpckldq  xmm6, xmm7

-	punpcklqdq xmm4, xmm6

-	movdqa     [r4],xmm4

-	LOAD_5_PARA_POP

-	ret

-

-WELS_EXTERN WelsSampleSadFour8x16_sse2

-WelsSampleSadFour8x16_sse2:

-	;push ebx

-	;push edi

-	;mov    eax,    [esp+12]

-	;mov    ebx,    [esp+16]

-	;mov    edi,    [esp+20]

-	;mov    edx,    [esp+24]

-

-	%assign  push_num 0

-	LOAD_5_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	pxor   xmm4,   xmm4    ;sad pRefMb-i_stride_ref

-	pxor   xmm5,   xmm5    ;sad pRefMb+i_stride_ref

-	pxor   xmm6,   xmm6    ;sad pRefMb-1

-	pxor   xmm7,   xmm7    ;sad pRefMb+1

-	movq   xmm0,   [r0]

-	movhps xmm0,   [r0+r1]

-	sub    r2,    r3

-	movq   xmm3,   [r2]

-	movhps xmm3,   [r2+r3]

-	psadbw xmm3,   xmm0

-	paddw  xmm4,   xmm3

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	movq   xmm0,  [r0]

-	movhps xmm0,  [r0+r1]

-	psadbw xmm3,  xmm0

-	paddw  xmm4,  xmm3

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	movq   xmm0,  [r0]

-	movhps xmm0,  [r0+r1]

-	psadbw xmm3,  xmm0

-	paddw  xmm4,  xmm3

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	movq   xmm0,  [r0]

-	movhps xmm0,  [r0+r1]

-	psadbw xmm3,  xmm0

-	paddw  xmm4,  xmm3

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	movq   xmm0,  [r0]

-	movhps xmm0,  [r0+r1]

-	psadbw xmm3,  xmm0

-	paddw  xmm4,  xmm3

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	movq   xmm0,  [r0]

-	movhps xmm0,  [r0+r1]

-	psadbw xmm3,  xmm0

-	paddw  xmm4,  xmm3

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	movq   xmm0,  [r0]

-	movhps xmm0,  [r0+r1]

-	psadbw xmm3,  xmm0

-	paddw  xmm4,  xmm3

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	movq   xmm0,  [r0]

-	movhps xmm0,  [r0+r1]

-	psadbw xmm3,  xmm0

-	paddw  xmm4,  xmm3

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	;mov        edi,  [esp+28]

-	movhlps    xmm0, xmm4

-	paddw      xmm4, xmm0

-	movhlps    xmm0, xmm5

-	paddw      xmm5, xmm0

-	movhlps    xmm0, xmm6

-	paddw      xmm6, xmm0

-	movhlps    xmm0, xmm7

-	paddw      xmm7, xmm0

-	punpckldq  xmm4, xmm5

-	punpckldq  xmm6, xmm7

-	punpcklqdq xmm4, xmm6

-	movdqa     [r4],xmm4

-	LOAD_5_PARA_POP

-	ret

-

-

-WELS_EXTERN WelsSampleSadFour8x8_sse2

-WelsSampleSadFour8x8_sse2:

-	;push ebx

-	;push edi

-	;mov    eax,    [esp+12]

-	;mov    ebx,    [esp+16]

-	;mov    edi,    [esp+20]

-	;mov    edx,    [esp+24]

-

-	%assign  push_num 0

-	LOAD_5_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	pxor   xmm4,   xmm4    ;sad pRefMb-i_stride_ref

-	pxor   xmm5,   xmm5    ;sad pRefMb+i_stride_ref

-	pxor   xmm6,   xmm6    ;sad pRefMb-1

-	pxor   xmm7,   xmm7    ;sad pRefMb+1

-	movq   xmm0,   [r0]

-	movhps xmm0,   [r0+r1]

-	sub    r2,    r3

-	movq   xmm3,   [r2]

-	movhps xmm3,   [r2+r3]

-	psadbw xmm3,   xmm0

-	paddw  xmm4,   xmm3

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	movq   xmm0,  [r0]

-	movhps xmm0,  [r0+r1]

-	psadbw xmm3,  xmm0

-	paddw  xmm4,  xmm3

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	movq   xmm0,  [r0]

-	movhps xmm0,  [r0+r1]

-	psadbw xmm3,  xmm0

-	paddw  xmm4,  xmm3

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	movq   xmm0,  [r0]

-	movhps xmm0,  [r0+r1]

-	psadbw xmm3,  xmm0

-	paddw  xmm4,  xmm3

-

-

-	movq   xmm1,  [r2+r3-1]

-	movq   xmm3,  [r2+r3+1]

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	movhps xmm1,  [r2-1]

-	movhps xmm3,  [r2+1]

-

-	psadbw xmm1,  xmm0

-	paddw  xmm6,  xmm1

-	psadbw xmm3,  xmm0

-	paddw  xmm7,  xmm3

-

-	movq   xmm3,  [r2]

-	movhps xmm3,  [r2+r3]

-	psadbw xmm0,  xmm3

-	paddw  xmm5,  xmm0

-

-	;mov        edi,  [esp+28]

-	movhlps    xmm0, xmm4

-	paddw      xmm4, xmm0

-	movhlps    xmm0, xmm5

-	paddw      xmm5, xmm0

-	movhlps    xmm0, xmm6

-	paddw      xmm6, xmm0

-	movhlps    xmm0, xmm7

-	paddw      xmm7, xmm0

-	punpckldq  xmm4, xmm5

-	punpckldq  xmm6, xmm7

-	punpcklqdq xmm4, xmm6

-	movdqa     [r4],xmm4

-	LOAD_5_PARA_POP

-	ret

-

-WELS_EXTERN WelsSampleSadFour4x4_sse2

-WelsSampleSadFour4x4_sse2:

-	;push ebx

-	;push edi

-	;mov    eax,    [esp+12]

-	;mov    ebx,    [esp+16]

-	;mov    edi,    [esp+20]

-	;mov    edx,    [esp+24]

-

-	%assign  push_num 0

-	LOAD_5_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	movd   xmm0,   [r0]

-	movd   xmm1,   [r0+r1]

-	lea        r0,    [r0+2*r1]

-	movd       xmm2,   [r0]

-	movd       xmm3,   [r0+r1]

-	punpckldq  xmm0, xmm1

-	punpckldq  xmm2, xmm3

-	punpcklqdq xmm0, xmm2

-	sub        r2,  r3

-	movd       xmm1, [r2]

-	movd       xmm2, [r2+r3]

-	punpckldq  xmm1, xmm2

-	movd       xmm2, [r2+r3-1]

-	movd       xmm3, [r2+r3+1]

-

-	lea        r2,  [r2+2*r3]

-

-	movd       xmm4, [r2]

-	movd       xmm5, [r2-1]

-	punpckldq  xmm2, xmm5

-	movd       xmm5, [r2+1]

-	punpckldq  xmm3, xmm5

-

-	movd       xmm5, [r2+r3]

-	punpckldq  xmm4, xmm5

-

-	punpcklqdq xmm1, xmm4 ;-L

-

-	movd       xmm5, [r2+r3-1]

-	movd       xmm6, [r2+r3+1]

-

-	lea        r2,  [r2+2*r3]

-	movd       xmm7, [r2-1]

-	punpckldq  xmm5, xmm7

-	punpcklqdq xmm2, xmm5 ;-1

-	movd       xmm7, [r2+1]

-	punpckldq  xmm6, xmm7

-	punpcklqdq xmm3, xmm6 ;+1

-	movd       xmm6, [r2]

-	movd       xmm7, [r2+r3]

-	punpckldq  xmm6, xmm7

-	punpcklqdq xmm4, xmm6 ;+L

-	psadbw     xmm1, xmm0

-	psadbw     xmm2, xmm0

-	psadbw     xmm3, xmm0

-	psadbw     xmm4, xmm0

-

-	movhlps    xmm0, xmm1

-	paddw      xmm1, xmm0

-	movhlps    xmm0, xmm2

-	paddw      xmm2, xmm0

-	movhlps    xmm0, xmm3

-	paddw      xmm3, xmm0

-	movhlps    xmm0, xmm4

-	paddw      xmm4, xmm0

-	;mov        edi,  [esp+28]

-	punpckldq  xmm1, xmm4

-	punpckldq  xmm2, xmm3

-	punpcklqdq xmm1, xmm2

-	movdqa     [r4],xmm1

-	LOAD_5_PARA_POP

-	ret

-

-;***********************************************************************

-;

-;Pixel_sad_4_wxh_sse2 END

-;

-;***********************************************************************

-

-WELS_EXTERN WelsSampleSad4x4_mmx

-

-align 16

-;***********************************************************************

-;   int32_t __cdecl WelsSampleSad4x4_mmx (uint8_t *, int32_t, uint8_t *, int32_t )

-;***********************************************************************

-WelsSampleSad4x4_mmx:

-    ;push    ebx

-	;%define pushsize     4

-	;%define pix1address	 esp+pushsize+4

-	;%define pix1stride   esp+pushsize+8

-	;%define pix2address  esp+pushsize+12

-	;%define pix2stride   esp+pushsize+16

-    ;mov		  eax, [pix1address]

-    ;mov		  ebx, [pix1stride ]

-    ;mov		  ecx, [pix2address]

-    ;mov		  edx, [pix2stride ]

-

-    %assign  push_num 0

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	movd	  mm0, [r0]

-	movd	  mm1, [r0+r1]

-	punpckldq mm0, mm1

-

-	movd      mm3, [r2]

-	movd      mm4, [r2+r3]

-	punpckldq mm3, mm4

-	psadbw    mm0, mm3

-

-	lea       r0, [r0+2*r1]

-	lea       r2, [r2+2*r3]

-

-	movd      mm1, [r0]

-	movd      mm2, [r0+r1]

-	punpckldq mm1, mm2

-

-	movd      mm3, [r2]

-	movd      mm4, [r2+r3]

-	punpckldq mm3, mm4

-	psadbw    mm1, mm3

-	paddw     mm0, mm1

-

-    movd      retrd, mm0

-

-	WELSEMMS

-    LOAD_4_PARA_POP

-    ret

--- a/codec/encoder/targets.mk

+++ b/codec/encoder/targets.mk

@@ -41,7 +41,6 @@

 	$(ENCODER_SRCDIR)/./core/asm/intra_pred.asm\

 	$(ENCODER_SRCDIR)/./core/asm/memzero.asm\

 	$(ENCODER_SRCDIR)/./core/asm/quant.asm\

-	$(ENCODER_SRCDIR)/./core/asm/satd_sad.asm\

 	$(ENCODER_SRCDIR)/./core/asm/score.asm\

 ENCODER_OBJS += $(ENCODER_ASM_SRCS:.asm=.o)

--- a/codec/processing/build/win32/WelsVP_2008.vcproj

+++ b/codec/processing/build/win32/WelsVP_2008.vcproj

@@ -594,47 +594,7 @@

 				</FileConfiguration>

 			</File>

 			<File

-				RelativePath="..\..\src\asm\intra_pred.asm"

-				>

-				<FileConfiguration

-					Name="Debug|Win32"

-					>

-					<Tool

-						Name="VCCustomBuildTool"

-						CommandLine="nasm -I$(InputDir) -I$(InputDir)/../../../common/ -f win32 -DPREFIX  -DX86_32 -o $(IntDir)\$(InputName).obj $(InputPath)
"

-						Outputs="$(IntDir)\$(InputName).obj"

-					/>

-				</FileConfiguration>

-				<FileConfiguration

-					Name="Debug|x64"

-					>

-					<Tool

-						Name="VCCustomBuildTool"

-						CommandLine="nasm -I$(InputDir) -I$(InputDir)/../../../common/ -f win64 -O3 -DWIN64 -o $(IntDir)\$(InputName).obj $(InputPath)
"

-						Outputs="$(IntDir)\$(InputName).obj"

-					/>

-				</FileConfiguration>

-				<FileConfiguration

-					Name="Release|Win32"

-					>

-					<Tool

-						Name="VCCustomBuildTool"

-						CommandLine="nasm -I$(InputDir) -I$(InputDir)/../../../common/ -f win32 -DPREFIX  -DX86_32 -o $(IntDir)\$(InputName).obj $(InputPath)
"

-						Outputs="$(IntDir)\$(InputName).obj"

-					/>

-				</FileConfiguration>

-				<FileConfiguration

-					Name="Release|x64"

-					>

-					<Tool

-						Name="VCCustomBuildTool"

-						CommandLine="nasm -I$(InputDir) -I$(InputDir)/../../../common/ -f win64 -O3 -DWIN64 -o $(IntDir)\$(InputName).obj $(InputPath)
"

-						Outputs="$(IntDir)\$(InputName).obj"

-					/>

-				</FileConfiguration>

-			</File>

-			<File

-				RelativePath="..\..\src\asm\sad.asm"

+				RelativePath="..\..\..\common\satd_sad.asm"

 				>

 				<FileConfiguration

 					Name="Debug|Win32"

--- a/codec/processing/src/asm/intra_pred.asm

+++ /dev/null

@@ -1,1505 +1,0 @@

-;*!

-;* \copy

-;*     Copyright (c)  2009-2013, Cisco Systems

-;*     All rights reserved.

-;*

-;*     Redistribution and use in source and binary forms, with or without

-;*     modification, are permitted provided that the following conditions

-;*     are met:

-;*

-;*        * Redistributions of source code must retain the above copyright

-;*          notice, this list of conditions and the following disclaimer.

-;*

-;*        * Redistributions in binary form must reproduce the above copyright

-;*          notice, this list of conditions and the following disclaimer in

-;*          the documentation and/or other materials provided with the

-;*          distribution.

-;*

-;*     THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS

-;*     "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT

-;*     LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS

-;*     FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE

-;*     COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,

-;*     INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,

-;*     BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;

-;*     LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER

-;*     CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT

-;*     LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN

-;*     ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE

-;*     POSSIBILITY OF SUCH DAMAGE.

-;*

-;*

-;*  intra_pred.asm

-;*

-;*  Abstract

-;*      sse2 function for intra predict operations

-;*

-;*  History

-;*      18/09/2009 Created

-;*

-;*

-;*************************************************************************/

-%include "asm_inc.asm"

-

-;***********************************************************************

-; Local Data (Read Only)

-;***********************************************************************

-

-%ifdef FORMAT_COFF

-SECTION .rodata pData

-%else

-SECTION .rodata align=16

-%endif

-

-align 16

-sse2_plane_inc_minus dw -7, -6, -5, -4, -3, -2, -1, 0

-align 16

-sse2_plane_inc dw 1, 2, 3, 4, 5, 6, 7, 8

-align 16

-sse2_plane_dec dw 8, 7, 6, 5, 4, 3, 2, 1

-

-; for chroma plane mode

-sse2_plane_inc_c dw 1, 2, 3, 4

-sse2_plane_dec_c dw 4, 3, 2, 1

-align 16

-sse2_plane_mul_b_c dw -3, -2, -1, 0, 1, 2, 3, 4

-

-align 16

-mmx_01bytes:		times 16	db 1

-;align 16

-;sse_0x0004bytes:	times 8		dw 4

-;ALIGN 16

-;sse_f000 db  255,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0

-

-align 16

-mmx_0x02: dw 0x02, 0x00, 0x00, 0x00

-

-

-;***********************************************************************

-; macros

-;***********************************************************************

-;dB 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1

-;%1 will keep the last result

-%macro SSE_DB_1_2REG 2

-      pxor %1, %1

-      pcmpeqw %2, %2

-      psubb %1, %2

-%endmacro

-

-;xmm0, xmm1, xmm2, eax, ecx

-;lower 64 bits of xmm0 save the result

-%macro SSE2_PRED_H_4X4_TWO_LINE 5

-    movd		%1,	[%4-1]

-	movdqa		%3,	%1

-	punpcklbw	%1,	%3

-	movdqa		%3,	%1

-	punpcklbw	%1,	%3

-

-	;add			%4,	%5

-	movd		%2,	[%4+%5-1]

-	movdqa		%3,	%2

-	punpcklbw	%2,	%3

-	movdqa		%3,	%2

-	punpcklbw	%2,	%3

-	punpckldq	%1,	%2

-%endmacro

-

-%macro  SUMW_HORIZON1 2

-	movdqa      %2, %1

-	psrldq      %2, 8

-	paddusw     %1, %2

-	movdqa      %2, %1

-	psrldq      %2, 4

-	paddusw     %1, %2

-	movdqa      %2, %1

-	psrldq      %2, 2

-	paddusw     %1, %2

-%endmacro

-

-%macro	LOAD_COLUMN 6

-		movd	%1,	[%5]

-		movd	%2,	[%5+%6]

-		punpcklbw %1,	%2

-		lea		%5,	[%5+2*%6]

-		movd	%3,	[%5]

-		movd	%2,	[%5+%6]

-		punpcklbw %3,	%2

-		punpcklwd %1,	%3

-		lea		%5,	[%5+2*%6]

-		movd	%4,	[%5]

-		movd	%2,	[%5+%6]

-		punpcklbw %4,	%2

-		lea		%5,	[%5+2*%6]

-		movd	%3,	[%5]

-		movd	%2,	[%5+%6]

-		lea		%5,	[%5+2*%6]

-		punpcklbw %3,	%2

-		punpcklwd %4,	%3

-		punpckhdq %1,	%4

-%endmacro

-

-%macro  SUMW_HORIZON 3

-	movhlps		%2, %1			; x2 = xx xx xx xx d7 d6 d5 d4

-	paddw		%1, %2			; x1 = xx xx xx xx d37 d26 d15 d04

-	punpcklwd	%1, %3			; x1 =  d37  d26 d15 d04

-	movhlps		%2, %1			; x2 = xxxx xxxx d37 d26

-	paddd		%1, %2			; x1 = xxxx xxxx d1357 d0246

-	pshuflw		%2, %1, 0x4e	; x2 = xxxx xxxx d0246 d1357

-	paddd		%1, %2			; x1 = xxxx xxxx xxxx  d01234567

-%endmacro

-

-

-%macro  COPY_16_TIMES 2

-		movdqa		%2,	[%1-16]

-		psrldq		%2,	15

-		pmuludq		%2,	[mmx_01bytes]

-		pshufd		%2,	%2, 0

-%endmacro

-

-%macro  COPY_16_TIMESS 3

-		movdqa		%2,	[%1+%3-16]

-		psrldq		%2,	15

-		pmuludq		%2,	[mmx_01bytes]

-		pshufd		%2,	%2, 0

-%endmacro

-

-%macro	LOAD_COLUMN_C 6

-		movd	%1,	[%5]

-		movd	%2,	[%5+%6]

-		punpcklbw %1,%2

-		lea		%5,	[%5+2*%6]

-		movd	%3,	[%5]

-		movd	%2,	[%5+%6]

-		punpcklbw %3,	%2

-		punpckhwd %1,	%3

-		lea		%5,	[%5+2*%6]

-%endmacro

-

-%macro LOAD_2_LEFT_AND_ADD 0

-        lea         r1, [r1+2*r2]

-        movzx		r4, byte [r1-0x01]

-        add			r3, r4

-        movzx		r4, byte [r1+r2-0x01]

-        add			r3, r4

-%endmacro

-

-;***********************************************************************

-; Code

-;***********************************************************************

-

-SECTION .text

-WELS_EXTERN WelsI4x4LumaPredH_sse2

-WELS_EXTERN WelsI4x4LumaPredDDR_mmx

-WELS_EXTERN WelsI4x4LumaPredDc_sse2

-WELS_EXTERN WelsI16x16LumaPredPlane_sse2

-

-ALIGN 16

-;***********************************************************************

-;   void __cdecl WelsI4x4LumaPredH_sse2(uint8_t *pred, uint8_t *pRef, int32_t stride)

-;

-;	pred must align to 16

-;***********************************************************************

-WelsI4x4LumaPredH_sse2:

-	push r3

-	%assign push_num 1

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	movzx		r3,	byte [r1-1]

-	movd		xmm0,	r3d

-	pmuludq		xmm0,	[mmx_01bytes]

-

-	movzx		r3,	byte [r1+r2-1]

-	movd		xmm1,	r3d

-	pmuludq		xmm1,	[mmx_01bytes]

-

-	unpcklps	xmm0,	xmm1

-

-	lea			r1,	[r1+r2*2]

-	movzx		r3,	byte [r1-1]

-	movd		xmm2,	r3d

-	pmuludq		xmm2,	[mmx_01bytes]

-

-	movzx		r3,	byte [r1+r2-1]

-	movd		xmm3,	r3d

-	pmuludq		xmm3,	[mmx_01bytes]

-

-	unpcklps	xmm2,	xmm3

-	unpcklpd	xmm0,	xmm2

-

-	movdqa		[r0],	xmm0

-	pop r3

-	ret

-

-;***********************************************************************

-; void WelsI16x16LumaPredPlane_sse2(uint8_t *pred, uint8_t *pRef, int32_t stride);

-;***********************************************************************

-WelsI16x16LumaPredPlane_sse2:

-		;%define pushsize	4

-		;push	esi

-		;mov		esi,	[esp + pushsize + 8]

-		;mov		ecx,	[esp + pushsize + 12]

-		push r3

-		push r4

-		%assign push_num 2

-		LOAD_3_PARA

-		%ifndef X86_32

-		movsx r2, r2d

-		%endif

-		sub		r1,	1

-		sub		r1,	r2

-

-		;for H

-		pxor	xmm7,	xmm7

-		movq	xmm0,	[r1]

-		movdqa	xmm5,	[sse2_plane_dec]

-		punpcklbw xmm0,	xmm7

-		pmullw	xmm0,	xmm5

-		movq	xmm1,	[r1 + 9]

-		movdqa	xmm6,	[sse2_plane_inc]

-		punpcklbw xmm1,	xmm7

-		pmullw	xmm1,	xmm6

-		psubw	xmm1,	xmm0

-

-		SUMW_HORIZON	xmm1,xmm0,xmm2

-		movd    r3d,	xmm1		; H += (i + 1) * (top[8 + i] - top[6 - i]);

-		movsx	r3,	r3w

-		imul	r3,	5

-		add		r3,	32

-		sar		r3,	6			; b = (5 * H + 32) >> 6;

-		SSE2_Copy8Times	xmm1, r3d	; xmm1 = b,b,b,b,b,b,b,b

-

-		movzx	r4,	BYTE [r1+16]

-		sub	r1, 3

-		LOAD_COLUMN		xmm0, xmm2, xmm3, xmm4, r1, r2

-

-		add		r1,	3

-		movzx	r3,	BYTE [r1+8*r2]

-		add		r4,	r3

-		shl		r4,	4			;	a = (left[15*stride] + top[15]) << 4;

-

-		sub	r1, 3

-		add		r1,	r2

-		LOAD_COLUMN		xmm7, xmm2, xmm3, xmm4, r1, r2

-		pxor	xmm4,	xmm4

-		punpckhbw xmm0,	xmm4

-		pmullw	xmm0,	xmm5

-		punpckhbw xmm7,	xmm4

-		pmullw	xmm7,	xmm6

-		psubw	xmm7,	xmm0

-

-		SUMW_HORIZON   xmm7,xmm0,xmm2

-		movd    r3d,   xmm7			; V

-		movsx	r3,	r3w

-		imul	r3,	5

-		add		r3,	32

-		sar		r3,	6				; c = (5 * V + 32) >> 6;

-		SSE2_Copy8Times	xmm4, r3d		; xmm4 = c,c,c,c,c,c,c,c

-

-		;mov		esi,	[esp + pushsize + 4]

-		add		r4,	16

-		imul	r3,	-7

-		add		r3,	r4				; s = a + 16 + (-7)*c

-		SSE2_Copy8Times	xmm0, r3d		; xmm0 = s,s,s,s,s,s,s,s

-

-		xor		r3,	r3

-		movdqa	xmm5,	[sse2_plane_inc_minus]

-

-get_i16x16_luma_pred_plane_sse2_1:

-		movdqa	xmm2,	xmm1

-		pmullw	xmm2,	xmm5

-		paddw	xmm2,	xmm0

-		psraw	xmm2,	5

-		movdqa	xmm3,	xmm1

-		pmullw	xmm3,	xmm6

-		paddw	xmm3,	xmm0

-		psraw	xmm3,	5

-		packuswb xmm2,	xmm3

-		movdqa	[r0],	xmm2

-		paddw	xmm0,	xmm4

-		add		r0,	16

-		inc		r3

-		cmp		r3,	16

-		jnz get_i16x16_luma_pred_plane_sse2_1

-		pop r4

-		pop r3

-		ret

-

-;***********************************************************************

-; void WelsI16x16LumaPredH_sse2(uint8_t *pred, uint8_t *pRef, int32_t stride);

-;***********************************************************************

-

-%macro SSE2_PRED_H_16X16_ONE_LINE 0

-	add r0, 16

-	add r1, r2

-	movzx r3, byte [r1]

-	SSE2_Copy16Times xmm0, r3d

-	movdqa [r0], xmm0

-%endmacro

-

-WELS_EXTERN WelsI16x16LumaPredH_sse2

-WelsI16x16LumaPredH_sse2:

-	push r3

-	%assign push_num 1

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	dec r1

-	movzx r3, byte [r1]

-	SSE2_Copy16Times xmm0, r3d

-	movdqa [r0], xmm0

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	SSE2_PRED_H_16X16_ONE_LINE

-	pop r3

-    ret

-

-;***********************************************************************

-; void WelsI16x16LumaPredV_sse2(uint8_t *pred, uint8_t *pRef, int32_t stride);

-;***********************************************************************

-WELS_EXTERN WelsI16x16LumaPredV_sse2

-WelsI16x16LumaPredV_sse2:

-    ;mov     edx, [esp+4]    ; pred

-    ;mov     eax, [esp+8]	; pRef

-    ;mov     ecx, [esp+12]   ; stride

-    %assign push_num 0

-    LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-    sub     r1, r2

-    movdqa  xmm0, [r1]

-

-    movdqa  [r0], xmm0

-    movdqa  [r0+10h], xmm0

-    movdqa  [r0+20h], xmm0

-    movdqa  [r0+30h], xmm0

-    movdqa  [r0+40h], xmm0

-    movdqa  [r0+50h], xmm0

-    movdqa  [r0+60h], xmm0

-    movdqa  [r0+70h], xmm0

-    movdqa  [r0+80h], xmm0

-    movdqa  [r0+90h], xmm0

-    movdqa  [r0+160], xmm0

-    movdqa  [r0+176], xmm0

-    movdqa  [r0+192], xmm0

-    movdqa  [r0+208], xmm0

-    movdqa  [r0+224], xmm0

-    movdqa  [r0+240], xmm0

-

-    ret

-

-;***********************************************************************

-; void WelsIChromaPredPlane_sse2(uint8_t *pred, uint8_t *pRef, int32_t stride);

-;***********************************************************************

-WELS_EXTERN WelsIChromaPredPlane_sse2

-WelsIChromaPredPlane_sse2:

-		;%define pushsize	4

-		;push	esi

-		;mov		esi,	[esp + pushsize + 8]	;pRef

-		;mov		ecx,	[esp + pushsize + 12]	;stride

-		push r3

-		push r4

-		%assign push_num 2

-		LOAD_3_PARA

-		%ifndef X86_32

-		movsx r2, r2d

-		%endif

-		sub		r1,	1

-		sub		r1,	r2

-

-		pxor	mm7,	mm7

-		movq	mm0,	[r1]

-		movq	mm5,	[sse2_plane_dec_c]

-		punpcklbw mm0,	mm7

-		pmullw	mm0,	mm5

-		movq	mm1,	[r1 + 5]

-		movq	mm6,	[sse2_plane_inc_c]

-		punpcklbw mm1,	mm7

-		pmullw	mm1,	mm6

-		psubw	mm1,	mm0

-

-		movq2dq xmm1,   mm1

-		pxor    xmm2,   xmm2

-		SUMW_HORIZON	xmm1,xmm0,xmm2

-		movd    r3d,	xmm1

-		movsx	r3,	r3w

-		imul	r3,	17

-		add		r3,	16

-		sar		r3,	5			; b = (17 * H + 16) >> 5;

-		SSE2_Copy8Times	xmm1, r3d	; mm1 = b,b,b,b,b,b,b,b

-

-		movzx	r3,	BYTE [r1+8]

-		sub	r1, 3

-		LOAD_COLUMN_C	mm0, mm2, mm3, mm4, r1, r2

-

-		add		r1,	3

-		movzx	r4,	BYTE [r1+4*r2]

-		add		r4,	r3

-		shl		r4,	4			; a = (left[7*stride] + top[7]) << 4;

-

-		sub	r1, 3

-		add		r1,	r2

-		LOAD_COLUMN_C	mm7, mm2, mm3, mm4, r1, r2

-		pxor	mm4,	mm4

-		punpckhbw mm0,	mm4

-		pmullw	mm0,	mm5

-		punpckhbw mm7,	mm4

-		pmullw	mm7,	mm6

-		psubw	mm7,	mm0

-

-		movq2dq xmm7,   mm7

-		pxor    xmm2,   xmm2

-		SUMW_HORIZON	xmm7,xmm0,xmm2

-		movd    r3d,    xmm7			; V

-		movsx	r3,	r3w

-		imul	r3,	17

-		add		r3,	16

-		sar		r3,	5				; c = (17 * V + 16) >> 5;

-		SSE2_Copy8Times	xmm4, r3d	; mm4 = c,c,c,c,c,c,c,c

-

-		;mov		esi,	[esp + pushsize + 4]

-		add		r4,	16

-		imul	r3,	-3

-		add		r3,	r4		; s = a + 16 + (-3)*c

-		SSE2_Copy8Times	xmm0, r3d	; xmm0 = s,s,s,s,s,s,s,s

-

-		xor		r3,	r3

-		movdqa	xmm5,	[sse2_plane_mul_b_c]

-

-get_i_chroma_pred_plane_sse2_1:

-		movdqa	xmm2,	xmm1

-		pmullw	xmm2,	xmm5

-		paddw	xmm2,	xmm0

-		psraw	xmm2,	5

-		packuswb xmm2,	xmm2

-		movq	[r0],	xmm2

-		paddw	xmm0,	xmm4

-		add		r0,	8

-		inc		r3

-		cmp		r3,	8

-		jnz get_i_chroma_pred_plane_sse2_1

-		pop r4

-		pop r3

-		WELSEMMS

-		ret

-

-ALIGN 16

-;***********************************************************************

-;	0 |1 |2 |3 |4 |

-;	6 |7 |8 |9 |10|

-;	11|12|13|14|15|

-;	16|17|18|19|20|

-;	21|22|23|24|25|

-;	7 is the start pixel of current 4x4 block

-;	pred[7] = ([6]+[0]*2+[1]+2)/4

-;

-;   void __cdecl WelsI4x4LumaPredDDR_mmx(uint8_t *pred,uint8_t *pRef,int32_t stride)

-;

-;***********************************************************************

-WelsI4x4LumaPredDDR_mmx:

-	;mov			edx,[esp+4]			;pred

-	;mov         eax,[esp+8]			;pRef

-	;mov			ecx,[esp+12]		;stride

-	%assign push_num 0

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	movq        mm1,[r1+r2-8]		;get value of 11,decreasing 8 is trying to improve the performance of movq mm1[8] = 11

-	movq        mm2,[r1-8]			;get value of 6 mm2[8] = 6

-	sub		r1, r2			;mov eax to above line of current block(postion of 1)

-	punpckhbw   mm2,[r1-8]			;mm2[8](high 8th byte of mm2) = [0](value of 0), mm2[7]= [6]

-	movd        mm3,[r1]			;get value 1, mm3[1] = [1],mm3[2]=[2],mm3[3]=[3]

-	punpckhwd   mm1,mm2				;mm1[8]=[0],mm1[7]=[6],mm1[6]=[11]

-	psllq       mm3,18h				;mm3[5]=[1]

-	psrlq       mm1,28h				;mm1[3]=[0],mm1[2]=[6],mm1[1]=[11]

-	por         mm3,mm1				;mm3[6]=[3],mm3[5]=[2],mm3[4]=[1],mm3[3]=[0],mm3[2]=[6],mm3[1]=[11]

-	movq        mm1,mm3				;mm1[6]=[3],mm1[5]=[2],mm1[4]=[1],mm1[3]=[0],mm1[2]=[6],mm1[1]=[11]

-	lea  	    r1,[r1+r2*2-8h]		;set eax point to 12

-	movq        mm4,[r1+r2]		;get value of 16, mm4[8]=[16]

-	psllq       mm3,8				;mm3[7]=[3],mm3[6]=[2],mm3[5]=[1],mm3[4]=[0],mm3[3]=[6],mm3[2]=[11],mm3[1]=0

-	psrlq       mm4,38h				;mm4[1]=[16]

-	por         mm3,mm4				;mm3[7]=[3],mm3[6]=[2],mm3[5]=[1],mm3[4]=[0],mm3[3]=[6],mm3[2]=[11],mm3[1]=[16]

-	movq        mm2,mm3				;mm2[7]=[3],mm2[6]=[2],mm2[5]=[1],mm2[4]=[0],mm2[3]=[6],mm2[2]=[11],mm2[1]=[16]

-	movq        mm4,[r1+r2*2]		;mm4[8]=[21]

-	psllq       mm3,8				;mm3[8]=[3],mm3[7]=[2],mm3[6]=[1],mm3[5]=[0],mm3[4]=[6],mm3[3]=[11],mm3[2]=[16],mm3[1]=0

-	psrlq       mm4,38h				;mm4[1]=[21]

-	por         mm3,mm4				;mm3[8]=[3],mm3[7]=[2],mm3[6]=[1],mm3[5]=[0],mm3[4]=[6],mm3[3]=[11],mm3[2]=[16],mm3[1]=[21]

-	movq        mm4,mm3				;mm4[8]=[3],mm4[7]=[2],mm4[6]=[1],mm4[5]=[0],mm4[4]=[6],mm4[3]=[11],mm4[2]=[16],mm4[1]=[21]

-	pavgb       mm3,mm1				;mm3=([11]+[21]+1)/2

-	pxor        mm1,mm4				;find odd value in the lowest bit of each byte

-	pand        mm1,[mmx_01bytes]	;set the odd bit

-	psubusb     mm3,mm1				;decrease 1 from odd bytes

-	pavgb       mm2,mm3				;mm2=(([11]+[21]+1)/2+1+[16])/2

-

-	movd        [r0+12],mm2

-	psrlq       mm2,8

-	movd        [r0+8],mm2

-	psrlq       mm2,8

-	movd        [r0+4],mm2

-	psrlq       mm2,8

-	movd        [r0],mm2

-	WELSEMMS

-	ret

-

-ALIGN 16

-;***********************************************************************

-;	0 |1 |2 |3 |4 |

-;	5 |6 |7 |8 |9 |

-;	10|11|12|13|14|

-;	15|16|17|18|19|

-;	20|21|22|23|24|

-;	6 is the start pixel of current 4x4 block

-;	pred[6] = ([1]+[2]+[3]+[4]+[5]+[10]+[15]+[20]+4)/8

-;

-;   void __cdecl WelsI4x4LumaPredDc_sse2(uint8_t *pred,uint8_t *pRef,int32_t stride)

-;

-;***********************************************************************

-WelsI4x4LumaPredDc_sse2:

-	push r3

-	push r4

-	%assign push_num 2

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	movzx		r4,	byte [r1-1h]

-	sub			r1,	r2

-	movd		xmm0,	[r1]

-	pxor		xmm1,	xmm1

-	psadbw		xmm0,	xmm1

-	xor r3, r3

-	movd		r3d,	xmm0

-	add			r3,	r4

-	movzx		r4,	byte [r1+r2*2-1h]

-	add			r3,	r4

-

-	lea			r1,	[r1+r2*2-1]

-	movzx		r4,	byte [r1+r2]

-	add			r3,	r4

-

-	movzx		r4,	byte [r1+r2*2]

-	add			r3,	r4

-	add			r3,	4

-	sar			r3,	3

-	imul		r3,	0x01010101

-

-	movd		xmm0,	r3d

-	pshufd		xmm0,	xmm0,	0

-	movdqa		[r0],	xmm0

-	pop r4

-	pop r3

-	ret

-

-ALIGN 16

-;***********************************************************************

-;	void __cdecl WelsIChromaPredH_mmx(uint8_t *pred, uint8_t *pRef, int32_t stride)

-;   copy 8 pixel of 8 line from left

-;***********************************************************************

-%macro MMX_PRED_H_8X8_ONE_LINE 4

-	movq		%1,		[%3-8]

-	psrlq		%1,		38h

-

-	;pmuludq		%1,		[mmx_01bytes]		;extend to 4 bytes

-	pmullw		%1,		[mmx_01bytes]

-	pshufw		%1,		%1,	0

-	movq		[%4],	%1

-%endmacro

-

-%macro MMX_PRED_H_8X8_ONE_LINEE 4

-	movq		%1,		[%3+r2-8]

-	psrlq		%1,		38h

-

-	;pmuludq		%1,		[mmx_01bytes]		;extend to 4 bytes

-	pmullw		%1,		[mmx_01bytes]

-	pshufw		%1,		%1,	0

-	movq		[%4],	%1

-%endmacro

-

-WELS_EXTERN WelsIChromaPredH_mmx

-WelsIChromaPredH_mmx:

-	;mov			edx,	[esp+4]			;pred

-	;mov         eax,	[esp+8]			;pRef

-	;mov			ecx,	[esp+12]		;stride

-	%assign push_num 0

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	movq		mm0,	[r1-8]

-	psrlq		mm0,	38h

-

-	;pmuludq		mm0,	[mmx_01bytes]		;extend to 4 bytes

-	pmullw		mm0,		[mmx_01bytes]

-	pshufw		mm0,	mm0,	0

-	movq		[r0],	mm0

-

-	MMX_PRED_H_8X8_ONE_LINEE	mm0, mm1, r1,r0+8

-

-	lea			r1,[r1+r2*2]

-	MMX_PRED_H_8X8_ONE_LINE	mm0, mm1, r1,r0+16

-

-	MMX_PRED_H_8X8_ONE_LINEE	mm0, mm1, r1,r0+24

-

-	lea			r1,[r1+r2*2]

-	MMX_PRED_H_8X8_ONE_LINE	mm0, mm1, r1,r0+32

-

-	MMX_PRED_H_8X8_ONE_LINEE	mm0, mm1, r1,r0+40

-

-	lea			r1,[r1+r2*2]

-	MMX_PRED_H_8X8_ONE_LINE	mm0, mm1, r1,r0+48

-

-	MMX_PRED_H_8X8_ONE_LINEE	mm0, mm1, r1,r0+56

-	WELSEMMS

-	ret

-

-ALIGN 16

-;***********************************************************************

-;	void __cdecl WelsI4x4LumaPredV_sse2(uint8_t *pred, uint8_t *pRef, int32_t stride)

-;   copy pixels from top 4 pixels

-;***********************************************************************

-WELS_EXTERN WelsI4x4LumaPredV_sse2

-WelsI4x4LumaPredV_sse2:

-	%assign push_num 0

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	sub			r1,	r2

-	movd		xmm0,	[r1]

-	pshufd		xmm0,	xmm0,	0

-	movdqa		[r0],	xmm0

-	ret

-

-ALIGN 16

-;***********************************************************************

-;	void __cdecl WelsIChromaPredV_sse2(uint8_t *pred, uint8_t *pRef, int32_t stride)

-;   copy 8 pixels from top 8 pixels

-;***********************************************************************

-WELS_EXTERN WelsIChromaPredV_sse2

-WelsIChromaPredV_sse2:

-	%assign push_num 0

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	sub		r1,		r2

-	movq		xmm0,		[r1]

-	movdqa		xmm1,		xmm0

-	punpcklqdq	xmm0,		xmm1

-	movdqa		[r0],		xmm0

-	movdqa		[r0+16],	xmm0

-	movdqa		[r0+32],	xmm0

-	movdqa		[r0+48],	xmm0

-	ret

-

-	ALIGN 16

-;***********************************************************************

-;	lt|t0|t1|t2|t3|

-;	l0|

-;	l1|

-;	l2|

-;	l3|

-;	t3 will never been used

-;   destination:

-;	|a |b |c |d |

-;	|e |f |a |b |

-;	|g |h |e |f |

-;	|i |j |g |h |

-

-;   a = (1 + lt + l0)>>1

-;   e = (1 + l0 + l1)>>1

-;   g = (1 + l1 + l2)>>1

-;   i = (1 + l2 + l3)>>1

-

-;   d = (2 + t0 + (t1<<1) + t2)>>2

-;   c = (2 + lt + (t0<<1) + t1)>>2

-;   b = (2 + l0 + (lt<<1) + t0)>>2

-

-;   f = (2 + l1 + (l0<<1) + lt)>>2

-;   h = (2 + l2 + (l1<<1) + l0)>>2

-;   j = (2 + l3 + (l2<<1) + l1)>>2

-;   [b a f e h g j i] + [d c b a] --> mov to memory

-;

-;   void WelsI4x4LumaPredHD_mmx(uint8_t *pred,uint8_t *pRef,int32_t stride)

-;***********************************************************************

-WELS_EXTERN WelsI4x4LumaPredHD_mmx

-WelsI4x4LumaPredHD_mmx:

-	%assign push_num 0

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	sub         r1, r2

-	movd        mm0, [r1-1]            ; mm0 = [xx xx xx xx t2 t1 t0 lt]

-	psllq       mm0, 20h                ; mm0 = [t2 t1 t0 lt xx xx xx xx]

-

-	movd        mm1, [r1+2*r2-4]

-	punpcklbw   mm1, [r1+r2-4]        ; mm1[7] = l0, mm1[6] = l1

-	lea         r1, [r1+2*r2]

-	movd        mm2, [r1+2*r2-4]

-	punpcklbw   mm2, [r1+r2-4]        ; mm2[7] = l2, mm2[6] = l3

-	punpckhwd   mm2, mm1                ; mm2 = [l0 l1 l2 l3 xx xx xx xx]

-	psrlq       mm2, 20h

-	pxor        mm0, mm2                ; mm0 = [t2 t1 t0 lt l0 l1 l2 l3]

-

-	movq        mm1, mm0

-	psrlq       mm1, 10h                ; mm1 = [xx xx t2 t1 t0 lt l0 l1]

-	movq        mm2, mm0

-	psrlq       mm2, 8h                 ; mm2 = [xx t2 t1 t0 lt l0 l1 l2]

-	movq        mm3, mm2

-	movq        mm4, mm1

-	pavgb       mm1, mm0

-

-	pxor        mm4, mm0				; find odd value in the lowest bit of each byte

-	pand        mm4, [mmx_01bytes]	    ; set the odd bit

-	psubusb     mm1, mm4				; decrease 1 from odd bytes

-

-	pavgb       mm2, mm1                ; mm2 = [xx xx d  c  b  f  h  j]

-

-	movq        mm4, mm0

-	pavgb       mm3, mm4                ; mm3 = [xx xx xx xx a  e  g  i]

-	punpcklbw   mm3, mm2                ; mm3 = [b  a  f  e  h  g  j  i]

-

-	psrlq       mm2, 20h

-	psllq       mm2, 30h                ; mm2 = [d  c  0  0  0  0  0  0]

-	movq        mm4, mm3

-	psrlq       mm4, 10h                ; mm4 = [0  0  b  a  f  e  h  j]

-	pxor        mm2, mm4                ; mm2 = [d  c  b  a  xx xx xx xx]

-	psrlq       mm2, 20h                ; mm2 = [xx xx xx xx  d  c  b  a]

-

-	movd        [r0], mm2

-	movd        [r0+12], mm3

-	psrlq       mm3, 10h

-	movd        [r0+8], mm3

-	psrlq       mm3, 10h

-	movd        [r0+4], mm3

-	WELSEMMS

-	ret

-

-ALIGN 16

-;***********************************************************************

-;	lt|t0|t1|t2|t3|

-;	l0|

-;	l1|

-;	l2|

-;	l3|

-;	t3 will never been used

-;   destination:

-;	|a |b |c |d |

-;	|c |d |e |f |

-;	|e |f |g |g |

-;	|g |g |g |g |

-

-;   a = (1 + l0 + l1)>>1

-;   c = (1 + l1 + l2)>>1

-;   e = (1 + l2 + l3)>>1

-;   g = l3

-

-;   b = (2 + l0 + (l1<<1) + l2)>>2

-;   d = (2 + l1 + (l2<<1) + l3)>>2

-;   f = (2 + l2 + (l3<<1) + l3)>>2

-

-;   [g g f e d c b a] + [g g g g] --> mov to memory

-;

-;   void WelsI4x4LumaPredHU_mmx(uint8_t *pred,uint8_t *pRef,int32_t stride)

-;***********************************************************************

-WELS_EXTERN WelsI4x4LumaPredHU_mmx

-WelsI4x4LumaPredHU_mmx:

-	%assign push_num 0

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	movd        mm0, [r1-4]            ; mm0[3] = l0

-	punpcklbw   mm0, [r1+r2-4]        ; mm0[7] = l1, mm0[6] = l0

-	lea         r1, [r1+2*r2]

-	movd        mm2, [r1-4]            ; mm2[3] = l2

-	movd        mm4, [r1+r2-4]        ; mm4[3] = l3

-	punpcklbw   mm2, mm4

-	punpckhwd   mm0, mm2                ; mm0 = [l3 l2 l1 l0 xx xx xx xx]

-

-	psrlq       mm4, 18h

-	psllq       mm4, 38h                ; mm4 = [l3 xx xx xx xx xx xx xx]

-	psrlq       mm0, 8h

-	pxor        mm0, mm4                ; mm0 = [l3 l3 l2 l1 l0 xx xx xx]

-

-	movq        mm1, mm0

-	psllq       mm1, 8h                 ; mm1 = [l3 l2 l1 l0 xx xx xx xx]

-	movq        mm3, mm1                ; mm3 = [l3 l2 l1 l0 xx xx xx xx]

-	pavgb       mm1, mm0                ; mm1 = [g  e  c  a  xx xx xx xx]

-

-	movq        mm2, mm0

-	psllq       mm2, 10h                ; mm2 = [l2 l1 l0 xx xx xx xx xx]

-	movq        mm5, mm2

-	pavgb       mm2, mm0

-

-	pxor        mm5, mm0				; find odd value in the lowest bit of each byte

-	pand        mm5, [mmx_01bytes]	    ; set the odd bit

-	psubusb     mm2, mm5				; decrease 1 from odd bytes

-

-	pavgb       mm2, mm3                ; mm2 = [f  d  b  xx xx xx xx xx]

-

-	psrlq       mm2, 8h

-	pxor        mm2, mm4                ; mm2 = [g  f  d  b  xx xx xx xx]

-

-	punpckhbw   mm1, mm2                ; mm1 = [g  g  f  e  d  c  b  a]

-	punpckhbw   mm4, mm4                ; mm4 = [g  g  xx xx xx xx xx xx]

-	punpckhbw   mm4, mm4                ; mm4 = [g  g  g  g  xx xx xx xx]

-

-	psrlq       mm4, 20h

-	movd        [r0+12], mm4

-

-	movd        [r0], mm1

-	psrlq       mm1, 10h

-	movd        [r0+4], mm1

-	psrlq       mm1, 10h

-	movd        [r0+8], mm1

-	WELSEMMS

-	ret

-

-

-

-ALIGN 16

-;***********************************************************************

-;	lt|t0|t1|t2|t3|

-;	l0|

-;	l1|

-;	l2|

-;	l3|

-;	l3 will never been used

-;   destination:

-;	|a |b |c |d |

-;	|e |f |g |h |

-;	|i |a |b |c |

-;	|j |e |f |g |

-

-;   a = (1 + lt + t0)>>1

-;   b = (1 + t0 + t1)>>1

-;   c = (1 + t1 + t2)>>1

-;   d = (1 + t2 + t3)>>1

-

-;   e = (2 + l0 + (lt<<1) + t0)>>2

-;   f = (2 + lt + (t0<<1) + t1)>>2

-;   g = (2 + t0 + (t1<<1) + t2)>>2

-

-;   h = (2 + t1 + (t2<<1) + t3)>>2

-;   i = (2 + lt + (l0<<1) + l1)>>2

-;   j = (2 + l0 + (l1<<1) + l2)>>2

-;

-;   void WelsI4x4LumaPredVR_mmx(uint8_t *pred,uint8_t *pRef,int32_t stride)

-;***********************************************************************

-WELS_EXTERN WelsI4x4LumaPredVR_mmx

-WelsI4x4LumaPredVR_mmx:

-	%assign push_num 0

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	sub         r1, r2

-	movq        mm0, [r1-1]            ; mm0 = [xx xx xx t3 t2 t1 t0 lt]

-	psllq       mm0, 18h                ; mm0 = [t3 t2 t1 t0 lt xx xx xx]

-

-	movd        mm1, [r1+2*r2-4]

-	punpcklbw   mm1, [r1+r2-4]        ; mm1[7] = l0, mm1[6] = l1

-	lea         r1, [r1+2*r2]

-	movq        mm2, [r1+r2-8]        ; mm2[7] = l2

-	punpckhwd   mm2, mm1                ; mm2 = [l0 l1 l2 xx xx xx xx xx]

-	psrlq       mm2, 28h

-	pxor        mm0, mm2                ; mm0 = [t3 t2 t1 t0 lt l0 l1 l2]

-

-	movq        mm1, mm0

-	psllq       mm1, 8h                 ; mm1 = [t2 t1 t0 lt l0 l1 l2 xx]

-	pavgb       mm1, mm0                ; mm1 = [d  c  b  a  xx xx xx xx]

-

-	movq        mm2, mm0

-	psllq       mm2, 10h                ; mm2 = [t1 t0 lt l0 l1 l2 xx xx]

-	movq        mm3, mm2

-	pavgb       mm2, mm0

-

-	pxor        mm3, mm0				; find odd value in the lowest bit of each byte

-	pand        mm3, [mmx_01bytes]	    ; set the odd bit

-	psubusb     mm2, mm3				; decrease 1 from odd bytes

-

-	movq        mm3, mm0

-	psllq       mm3, 8h                 ; mm3 = [t2 t1 t0 lt l0 l1 l2 xx]

-	pavgb       mm3, mm2                ; mm3 = [h  g  f  e  i  j  xx xx]

-	movq        mm2, mm3

-

-	psrlq       mm1, 20h                ; mm1 = [xx xx xx xx d  c  b  a]

-	movd        [r0], mm1

-

-	psrlq       mm2, 20h                ; mm2 = [xx xx xx xx h  g  f  e]

-	movd        [r0+4], mm2

-

-	movq        mm4, mm3

-	psllq       mm4, 20h

-	psrlq       mm4, 38h                ; mm4 = [xx xx xx xx xx xx xx i]

-

-	movq        mm5, mm3

-	psllq       mm5, 28h

-	psrlq       mm5, 38h                ; mm5 = [xx xx xx xx xx xx xx j]

-

-	psllq       mm1, 8h

-	pxor        mm4, mm1                ; mm4 = [xx xx xx xx c  b  a  i]

-	movd        [r0+8], mm4

-

-	psllq       mm2, 8h

-	pxor        mm5, mm2                ; mm5 = [xx xx xx xx g  f  e  j]

-	movd        [r0+12], mm5

-	WELSEMMS

-	ret

-

-ALIGN 16

-;***********************************************************************

-;	lt|t0|t1|t2|t3|t4|t5|t6|t7

-;	l0|

-;	l1|

-;	l2|

-;	l3|

-;	lt,t0,t1,t2,t3 will never been used

-;   destination:

-;	|a |b |c |d |

-;	|b |c |d |e |

-;	|c |d |e |f |

-;	|d |e |f |g |

-

-;   a = (2 + t0 + t2 + (t1<<1))>>2

-;   b = (2 + t1 + t3 + (t2<<1))>>2

-;   c = (2 + t2 + t4 + (t3<<1))>>2

-;   d = (2 + t3 + t5 + (t4<<1))>>2

-

-;   e = (2 + t4 + t6 + (t5<<1))>>2

-;   f = (2 + t5 + t7 + (t6<<1))>>2

-;   g = (2 + t6 + t7 + (t7<<1))>>2

-

-;   [g f e d c b a] --> mov to memory

-;

-;   void WelsI4x4LumaPredDDL_mmx(uint8_t *pred,uint8_t *pRef,int32_t stride)

-;***********************************************************************

-WELS_EXTERN WelsI4x4LumaPredDDL_mmx

-WelsI4x4LumaPredDDL_mmx:

-	%assign push_num 0

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	sub         r1, r2

-	movq        mm0, [r1]              ; mm0 = [t7 t6 t5 t4 t3 t2 t1 t0]

-	movq        mm1, mm0

-	movq        mm2, mm0

-

-	movq        mm3, mm0

-	psrlq       mm3, 38h

-	psllq       mm3, 38h                ; mm3 = [t7 xx xx xx xx xx xx xx]

-

-	psllq       mm1, 8h                 ; mm1 = [t6 t5 t4 t3 t2 t1 t0 xx]

-	psrlq       mm2, 8h

-	pxor        mm2, mm3                ; mm2 = [t7 t7 t6 t5 t4 t3 t2 t1]

-

-	movq        mm3, mm1

-	pavgb       mm1, mm2

-	pxor        mm3, mm2				; find odd value in the lowest bit of each byte

-	pand        mm3, [mmx_01bytes]	    ; set the odd bit

-	psubusb     mm1, mm3				; decrease 1 from odd bytes

-

-	pavgb       mm0, mm1                ; mm0 = [g f e d c b a xx]

-

-	psrlq       mm0, 8h

-	movd        [r0], mm0

-	psrlq       mm0, 8h

-	movd        [r0+4], mm0

-	psrlq       mm0, 8h

-	movd        [r0+8], mm0

-	psrlq       mm0, 8h

-	movd        [r0+12], mm0

-	WELSEMMS

-	ret

-

-

-ALIGN 16

-;***********************************************************************

-;	lt|t0|t1|t2|t3|t4|t5|t6|t7

-;	l0|

-;	l1|

-;	l2|

-;	l3|

-;	lt,t0,t1,t2,t3 will never been used

-;   destination:

-;	|a |b |c |d |

-;	|e |f |g |h |

-;	|b |c |d |i |

-;	|f |g |h |j |

-

-;   a = (1 + t0 + t1)>>1

-;   b = (1 + t1 + t2)>>1

-;   c = (1 + t2 + t3)>>1

-;   d = (1 + t3 + t4)>>1

-;   i = (1 + t4 + t5)>>1

-

-;   e = (2 + t0 + (t1<<1) + t2)>>2

-;   f = (2 + t1 + (t2<<1) + t3)>>2

-;   g = (2 + t2 + (t3<<1) + t4)>>2

-;   h = (2 + t3 + (t4<<1) + t5)>>2

-;   j = (2 + t4 + (t5<<1) + t6)>>2

-

-;   [i d c b a] + [j h g f e] --> mov to memory

-;

-;   void WelsI4x4LumaPredVL_mmx(uint8_t *pred,uint8_t *pRef,int32_t stride)

-;***********************************************************************

-WELS_EXTERN WelsI4x4LumaPredVL_mmx

-WelsI4x4LumaPredVL_mmx:

-	%assign push_num 0

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	sub         r1, r2

-	movq        mm0, [r1]              ; mm0 = [t7 t6 t5 t4 t3 t2 t1 t0]

-	movq        mm1, mm0

-	movq        mm2, mm0

-

-	psrlq       mm1, 8h                 ; mm1 = [xx t7 t6 t5 t4 t3 t2 t1]

-	psrlq       mm2, 10h                ; mm2 = [xx xx t7 t6 t5 t4 t3 t2]

-

-	movq        mm3, mm1

-	pavgb       mm3, mm0                ; mm3 = [xx xx xx i  d  c  b  a]

-

-	movq        mm4, mm2

-	pavgb       mm2, mm0

-	pxor        mm4, mm0				; find odd value in the lowest bit of each byte

-	pand        mm4, [mmx_01bytes]	    ; set the odd bit

-	psubusb     mm2, mm4				; decrease 1 from odd bytes

-

-	pavgb       mm2, mm1                ; mm2 = [xx xx xx j  h  g  f  e]

-

-	movd        [r0], mm3

-	psrlq       mm3, 8h

-	movd        [r0+8], mm3

-

-	movd        [r0+4], mm2

-	psrlq       mm2, 8h

-	movd        [r0+12], mm2

-	WELSEMMS

-	ret

-

-ALIGN 16

-;***********************************************************************

-;

-;   void WelsIChromaPredDc_sse2(uint8_t *pred, uint8_t *pRef, int32_t stride)

-;***********************************************************************

-WELS_EXTERN WelsIChromaPredDc_sse2

-WelsIChromaPredDc_sse2:

-	push r3

-	push r4

-	%assign push_num 2

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	sub         r1, r2

-	movq        mm0, [r1]

-

-	movzx		r3, byte [r1+r2-0x01] ; l1

-	lea         	r1, [r1+2*r2]

-	movzx		r4, byte [r1-0x01]     ; l2

-	add		r3, r4

-	movzx		r4, byte [r1+r2-0x01] ; l3

-	add		r3, r4

-	lea         	r1, [r1+2*r2]

-	movzx		r4, byte [r1-0x01]     ; l4

-	add		r3, r4

-	movd        	mm1, r3d                 ; mm1 = l1+l2+l3+l4

-

-	movzx		r3, byte [r1+r2-0x01] ; l5

-	lea         	r1, [r1+2*r2]

-	movzx		r4, byte [r1-0x01]     ; l6

-	add		r3, r4

-	movzx		r4, byte [r1+r2-0x01] ; l7

-	add		r3, r4

-	lea         	r1, [r1+2*r2]

-	movzx		r4, byte [r1-0x01]     ; l8

-	add		r3, r4

-	movd        	mm2, r3d                 ; mm2 = l5+l6+l7+l8

-

-	movq        mm3, mm0

-	psrlq       mm0, 0x20

-	psllq       mm3, 0x20

-	psrlq       mm3, 0x20

-	pxor		mm4, mm4

-	psadbw		mm0, mm4

-	psadbw		mm3, mm4                 ; sum1 = mm3+mm1, sum2 = mm0, sum3 = mm2

-

-	paddq       mm3, mm1

-	movq        mm1, mm2

-	paddq       mm1, mm0;                ; sum1 = mm3, sum2 = mm0, sum3 = mm2, sum4 = mm1

-

-	movq        mm4, [mmx_0x02]

-

-	paddq       mm0, mm4

-	psrlq       mm0, 0x02

-

-	paddq       mm2, mm4

-	psrlq       mm2, 0x02

-

-	paddq       mm3, mm4

-	paddq       mm3, mm4

-	psrlq       mm3, 0x03

-

-	paddq       mm1, mm4

-	paddq       mm1, mm4

-	psrlq       mm1, 0x03

-

-	pmuludq     mm0, [mmx_01bytes]

-	pmuludq     mm3, [mmx_01bytes]

-	psllq       mm0, 0x20

-	pxor        mm0, mm3                 ; mm0 = m_up

-

-	pmuludq     mm2, [mmx_01bytes]

-	pmuludq     mm1, [mmx_01bytes]

-	psllq       mm1, 0x20

-	pxor        mm1, mm2                 ; mm2 = m_down

-

-	movq        [r0], mm0

-	movq        [r0+0x08], mm0

-	movq        [r0+0x10], mm0

-	movq        [r0+0x18], mm0

-

-	movq        [r0+0x20], mm1

-	movq        [r0+0x28], mm1

-	movq        [r0+0x30], mm1

-	movq        [r0+0x38], mm1

-

-	pop r4

-	pop r3

-	WELSEMMS

-	ret

-

-

-

-ALIGN 16

-;***********************************************************************

-;

-;   void WelsI16x16LumaPredDc_sse2(uint8_t *pred, uint8_t *pRef, int32_t stride)

-;***********************************************************************

-WELS_EXTERN WelsI16x16LumaPredDc_sse2

-WelsI16x16LumaPredDc_sse2:

-	push r3

-	push r4

-	%assign push_num 2

-	LOAD_3_PARA

-	%ifndef X86_32

-	movsx r2, r2d

-	%endif

-	sub         r1, r2

-	movdqa      xmm0, [r1]             ; read one row

-	pxor		xmm1, xmm1

-	psadbw		xmm0, xmm1

-	movdqa      xmm1, xmm0

-	psrldq      xmm1, 0x08

-	pslldq      xmm0, 0x08

-	psrldq      xmm0, 0x08

-	paddw       xmm0, xmm1

-

-	movzx		r3, byte [r1+r2-0x01]

-	movzx		r4, byte [r1+2*r2-0x01]

-	add		r3, r4

-	lea         r1, [r1+r2]

-	LOAD_2_LEFT_AND_ADD

-	LOAD_2_LEFT_AND_ADD

-	LOAD_2_LEFT_AND_ADD

-	LOAD_2_LEFT_AND_ADD

-	LOAD_2_LEFT_AND_ADD

-	LOAD_2_LEFT_AND_ADD

-	LOAD_2_LEFT_AND_ADD

-	add         r3, 0x10

-	movd        xmm1, r3d

-	paddw       xmm0, xmm1

-	psrld       xmm0, 0x05

-	pmuludq     xmm0, [mmx_01bytes]

-	pshufd      xmm0, xmm0, 0

-

-	movdqa      [r0], xmm0

-	movdqa      [r0+0x10], xmm0

-	movdqa      [r0+0x20], xmm0

-	movdqa      [r0+0x30], xmm0

-	movdqa      [r0+0x40], xmm0

-	movdqa      [r0+0x50], xmm0

-	movdqa      [r0+0x60], xmm0

-	movdqa      [r0+0x70], xmm0

-	movdqa      [r0+0x80], xmm0

-	movdqa      [r0+0x90], xmm0

-	movdqa      [r0+0xa0], xmm0

-	movdqa      [r0+0xb0], xmm0

-	movdqa      [r0+0xc0], xmm0

-	movdqa      [r0+0xd0], xmm0

-	movdqa      [r0+0xe0], xmm0

-	movdqa      [r0+0xf0], xmm0

-

-	pop r4

-	pop r3

-	ret

-

-;***********************************************************************

-;

-;int32_t WelsSmpleSatdThree4x4_sse2( uint8_t *pDec, int32_t iLineSizeDec, uint8_t *pEnc, int32_t iLinesizeEnc,

-;                             uint8_t* pRed, int32_t* pBestMode, int32_t, int32_t, int32_t);

-;

-;***********************************************************************

-%ifdef X86_ASM

-WELS_EXTERN WelsSmpleSatdThree4x4_sse2

-align 16

-WelsSmpleSatdThree4x4_sse2:

-	push      ebx

-	push      esi

-	push      edi

-	mov       eax,  [esp+24];p_enc

-	mov       ebx,  [esp+28];linesize_enc

-

-	; load source 4x4 samples and Hadamard transform

-    movd      xmm0, [eax]

-    movd      xmm1, [eax+ebx]

-    lea       eax , [eax+2*ebx]

-    movd      xmm2, [eax]

-    movd      xmm3, [eax+ebx]

-    punpckldq xmm0, xmm2

-    punpckldq xmm1, xmm3

-

-    pxor      xmm6, xmm6

-    punpcklbw xmm0, xmm6

-    punpcklbw xmm1, xmm6

-

-    movdqa    xmm2, xmm0

-    paddw     xmm0, xmm1

-    psubw     xmm2, xmm1

-    SSE2_XSawp  qdq, xmm0, xmm2, xmm3

-

-    movdqa    xmm4, xmm0

-    paddw     xmm0, xmm3

-    psubw     xmm4, xmm3

-

-    movdqa    xmm2, xmm0

-    punpcklwd xmm0, xmm4

-    punpckhwd xmm4, xmm2

-

-	SSE2_XSawp  dq,  xmm0, xmm4, xmm3

-	SSE2_XSawp  qdq, xmm0, xmm3, xmm5

-

-    movdqa    xmm7, xmm0

-    paddw     xmm0, xmm5

-    psubw     xmm7, xmm5

-

-	SSE2_XSawp  qdq,  xmm0, xmm7, xmm1

-

-    ; Hadamard transform results are saved in xmm0 and xmm2

-    movdqa    xmm2, xmm0

-    paddw     xmm0, xmm1

-    psubw     xmm2, xmm1

-

-	; load top boundary samples: [a b c d]

-    mov       eax,  [esp+16];p_dec

-	sub		  eax,	[esp+20];linesize_dec

-	movzx     ecx,  byte [eax]

-	movzx     edx,  byte [eax+1]

-	movzx     esi,  byte [eax+2]

-	movzx     edi,  byte [eax+3]

-

-	; get the transform results of top boundary samples: [a b c d]

-	add       edx, ecx ; edx = a + b

-	add       edi, esi ; edi = c + d

-	add       ecx, ecx ; ecx = a + a

-	add       esi, esi ; esi = c + c

-	sub       ecx, edx ; ecx = a + a - a - b = a - b

-	sub       esi, edi ; esi = c + c - c - d = c - d

-	add       edi, edx ; edi = (a + b) + (c + d)

-	add       edx, edx

-	sub       edx, edi ; edx = (a + b) - (c + d)

-	add       esi, ecx ; esi = (a - b) + (c - d)

-	add       ecx, ecx

-	sub       ecx, esi ; ecx = (a - b) - (c - d) ; [edi edx ecx esi]

-

-	movdqa    xmm6, xmm0

-	movdqa    xmm7, xmm2

-	movd      xmm5, edi ; store the edi for DC mode

-	pxor      xmm3, xmm3

-	pxor      xmm4, xmm4

-	pinsrw    xmm3, edi, 0

-	pinsrw    xmm3, esi, 4

-	psllw     xmm3, 2

-	pinsrw    xmm4, edx, 0

-	pinsrw    xmm4, ecx, 4

-	psllw     xmm4, 2

-

-	; get the satd of H

-	psubw     xmm0, xmm3

-	psubw     xmm2, xmm4

-

-	WELS_AbsW  xmm0, xmm1

-	WELS_AbsW  xmm2, xmm1

-    paddusw        xmm0, xmm2

-    SUMW_HORIZON1  xmm0, xmm1 ; satd of V is stored in xmm0

-

-	; load left boundary samples: [a b c d]'

-    mov       eax,  [esp+16]

-	mov       ebx,  [esp+20]

-	movzx     ecx,  byte [eax-1]

-	movzx     edx,  byte [eax+ebx-1]

-	lea       eax , [eax+2*ebx]

-	movzx     esi,  byte [eax-1]

-	movzx     edi,  byte [eax+ebx-1]

-

-	; get the transform results of left boundary samples: [a b c d]'

-	add       edx, ecx ; edx = a + b

-	add       edi, esi ; edi = c + d

-	add       ecx, ecx ; ecx = a + a

-	add       esi, esi ; esi = c + c

-	sub       ecx, edx ; ecx = a + a - a - b = a - b

-	sub       esi, edi ; esi = c + c - c - d = c - d

-	add       edi, edx ; edi = (a + b) + (c + d)

-	add       edx, edx

-	sub       edx, edi ; edx = (a + b) - (c + d)

-	add       esi, ecx ; esi = (a - b) + (c - d)

-	add       ecx, ecx

-	sub       ecx, esi ; ecx = (a - b) - (c - d) ; [edi edx ecx esi]'

-

-	; store the transform results in xmm3

-    movd      xmm3, edi

-	pinsrw    xmm3, edx, 1

-	pinsrw    xmm3, ecx, 2

-	pinsrw    xmm3, esi, 3

-	psllw     xmm3, 2

-

-	; get the satd of V

-	movdqa    xmm2, xmm6

-	movdqa    xmm4, xmm7

-	psubw     xmm2, xmm3

-	WELS_AbsW  xmm2, xmm1

-	WELS_AbsW  xmm4, xmm1

-    paddusw        xmm2, xmm4

-    SUMW_HORIZON1  xmm2, xmm1 ; satd of H is stored in xmm2

-

-	; DC result is stored in xmm1

-	add       edi, 4

-	movd      xmm1, edi

-	paddw     xmm1, xmm5

-	psrlw     xmm1, 3

-	movdqa    xmm5, xmm1

-	psllw     xmm1, 4

-

-    ; get the satd of DC

-    psubw          xmm6, xmm1

-    WELS_AbsW  xmm6, xmm1

-	WELS_AbsW  xmm7, xmm1

-    paddusw        xmm6, xmm7

-    SUMW_HORIZON1  xmm6, xmm1 ; satd of DC is stored in xmm6

-

-    ; comparing order: DC H V

-    mov       edx, [esp+32]

-    movd      eax, xmm6

-    movd      edi, xmm2

-    movd      esi, xmm0

-    and       eax, 0xffff

-    shr       eax, 1

-    and       edi, 0xffff

-    shr       edi, 1

-    and       esi, 0xffff

-    shr       esi, 1

-    add       eax, [esp+40]

-    add       edi, [esp+44]

-    add       esi, [esp+48]

-    cmp       ax, di

-    jg near   not_dc

-    cmp       ax, si

-    jg near   not_dc_h

-

-    ; for DC mode

-    movd      ebx, xmm5

-    imul      ebx, 0x01010101

-    movd	  xmm5, ebx

-	pshufd    xmm5, xmm5, 0

-	movdqa    [edx], xmm5

-	mov       ebx, [esp+36]

-	mov       dword [ebx], 0x02

-	pop       edi

-    pop       esi

-    pop       ebx

-    ret

-

-not_dc:

-    cmp       di, si

-    jg near   not_dc_h

-

-    ; for H mode

-    SSE_DB_1_2REG  xmm6, xmm7

-    mov       eax,  [esp+16]

-	mov       ebx,  [esp+20]

-    movzx     ecx,  byte [eax-1]

-	movd      xmm0, ecx

-    pmuludq   xmm0, xmm6

-

-	movzx     ecx,  byte [eax+ebx-1]

-	movd      xmm1, ecx

-    pmuludq   xmm1, xmm6

-%if 1

-    punpckldq xmm0, xmm1

-%else

-	unpcklps  xmm0,	xmm1

-%endif

-	lea       eax,	[eax+ebx*2]

-	movzx	  ecx,	byte [eax-1]

-	movd	  xmm2,	ecx

-    pmuludq   xmm2, xmm6

-

-	movzx	  ecx,	byte [eax+ebx-1]

-	movd	  xmm3,	ecx

-    pmuludq   xmm3, xmm6

-%if 1

-    punpckldq  xmm2, xmm3

-    punpcklqdq xmm0, xmm2

-%else

-	unpcklps  xmm2,	xmm3

-	unpcklpd  xmm0,	xmm2

-%endif

-	movdqa	  [edx],xmm0

-

-	mov       eax, edi

-    mov       ebx, [esp+36]

-	mov       dword [ebx], 0x01

-

-    pop       edi

-    pop       esi

-    pop       ebx

-    ret

-not_dc_h:

-    ; for V mode

-    mov       eax,  [esp+16]

-    sub		  eax,	[esp+20]

-	movd	  xmm0,	[eax]

-	pshufd	  xmm0,	xmm0, 0

-	movdqa	  [edx],xmm0

-

-	mov       eax, esi

-    mov       ebx, [esp+36]

-	mov       dword [ebx], 0x00

-

-    pop       edi

-    pop       esi

-    pop       ebx

-    ret

-%endif

-

--- a/codec/processing/src/asm/sad.asm

+++ /dev/null

@@ -1,220 +1,0 @@

-;*!

-;* \copy

-;*     Copyright (c)  2009-2013, Cisco Systems

-;*     All rights reserved.

-;*

-;*     Redistribution and use in source and binary forms, with or without

-;*     modification, are permitted provided that the following conditions

-;*     are met:

-;*

-;*        * Redistributions of source code must retain the above copyright

-;*          notice, this list of conditions and the following disclaimer.

-;*

-;*        * Redistributions in binary form must reproduce the above copyright

-;*          notice, this list of conditions and the following disclaimer in

-;*          the documentation and/or other materials provided with the

-;*          distribution.

-;*

-;*     THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS

-;*     "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT

-;*     LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS

-;*     FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE

-;*     COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,

-;*     INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,

-;*     BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;

-;*     LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER

-;*     CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT

-;*     LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN

-;*     ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE

-;*     POSSIBILITY OF SUCH DAMAGE.

-;*

-;*

-;*  sad.asm

-;*

-;*  Abstract

-;*      WelsSampleSad8x8_sse21

-;*

-;*  History

-;*      8/5/2009 Created

-;*

-;*

-;*************************************************************************/

-

-%include "asm_inc.asm"

-

-;***********************************************************************

-; Macros and other preprocessor constants

-;***********************************************************************

-%macro CACHE_SPLIT_CHECK 3 ; address, width, cacheline

-and    %1,  0x1f|(%3>>1)

-cmp    %1,  (32-%2)|(%3>>1)

-%endmacro

-

-%macro SSE2_GetSad8x4 0

-	movq   xmm0,   [r0]

-	movq   xmm1,   [r0+r1]

-	lea    r0,     [r0+2*r1]

-	movhps xmm0,   [r0]

-	movhps xmm1,   [r0+r1]

-

-	movq   xmm2,   [r2]

-	movq   xmm3,   [r2+r3]

-	lea    r2,     [r2+2*r3]

-	movhps xmm2,   [r2]

-	movhps xmm3,   [r2+r3]

-	psadbw xmm0,   xmm2

-	psadbw xmm1,   xmm3

-	paddw  xmm6,   xmm0

-	paddw  xmm6,   xmm1

-%endmacro

-

-

-;***********************************************************************

-; Code

-;***********************************************************************

-SECTION .text

-

-WELS_EXTERN WelsSampleSad8x8_sse21

-WelsSampleSad8x8_sse21:

-    ;mov    ecx,    [esp+12]

-	;mov    edx,    ecx

-    ;CACHE_SPLIT_CHECK edx, 8, 64

-	;jle    near   .pixel_sad_8x8_nsplit

-	;push   ebx

-	;push   edi

-	;mov    eax,    [esp+12]

-	;mov    ebx,    [esp+16]

-

-	%assign  push_num 0

-	mov		r2,  arg3

-	push	r2

-	CACHE_SPLIT_CHECK r2, 8, 64

-	jle    near   .pixel_sad_8x8_nsplit

-	pop		r2

-%ifdef X86_32

-	push	r3

-	push	r4

-	push	r5

-%endif

-	%assign  push_num 3

-	mov		r0,  arg1

-	mov		r1,  arg2

-	SIGN_EXTENTION r1, r1d

-    pxor   xmm7,   xmm7

-

-    ;ecx r2, edx r4, edi r5

-

-    mov    r5,    r2

-    and    r5,    0x07

-    sub    r2,    r5

-    mov    r4,    8

-    sub    r4,    r5

-

-    shl    r5,    3

-    shl    r4,    3

-    movd   xmm5,   r5d

-    movd   xmm6,   r4d

-	mov    r5,    8

-	add    r5,    r2

-    mov    r3,    arg4

-	SIGN_EXTENTION r3, r3d

-    movq   xmm0,   [r0]

-	movhps xmm0,   [r0+r1]

-

-	movq   xmm1,   [r2]

-	movq   xmm2,   [r5]

-	movhps xmm1,   [r2+r3]

-	movhps xmm2,   [r5+r3]

-	psrlq  xmm1,   xmm5

-	psllq  xmm2,   xmm6

-	por    xmm1,   xmm2

-

-	psadbw xmm0,   xmm1

-	paddw  xmm7,   xmm0

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	lea    r5,    [r5+2*r3]

-

-    movq   xmm0,   [r0]

-	movhps xmm0,   [r0+r1]

-

-	movq   xmm1,   [r2]

-	movq   xmm2,   [r5]

-	movhps xmm1,   [r2+r3]

-	movhps xmm2,   [r5+r3]

-	psrlq  xmm1,   xmm5

-	psllq  xmm2,   xmm6

-	por    xmm1,   xmm2

-

-	psadbw xmm0,   xmm1

-	paddw  xmm7,   xmm0

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	lea    r5,    [r5+2*r3]

-

-    movq   xmm0,   [r0]

-	movhps xmm0,   [r0+r1]

-

-	movq   xmm1,   [r2]

-	movq   xmm2,   [r5]

-	movhps xmm1,   [r2+r3]

-	movhps xmm2,   [r5+r3]

-	psrlq  xmm1,   xmm5

-	psllq  xmm2,   xmm6

-	por    xmm1,   xmm2

-

-	psadbw xmm0,   xmm1

-	paddw  xmm7,   xmm0

-

-	lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-	lea    r5,    [r5+2*r3]

-

-    movq   xmm0,   [r0]

-	movhps xmm0,   [r0+r1]

-

-	movq   xmm1,   [r2]

-	movq   xmm2,   [r5]

-	movhps xmm1,   [r2+r3]

-	movhps xmm2,   [r5+r3]

-	psrlq  xmm1,   xmm5

-	psllq  xmm2,   xmm6

-	por    xmm1,   xmm2

-

-	psadbw xmm0,   xmm1

-	paddw  xmm7,   xmm0

-

-    movhlps    xmm0, xmm7

-	paddw      xmm0, xmm7

-	movd       retrd,  xmm0

-%ifdef X86_32

-	pop	 r5

-	pop	 r4

-	pop	 r3

-%endif

-	jmp        .return

-

-.pixel_sad_8x8_nsplit:

-    ;push   ebx

-    ;mov    eax,    [esp+8]

-	;mov    ebx,    [esp+12]

-	;mov    edx,    [esp+20]

-

-	pop r2

-	%assign  push_num 0

-	LOAD_4_PARA

-	SIGN_EXTENTION r1, r1d

-	SIGN_EXTENTION r3, r3d

-	pxor   xmm6,   xmm6

-	SSE2_GetSad8x4

-    lea    r0,    [r0+2*r1]

-	lea    r2,    [r2+2*r3]

-    SSE2_GetSad8x4

-    movhlps    xmm0, xmm6

-	paddw      xmm0, xmm6

-	movd       retrd,  xmm0

-	LOAD_4_PARA_POP

-.return:

-	ret

\ No newline at end of file

--- a/codec/processing/targets.mk

+++ b/codec/processing/targets.mk

@@ -23,8 +23,6 @@

 PROCESSING_ASM_SRCS=\

 	$(PROCESSING_SRCDIR)/./src/asm/denoisefilter.asm\

 	$(PROCESSING_SRCDIR)/./src/asm/downsample_bilinear.asm\

-	$(PROCESSING_SRCDIR)/./src/asm/intra_pred.asm\

-	$(PROCESSING_SRCDIR)/./src/asm/sad.asm\

 	$(PROCESSING_SRCDIR)/./src/asm/vaa.asm\

 PROCESSING_OBJS += $(PROCESSING_ASM_SRCS:.asm=.o)