Subversion Repositories Kolibri OS

/*

 *  Copyright 2000-2011 Intel Corporation All Rights Reserved

 *

 *  Licensed under the Apache License, Version 2.0 (the "License");

 *  you may not use this file except in compliance with the License.

 *  You may obtain a copy of the License at

 *

 *      http://www.apache.org/licenses/LICENSE-2.0

 *

 *  Unless required by applicable law or agreed to in writing, software

 *  distributed under the License is distributed on an "AS IS" BASIS,

 *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

 *  See the License for the specific language governing permissions and

 *  limitations under the License.

 */

//  114    // Total instruction count

//    1    // Total kernel count

.kernel PA_DN_422CP

.code

// FileName:	DN_PA_Core.asm

// Author:		Vivek Kumar

// Description:	Tasks for DN only case (16x8 block) for Packed format

// FileName:	DN.asm

// Author:		Vivek Kumar

// Description:	Tasks for DN only case (16x8 block)

// Module name: common.inc

//

// Common header file for all Video-Processing kernels

//

.default_execution_size (16)

.default_register_type  :ub

.reg_count_total        128

.reg_count_payload      7

//========== Common constants ==========

//========== Macros ==========

//Fast Jump, For more details see "Set_Layer_N.asm"

//========== Defines ====================

//========== Static Parameters (Common To All) ==========

//r1

//r2

                                    //  e.g.            byte0   byte1  byte2

                                    // YUYV               0       1      3

                                    // YVYU               0       3      1

//Color Pipe (IECP) parameters

//ByteCopy

//r4

                                    //  e.g.              byte0           byte1           byte2

                                    // YUYV                 0               1               3

                                    // YVYU                 0               3               1

//========== Inline parameters (Common To All) ===========

//============== Binding Index Table===========

//Common between DNDI and DNUV

//================= Common Message Descriptor =====

// Message descriptor for thread spawning

// Message Descriptors

//                = 000 0001 (min message len 1 ) 0,0000 (resp len 0   -add later)

//                  0000,0000,0000

//                  0001(Spawn a root thread),0001 (Root thread spawn thread)

//                = 0x02000011

// Thread Spawner Message Descriptor

// Message descriptor for atomic operation add

// Message Descriptors

//                = 000 0110 (min message len 6 ) 0,0000 (resp len 0   -add later)

//                  1(header present)001,10(typed atomic operation)0(return enabled)0(slot group, low 8 bits),0111 (AOP_Add)

//                  0000,0000 (Binding table index, added later)

//                = 0x02000011

// Atomic Operation Add Message Descriptor

// Message descriptor for dataport media write

        // Message Descriptors

                //                = 000 0001 (min message len 1 - add later) 00000 (resp len 0)

                //                  1 (header present 1) 0 1010 (media block write) 000000

                //                  00000000 (binding table index - set later)

                //                = 0x020A8000

// Message Length defines

// Response Length defines

// Block Width and Height Size defines

// Extended Message Descriptors

// Common message descriptors:

//===================== Math Function Control ===================================

//============ Message Registers ===============

                             // buf4 starts from r28

//#define mMSGHDR_EOT  r43    // Dummy Message Register for EOT

.declare    mubMSGPAYLOAD  Base=r30 ElementSize=1 SrcRegion=<16;16,1> Type=ub

.declare    muwMSGPAYLOAD  Base=r30 ElementSize=2 SrcRegion=<16;16,1> Type=uw

.declare    mudMSGPAYLOAD  Base=r30 ElementSize=4 SrcRegion=<8;8,1> Type=ud

.declare    mfMSGPAYLOAD   Base=r30 ElementSize=4 SrcRegion=<8;8,1> Type=f

//=================== End of thread instruction ===========================

//=====================Pointers Used=====================================

//=======================================================================

//r9-r17

// Define temp space for any usages

// Common Buffers

// temp space for rotation

.declare fROBUF		  Base=r9.0		ElementSize=4		SrcRegion=<8;8,1>		  DstRegion=<1>		Type=f

.declare udROBUF		Base=r9.0		ElementSize=4		SrcRegion=<8;8,1>		  DstRegion=<1>		Type=ud

.declare uwROBUF		Base=r9.0		ElementSize=2		SrcRegion=<16;16,1>		DstRegion=<1>		Type=uw

.declare ubROBUF		Base=r9.0		ElementSize=1		SrcRegion=<16;16,1>		DstRegion=<1>		Type=ub

.declare ub4ROBUF 	Base=r9.0		ElementSize=1		SrcRegion=<32;8,4>		DstRegion=<4>		Type=ub

// End of common.inc

// FileName:    DNDI.inc

// Author:      Vivek Kumar

// Description: Include file for DN, DI and DNDI

// Inputs:      DI_ENABLE, DN_ENABLE, DN_PLANAR, DN_PACKED

// End of common.inc

//Interface:

//Static Parameters:

//r1

//====================== Binding table (Explicit To DNDI)=========================================

.declare mudMSGHDR_DNDI     Base=r18      ElementSize=4    Type=ud

.declare mdMSGHDR_DNDI      Base=r18      ElementSize=4    Type=d

.declare mwMSGHDR_DNDI      Base=r18      ElementSize=2    Type=w

.declare mudMSGHDR_STMM     Base=r20      ElementSize=4    Type=ud

.declare mudMSGHDR_HIST     Base=r22      ElementSize=4    Type=ud

.declare mudMSGHDR_ENC_STATS Base=r24 ElementSize=4   Type=ud

.declare muwMSGHDR_ENC_STATS Base=r24 ElementSize=2   Type=uw

.declare mubMSGHDR_ENC_STATS Base=r24 ElementSize=1   Type=ub

.declare mudMSGHDR_DN_OUT   Base=r31.0      ElementSize=4  Type=ud

.declare mdMSGHDR_DN_OUT    Base=r31.0      ElementSize=4  Type=d

.declare mubMSGHDR_DN_OUT   Base=r31.0      ElementSize=1  Type=ub

.declare mudMSGHDR_UVCOPY   Base=r36      ElementSize=4  Type=ud

.declare mdMSGHDR_UVCOPY    Base=r36      ElementSize=4  Type=d

.declare mudMSGHDR_UCOPY    Base=r36       ElementSize=4  Type=ud

.declare mudMSGHDR_VCOPY    Base=r38       ElementSize=4  Type=ud

.declare mudMSGHDR_DI_OUT1  Base=r18.0      ElementSize=4     Type=ud

.declare mubMSGHDR_DI_OUT1  Base=r18.0      ElementSize=1     Type=ub

.declare mudMSGHDR_DI_OUT2  Base=r23.0      ElementSize=4     Type=ud

.declare mubMSGHDR_DI_OUT2  Base=r23.0      ElementSize=1     Type=ub

//r45

//Use r45 as message header, so no need to "mov" the data.

.declare mudDN_Y_OUT        Base=r45.0 ElementSize=4 SrcRegion=<8;8,1>   DstRegion=<1> Type=ud

// Message response (Denoised & DI-ed pixels & statistics); Use buffer 5

.declare udDNDI_RESP        Base=r46.0 ElementSize=4 SrcRegion=<8;8,1>   DstRegion=<1> Type=ud

.declare uwDNDI_RESP        Base=r46.0 ElementSize=2 SrcRegion=<16;16,1> DstRegion=<1> Type=uw

.declare ubDNDI_RESP        Base=r46.0 ElementSize=1 SrcRegion=<16;16,1> DstRegion=<1> Type=ub

// Message response (UV Copy); Use buffer 5

.declare udDNDI_UV_RESP     Base=r58.0 ElementSize=4 SrcRegion=<8;8,1>  DstRegion=<1> Type=ud

.declare ubDNDI_UV_RESP     Base=r58.0 ElementSize=1 SrcRegion=<16;16,1>    DstRegion=<1> Type=ub

//Temp GRFs: For 42X to 422 Conversion

.declare uwDNDI_UVCOPY_TEMP Base=r10.0 ElementSize=2 SrcRegion=<16;16,1>    DstRegion=<1> Type=uw       //8 GRFs

.declare ubDNDI_UVCOPY_TEMP Base=r10.0 ElementSize=1 SrcRegion=<16;16,1>    DstRegion=<1> Type=ub       //8 GRFs

//---------------------------------------------------------------------------

// Message descriptors

//---------------------------------------------------------------------------

// Extended message descriptor

    // Message descriptor   for sampler read

    //                    = 000 0010 (message len 2) 00000 (resp len - set later, 12 or 5 or 11)

    //                      1 (header present 1) 0 11 (SIMD32/64 mode)

    //                      1000 (message type) 0000 (DI state index)

    //                      00000000 (binding table index - set later)

    //                    = 0x040b8000

// Attention: The Message Length is The Number of GRFs with Data Only, without the Header

//---------------------------------------------------------------------------

// VDI Return Data format

//---------------------------------------------------------------------------

// Defines for DI enabled

// Defines for DI disabled

// FileName:	DNDI_Command.asm

// Author:		Vivek Kumar

// Description:	Sends a message to the VDI to process one DN (16x8) or DNDI (16x4) block

// Prepare the DNDI send command

mov (8)		mudMSGHDR_DNDI(0)<1>			r0.0<8;8,1>:ud					// message header

mov (1)		mwMSGHDR_DNDI(1,4)<1>			r7.0<0;1,0>:w		{ NoDDClr }		// horizontal origin	// Do we need to add offset here? -vK

mov (1)		mwMSGHDR_DNDI(1,12)<1>			r7.1<0;1,0>:w		{ NoDDChk }		// vertical origin		// Can these 2 be combined? - vK

send (8)	udDNDI_RESP(0)<1>	r18	0x2	0x49E8003:ud

// On Gen6, with VDI walker, use the XY pair returned rather than programmed above

// VDI_RETURNED_XY is ordered XY in case of walker enables and the same as programmed in case of walker disabled

mov (2) 	r7.0<1>:w     uwDNDI_RESP(4,14)<2;2,1>	// horizontal/Vertial origin in W.14 and W.15

// FileName:	DN_Hist_Save.asm

// Author:		Vivek Kumar

// Description:	Saves DN history data to statistics surface

// Write denoise history to memory

mov (8)    r27<1>:ud				r0.0<8;8,1>:ud                   			// message header

	mov (2)    mudMSGHDR_HIST(1)<1>    	udDNDI_RESP(4,0)<2;2,1>    	// Move denoise history to MRF (4x2)

shr (2)    r27.0<1>:ud		r7.0<2;2,1>:w				2:w                                 	// X,Y origin / 4

add (1)    r27.0<1>:ud		r27.0<0;1,0>:ud			r1.12<0;1,0>:uw		{ NoDDClr }  	// Add pitch to X origin

mov (1)    r27.2<1>:ud		0x10003:ud									{ NoDDChk }  	// block width and height

mov (8)		mudMSGHDR_HIST(0)<1>		r27.0<8;8,1>:ud

send (8)	null<1>:d	r22	0x5		0x40A8021:ud

// FileName:	DNDI_Enc_Stats_Save.asm

// Author:		Vivek Kumar

// Description:	Saves Encoder Statistics data to statistics surface in case of DI enabled (for 16x4 block)

// Write encoder statistics to memory

//Currently enable this only on Gen6 validation

mov (8)		mudMSGHDR_ENC_STATS(1)<1>	0x0:ud						// Init payload MRF

mov (8)		mudMSGHDR_ENC_STATS(0)<1>	r0.0<8;8,1>:ud				// message header

shr (1)		mudMSGHDR_ENC_STATS(0,0)<1>		r7.0<0;1,0>:w            	1:w  	{ NoDDClr }			//enable the flag after testing on si           			{ NoDDClr }	// X origin / 2

mul (1)		acc0.1<1>:ud					r7.1<0;1,0>:w				3:w																							// Y origin * 3

shr (1)		mudMSGHDR_ENC_STATS(0,1)<1>		acc0.1<0;1,0>:ud			2:w		{ NoDDClr, NoDDChk }		//enable the flag after testing on si			   { NoDDClr, NoDDChk } // Y origin * 3/4

mov (1)		mudMSGHDR_ENC_STATS(0,2)<1>		0x50003:ud				{ NoDDChk }			//enable the flag after testing on si						{ NoDDChk } // block width and height (8x3)

add (2)		mudMSGHDR_ENC_STATS(0,0)<1>		mudMSGHDR_ENC_STATS(0,0)<2;2,1>       	r1.12<2;2,1>:uw					// Add pitch to X,Y origin

	//Data block for Encoder Statistics

	//----------------------------------------------------

	//|  0  |   1  |   2   |  3  |  4  |  5  |  6  |  7  | Bytes

	//----------------------------------------------------

	//| BNE |   X  |   X   |  X  |           X           |

	//----------------------------------------------------

	//|     X      |     SVCM    |           X           |

	//----------------------------------------------------

	//|   SHCM     |     STAD    |           X           |

	//----------------------------------------------------

	//|            X             |           X           |

	//----------------------------------------------------

	//|     X      |     SVCM    |           X           |

	//----------------------------------------------------

	//|   SHCM     |     STAD    |           X           |

	//----------------------------------------------------

	mov (1)		mubMSGHDR_ENC_STATS(1,0)<1>		ubDNDI_RESP(4,8)<0;1,0>    		{ NoDDClr }				// Move encoder statistics to MRF

	mov (1)		muwMSGHDR_ENC_STATS(1,3)<1>		uwDNDI_RESP(4,11)<0;1,0>    	{ NoDDClr, NoDDChk }			// Move encoder statistics to MRF

	mov (2)		muwMSGHDR_ENC_STATS(1,4)<1>		uwDNDI_RESP(4,12)<2;2,1>    	{ NoDDClr, NoDDChk }			// Move encoder statistics to MRF

	mov (1)		muwMSGHDR_ENC_STATS(1,9)<1>		uwDNDI_RESP(4,8)<0;1,0>    		{ NoDDClr, NoDDChk }			// Move encoder statistics to MRF

	mov (2)		muwMSGHDR_ENC_STATS(1,10)<1>	uwDNDI_RESP(4,9)<2;2,1>    		{ NoDDChk }				// Move encoder statistics to MRF

send (8)   null<1>:d    r24    0x5    0x40A8021:ud

// FileName:	DN_Save_PA.asm

// Author:		Vivek Kumar

// Description:	Save one 16x8 blocks of DN output in Packed format

add (4)		a0.4<1>:uw   r2.28<4;4,1>:ub   1024:w    // Initial Y,U,V offset in YUV422 block; it starts at m14

mov (8)		mudMSGHDR_DN_OUT(0)<1>		r0<8;8,1>:ud            					// message header

shl (1)     mdMSGHDR_DN_OUT(0,0)<1>		r7.0<0;1,0>:w     1:w  		{ NoDDClr }     // X origin * 2 (422 output)

mov (1)     mdMSGHDR_DN_OUT(0,1)<1>		r7.1<0;1,0>:w          		{ NoDDClr, NoDDChk }  // Y origin

mov (1)     mudMSGHDR_DN_OUT(0,2)<1>	0x7001F:ud	{ NoDDChk }     // block width and height (32x8)

	mov (16)    r[a0.4,0]<2>:ub   ubDNDI_RESP(0,0)<16;16,1>    	{ NoDDClr }   	// copy line of Y directly to memory as optimization

	mov (16)    r[a0.4,32]<2>:ub   ubDNDI_RESP(0,16)<16;16,1>    	{ NoDDClr }   	// copy line of Y directly to memory as optimization

	mov (16)    r[a0.4,64]<2>:ub   ubDNDI_RESP(0,32)<16;16,1>    	{ NoDDClr }   	// copy line of Y directly to memory as optimization

	mov (16)    r[a0.4,96]<2>:ub   ubDNDI_RESP(0,48)<16;16,1>    	{ NoDDClr }   	// copy line of Y directly to memory as optimization

	mov (16)    r[a0.4,128]<2>:ub   ubDNDI_RESP(0,64)<16;16,1>    	{ NoDDClr }   	// copy line of Y directly to memory as optimization

	mov (16)    r[a0.4,160]<2>:ub   ubDNDI_RESP(0,80)<16;16,1>    	{ NoDDClr }   	// copy line of Y directly to memory as optimization

	mov (16)    r[a0.4,192]<2>:ub   ubDNDI_RESP(0,96)<16;16,1>    	{ NoDDClr }   	// copy line of Y directly to memory as optimization

	mov (16)    r[a0.4,224]<2>:ub   ubDNDI_RESP(0,112)<16;16,1>    	{ NoDDClr }   	// copy line of Y directly to memory as optimization

	mov (8)     r[a0.5,0]<4>:ub   ubDNDI_RESP(5,1)<16;8,2>  { NoDDClr, NoDDChk }  // copy line of U directly to memory as optimization

	mov (8)     r[a0.6,0]<4>:ub   ubDNDI_RESP(5,0)<16;8,2>    { NoDDChk }  	// copy line of V directly to memory as optimization

	mov (8)     r[a0.5,32]<4>:ub   ubDNDI_RESP(5,17)<16;8,2>  { NoDDClr, NoDDChk }  // copy line of U directly to memory as optimization

	mov (8)     r[a0.6,32]<4>:ub   ubDNDI_RESP(5,16)<16;8,2>    { NoDDChk }  	// copy line of V directly to memory as optimization

	mov (8)     r[a0.5,64]<4>:ub   ubDNDI_RESP(5,33)<16;8,2>  { NoDDClr, NoDDChk }  // copy line of U directly to memory as optimization

	mov (8)     r[a0.6,64]<4>:ub   ubDNDI_RESP(5,32)<16;8,2>    { NoDDChk }  	// copy line of V directly to memory as optimization

	mov (8)     r[a0.5,96]<4>:ub   ubDNDI_RESP(5,49)<16;8,2>  { NoDDClr, NoDDChk }  // copy line of U directly to memory as optimization

	mov (8)     r[a0.6,96]<4>:ub   ubDNDI_RESP(5,48)<16;8,2>    { NoDDChk }  	// copy line of V directly to memory as optimization

	mov (8)     r[a0.5,128]<4>:ub   ubDNDI_RESP(5,65)<16;8,2>  { NoDDClr, NoDDChk }  // copy line of U directly to memory as optimization

	mov (8)     r[a0.6,128]<4>:ub   ubDNDI_RESP(5,64)<16;8,2>    { NoDDChk }  	// copy line of V directly to memory as optimization

	mov (8)     r[a0.5,160]<4>:ub   ubDNDI_RESP(5,81)<16;8,2>  { NoDDClr, NoDDChk }  // copy line of U directly to memory as optimization

	mov (8)     r[a0.6,160]<4>:ub   ubDNDI_RESP(5,80)<16;8,2>    { NoDDChk }  	// copy line of V directly to memory as optimization

	mov (8)     r[a0.5,192]<4>:ub   ubDNDI_RESP(5,97)<16;8,2>  { NoDDClr, NoDDChk }  // copy line of U directly to memory as optimization

	mov (8)     r[a0.6,192]<4>:ub   ubDNDI_RESP(5,96)<16;8,2>    { NoDDChk }  	// copy line of V directly to memory as optimization

	mov (8)     r[a0.5,224]<4>:ub   ubDNDI_RESP(5,113)<16;8,2>  { NoDDClr, NoDDChk }  // copy line of U directly to memory as optimization

	mov (8)     r[a0.6,224]<4>:ub   ubDNDI_RESP(5,112)<16;8,2>    { NoDDChk }  	// copy line of V directly to memory as optimization

//send out data through data port

send (8)    null<1>:d    r31.0		0x5    0x120A8018:ud

// FileName:	DN_Save_422CP_16x8.asm

// Author:		Vivek Kumar

// Description:	Save one 16x8 blocks of DN output to the color pipe in 4-2-2 format

.declare mubMSGHDR_DN_OUT_2   Base=r36.0      ElementSize=1  Type=ub

mov (8)		mudMSGHDR_DN_OUT(0)<1>		r0<8;8,1>:ud            			// message header

shl (1)     mdMSGHDR_DN_OUT(0,0)<1>		r7.0<0;1,0>:w     1:w  		{ NoDDClr }            // X origin * 2 (422 output)

mov (1)     mdMSGHDR_DN_OUT(0,1)<1>		r7.1<0;1,0>:w          		{ NoDDClr, NoDDChk }   // Y origin

mov (1)     mudMSGHDR_DN_OUT(0,2)<1>	0x7000F:ud	{ NoDDClr, NoDDChk }            // block width and height (16x8)

//M0.3	- 0 - CP Enable, 1 - Area of Interest, 3:2 Message Format(TBD), 4:3 - Ignored, 31:5 CP state pointer

//Compose area-of-interest bit + color pipe state pointer

or (1)		mudMSGHDR_DN_OUT(0,3)<1>		r2.4<0;1,0>:ud 	r7.26<0;1,0>:b		{ NoDDChk }

// First 8 x 8 Block

	mov (8)		mubMSGHDR_DN_OUT(1)<2>			ubDNDI_RESP(0,0)<8;8,1>				{ NoDDClr } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT(1,16)<2>		ubDNDI_RESP(0,16)<8;8,1>			{ NoDDClr, NoDDChk } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT(2)<2>			ubDNDI_RESP(0,32)<8;8,1>				{ NoDDClr } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT(2,16)<2>		ubDNDI_RESP(0,48)<8;8,1>			{ NoDDClr, NoDDChk } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT(3)<2>			ubDNDI_RESP(0,64)<8;8,1>				{ NoDDClr } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT(3,16)<2>		ubDNDI_RESP(0,80)<8;8,1>			{ NoDDClr, NoDDChk } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT(4)<2>			ubDNDI_RESP(0,96)<8;8,1>				{ NoDDClr } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT(4,16)<2>		ubDNDI_RESP(0,112)<8;8,1>			{ NoDDClr, NoDDChk } 	// copy line of Y directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(1,1)<4>   	ubDNDI_RESP(5,1)<8;4,2>			{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(1,17)<4>   	ubDNDI_RESP(5,17)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(1,3)<4>   	ubDNDI_RESP(5,0)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(1,19)<4>   	ubDNDI_RESP(5,16)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(2,1)<4>   	ubDNDI_RESP(5,33)<8;4,2>			{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(2,17)<4>   	ubDNDI_RESP(5,49)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(2,3)<4>   	ubDNDI_RESP(5,32)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(2,19)<4>   	ubDNDI_RESP(5,48)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(3,1)<4>   	ubDNDI_RESP(5,65)<8;4,2>			{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(3,17)<4>   	ubDNDI_RESP(5,81)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(3,3)<4>   	ubDNDI_RESP(5,64)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(3,19)<4>   	ubDNDI_RESP(5,80)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(4,1)<4>   	ubDNDI_RESP(5,97)<8;4,2>			{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(4,17)<4>   	ubDNDI_RESP(5,113)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(4,3)<4>   	ubDNDI_RESP(5,96)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT(4,19)<4>   	ubDNDI_RESP(5,112)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

// Second 8 x 8 Block

mov	(8)	r36.0<1>:ud		r31.0<8;8,1>:ud

add	(1)	r36.0<1>:ud		r36.0<0;1,0>:w		0x10:w

	mov (8)		mubMSGHDR_DN_OUT_2(1)<2>		ubDNDI_RESP(0,8)<8;8,1>			{ NoDDClr } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT_2(1,16)<2>	ubDNDI_RESP(0,24)<8;8,1>			{ NoDDClr, NoDDChk } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT_2(2)<2>		ubDNDI_RESP(0,40)<8;8,1>			{ NoDDClr } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT_2(2,16)<2>	ubDNDI_RESP(0,56)<8;8,1>			{ NoDDClr, NoDDChk } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT_2(3)<2>		ubDNDI_RESP(0,72)<8;8,1>			{ NoDDClr } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT_2(3,16)<2>	ubDNDI_RESP(0,88)<8;8,1>			{ NoDDClr, NoDDChk } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT_2(4)<2>		ubDNDI_RESP(0,104)<8;8,1>			{ NoDDClr } 	// copy line of Y directly to memory as optimization

	mov (8)		mubMSGHDR_DN_OUT_2(4,16)<2>	ubDNDI_RESP(0,120)<8;8,1>			{ NoDDClr, NoDDChk } 	// copy line of Y directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(1,1)<4>   	ubDNDI_RESP(5,9)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(1,17)<4>   	ubDNDI_RESP(5,25)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(1,3)<4>   	ubDNDI_RESP(5,8)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(1,19)<4>   	ubDNDI_RESP(5,24)<8;4,2>		{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(2,1)<4>   	ubDNDI_RESP(5,41)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(2,17)<4>   	ubDNDI_RESP(5,57)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(2,3)<4>   	ubDNDI_RESP(5,40)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(2,19)<4>   	ubDNDI_RESP(5,56)<8;4,2>		{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(3,1)<4>   	ubDNDI_RESP(5,73)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(3,17)<4>   	ubDNDI_RESP(5,89)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(3,3)<4>   	ubDNDI_RESP(5,72)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(3,19)<4>   	ubDNDI_RESP(5,88)<8;4,2>		{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(4,1)<4>   	ubDNDI_RESP(5,105)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(4,17)<4>   	ubDNDI_RESP(5,121)<8;4,2>		{ NoDDClr, NoDDChk } 	// copy line of U directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(4,3)<4>   	ubDNDI_RESP(5,104)<8;4,2>			{ NoDDChk }    	// copy line of V directly to memory as optimization

	mov (4)     mubMSGHDR_DN_OUT_2(4,19)<4>   	ubDNDI_RESP(5,120)<8;4,2>		{ NoDDChk }    	// copy line of V directly to memory as optimization

//send out data through data port

send (8)    null<1>:d    r31.0		0x5    0xA0A801B:ud

send (8)    null<1>:d    r36.0	0x5    0xA0A801B:ud

//End of Thread message

mov (8) r127<1>:ud r0.0<8;8,1>:ud

 send (1) null<1>:d r127 0x27 0x02000010

.end_code

.end_kernel