/* * Copyright 2000-2011 Intel Corporation All Rights Reserved * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including * without limitation the rights to use, copy, modify, merge, publish, * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. * IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * * This file was originally licensed under the following license * * 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. */ // 120 // Total instruction count // 1 // Total kernel count .kernel PL3_DNDI_422CP .code // FileName: DNDI_PL_Core.asm // Author: Tatiya, Rupesh // FileName: DNDI_Core.asm // Author: Vivek Kumar // Description: Tasks for DN+DI case (16x4 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===================================== //======================================================================= //r11-r17 // Define temp space for any usages // Common Buffers // temp space for rotation .declare fROBUF Base=r11.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=f .declare udROBUF Base=r11.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=ud .declare uwROBUF Base=r11.0 ElementSize=2 SrcRegion=<16;16,1> DstRegion=<1> Type=uw .declare ubROBUF Base=r11.0 ElementSize=1 SrcRegion=<16;16,1> DstRegion=<1> Type=ub .declare ub4ROBUF Base=r11.0 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> r9.0<0;1,0>:w { NoDDClr } // horizontal origin // Do we need to add offset here? -vK mov (1) mwMSGHDR_DNDI(1,12)<1> r9.1<0;1,0>:w { NoDDChk } // vertical origin // Can these 2 be combined? - vK send (8) udDNDI_RESP(0)<1> r18 0x2 0x4BE8003: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) r9.0<1>:w uwDNDI_RESP(9,14)<2;2,1> // horizontal/Vertial origin in W.14 and W.15 // FileName: DI_STMM_Save.asm // Author: Vivek Kumar // Description: Saves DI STMM Data to statistics surface in case of DI enabled (for 16x4 block) // Write STMM to memory mov (8) mudMSGHDR_STMM(0)<1> r0.0<8;8,1>:ud // message header mov (8) mudMSGHDR_STMM(1)<1> udDNDI_RESP(8,0) // Move STMM to MRF shr (1) mudMSGHDR_STMM(0,0)<1> r9.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2 mov (1) mudMSGHDR_STMM(0,1)<1> r9.1<0;1,0>:w { NoDDClr, NoDDChk } // Y origin mov (1) mudMSGHDR_STMM(0,2)<1> 0x30007:ud { NoDDChk } // block width and height (8x4) send (8) null<1>:d r20 0x5 0x40A8021:ud // 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 (1) mudMSGHDR_HIST(1)<1> udDNDI_RESP(9,0)<0;1,0> // Move denoise history to MRF (4x1) shr (2) r27.0<1>:ud r9.0<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 0x3: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> r9.0<0;1,0>:w 1:w { NoDDClr } //enable the flag after testing on si { NoDDClr } // X origin / 2 mul (1) acc0.1<1>:ud r9.1<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> 0x20007: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 | MCNT | FCNT | TCNT | X | X | X | X | //---------------------------------------------------- //| DcTpT | SVCM | DcBpT | DcTpB | //---------------------------------------------------- //| SHCM | STAD | DcTcB | DcBpB | //---------------------------------------------------- mov (1) mudMSGHDR_ENC_STATS(1,0)<1> udDNDI_RESP(9,1)<0;1,0> { NoDDClr } // Move encoder statistics to MRF mov (2) mudMSGHDR_ENC_STATS(1,3)<2> udDNDI_RESP(9,3)<2;2,1> { NoDDClr, NoDDChk } // Move encoder statistics to MRF mov (2) mudMSGHDR_ENC_STATS(1,2)<2> udDNDI_RESP(9,5)<2;2,1> { NoDDChk } // Move encoder statistics to MRF send (8) null<1>:d r24 0x5 0x40A8021:ud // FileName: DN_Load_UV_IMC3_16x4.asm // Author: Vivek Kumar // Description: Read UV for 16x4 block through DATAPORT // FileName: UVCopy_Load_16x4.asm // Author: Vivek Kumar // Description: Read UV for 16x4 block through DATAPORT add (2) r27.0<1>:d r9.0<2;2,1>:w r4.4<2;2,1>:w // Source Y Block origin asr (2) r27.0<1>:d r27.0<2;2,1>:d 1:w { NoDDClr } // U/V block origin should be half of Y's mov (1) r27.2<1>:ud 0x10007:ud { NoDDChk } // U/V block width and height (8x2) mov (8) r36<1>:ud r27.0<8;8,1>:ud mov (8) r38<1>:ud r27.0<8;8,1>:ud send (8) udDNDI_UV_RESP(0)<1> r36 0x4 0x2190001:ud send (8) udDNDI_UV_RESP(1)<1> r38 0x4 0x2190002:ud // FileName: DN_Save_Y_16x4.asm // Author: Vivek Kumar // Description: Save one 16x4 blocks of Y channel of DN output for reference // check top/bottom field first cmp.e.f0.0 (1) null<1>:w r1.28<0;1,0>:ub 1:w mov (8) mudMSGHDR_DN_OUT(0)<1> r0<8;8,1>:ud // message header mov (2) mdMSGHDR_DN_OUT(0,0)<1> r9.0<2;2,1>:w { NoDDClr } // X origin * 2 (422 output) mov (1) mudMSGHDR_DN_OUT(0,2)<1> 0x3000F:ud { NoDDChk } // block width and height (32x8) (f0.0) jmpi (1) TOP_FIELD_FIRST BOTTOM_FIELD_FIRST: mov (4) mudMSGHDR_DN_OUT(1,0)<1> udDNDI_RESP(10,0)<4;4,1> { NoDDClr } // 2nd field luma from current frame (line 0,2) mov (4) mudMSGHDR_DN_OUT(1,4)<1> udDNDI_RESP(4,4)<4;4,1> { NoDDChk } // 1st field luma from current frame (line 1,3) mov (4) mudMSGHDR_DN_OUT(2,0)<1> udDNDI_RESP(10,4)<4;4,1> { NoDDClr } // 2nd field luma from current frame (line 0,2) mov (4) mudMSGHDR_DN_OUT(2,4)<1> udDNDI_RESP(5,4)<4;4,1> { NoDDChk } // 1st field luma from current frame (line 1,3) jmpi (1) SAVE_DN_CURR TOP_FIELD_FIRST: mov (4) mudMSGHDR_DN_OUT(1,0)<1> udDNDI_RESP(4,0)<4;4,1> { NoDDClr } // 2nd field luma from current frame (line 0,2) mov (4) mudMSGHDR_DN_OUT(1,4)<1> udDNDI_RESP(10,0)<4;4,1> { NoDDChk } // 1st field luma from current frame (line 1,3) mov (4) mudMSGHDR_DN_OUT(2,0)<1> udDNDI_RESP(5,0)<4;4,1> { NoDDClr } // 2nd field luma from current frame (line 0,2) mov (4) mudMSGHDR_DN_OUT(2,4)<1> udDNDI_RESP(10,4)<4;4,1> { NoDDChk } // 1st field luma from current frame (line 1,3) SAVE_DN_CURR: //send out data through data port send (8) null<1>:d r31.0 0x5 0x60A8018:ud // FileName: DI_Save_422CP_16x4.asm // Author: Vivek Kumar // Description: Save two 16x4 blocks of DI output in 422 format to Color Pipe (IECP) .declare mubMSGHDR_DI_OUT1_1 Base=r18.0 ElementSize=1 Type=ub .declare mubMSGHDR_DI_OUT1_2 Base=r21.0 ElementSize=1 Type=ub .declare mubMSGHDR_DI_OUT2_1 Base=r24.0 ElementSize=1 Type=ub .declare mubMSGHDR_DI_OUT2_2 Base=r27.0 ElementSize=1 Type=ub mov (8) r27.0<1>:ud r0.0<8;8,1>:ud shl (1) r27.0<1>:ud r9.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled mov (1) r27.1<1>:ud r9.1<0;1,0>:w { NoDDClr, NoDDChk } // Block origin mov (1) r27.2<1>:ud 0x3000F: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) r27.3<1>:ud r2.4<0;1,0>:ud r9.26<0;1,0>:b { NoDDChk } //prepare the message headers mov (8) r18.0<1>:ud r27<8;8,1>:ud mov (8) r24.0<1>:ud r27<8;8,1>:ud // Pack 2nd field Y; First 8x4 block mov (8) mubMSGHDR_DI_OUT1_1(1)<2> ubDNDI_RESP(0,0)<8;8,1> { NoDDClr } // copy line of Y directly to memory as optimization mov (8) mubMSGHDR_DI_OUT1_1(1,16)<2> ubDNDI_RESP(0,16)<8;8,1> { NoDDClr, NoDDChk } // copy line of Y directly to memory as optimization mov (8) mubMSGHDR_DI_OUT1_1(2)<2> ubDNDI_RESP(0,32)<8;8,1> { NoDDClr } // copy line of Y directly to memory as optimization mov (8) mubMSGHDR_DI_OUT1_1(2,16)<2> ubDNDI_RESP(0,48)<8;8,1> { NoDDClr, NoDDChk } // copy line of Y directly to memory as optimization // Pack 2nd field U, V; First 8x4 block mov (4) mubMSGHDR_DI_OUT1_1(1,1)<4> ubDNDI_RESP(2,1)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_1(1,17)<4> ubDNDI_RESP(2,17)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_1(1,3)<4> ubDNDI_RESP(2,0)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_1(1,19)<4> ubDNDI_RESP(2,16)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_1(2,1)<4> ubDNDI_RESP(2,33)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_1(2,17)<4> ubDNDI_RESP(2,49)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_1(2,3)<4> ubDNDI_RESP(2,32)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_1(2,19)<4> ubDNDI_RESP(2,48)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization // Pack 2nd field Y; Second 8x4 block mov (8) r21.0<1>:ud r18.0<8;8,1>:ud add (1) r21.0<1>:ud r21.0<0;1,0>:w 0x10:w mov (8) mubMSGHDR_DI_OUT1_2(1)<2> ubDNDI_RESP(0,8)<8;8,1> { NoDDClr } // copy line of Y directly to memory as optimization mov (8) mubMSGHDR_DI_OUT1_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_DI_OUT1_2(2)<2> ubDNDI_RESP(0,40)<8;8,1> { NoDDClr } // copy line of Y directly to memory as optimization mov (8) mubMSGHDR_DI_OUT1_2(2,16)<2> ubDNDI_RESP(0,56)<8;8,1> { NoDDClr, NoDDChk } // copy line of Y directly to memory as optimization // Pack 2nd field U, V; Second 8x4 block mov (4) mubMSGHDR_DI_OUT1_2(1,1)<4> ubDNDI_RESP(2,9)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_2(1,17)<4> ubDNDI_RESP(2,25)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_2(1,3)<4> ubDNDI_RESP(2,8)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_2(1,19)<4> ubDNDI_RESP(2,24)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_2(2,1)<4> ubDNDI_RESP(2,41)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_2(2,17)<4> ubDNDI_RESP(2,57)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_2(2,3)<4> ubDNDI_RESP(2,40)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT1_2(2,19)<4> ubDNDI_RESP(2,56)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization send (8) null<1>:d r18.0 0x5 0x60A801B:ud send (8) null<1>:d r21.0 0x5 0x60A801B:ud // Pack 1st field Y; 1st 8x4 block mov (8) mubMSGHDR_DI_OUT2_1(1)<2> ubDNDI_RESP(4,0)<8;8,1> { NoDDClr } // copy line of Y directly to memory as optimization mov (8) mubMSGHDR_DI_OUT2_1(1,16)<2> ubDNDI_RESP(4,16)<8;8,1> { NoDDClr, NoDDChk } // copy line of Y directly to memory as optimization mov (8) mubMSGHDR_DI_OUT2_1(2)<2> ubDNDI_RESP(4,32)<8;8,1> { NoDDClr } // copy line of Y directly to memory as optimization mov (8) mubMSGHDR_DI_OUT2_1(2,16)<2> ubDNDI_RESP(4,48)<8;8,1> { NoDDClr, NoDDChk } // copy line of Y directly to memory as optimization // Pack 1st field U,V; 1st 8x4 block mov (4) mubMSGHDR_DI_OUT2_1(1,1)<4> ubDNDI_RESP(6,1)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_1(1,17)<4> ubDNDI_RESP(6,17)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_1(1,3)<4> ubDNDI_RESP(6,0)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_1(1,19)<4> ubDNDI_RESP(6,16)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_1(2,1)<4> ubDNDI_RESP(6,33)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_1(2,17)<4> ubDNDI_RESP(6,49)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_1(2,3)<4> ubDNDI_RESP(6,32)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_1(2,19)<4> ubDNDI_RESP(6,48)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization // Pack 1st field Y; 2nd 8x4 block mov (8) r27.0<1>:ud r24.0<8;8,1>:ud add (1) r27.0<1>:ud r27.0<0;1,0>:w 0x10:w mov (8) mubMSGHDR_DI_OUT2_2(1)<2> ubDNDI_RESP(4,8)<8;8,1> { NoDDClr } // copy line of Y directly to memory as optimization mov (8) mubMSGHDR_DI_OUT2_2(1,16)<2> ubDNDI_RESP(4,24)<8;8,1> { NoDDClr, NoDDChk } // copy line of Y directly to memory as optimization mov (8) mubMSGHDR_DI_OUT2_2(2)<2> ubDNDI_RESP(4,40)<8;8,1> { NoDDClr } // copy line of Y directly to memory as optimization mov (8) mubMSGHDR_DI_OUT2_2(2,16)<2> ubDNDI_RESP(4,56)<8;8,1> { NoDDClr, NoDDChk } // copy line of Y directly to memory as optimization // Pack 1st field U, V; 2nd 8x4 block mov (4) mubMSGHDR_DI_OUT2_2(1,1)<4> ubDNDI_RESP(6,9)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_2(1,17)<4> ubDNDI_RESP(6,25)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_2(1,3)<4> ubDNDI_RESP(6,8)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_2(1,19)<4> ubDNDI_RESP(6,24)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_2(2,1)<4> ubDNDI_RESP(6,41)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_2(2,17)<4> ubDNDI_RESP(6,57)<8;4,2> { NoDDClr, NoDDChk } // copy line of U directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_2(2,3)<4> ubDNDI_RESP(6,40)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization mov (4) mubMSGHDR_DI_OUT2_2(2,19)<4> ubDNDI_RESP(6,56)<8;4,2> { NoDDChk } // copy line of V directly to memory as optimization send (8) null<1>:d r24.0 0x5 0x60A801E:ud send (8) null<1>:d r27.0 0x5 0x60A801E:ud // FileName: DN_Save_UV_IMC3_16x4.asm // Author: Vivek Kumar // Description: Save UV for 16x4 block through DATAPORT // FileName: UVCopy_Save_16x4.asm // Author: Vivek Kumar // Description: Save UV for 16x4 block through DATAPORT //Reuse the header from Load component mov (4) mudMSGHDR_UCOPY(1)<1> udDNDI_UV_RESP(0)<4;4,1> mov (4) mudMSGHDR_VCOPY(1)<1> udDNDI_UV_RESP(1)<4;4,1> send (4) null<1>:d r36 0x5 0x40A8019:ud send (4) null<1>:d r38 0x5 0x40A801A: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