In addtion, add the license header.
You need the latest intel-gen4asm to build these shaders
Signed-off-by: Xiang, Haihao <haihao.xiang@intel.com>
+/*
+ * 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.
+ */
// 22 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 20 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 33 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 2 // Total instruction count
// 1 // Total kernel count
DI_Save_NV12_16x4.g4a \
DI_Save_PA_16x4.g4a \
EOT.g4a \
+ NV12_DNDI_422CP.g4a \
+ NV12_DNDI_PA.g4a \
+ NV12_DNUV_NV12.g4a \
+ NV12_DN_422CP.g4a \
+ NV12_DN_NV12.g4a \
PA_AVS_Buf_0.g4a \
PA_AVS_Buf_1.g4a \
PA_AVS_Buf_2.g4a \
PA_AVS_Buf_3.g4a \
+ PA_DI_422CP.g4a \
+ PA_DI_PA.g4a \
+ PA_DNDI_422CP.g4a \
+ PA_DNDI_PA.g4a \
+ PA_DNUV_PA.g4a \
+ PA_DN_422CP.g4a \
+ PA_DN_PA.g4a \
PL2_AVS_Buf_0.g4a \
PL2_AVS_Buf_1.g4a \
PL2_AVS_Buf_2.g4a \
PL3_AVS_Buf_1.g4a \
PL3_AVS_Buf_2.g4a \
PL3_AVS_Buf_3.g4a \
+ PL3_DNDI_422CP.g4a \
+ PL3_DNDI_PA.g4a \
+ PL3_DNUV_PL3.g4a \
+ PL3_DN_422CP.g4a \
+ PL3_DN_PL3.g4a \
+ PL_DI_422CP.g4a \
+ PL_DI_PA.g4a \
Save_AVS_PA.g4a \
Save_AVS_PL3.g4a \
Save_AVS_NV12.g4a \
--- /dev/null
+/*
+ * 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.
+ */
+// 116 // Total instruction count
+// 1 // Total kernel count
+
+.kernel NV12_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=====================================
+
+
+//=======================================================================
+
+
+//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 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) r7.0<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> r7.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2
+mov (1) mudMSGHDR_STMM(0,1)<1> r7.1<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 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 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> 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> 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_NV12_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 r7.0<2;2,1>:w r4.4<2;2,1>:w // Source Y Block origin
+ asr (1) r27.1<1>:d r27.1<0;1,0>:d 1:w { NoDDClr } // U/V block origin should be half of Y's
+ mov (1) r27.2<1>:ud 0x1000F:ud { NoDDChk } // U/V block width and height (8x4)
+ mov (8) mudMSGHDR_UVCOPY(0)<1> r27.0<8;8,1>:ud
+ send (8) udDNDI_UV_RESP(0)<1> r36 0x4 0x2190001: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> r7.0<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 r7.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
+mov (1) r27.1<1>:ud r7.1<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 r7.26<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_NV12_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 (8) mudMSGHDR_UVCOPY(1)<1> udDNDI_UV_RESP(0)<8;8,1>
+ send (8) null<1>:d r36 0x5 0x40A8019: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
--- /dev/null
+/*
+ * 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.
+ */
+// 86 // Total instruction count
+// 1 // Total kernel count
+
+.kernel NV12_DNDI_PA
+.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=====================================
+
+
+//=======================================================================
+
+
+//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 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) r7.0<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> r7.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2
+mov (1) mudMSGHDR_STMM(0,1)<1> r7.1<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 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 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> 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> 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_NV12_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 r7.0<2;2,1>:w r4.4<2;2,1>:w // Source Y Block origin
+ asr (1) r27.1<1>:d r27.1<0;1,0>:d 1:w { NoDDClr } // U/V block origin should be half of Y's
+ mov (1) r27.2<1>:ud 0x1000F:ud { NoDDChk } // U/V block width and height (8x4)
+ mov (8) mudMSGHDR_UVCOPY(0)<1> r27.0<8;8,1>:ud
+ send (8) udDNDI_UV_RESP(0)<1> r36 0x4 0x2190001: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> r7.0<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_PA_16x4.asm
+// Author: Vivek Kumar
+// Description: Save two 16x4 blocks of DI output in Packed format
+
+
+add (4) a0.4<1>:uw r2.28<4;4,1>:ub 608:w // Initial Y,U,V offset in YUV422 block; it starts at m20
+
+mov (8) r27.0<1>:ud r0.0<8;8,1>:ud
+shl (1) r27.0<1>:d r7.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
+mov (1) r27.1<1>:d r7.1<0;1,0>:w { NoDDClr, NoDDChk } // Block origin
+mov (1) r27.2<1>:ud 0x3001F:ud { NoDDChk } // Block width and height (32x8)
+
+//prepare the message headers
+mov (8) r18.0<1>:ud r27<8;8,1>:ud
+mov (8) r23.0<1>:ud r27<8;8,1>:ud
+
+// Pack 2nd field Y
+ mov (16) r[a0.4, 0]<2> ubDNDI_RESP(0,0) { NoDDClr }
+ mov (16) r[a0.4, 32]<2> ubDNDI_RESP(0,16) { NoDDClr }
+ mov (16) r[a0.4, 64]<2> ubDNDI_RESP(0,32) { NoDDClr }
+ mov (16) r[a0.4, 96]<2> ubDNDI_RESP(0,48) { NoDDClr }
+// Pack 2nd field U
+ mov (8) r[a0.5, 0]<4> ubDNDI_RESP(2,1)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 32]<4> ubDNDI_RESP(2,17)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 64]<4> ubDNDI_RESP(2,33)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 96]<4> ubDNDI_RESP(2,49)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+// Pack 2nd field V
+ mov (8) r[a0.6, 0]<4> ubDNDI_RESP(2,0)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 32]<4> ubDNDI_RESP(2,16)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 64]<4> ubDNDI_RESP(2,32)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 96]<4> ubDNDI_RESP(2,48)<16;8,2> { NoDDChk } //Vpixels
+
+// Pack 1st field Y
+ mov (16) r[a0.4, 160]<2> ubDNDI_RESP(4,0) { NoDDClr }
+ mov (16) r[a0.4, 192]<2> ubDNDI_RESP(4,16) { NoDDClr }
+ mov (16) r[a0.4, 224]<2> ubDNDI_RESP(4,32) { NoDDClr }
+ mov (16) r[a0.4, 256]<2> ubDNDI_RESP(4,48) { NoDDClr }
+// Pack 1st field U
+ mov (8) r[a0.5, 160]<4> ubDNDI_RESP(6,1)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 192]<4> ubDNDI_RESP(6,17)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 224]<4> ubDNDI_RESP(6,33)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 256]<4> ubDNDI_RESP(6,49)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+// Pack 1st field V
+ mov (8) r[a0.6, 160]<4> ubDNDI_RESP(6,0)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 192]<4> ubDNDI_RESP(6,16)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 224]<4> ubDNDI_RESP(6,32)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 256]<4> ubDNDI_RESP(6,48)<16;8,2> { NoDDChk } //Vpixels
+
+//save the previous frame
+send (8) null<1>:d r18.0 0x5 0xA0A801B:ud
+
+//save the current frame
+send (8) null<1>:d r23.0 0x5 0xA0A801E:ud
+
+
+
+// FileName: DN_Save_UV_NV12_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 (8) mudMSGHDR_UVCOPY(1)<1> udDNDI_UV_RESP(0)<8;8,1>
+ send (8) null<1>:d r36 0x5 0x40A8019: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
--- /dev/null
+/*
+ * 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.
+ */
+// 1153 // Total instruction count
+// 1 // Total kernel count
+
+
+.kernel NV12_DNUV_NV12
+.code
+
+
+
+//Module : DN_UV_Setup
+//Author : Tatiya, Rupesh
+//Description : Initial Set-up for DN_UV
+
+
+
+
+// Module name : ChromaDenoise.inc
+// Author : Tatiya, Rupesh
+
+
+
+
+// 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
+
+
+//Interface:
+//Static Parameters:
+//r1
+
+
+//======================================================
+//Interface for serpent mode Chroma Denoise, added by Le
+//======================================================
+//r1
+
+
+//noise history thresholds (low and high)
+
+
+//temporal difference thresholds (high and low)
+
+
+//noise history thresholds (low and high)
+//#define ubNoiseHistMaxHigh r1.22
+//#define ubNoiseHistMaxLow r1.23
+//#define ubNoiseHistDeltaHigh r1.24
+//#define ubNoiseHistDeltaLow r1.25
+
+//Gaussian thresholds
+
+
+//temporal difference thresholds (default)
+
+
+//r2
+//history thresholds (default)
+
+
+//denoise factor (0-63)
+
+
+//====================== Binding table (Explicit To DNUV)=========================================
+//Used by DN_UV kernels
+
+
+ //Pointer to Current Frame UV
+
+
+//r1-r6
+ //CURBE GRFs used as TEMP : Used for max computation and storing max temporarily. : r1-r6
+
+
+ .declare ubCURBE_TEMP Base=r1.0 ElementSize=1 Type=ub
+ .declare uwCURBE_TEMP Base=r1.0 ElementSize=2 Type=uw
+ .declare wCURBE_TEMP Base=r1.0 ElementSize=2 Type=w
+ .declare fCURBE_TEMP Base=r1.0 ElementSize=4 Type=f
+ .declare udCURBE_TEMP Base=r1.0 ElementSize=4 Type=ud
+ .declare uwMAX_ABS_DIFF Base=r5.0 ElementSize=2 Type=uw
+
+ //r1
+
+
+ //r3
+
+
+ //r4
+
+//r7
+ //All of the following has to defined in Same GRF for optimal performance.
+
+
+//r8-24
+ //Previous Frame UV
+
+ .declare udPREV_UV Base=r8.0 ElementSize=4 Type=ud
+ .declare ubPREV_UV Base=r8.0 ElementSize=1 Type=ub
+
+
+//r25-48
+ //TEMP Space for any Usage.
+
+
+//=========================================================================
+//Definations and declarations for serpent mode Chroma Denoise, added by Le
+//=========================================================================
+
+
+ .declare udGNE_UV Base=r24.0 ElementSize=4 Type=ud
+ .declare fGNE_UV Base=r24.0 ElementSize=4 Type=f
+ .declare ubGNE_UV Base=r24.0 ElementSize=1 Type=ub
+
+ .declare udMSGHDR_BNE_SERP Base=r25.0 ElementSize=4 Type=ud
+ .declare udMSGSRC_BNE_SERP Base=r26.0 ElementSize=4 Type=ud
+
+
+ .declare ubDN_UV_Thresholds Base=r26.0 ElementSize=1 Type=ub
+ .declare ubDN_UV_Thresholds_Temp Base=r27.0 ElementSize=1 Type=ub
+ .declare udDN_UV_Thresholds Base=r26.0 ElementSize=4 Type=ud
+ .declare udDN_UV_Thresholds_Temp Base=r27.0 ElementSize=4 Type=ud
+ .declare fDN_UV_Thresholds Base=r26.0 ElementSize=4 Type=f
+ .declare fDN_UV_Thresholds_Temp Base=r27.0 ElementSize=4 Type=f
+
+
+//====================================================================================
+
+
+ //TEMP23: To hold V data for PL3 surfaces
+ .declare udCURR_V_TEMP Base=r25.0 ElementSize=4 Type=ud
+ .declare ubCURR_V_TEMP Base=r25.0 ElementSize=1 Type=ub
+
+ //GRFs to calculate Median: r25-r42
+ .declare ubMEDIAN_TEMP Base=r25.0 ElementSize=1 Type=ub
+
+ //18 GRFs to hold difference : r25-r42
+ .declare wDIFF Base=r25.0 ElementSize=2 Type=w
+ .declare uwDIFF Base=r25.0 ElementSize=2 Type=uw
+
+ //Temporal Diff
+ .declare wDIFF_TEMPORAL Base=r25.0 ElementSize=2 Type=w
+ .declare ubDIFF_TEMPORAL Base=r25.0 ElementSize=1 Type=ub
+
+ //4 GRFs to hold Sobel Value : r43-46
+ .declare wSOBEL_X Base=r43.0 ElementSize=2 Type=w
+ .declare uwSOBEL Base=r43.0 ElementSize=2 Type=uw
+
+
+ //2 GRFs to hold SOAD temporarily: r47-48
+ .declare uwSOAD Base=r47.0 ElementSize=2 Type=uw
+
+ //Temp GRFs to hold extra YUYV pixels: r43-r48
+ .declare ubTEMP5 Base=r43.0 ElementSize=1 Type=ub
+
+ //Temp GRFs in Median Calculation: r47-r48
+ .declare ubTEMP1 Base=r47.0 ElementSize=1 Type=ub
+
+ .declare uwTEMP0 Base=r48.0 ElementSize=2 Type=uw
+ .declare ubTEMP0 Base=r48.0 ElementSize=1 Type=ub
+
+ //Temp Space to store Median : r49-50
+
+ .declare ubMEDIAN Base=r49.0 ElementSize=1 Type=ub
+
+//r49
+
+
+//r50
+ //Message Source
+
+
+//r51
+ //DN_UV History Surface
+
+ .declare udHIST_UV Base=r51.0 ElementSize=4 Type=ud
+ .declare ubHIST_UV Base=r51.0 ElementSize=1 Type=ub
+
+//r52 - r91
+ //r52
+ //Current Frame UV
+
+
+ .declare udCURR_UV Base=r52.0 ElementSize=4 Type=ud
+ .declare ubCURR_UV Base=r52.0 ElementSize=1 Type=ub
+
+ //r54
+ //CURBE COPY
+
+
+ //r55
+
+
+ .declare uwSOAD_MIN_8x4 Base=r56.0 ElementSize=2 Type=uw
+
+ //r61
+
+
+ //r62
+
+
+ //History Surface Temp Origin
+
+
+ //r63
+ //Current Frame Y Temp Origin
+
+
+ //BNE Surface Origin
+
+
+ //r70
+
+ .declare uwDIFF_TEMPORAL_SUM4x4 Base=r70.0 ElementSize=2 Type=uw //4 GRFs
+
+ //r74-91 : For Saving Dest UV (PL2/PL3)
+
+
+ .declare ubMSGPAYLOAD_UV0 Base=r75.0 ElementSize=1 Type=ub
+
+
+ .declare ubMSGPAYLOAD_U Base=r75.0 ElementSize=1 Type=ub
+
+
+ .declare ubMSGPAYLOAD_UV1 Base=r84.0 ElementSize=1 Type=ub
+
+
+ .declare ubMSGPAYLOAD_V Base=r84.0 ElementSize=1 Type=ub
+
+ //r90
+
+ .declare uwDIFF_TEMPORAL_SUM4x4_FINAL Base=r90.0 ElementSize=2 Type=uw //2 GRFs
+
+//r92-127
+ //Current Frame Y
+
+
+ //r92
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_0 Base=r92 ElementSize=2 Type=uw
+ //r101
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_1 Base=r101 ElementSize=2 Type=uw
+ //r110
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_2 Base=r110 ElementSize=2 Type=uw
+ //r119
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_3 Base=r119 ElementSize=2 Type=uw
+
+ .declare udCURR_Y0 Base=r93.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y0 Base=r93.0 ElementSize=1 Type=ub
+ .declare udCURR_Y1 Base=r102.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y1 Base=r102.0 ElementSize=1 Type=ub
+ .declare udCURR_Y2 Base=r111.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y2 Base=r111.0 ElementSize=1 Type=ub
+ .declare udCURR_Y3 Base=r120.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y3 Base=r120.0 ElementSize=1 Type=ub
+
+ //r92: To hold U data for PL3 surfaces
+ .declare udCURR_U_TEMP Base=r92.0 ElementSize=4 Type=ud
+ .declare ubCURR_U_TEMP Base=r92.0 ElementSize=1 Type=ub
+
+ //r112: To hold U data for PL3 surfaces
+ .declare udPREV_U_TEMP Base=r112.0 ElementSize=4 Type=ud
+ .declare ubPREV_U_TEMP Base=r112.0 ElementSize=1 Type=ub
+
+ //r120: To hold U data for PL3 surfaces
+ .declare udPREV_V_TEMP Base=r120.0 ElementSize=4 Type=ud
+ .declare ubPREV_V_TEMP Base=r120.0 ElementSize=1 Type=ub
+
+
+ // Initialize message source with r0.
+ mov (8) r50.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r92.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r101.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r110.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r119.0<1>:ud r0.0<8;8,1>:ud
+
+
+
+//Module Name : DN_UV_PL2_Load_Curr_Frame_UV
+//Author : Tatiya, Rupesh
+//Description : Loads Current Frame UV data for PL2 input.
+
+
+
+//Module name : DN_UV_Load_Curr_Frame_UV
+//Author : Tatiya, Rupesh
+//Description : Loads Current Frame (UV only).
+// We need 4 extra rows (2 per field) and 2 extra pixel (1 each side) for both U and V each.
+// The processing size is 16x16 U and V each. So we need : U size - 18x20, V size - 18x20, UV size - 36x20, YUYV size - 72x20.
+
+
+
+
+//36x20 interleaved UV block is partitioned as follows:
+// <------ 18 --------> <--------18 ------->
+// -----------------------------------------
+// | 20x8 A !| 20x8 D !
+// | (overlapped) !| (overlapped) !
+// |-------------------!|------------------!
+// | 20x8 B !| 20x8 E !
+// | (overlapped) !| (overlapped) !
+// |-------------------!|------------------!
+// | 20x4 C !| 20x4 F !
+// | (overlapped) !| (overlapped) !
+// -----------------------------------------
+//
+// Cordinates: (x-2, y-2), (x+14, y-2), (x-2, y+6), (x+14, y+6), (x-2, y+14), (x+14, y+14)
+
+ //UV surface origin: (ORIX, ORIY/2)
+ add (2) r7.4<1>:w r7.0<2;2,1>:w r4.4<2;2,1>:w { AccWrEn } // Source Block origin
+ shr (1) r7.5<1>:w acc0.5<0;1,0>:w 1:w
+ mov (2) acc0.0<1>:d r7.4<2;2,1>:w
+
+ //A
+ add (2) r50.0<1>:d acc0.0<2;2,1>:d -2:d { AccWrEn }
+ mov (1) r50.2<1>:ud 0x70013:ud
+ send (8) udCURR_UV(0)<1> r50 0x4 0x2890004:ud
+
+ //B
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d 8:d
+ send (8) udCURR_UV(8)<1> r50 0x4 0x2890004:ud
+
+ //C
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d 16:d
+ mov (1) r50.2<1>:ud 0x30013:ud
+ send (8) udCURR_UV(16)<1> r50 0x4 0x2490004:ud
+
+ //D
+ add (1) r50.0<1>:d acc0.0<0;1,0>:d 16:d { AccWrEn }
+ mov (1) r50.1<1>:d acc0.1<0;1,0>:d
+ mov (1) r50.2<1>:ud 0x70013:ud
+ send (8) udCURR_UV(20)<1> r50 0x4 0x2890004:ud
+
+ //E
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d 8:d
+ send (8) udCURR_UV(28)<1> r50 0x4 0x2890004:ud
+
+ //F
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d 16:d
+ mov (1) r50.2<1>:ud 0x30013:ud
+ send (8) udCURR_UV(36)<1> r50 0x4 0x2490004:ud
+
+ //History Origin, Current Y origin and BNE surface origin - all are in inline GRF. Use , . -rT.
+
+ //Calculate Origin For History Surface: (ORIX/4, ORIY/8)
+ mov (16) acc0.0<1>:w r7.0<0;2,1>:w
+ shr (1) r7.2<1>:w acc0.2<0;1,0>:w 2:w
+ shr (1) r7.3<1>:w acc0.3<0;1,0>:w 3:w
+
+ //Calculate Origin For BNE Surface: (ORIX/8, ORIY/16)
+ shr (1) r7.6<1>:w acc0.6<0;1,0>:w 3:w
+ shr (1) r7.7<1>:w acc0.7<0;1,0>:w 4:w
+
+
+
+//Module Name : DN_UV_PL2_Load_Prev_Frame_UV
+//Author : Tatiya, Rupesh
+//Description : Loads Pevious Frame UV data for PL2 input.
+
+
+
+//Module Name : DN_UV_Load_Prev_Frame_UV
+//Author : Tatiya, Rupesh
+//Description : Loads Prev Frame (UV only). U size - 16x16, V size - 16x16, UV size - 32x16, YUYV size - 64x16.
+
+
+
+
+ mov (2) r50.0<1>:d r7.4<2;2,1>:w { AccWrEn } // Source lock origin
+ mov (1) r50.2<1>:ud 0xF000F:ud // U/V block width and height (16x16)
+ send (8) udPREV_UV(0)<1> r50 0x4 0x2890001:ud
+
+ add (1) r50.0<1>:ud acc0.0<0;1,0>:d 16:w // Add 16 to X origin
+ send (8) udPREV_UV(8)<1> r50 0x4 0x2890001:ud
+
+
+ //TODO - See if History loading can be combined with Prev Frame Load. - rT
+
+
+//Module name : DN_UV_Load_Hist_UV
+//Author : Tatiya, Rupesh
+//Description : Load DN History for UV denoise. 4x4 for each U & V.
+
+
+
+
+ mov (2) r50.0<1>:d r7.2<2;2,1>:w
+ mov (1) r50.2<1>:ud 0x30007:ud
+ send (8) udHIST_UV(0)<1> r50 0x4 0x2190022:ud
+
+
+
+//Module Name : DN_UV_420_Load_Curr_Frame_Y
+//Author : Tatiya, Rupesh
+//Description : Load Curr Frame Y data for 420 Input
+
+
+
+//Module Name : DN_UV_Load_Curr_Frame_Y
+//Author : Tatiya, Rupesh
+//Description : Loads Y of Current frame.
+
+
+
+
+ //For 16x16 U and 16x16 V for 420, we need to read 32x32 Y.
+
+ mov (8) acc0.0<1>:ud r0.0<8;8,1>:ud
+ mov (1) acc0.2<1>:ud 0xF000F:ud
+ add (2) acc0.0<1>:ud r7.0<2;2,1>:w r4.4<2;2,1>:w
+
+ mov (8) r92.0<1>:ud acc0.0<8;8,1>:ud
+
+ mov (8) r101.0<1>:ud acc0.0<8;8,1>:ud
+ mov (8) r110.0<1>:ud acc0.0<8;8,1>:ud
+ mov (8) r119.0<1>:ud acc0.0<8;8,1>:ud
+
+ add (1) r101.1<1>:d acc0.1<0;1,0>:d 16:d
+
+ add (1) r110.0<1>:d acc0.0<0;1,0>:d 16:d
+
+ add (2) r119.0<1>:d acc0.0<2;2,1>:d 16:d
+
+ send (8) udCURR_Y0(0)<1> r92 0x4 0x2890003:ud
+ send (8) udCURR_Y1(0)<1> r101 0x4 0x2890003:ud
+ send (8) udCURR_Y2(0)<1> r110 0x4 0x2890003:ud
+ send (8) udCURR_Y3(0)<1> r119 0x4 0x2890003:ud
+
+
+
+//Module Name : DN_UV_Noise_Detection_UV
+//Author : Tatiya, Rupesh
+//Description : Performs noise detection on 16x16 U and 16x16 V each.
+
+
+
+//Module Name : DN_UV_Move_CURBE_Inline_UV.asm
+//Author : Tatiya, Rupesh
+
+
+
+
+ //Mov CURBE data to another space - so that it can be used as Temp Space --> r1 - r6
+ mov (4) r54.28<1>:ub r2.28<4;4,1>:ub //Dest. YUY2 offset
+ mov (2) r54.5<1>:ud r4.0<4;2,2>:ud //Src YUY2 offset and Origin offset
+ mov (4) r55.28<1>:ub r1.0<4;4,1>:ub
+
+ mov (8) r61.20<1>:ub r1.4<8;8,1>:ub
+ mov (4) r61.28<1>:ub r1.12<4;4,1>:ub
+
+ //Move Inline Data to another space - so that it can be used as Temp Space --> r7
+ mov (4) r62.10<1>:w r7.0<4;4,1>:w
+ mov (4) r63.10<1>:w r7.4<4;4,1>:w
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ mov (1) a0.0:uw 1664:uw
+ mov (1) a0.1:uw 1816:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 1792:uw
+ mov (1) a0.1:uw 1820:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 1920:uw
+ mov (1) a0.1:uw 1848:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2048:uw
+ mov (1) a0.1:uw 1852:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ mov (1) a0.0:uw 2304:uw
+ mov (1) a0.1:uw 1880:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2432:uw
+ mov (1) a0.1:uw 1884:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2560:uw
+ mov (1) a0.1:uw 1912:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2688:uw
+ mov (1) a0.1:uw 1916:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+//Module : DN_UV_Noise_Reduction_UV
+//Author : Tatiya, Rupesh
+//Description : Performs Noise Reduction on 16x16 U and 16x16 V.
+//Tasks : 1. Update weight history
+// 2. Find if it block is motion block
+// 3. Compute Denoised Pixel.
+
+
+
+
+//History is 1+1 byte every 4x4 U and 4x4 V.
+
+ cmp.l.f0.0 (16) null<1>:w ubHIST_UV(0,0)<16;16,1> r61.20<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w ubHIST_UV(0,0)<16;16,1> r61.22<0;2,1>:ub
+
+ mov (16) uwCURBE_TEMP(0)<1> 0:w
+ mov (16) uwCURBE_TEMP(1)<1> 0:w
+
+ //Compute diff betn curr and prev. - First 16 lines
+ // 8 lines here
+ add (16) wDIFF_TEMPORAL(0)<1> ubCURR_UV(2,2)<16;16,1> -ubPREV_UV(0,0)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(1)<1> ubCURR_UV(3,2)<16;16,1> -ubPREV_UV(0,16)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(2)<1> ubCURR_UV(4,2)<16;16,1> -ubPREV_UV(0,32)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(3)<1> ubCURR_UV(5,2)<16;16,1> -ubPREV_UV(0,48)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(4)<1> ubCURR_UV(6,2)<16;16,1> -ubPREV_UV(0,64)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(5)<1> ubCURR_UV(7,2)<16;16,1> -ubPREV_UV(0,80)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(6)<1> ubCURR_UV(8,2)<16;16,1> -ubPREV_UV(0,96)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(7)<1> ubCURR_UV(9,2)<16;16,1> -ubPREV_UV(0,112)<16;16,1> //Diff UV interleaved
+
+ //Update WT HIST
+ (-f0.0) shr (16) uwCURBE_TEMP(0)<1> ubHIST_UV(0,0)<16;16,1> 1:w
+ (f1.0) add (16) uwCURBE_TEMP(2)<1> ubHIST_UV(0,0)<16;16,1> r61.24<0;2,1>:ub
+ (f0.0) mov (16) uwCURBE_TEMP(2)<1> r61.20<0;2,1>:ub
+ (-f0.0.anyv) mov (16) uwCURBE_TEMP(2)<1> ubHIST_UV(0,0)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w ubHIST_UV(0,16)<16;16,1> r61.20<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w ubHIST_UV(0,16)<16;16,1> r61.22<0;2,1>:ub
+
+ //Compute diff betn curr and prev. - First 16 lines
+ // 8 more lines here
+ add (16) wDIFF_TEMPORAL(8)<1> ubCURR_UV(10,2)<16;16,1> -ubPREV_UV(0,128)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(9)<1> ubCURR_UV(11,2)<16;16,1> -ubPREV_UV(0,144)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(10)<1> ubCURR_UV(12,2)<16;16,1> -ubPREV_UV(0,160)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(11)<1> ubCURR_UV(13,2)<16;16,1> -ubPREV_UV(0,176)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(12)<1> ubCURR_UV(14,2)<16;16,1> -ubPREV_UV(0,192)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(13)<1> ubCURR_UV(15,2)<16;16,1> -ubPREV_UV(0,208)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(14)<1> ubCURR_UV(16,2)<16;16,1> -ubPREV_UV(0,224)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(15)<1> ubCURR_UV(17,2)<16;16,1> -ubPREV_UV(0,240)<16;16,1> //Diff UV interleaved
+
+ (-f0.0) shr (16) uwCURBE_TEMP(1)<1> ubHIST_UV(0,16)<16;16,1> 1:w
+ (f1.0) add (16) uwCURBE_TEMP(3)<1> ubHIST_UV(0,16)<16;16,1> r61.24<0;2,1>:ub
+ (f0.0) mov (16) uwCURBE_TEMP(3)<1> r61.20<0;2,1>:ub
+ (-f0.0.anyv) mov(16) uwCURBE_TEMP(3)<1> ubHIST_UV(0,16)<16;16,1>
+
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(0)<16;16,1> (abs)wDIFF_TEMPORAL(1)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(2)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(3)<16;16,1>
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(4)<16;16,1> (abs)wDIFF_TEMPORAL(5)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(6)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(7)<16;16,1>
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(8)<16;16,1> (abs)wDIFF_TEMPORAL(9)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(10)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(2)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(11)<16;16,1>
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(12)<16;16,1> (abs)wDIFF_TEMPORAL(13)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(14)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(3)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(15)<16;16,1>
+
+//Compute diff betn curr and prev. - Second 16 lines
+//13 lines.
+ add (16) wDIFF_TEMPORAL(16)<1> ubCURR_UV(22,2)<16;16,1> -ubPREV_UV(8,0)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(17)<1> ubCURR_UV(23,2)<16;16,1> -ubPREV_UV(8,16)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(18)<1> ubCURR_UV(24,2)<16;16,1> -ubPREV_UV(8,32)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(19)<1> ubCURR_UV(25,2)<16;16,1> -ubPREV_UV(8,48)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(20)<1> ubCURR_UV(26,2)<16;16,1> -ubPREV_UV(8,64)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(21)<1> ubCURR_UV(27,2)<16;16,1> -ubPREV_UV(8,80)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(22)<1> ubCURR_UV(28,2)<16;16,1> -ubPREV_UV(8,96)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(23)<1> ubCURR_UV(29,2)<16;16,1> -ubPREV_UV(8,112)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(24)<1> ubCURR_UV(30,2)<16;16,1> -ubPREV_UV(8,128)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(25)<1> ubCURR_UV(31,2)<16;16,1> -ubPREV_UV(8,144)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(26)<1> ubCURR_UV(32,2)<16;16,1> -ubPREV_UV(8,160)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(27)<1> ubCURR_UV(33,2)<16;16,1> -ubPREV_UV(8,176)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(28)<1> ubCURR_UV(34,2)<16;16,1> -ubPREV_UV(8,192)<16;16,1> //Diff UV interleaved
+
+//3 more lines
+ add (16) wCURBE_TEMP(4)<1> ubCURR_UV(35,2)<16;16,1> -ubPREV_UV(8,208)<16;16,1> //Diff UV interleaved
+ add (16) wCURBE_TEMP(5)<1> ubCURR_UV(36,2)<16;16,1> -ubPREV_UV(8,224)<16;16,1> //Diff UV interleaved
+ add (16) wCURBE_TEMP(6)<1> ubCURR_UV(37,2)<16;16,1> -ubPREV_UV(8,240)<16;16,1> //Diff UV interleaved
+
+ //16x4 to 8x4 - First 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> uwDIFF_TEMPORAL_SUM4x4(2,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(2,2)<4;2,1>
+
+ //8x4 to 4x4 - First 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4_FINAL(0)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1> { AccWrEn }
+
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(16)<16;16,1> (abs)wDIFF_TEMPORAL(17)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(18)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(19)<16;16,1>
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(20)<16;16,1> (abs)wDIFF_TEMPORAL(21)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(22)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(23)<16;16,1>
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(24)<16;16,1> (abs)wDIFF_TEMPORAL(25)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(26)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(2)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(27)<16;16,1>
+
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(28)<16;16,1> (abs)wCURBE_TEMP(4)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wCURBE_TEMP(5)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(3)<1> acc0.0<16;16,1>:uw (abs)wCURBE_TEMP(6)<16;16,1>
+
+ //Find if block is motion block - First 16 lines
+ cmp.g.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_FINAL(0)<16;16,1> r61.26<0;2,1>:ub
+
+ //Move TEMPORAL_SUM4x4 for SIMD16 use later.
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,0)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,2)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,4)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,6)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,8)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,10)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,12)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,14)<0;2,1>
+
+ //Pick Appropriate Weight History Based on motion. - First 16 lines
+ (-f0.0) mov (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(2)<16;16,1>
+
+ //Actual DN - First 16 lines
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(0)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(0)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(2,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(2,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(2,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,0)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,8)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(0)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(0)<1> ubCURR_UV(2,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(0)<1> wDIFF_TEMPORAL(0)<16;16,1> ubCURR_UV(2,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(1)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(1)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(3,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(3,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(3,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,16)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,24)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(1)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(1)<1> ubCURR_UV(3,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(1)<1> wDIFF_TEMPORAL(1)<16;16,1> ubCURR_UV(3,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(2)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(2)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(4,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(4,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(4,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,32)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,40)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(2)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(2)<1> ubCURR_UV(4,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(2)<1> wDIFF_TEMPORAL(2)<16;16,1> ubCURR_UV(4,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(3)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(3)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(5,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(5,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(5,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,48)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,56)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(3)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(3)<1> ubCURR_UV(5,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(3)<1> wDIFF_TEMPORAL(3)<16;16,1> ubCURR_UV(5,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(4)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(4)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(6,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(6,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(6,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,64)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,72)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(4)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(4)<1> ubCURR_UV(6,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(4)<1> wDIFF_TEMPORAL(4)<16;16,1> ubCURR_UV(6,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(5)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(5)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(7,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(7,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(7,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,80)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,88)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(5)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(5)<1> ubCURR_UV(7,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(5)<1> wDIFF_TEMPORAL(5)<16;16,1> ubCURR_UV(7,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(6)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(6)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(8,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(8,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(8,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,96)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,104)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(6)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(6)<1> ubCURR_UV(8,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(6)<1> wDIFF_TEMPORAL(6)<16;16,1> ubCURR_UV(8,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(7)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(7)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(9,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(9,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(9,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,112)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,120)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(7)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(7)<1> ubCURR_UV(9,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(7)<1> wDIFF_TEMPORAL(7)<16;16,1> ubCURR_UV(9,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(8)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(8)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(10,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(10,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(10,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,128)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,136)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(8)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(8)<1> ubCURR_UV(10,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(8)<1> wDIFF_TEMPORAL(8)<16;16,1> ubCURR_UV(10,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(9)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(9)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(11,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(11,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(11,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,144)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,152)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(9)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(9)<1> ubCURR_UV(11,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(9)<1> wDIFF_TEMPORAL(9)<16;16,1> ubCURR_UV(11,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(10)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(10)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(12,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(12,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(12,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,160)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,168)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(10)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(10)<1> ubCURR_UV(12,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(10)<1> wDIFF_TEMPORAL(10)<16;16,1> ubCURR_UV(12,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(11)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(11)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(13,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(13,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(13,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,176)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,184)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(11)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(11)<1> ubCURR_UV(13,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(11)<1> wDIFF_TEMPORAL(11)<16;16,1> ubCURR_UV(13,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(12)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(12)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(14,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(14,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(14,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,192)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,200)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(12)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(12)<1> ubCURR_UV(14,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(12)<1> wDIFF_TEMPORAL(12)<16;16,1> ubCURR_UV(14,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(13)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(13)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(15,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(15,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(15,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,208)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,216)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(13)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(13)<1> ubCURR_UV(15,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(13)<1> wDIFF_TEMPORAL(13)<16;16,1> ubCURR_UV(15,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(14)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(14)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(16,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(16,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(16,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,224)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,232)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(14)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(14)<1> ubCURR_UV(16,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(14)<1> wDIFF_TEMPORAL(14)<16;16,1> ubCURR_UV(16,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(15)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(15)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(17,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(17,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(17,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,240)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,248)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(15)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(15)<1> ubCURR_UV(17,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(15)<1> wDIFF_TEMPORAL(15)<16;16,1> ubCURR_UV(17,2)<16;16,1>
+
+
+ //16x4 to 8x4 - Second 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> uwDIFF_TEMPORAL_SUM4x4(2,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(2,2)<4;2,1>
+
+ //8x4 to 4x4 - Second 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4_FINAL(1)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1> { AccWrEn }
+
+ //Find if block is motion block - Second 16 lines
+ cmp.g.f1.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_FINAL(1)<16;16,1> r61.26<0;2,1>:ub
+
+ //Move TEMPORAL_SUM4x4 for SIMD16 use later.
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,0)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,2)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,4)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,6)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,8)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,10)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,12)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,14)<0;2,1>
+
+ //Pick Appropriate Weight History Based on motion. - Second 16 lines
+ (-f1.0) mov (16) uwCURBE_TEMP(1)<1> uwCURBE_TEMP(3)<16;16,1>
+
+ //Actual DN - Second 16 lines
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(16)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(16)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(22,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(22,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(22,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,0)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,8)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(16)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(16)<1> ubCURR_UV(22,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(16)<1> wDIFF_TEMPORAL(16)<16;16,1> ubCURR_UV(22,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(17)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(17)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(23,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(23,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(23,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,16)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,24)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(17)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(17)<1> ubCURR_UV(23,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(17)<1> wDIFF_TEMPORAL(17)<16;16,1> ubCURR_UV(23,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(18)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(18)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(24,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(24,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(24,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,32)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,40)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(18)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(18)<1> ubCURR_UV(24,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(18)<1> wDIFF_TEMPORAL(18)<16;16,1> ubCURR_UV(24,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(19)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(19)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(25,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(25,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(25,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,48)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,56)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(19)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(19)<1> ubCURR_UV(25,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(19)<1> wDIFF_TEMPORAL(19)<16;16,1> ubCURR_UV(25,2)<16;16,1>
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(20)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(20)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(26,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(26,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(26,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,64)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,72)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(20)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(20)<1> ubCURR_UV(26,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(20)<1> wDIFF_TEMPORAL(20)<16;16,1> ubCURR_UV(26,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(21)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(21)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(27,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(27,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(27,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,80)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,88)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(21)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(21)<1> ubCURR_UV(27,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(21)<1> wDIFF_TEMPORAL(21)<16;16,1> ubCURR_UV(27,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(22)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(22)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(28,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(28,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(28,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,96)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,104)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(22)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(22)<1> ubCURR_UV(28,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(22)<1> wDIFF_TEMPORAL(22)<16;16,1> ubCURR_UV(28,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(23)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(23)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(29,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(29,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(29,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,112)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,120)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(23)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(23)<1> ubCURR_UV(29,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(23)<1> wDIFF_TEMPORAL(23)<16;16,1> ubCURR_UV(29,2)<16;16,1>
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(24)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(24)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(30,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(30,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(30,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,128)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,136)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(24)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(24)<1> ubCURR_UV(30,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(24)<1> wDIFF_TEMPORAL(24)<16;16,1> ubCURR_UV(30,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(25)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(25)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(31,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(31,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(31,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,144)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,152)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(25)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(25)<1> ubCURR_UV(31,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(25)<1> wDIFF_TEMPORAL(25)<16;16,1> ubCURR_UV(31,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(26)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(26)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(32,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(32,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(32,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,160)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,168)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(26)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(26)<1> ubCURR_UV(32,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(26)<1> wDIFF_TEMPORAL(26)<16;16,1> ubCURR_UV(32,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(27)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(27)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(33,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(33,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(33,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,176)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,184)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(27)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(27)<1> ubCURR_UV(33,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(27)<1> wDIFF_TEMPORAL(27)<16;16,1> ubCURR_UV(33,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(28)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(28)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(34,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(34,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(34,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,192)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,200)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(28)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(28)<1> ubCURR_UV(34,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(28)<1> wDIFF_TEMPORAL(28)<16;16,1> ubCURR_UV(34,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wCURBE_TEMP(4)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wCURBE_TEMP(4)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(35,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(35,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(35,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,208)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,216)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wCURBE_TEMP(4)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wCURBE_TEMP(4)<1> ubCURR_UV(35,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wCURBE_TEMP(4)<1> wCURBE_TEMP(4)<16;16,1> ubCURR_UV(35,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wCURBE_TEMP(5)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wCURBE_TEMP(5)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(36,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(36,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(36,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,224)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,232)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wCURBE_TEMP(5)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wCURBE_TEMP(5)<1> ubCURR_UV(36,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wCURBE_TEMP(5)<1> wCURBE_TEMP(5)<16;16,1> ubCURR_UV(36,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wCURBE_TEMP(6)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wCURBE_TEMP(6)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(37,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(37,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(37,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,240)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,248)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wCURBE_TEMP(6)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wCURBE_TEMP(6)<1> ubCURR_UV(37,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wCURBE_TEMP(6)<1> wCURBE_TEMP(6)<16;16,1> ubCURR_UV(37,2)<16;16,1>
+
+ //Pack Weight History WORD -> BYTE
+ mov (16) ubCURBE_TEMP(3,0)<1> ubCURBE_TEMP(0)<32;16,2>
+ mov (16) ubCURBE_TEMP(3,16)<1> ubCURBE_TEMP(1)<32;16,2>
+
+
+
+//Module Name : DN_UV_Compute_BNE_UV
+//Author : Tatiya, Rupesh
+//Description : Computes minimum SOAD for each 16x4 block.
+
+ cmp.l.f0.0 (8) null:w uwSOAD_MIN_8x4(0,12)<16;4,1> uwSOAD_MIN_8x4(2,12)<16;4,1>
+ (f0.0)sel (8) uwCURBE_TEMP(1,0)<1> uwSOAD_MIN_8x4(0,12)<16;4,1> uwSOAD_MIN_8x4(2,12)<16;4,1>
+
+ mov (8) ubCURBE_TEMP(1)<1> ubCURBE_TEMP(1)<16;8,2>
+
+
+
+//Module Name : DN_UV_PL2_Pack_Denoised_UV
+//Name : Tatiya, Rupesh
+//Description : Pack UV denoised data based on PL2 input.
+
+
+
+//Module Name : DN_UV_Pack_Denoised_UV
+//Name : Tatiya, Rupesh
+//Description : Pack UV denoised data based on PL2/PL3/PA.
+
+
+//First 16 lines.
+ mov (16) ubMSGPAYLOAD_UV0(0,0)<1> ubDIFF_TEMPORAL(0)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(0,16)<1> ubDIFF_TEMPORAL(1)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(1,0)<1> ubDIFF_TEMPORAL(2)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(1,16)<1> ubDIFF_TEMPORAL(3)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(2,0)<1> ubDIFF_TEMPORAL(4)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(2,16)<1> ubDIFF_TEMPORAL(5)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(3,0)<1> ubDIFF_TEMPORAL(6)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(3,16)<1> ubDIFF_TEMPORAL(7)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(4,0)<1> ubDIFF_TEMPORAL(8)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(4,16)<1> ubDIFF_TEMPORAL(9)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(5,0)<1> ubDIFF_TEMPORAL(10)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(5,16)<1> ubDIFF_TEMPORAL(11)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(6,0)<1> ubDIFF_TEMPORAL(12)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(6,16)<1> ubDIFF_TEMPORAL(13)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(7,0)<1> ubDIFF_TEMPORAL(14)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV0(7,16)<1> ubDIFF_TEMPORAL(15)<32;16,2>
+
+//Second 16 lines.
+//12 lines first
+ mov (16) ubMSGPAYLOAD_UV1(0,0)<1> ubDIFF_TEMPORAL(16)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(0,16)<1> ubDIFF_TEMPORAL(17)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(1,0)<1> ubDIFF_TEMPORAL(18)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(1,16)<1> ubDIFF_TEMPORAL(19)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(2,0)<1> ubDIFF_TEMPORAL(20)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(2,16)<1> ubDIFF_TEMPORAL(21)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(3,0)<1> ubDIFF_TEMPORAL(22)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(3,16)<1> ubDIFF_TEMPORAL(23)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(4,0)<1> ubDIFF_TEMPORAL(24)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(4,16)<1> ubDIFF_TEMPORAL(25)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(5,0)<1> ubDIFF_TEMPORAL(26)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(5,16)<1> ubDIFF_TEMPORAL(27)<32;16,2>
+
+ //3 lines next
+ mov (16) ubMSGPAYLOAD_UV1(6,0)<1> ubDIFF_TEMPORAL(28)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(6,16)<1> ubCURBE_TEMP(4)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(7,0)<1> ubCURBE_TEMP(5)<32;16,2>
+ mov (16) ubMSGPAYLOAD_UV1(7,16)<1> ubCURBE_TEMP(6)<32;16,2>
+
+
+
+//Module Name : DN_UV_420_Save_Curr_Frame_Y
+//Author : Tatiya, Rupesh
+//Description : Save Curr Frame Y data for 420 Input
+
+
+
+//Module Name : DN_UV_Load_Curr_Frame_Y
+//Author : Tatiya, Rupesh
+//Description : Saves Y or YUY2 of Current frame.
+
+
+
+
+ mov (8) acc0.0<1>:ud r0.0<8;8,1>:ud
+ mov (2) acc0.0<1>:d r62.10<2;2,1>:w
+
+ mov (1) acc0.2<1>:d 0xF000F:ud
+
+ mov (8) r92.0<1>:ud acc0.0<8;8,1>:ud
+
+ mov (8) r101.0<1>:ud acc0.0<8;8,1>:ud
+ mov (8) r110.0<1>:ud acc0.0<8;8,1>:ud
+ mov (8) r119.0<1>:ud acc0.0<8;8,1>:ud
+
+ add (1) r101.1<1>:d acc0.1<0;1,0>:d 16:d
+
+ add (1) r110.0<1>:d acc0.0<0;1,0>:d 16:d
+
+ add (2) r119.0<1>:d acc0.0<2;2,1>:d 16:d
+
+ send (8) null<1>:d r92 0x5 0x120A8018:ud
+ send (8) null<1>:d r101 0x5 0x120A8018:ud
+ send (8) null<1>:d r110 0x5 0x120A8018:ud
+ send (8) null<1>:d r119 0x5 0x120A8018:ud
+
+
+ //TODO - See if History saving can be combined with Curr Frame Save. - rT
+
+
+//Module Name : DN_UV_Save_Hist_UV
+//Author : Tatiya, Rupesh
+//Description : Saves DN history for UV data.
+
+ mov (8) r3.0<1>:ud r0.0<8;8,1>:ud
+ mov (2) r3.0<1>:d r62.12<2;2,1>:w
+ mov (1) r3.2<1>:d 0x30007:ud
+
+ send (8) null<1>:d r3 0x5 0x40A8021:ud
+
+
+
+//Module Name : DN_UV_Save_BNE_UV
+//Author : Tatiya, Rupesh
+//Description : Saves BNE values for 16x16 U and 16x16 V.
+
+ mov (8) r1.0<1>:ud r0.0<8;8,1>:ud
+ mov (2) r1.0<1>:d r63.12<2;2,1>:w
+ mov (1) r1.2<1>:d 0x10003:ud
+
+ send (8) null<1>:d r1 0x5 0x40A8023:ud
+
+
+
+//Module Name : DN_UV_PL2_Save_Curr_Frame_UV
+//Author : Tatiya, Rupesh
+
+
+
+//Module name : DN_UV_Save_Curr_Frame_UV
+//Author : Tatiya, Rupesh
+//Description : Saves Current Frame (UV only).
+
+
+
+
+ mov (8) r74<1>:ud r0.0<8;8,1>:ud
+ mov (8) r83<1>:ud r0.0<8;8,1>:ud
+
+ mov (1) r74.0<1>:d r62.10<0;1,0>:w
+ shr (1) r74.1<1>:d r62.11<0;1,0>:w 1:w
+ mov (1) r74.2<1>:d 0xF000F:ud
+
+ add (1) r83.0<1>:d r62.10<0;1,0>:w 16:d
+ shr (1) r83.1<1>:d r62.11<0;1,0>:w 1:w
+ mov (1) r83.2<1>:d 0xF000F:ud
+
+ send (8) null<1>:d r74 0x5 0x120A8019:ud
+ send (8) null<1>:d r83 0x5 0x120A8019:ud
+
+
+
+//End of Thread message
+
+mov (8) r127<1>:ud r0.0<8;8,1>:ud
+ send (1) null<1>:d r127 0x27 0x02000010
+
+
+ //All sub-routines here
+
+
+// Module Name : Noise_Detection
+// Author : Tatiya, Rupesh
+// Description : Performs noise detection on 32 pixels of U (8x4) and 32 pixels of V (8x4).
+
+DN_UV_NOISE_DETECTION_UV:
+
+// Find Field Block Median
+//
+// Purpose : Find the median value of the nine pixels in the same field
+// which are centered at current pixel.
+//
+// Works on 9 pixels centered at the current pixel
+// NOTE: pixels are within same field.
+// v4 - current pixel
+//
+// v2 v1 v0
+// * * * <--- Different field - not used
+// v5 v4 v3
+// * * * <--- Different field - not used
+// v8 v7 v6
+
+// Algorithm to find median modifies the data.
+// Copy the data needed to calculate median so the original source data stays intact.
+//
+
+//TODO - Change Interleaved implementation to separated one if - , does not work on predication. - rT
+
+//Delete Later - rT
+//mov (1) pCUR_UV:uw 52*32:uw
+
+// v0
+mov (16) ubMEDIAN_TEMP(0,0)<1> r[a0.0,0]<16;16,1>
+// v0
+mov (16) ubMEDIAN_TEMP(0,16)<1> r[a0.0,32]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(1,0)<1> r[a0.0,2]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(1,16)<1> r[a0.0,34]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(2,0)<1> r[a0.0,4]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(2,16)<1> r[a0.0,36]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(3,0)<1> r[a0.0,64]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(3,16)<1> r[a0.0,96]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(4,0)<1> r[a0.0,66]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(4,16)<1> r[a0.0,98]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(5,0)<1> r[a0.0,68]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(5,16)<1> r[a0.0,100]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(6,0)<1> r[a0.0,128]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(6,16)<1> r[a0.0,160]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(7,0)<1> r[a0.0,130]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(7,16)<1> r[a0.0,162]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(8,0)<1> r[a0.0,132]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(8,16)<1> r[a0.0,164]<16;16,1>
+
+//TODO - Optimize one instruction here.
+add (1) a0.0:uw a0.0<0;1,0>:uw 64:uw
+// v0
+mov (16) ubMEDIAN_TEMP(9,0)<1> r[a0.0,0]<16;16,1>
+// v0
+mov (16) ubMEDIAN_TEMP(9,16)<1> r[a0.0,32]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(10,0)<1> r[a0.0,2]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(10,16)<1> r[a0.0,34]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(11,0)<1> r[a0.0,4]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(11,16)<1> r[a0.0,36]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(12,0)<1> r[a0.0,64]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(12,16)<1> r[a0.0,96]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(13,0)<1> r[a0.0,66]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(13,16)<1> r[a0.0,98]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(14,0)<1> r[a0.0,68]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(14,16)<1> r[a0.0,100]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(15,0)<1> r[a0.0,128]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(15,16)<1> r[a0.0,160]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(16,0)<1> r[a0.0,130]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(16,16)<1> r[a0.0,162]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(17,0)<1> r[a0.0,132]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(17,16)<1> r[a0.0,164]<16;16,1>
+
+//TODO - Optimize one instruction here.
+add (1) a0.0:uw a0.0<0;1,0>:uw 64:uw
+
+// MedianSwap
+//
+// MedianSwap(inOutLeft, inOutRight)
+// {
+// if (inOutLeft > inOutRight)
+// {
+// temp = inOutLeft
+// inOutLeft = inOutRight
+// inOutRight = temp
+// }
+// }
+
+// MedianSwap(v1, v2) - U
+// MedianSwap(v4, v5) - U
+// MedianSwap(v1, v2) - V
+// MedianSwap(v4, v5) - V
+
+cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(1,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(5,0)<32;16,2>
+cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(1,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(1,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(1,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(1,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+(f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(5,0)<32;16,2>
+(f1.0) mov (16) ubMEDIAN_TEMP(1,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+(f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubTEMP1(0,0)<16;16,1>
+(f0.1) mov (16) ubMEDIAN_TEMP(5,0)<2> ubTEMP1(0,16)<16;16,1>
+(f1.0) mov (16) ubMEDIAN_TEMP(2,1)<2> ubTEMP1(1,0)<16;16,1>
+(f1.1) mov (16) ubMEDIAN_TEMP(5,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// MedianSwap(v7, v8) - U
+// MedianSwap(v0, v1) - U
+// MedianSwap(v7, v8) - V
+// MedianSwap(v0, v1) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(7,0)<32;16,2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(0,0)<32;16,2> ubMEDIAN_TEMP(1,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(7,1)<32;16,2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(0,1)<32;16,2> ubMEDIAN_TEMP(1,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(7,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(0,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(7,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(0,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(7,0)<2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(0,0)<2> ubMEDIAN_TEMP(1,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(7,1)<2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(0,1)<2> ubMEDIAN_TEMP(1,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(8,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(1,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(8,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(1,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v3, v4) - U
+ // MedianSwap(v6, v7) - U
+ // MedianSwap(v3, v4) - V
+ // MedianSwap(v6, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(3,0)<32;16,2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(6,0)<32;16,2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(3,1)<32;16,2> ubMEDIAN_TEMP(4,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(6,1)<32;16,2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(3,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(6,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(3,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(6,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(3,0)<2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(6,0)<2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(3,1)<2> ubMEDIAN_TEMP(4,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(6,1)<2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(7,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(4,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(7,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v1, v2) - U
+ // MedianSwap(v4, v5) - U
+ // MedianSwap(v1, v2) - V
+ // MedianSwap(v4, v5) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(1,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(1,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(1,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(1,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(1,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(1,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(5,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(2,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(5,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v7, v8) - U
+ // MedianSwap(v0, v3) - U
+ // MedianSwap(v7, v8) - V
+ // MedianSwap(v0, v3) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(7,0)<32;16,2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(0,0)<32;16,2> ubMEDIAN_TEMP(3,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(7,1)<32;16,2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(0,1)<32;16,2> ubMEDIAN_TEMP(3,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(7,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(0,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(7,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(0,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(7,0)<2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(0,0)<2> ubMEDIAN_TEMP(3,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(7,1)<2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(0,1)<2> ubMEDIAN_TEMP(3,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(8,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(3,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(8,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(3,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // NOTE:
+ // Compare v0 to v6 to find the minimum.
+ // Store the minimum for future use.
+ //TODO - Find if MIN is needed.
+ //cmp.l.f0.0 (16) null:w ubMEDIAN_TEMP(%1+0,0)<32;16,2> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //cmp.l.f1.0 (16) null:w ubMEDIAN_TEMP(%1+0,1)<32;16,2> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+ //(f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,0)<32;16,2>
+ //(f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,1)<32;16,2>
+ //(-f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //(-f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+
+ // MedianSwap(v5, v8) - U
+ // MedianSwap(v4, v7) - U
+ // MedianSwap(v5, v8) - V
+ // MedianSwap(v4, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(5,0)<32;16,2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(5,1)<32;16,2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(5,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(5,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(5,0)<2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(5,1)<2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(8,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(7,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(8,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(7,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// NOTE:
+// Compare v2 to v8 to find the maximum.
+// Store the maximum for future use.
+ //TODO - Find if MAX is needed.
+// cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(%1+2,0)<32;16,2> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+// cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(%1+2,1)<32;16,2> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+//(f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,0)<32;16,2>
+//(f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,1)<32;16,2>
+//(-f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+//(-f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+
+// MedianSwap(v3, v6) - U
+// MedianSwap(v1, v4) - U
+// MedianSwap(v3, v6) - V
+// MedianSwap(v1, v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(3,0)<32;16,2> ubMEDIAN_TEMP(6,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(1,0)<32;16,2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(3,1)<32;16,2> ubMEDIAN_TEMP(6,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(1,1)<32;16,2> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(6,0)<2> ubMEDIAN_TEMP(3,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(1,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(6,1)<2> ubMEDIAN_TEMP(3,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(1,1)<32;16,2>
+
+ // MedianSwap(v2,v5) - U
+ // MedianSwap(v4,v7) - U
+ // MedianSwap(v2,v5) - V
+ // MedianSwap(v4,v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(2,0)<32;16,2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(2,1)<32;16,2> ubMEDIAN_TEMP(5,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(2,1)<2> ubMEDIAN_TEMP(5,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(2,1)<2> ubTEMP1(0,16)<16;16,1>
+
+ // MedianSwap(v6,v4) - U
+ // MedianSwap(v6,v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(6,0)<32;16,2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(6,1)<32;16,2> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(6,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(6,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(10,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(14,0)<32;16,2>
+cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(10,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(10,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(10,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(10,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+(f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(14,0)<32;16,2>
+(f1.0) mov (16) ubMEDIAN_TEMP(10,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+(f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubTEMP1(0,0)<16;16,1>
+(f0.1) mov (16) ubMEDIAN_TEMP(14,0)<2> ubTEMP1(0,16)<16;16,1>
+(f1.0) mov (16) ubMEDIAN_TEMP(11,1)<2> ubTEMP1(1,0)<16;16,1>
+(f1.1) mov (16) ubMEDIAN_TEMP(14,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// MedianSwap(v7, v8) - U
+// MedianSwap(v0, v1) - U
+// MedianSwap(v7, v8) - V
+// MedianSwap(v0, v1) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(16,0)<32;16,2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(9,0)<32;16,2> ubMEDIAN_TEMP(10,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(16,1)<32;16,2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(9,1)<32;16,2> ubMEDIAN_TEMP(10,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(16,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(9,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(16,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(9,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(16,0)<2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(9,0)<2> ubMEDIAN_TEMP(10,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(16,1)<2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(9,1)<2> ubMEDIAN_TEMP(10,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(17,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(10,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(17,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(10,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v3, v4) - U
+ // MedianSwap(v6, v7) - U
+ // MedianSwap(v3, v4) - V
+ // MedianSwap(v6, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(12,0)<32;16,2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(15,0)<32;16,2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(12,1)<32;16,2> ubMEDIAN_TEMP(13,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(15,1)<32;16,2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(12,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(15,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(12,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(15,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(12,0)<2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(15,0)<2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(12,1)<2> ubMEDIAN_TEMP(13,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(15,1)<2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(16,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(13,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(16,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v1, v2) - U
+ // MedianSwap(v4, v5) - U
+ // MedianSwap(v1, v2) - V
+ // MedianSwap(v4, v5) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(10,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(10,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(10,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(10,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(10,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(10,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(14,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(11,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(14,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v7, v8) - U
+ // MedianSwap(v0, v3) - U
+ // MedianSwap(v7, v8) - V
+ // MedianSwap(v0, v3) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(16,0)<32;16,2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(9,0)<32;16,2> ubMEDIAN_TEMP(12,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(16,1)<32;16,2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(9,1)<32;16,2> ubMEDIAN_TEMP(12,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(16,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(9,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(16,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(9,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(16,0)<2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(9,0)<2> ubMEDIAN_TEMP(12,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(16,1)<2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(9,1)<2> ubMEDIAN_TEMP(12,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(17,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(12,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(17,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(12,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // NOTE:
+ // Compare v0 to v6 to find the minimum.
+ // Store the minimum for future use.
+ //TODO - Find if MIN is needed.
+ //cmp.l.f0.0 (16) null:w ubMEDIAN_TEMP(%1+0,0)<32;16,2> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //cmp.l.f1.0 (16) null:w ubMEDIAN_TEMP(%1+0,1)<32;16,2> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+ //(f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,0)<32;16,2>
+ //(f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,1)<32;16,2>
+ //(-f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //(-f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+
+ // MedianSwap(v5, v8) - U
+ // MedianSwap(v4, v7) - U
+ // MedianSwap(v5, v8) - V
+ // MedianSwap(v4, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(14,0)<32;16,2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(14,1)<32;16,2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(14,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(14,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(14,0)<2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(14,1)<2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(17,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(16,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(17,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(16,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// NOTE:
+// Compare v2 to v8 to find the maximum.
+// Store the maximum for future use.
+ //TODO - Find if MAX is needed.
+// cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(%1+2,0)<32;16,2> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+// cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(%1+2,1)<32;16,2> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+//(f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,0)<32;16,2>
+//(f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,1)<32;16,2>
+//(-f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+//(-f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+
+// MedianSwap(v3, v6) - U
+// MedianSwap(v1, v4) - U
+// MedianSwap(v3, v6) - V
+// MedianSwap(v1, v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(12,0)<32;16,2> ubMEDIAN_TEMP(15,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(10,0)<32;16,2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(12,1)<32;16,2> ubMEDIAN_TEMP(15,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(10,1)<32;16,2> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(15,0)<2> ubMEDIAN_TEMP(12,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(10,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(15,1)<2> ubMEDIAN_TEMP(12,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(10,1)<32;16,2>
+
+ // MedianSwap(v2,v5) - U
+ // MedianSwap(v4,v7) - U
+ // MedianSwap(v2,v5) - V
+ // MedianSwap(v4,v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(11,0)<32;16,2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(11,1)<32;16,2> ubMEDIAN_TEMP(14,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(11,1)<2> ubMEDIAN_TEMP(14,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(11,1)<2> ubTEMP1(0,16)<16;16,1>
+
+ // MedianSwap(v6,v4) - U
+ // MedianSwap(v6,v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(15,0)<32;16,2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(15,1)<32;16,2> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(15,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(15,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+// Sobel Value calculation for the current pixel v4
+// v2 v1 v0
+// * * * <--- Different field - not used
+// v5 v4 v3
+// * * * <--- Different field - not used
+// v8 v7 v6
+//
+// Gx = -v0 - 2*v3 - v6 + v2 + 2*v5 + v8
+// Gy = v0 + 2*v1 + v2 - v6 - 2*v7 - v8
+//
+// Sobel = (|Gx| + |Gy|) >> 3
+
+//TODO - Change Later - rT
+add (1) a0.0:uw a0.0<0;1,0>:uw -128:uw
+
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,64]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,132]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,0]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,128]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,4]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(0)<1> r[a0.0,68]<16;16,1>:ub 2:w
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,96]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,164]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,32]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,160]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,36]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(1)<1> r[a0.0,100]<16;16,1>:ub 2:w
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,128]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,196]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,64]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,192]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,68]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(2)<1> r[a0.0,132]<16;16,1>:ub 2:w
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,160]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,228]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,96]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,224]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,100]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(3)<1> r[a0.0,164]<16;16,1>:ub 2:w
+
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,2]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,0]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,132]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,4]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,128]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,130]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(0)<16;16,1>
+
+shr (16) uwSOBEL(0)<1> acc0.0<16;16,1>:uw 3:uw
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,34]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,32]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,164]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,36]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,160]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,162]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(1)<16;16,1>
+
+shr (16) uwSOBEL(1)<1> acc0.0<16;16,1>:uw 3:uw
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,66]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,64]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,196]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,68]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,192]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,194]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(2)<16;16,1>
+
+shr (16) uwSOBEL(2)<1> acc0.0<16;16,1>:uw 3:uw
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,98]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,96]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,228]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,100]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,224]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,226]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(3)<16;16,1>
+
+shr (16) uwSOBEL(3)<1> acc0.0<16;16,1>:uw 3:uw
+
+//Mov Median in CURBE_TEMP to free up temp space.
+mov (16) ubMEDIAN(0,0)<1> ubMEDIAN_TEMP(4,0)<16;16,1>
+mov (16) ubMEDIAN(0,16)<1> ubMEDIAN_TEMP(4,16)<16;16,1>
+mov (16) ubMEDIAN(0,32)<1> ubMEDIAN_TEMP(13,0)<16;16,1>
+mov (16) ubMEDIAN(0,48)<1> ubMEDIAN_TEMP(13,16)<16;16,1>
+
+// Find:
+// absDiff = abs(ubCurY - ubMedian)
+// Find the difference between pixel and median value.
+
+//Median is interleaved. So difference is also interleaved.
+
+//------------------------------------------------------------------------------------------
+//Process 16 U and 16 V pixels here and rest later.
+// first row - v0,v1,v2
+add (16) wDIFF(0)<1> r[a0.0,0]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(1)<1> r[a0.0,2]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(2)<1> r[a0.0,4]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+
+// second row - v3,v4,v5
+add (16) wDIFF(3)<1> r[a0.0,64]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(4)<1> r[a0.0,66]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(5)<1> r[a0.0,68]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+
+// third row - v6,v7,v8
+add (16) wDIFF(6)<1> r[a0.0,128]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(7)<1> r[a0.0,130]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(8)<1> r[a0.0,132]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+// first row - v0,v1,v2
+add (16) wDIFF(9)<1> r[a0.0,32]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(10)<1> r[a0.0,34]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(11)<1> r[a0.0,36]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+
+// second row - v3,v4,v5
+add (16) wDIFF(12)<1> r[a0.0,96]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(13)<1> r[a0.0,98]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(14)<1> r[a0.0,100]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+
+// third row - v6,v7,v8
+add (16) wDIFF(15)<1> r[a0.0,160]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(16)<1> r[a0.0,162]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(17)<1> r[a0.0,164]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+
+//TODO - Change Later - rT
+add (1) a0.0:uw a0.0<0;1,0>:uw 64:uw
+
+// Find sum of all absolute differences AND
+// maximum absolute difference for 16 U and 16 V here.
+//First 2 rows of 8x4
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(2)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(3)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(4)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(5)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(6)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(7)<16;16,1>
+ add (16) uwSOAD(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF(8)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//------------
+ //DIFF(0-7) is not needed here. Populate it.
+ // first row - v0,v1,v2
+ add (16) wDIFF(0)<1> r[a0.0,0]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(1)<1> r[a0.0,2]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(2)<1> r[a0.0,4]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+
+ // second row - v3,v4,v5
+ add (16) wDIFF(3)<1> r[a0.0,64]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(4)<1> r[a0.0,66]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(5)<1> r[a0.0,68]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+
+ // third row - v6,v7
+ add (16) wDIFF(6)<1> r[a0.0,128]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(7)<1> r[a0.0,130]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+//------------
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(0)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+
+//------------
+ //Load v8 - DIFF(8)
+ add (16) wDIFF(8)<1> r[a0.0,132]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+//------------
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(11)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(12)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(13)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(14)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(15)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(16)<16;16,1>
+ add (16) uwSOAD(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF(17)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//------------
+ //DIFF(0-7) is not needed here. Populate it.
+ // first row - v0,v1,v2
+ add (16) wDIFF(9)<1> r[a0.0,32]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(10)<1> r[a0.0,34]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(11)<1> r[a0.0,36]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+
+ // second row - v3,v4,v5
+ add (16) wDIFF(12)<1> r[a0.0,96]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(13)<1> r[a0.0,98]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(14)<1> r[a0.0,100]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+
+ // third row - v6,v7
+ add (16) wDIFF(15)<1> r[a0.0,160]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(16)<1> r[a0.0,162]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+//------------
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(1)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+
+//------------
+ //Load v8 - DIFF(8)
+ add (16) wDIFF(17)<1> r[a0.0,164]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+//------------
+
+//if ((sobel_edge_measure < m_SobelEdgeThreshold) && ((block_max-block_min) < m_LocalDiffThreshold))
+// if (sigma_mb_min > sigma)
+// sigma_mb_min = sigma;
+
+//NOTE: block_min is always zero as median is one of the value in 3x3 block. So no need o calculate it.
+// So just do -
+//if ((sobel_edge_measure < m_SobelEdgeThreshold) && ((block_max) < m_LocalDiffThreshold) && ( sigma < sigma_mb_min))
+// sigma_mb_min = sigma;
+
+//We are processing 32 bytes of U and 32 bytes of V - each of size 8x4.
+//Compare first 8 bytes with max possible (255).
+//Start above condition from second 8 bytes.
+
+//TODO - Change Later - rT
+// mov (1) pCUR_MIN_SOAD_8x4:uw 1752:uw //r54.24:ub
+
+//First row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(0)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(0)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(0)<16;16,1> 255:uw
+(f0.0) sel (16) uwSOBEL(0)<1> uwSOAD(0)<16;16,1> 255:uw
+
+//Second row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(1)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(1)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(1)<16;16,1> uwSOBEL(0)<16;16,1>
+(f0.0) mov (16) uwSOBEL(0)<1> uwSOAD(1)<16;16,1>
+
+// Find sum of all absolute differences AND
+// maximum absolute difference for 16 U and 16 V here.
+//Second 2 rows of 8x4
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(2)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(3)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(4)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(5)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(6)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(7)<16;16,1>
+ add (16) uwSOAD(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF(8)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(0)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(11)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(12)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(13)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(14)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(15)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(16)<16;16,1>
+ add (16) uwSOAD(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF(17)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(1)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+
+//Third row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(2)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(0)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(0)<16;16,1> uwSOBEL(0)<16;16,1>
+(f0.0) mov (16) uwSOBEL(0)<1> uwSOAD(0)<16;16,1>
+
+//Fourth row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(3)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(1)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(1)<16;16,1> uwSOBEL(0)<16;16,1>
+(f0.0) mov (16) uwSOBEL(0)<1> uwSOAD(1)<16;16,1>
+
+ cmp.l.f0.0 (8) null:uw uwSOBEL(0,0)<8;8,1> uwSOBEL(0,8)<8;8,1>
+(f0.0) sel (8) uwSOBEL(0)<1> uwSOBEL(0,0)<8;8,1> uwSOBEL(0,8)<8;8,1>
+
+ cmp.l.f0.0 (4) null:uw uwSOBEL(0,0)<4;4,1> uwSOBEL(0,4)<4;4,1>
+(f0.0) sel (4) uwSOBEL(0)<1> uwSOBEL(0,0)<4;4,1> uwSOBEL(0,4)<4;4,1>
+
+ cmp.l.f0.0 (2) null:uw uwSOBEL(0,0)<2;2,1> uwSOBEL(0,2)<2;2,1>
+(f0.0) sel (2) r[a0.1,0]<1>:uw uwSOBEL(0,0)<2;2,1> uwSOBEL(0,2)<2;2,1>
+
+
+
+
+
+
+// End of common.inc
+
+mov (1) ip:ud r7.7<0;1,0>:d
+
+
+.end_code
+.end_kernel
--- /dev/null
+/*
+ * 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.
+ */
+// 113 // Total instruction count
+// 1 // Total kernel count
+
+.kernel NV12_DN_422CP
+.code
+
+
+
+// FileName: DN_PL_Core.asm
+// Author: Vivek Kumar
+// Description: Tasks for DN only case (16x8 block) for planar 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 0x45E8003: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_Load_UV_NV12_16x8.asm
+// Author: Vivek Kumar
+// Description: Read UV for 16x8 block through DATAPORT
+
+
+
+// FileName: UVCopy_Load_16x8.asm
+// Author: Vivek Kumar
+// Description: Read UV for 16x8 block through DATAPORT
+
+//CHANGE : Read extra UV data to convert to 422. -rT
+//we are reading extra data in ALL cases irrespective of whether upsampling is reqd or not later on, to keep things simple.
+
+
+ add (2) r27.0<1>:d r7.0<2;2,1>:w r4.4<2;2,1>:w // Source Y Block origin
+ asr (1) r27.1<1>:d r27.1<0;1,0>:d 1:w { NoDDClr } // U/V block origin should be half of Y's
+ mov (1) r27.2<1>:ud 0x4000F:ud { NoDDChk } // U/V block width and height (8x5)
+ mov (8) mudMSGHDR_UVCOPY(0)<1> r27.0<8;8,1>:ud
+ send (8) udDNDI_UV_RESP(0)<1> r36 0x4 0x2390001:ud
+
+ //Update Header for Save
+ mov (1) mudMSGHDR_UVCOPY(0,2)<1> 0x3000F:ud // U/V block width and height (8x4)
+
+
+
+// FileName: DN_Save_Y_16x8.asm
+// Author: Vivek Kumar
+// Description: Save one 16x8 blocks of Y channel of DN output for reference
+
+
+mov (8) mudDN_Y_OUT(0,0)<1> r0<8;8,1>:ud // message header
+mov (2) mudDN_Y_OUT(0,0)<1> r7.0<2;2,1>:w { NoDDClr } // X origin
+mov (1) mudDN_Y_OUT(0,2)<1> 0x7000F:ud { NoDDChk } // block width and height (16x8)
+
+//send out data through data port
+send (8) null<1>:d mudDN_Y_OUT 0x5 0xA0A8018:ud
+
+
+
+// FileName: DN_Save_UV_NV12_16x8.asm
+// Author: Vivek Kumar
+// Description: Save UV for 16x8 block through DATAPORT
+
+
+
+// FileName: UVCopy_Save_16x8.asm
+// Author: Vivek Kumar
+// Description: Save UV for 16x8 block through DATAPORT
+
+
+//Reuse the header from Load component
+//Header is modified at the end of load - to be usable for save.
+
+ mov (8) mudMSGHDR_UVCOPY(1)<1> udDNDI_UV_RESP(0)<8;8,1>
+ mov (8) mudMSGHDR_UVCOPY(2)<1> udDNDI_UV_RESP(1)<8;8,1>
+ send (8) null<1>:d r36 0x5 0x60A8019:ud
+
+
+
+// FileName: DN_Upsample_UV_NV12_16x8.asm
+// Author: Tatiya, Rupesh
+// Description: Upconvert 420 UV to 422
+
+
+
+// FileName: UVCopy_Upsample_UV_16x8.asm
+// Author: Tatiya, Rupesh
+// Description: Convert 42X UV to 422 - to be used for IECP.
+
+
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(0) ubDNDI_UV_RESP(0,0)<16;16,1> ubDNDI_UV_RESP(0,0)<16;16,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(1) ubDNDI_UV_RESP(0,0)<16;16,1> ubDNDI_UV_RESP(0,16)<16;16,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(2) ubDNDI_UV_RESP(0,16)<16;16,1> ubDNDI_UV_RESP(0,16)<16;16,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(3) ubDNDI_UV_RESP(0,16)<16;16,1> ubDNDI_UV_RESP(0,32)<16;16,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(4) ubDNDI_UV_RESP(0,32)<16;16,1> ubDNDI_UV_RESP(0,32)<16;16,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(5) ubDNDI_UV_RESP(0,32)<16;16,1> ubDNDI_UV_RESP(0,48)<16;16,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(6) ubDNDI_UV_RESP(0,48)<16;16,1> ubDNDI_UV_RESP(0,48)<16;16,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(7) ubDNDI_UV_RESP(0,48)<16;16,1> ubDNDI_UV_RESP(0,64)<16;16,1>
+
+ mov (16) ubDNDI_RESP(5,1)<2> ubDNDI_UVCOPY_TEMP(0,0)<32;8,4> { NoDDClr } //Copy U data
+ mov (16) ubDNDI_RESP(5,0)<2> ubDNDI_UVCOPY_TEMP(0,2)<32;8,4> { NoDDChk } //Copy V data
+ mov (16) ubDNDI_RESP(5,33)<2> ubDNDI_UVCOPY_TEMP(2,0)<32;8,4> { NoDDClr } //Copy U data
+ mov (16) ubDNDI_RESP(5,32)<2> ubDNDI_UVCOPY_TEMP(2,2)<32;8,4> { NoDDChk } //Copy V data
+ mov (16) ubDNDI_RESP(5,65)<2> ubDNDI_UVCOPY_TEMP(4,0)<32;8,4> { NoDDClr } //Copy U data
+ mov (16) ubDNDI_RESP(5,64)<2> ubDNDI_UVCOPY_TEMP(4,2)<32;8,4> { NoDDChk } //Copy V data
+ mov (16) ubDNDI_RESP(5,97)<2> ubDNDI_UVCOPY_TEMP(6,0)<32;8,4> { NoDDClr } //Copy U data
+ mov (16) ubDNDI_RESP(5,96)<2> ubDNDI_UVCOPY_TEMP(6,2)<32;8,4> { NoDDChk } //Copy V data
+
+
+
+// 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
--- /dev/null
+/*
+ * 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.
+ */
+// 40 // Total instruction count
+// 1 // Total kernel count
+
+.kernel NV12_DN_NV12
+.code
+
+
+
+// FileName: DN_PL_Core.asm
+// Author: Vivek Kumar
+// Description: Tasks for DN only case (16x8 block) for planar 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 0x45E8003: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_Load_UV_NV12_16x8.asm
+// Author: Vivek Kumar
+// Description: Read UV for 16x8 block through DATAPORT
+
+
+
+// FileName: UVCopy_Load_16x8.asm
+// Author: Vivek Kumar
+// Description: Read UV for 16x8 block through DATAPORT
+
+//CHANGE : Read extra UV data to convert to 422. -rT
+//we are reading extra data in ALL cases irrespective of whether upsampling is reqd or not later on, to keep things simple.
+
+
+ add (2) r27.0<1>:d r7.0<2;2,1>:w r4.4<2;2,1>:w // Source Y Block origin
+ asr (1) r27.1<1>:d r27.1<0;1,0>:d 1:w { NoDDClr } // U/V block origin should be half of Y's
+ mov (1) r27.2<1>:ud 0x4000F:ud { NoDDChk } // U/V block width and height (8x5)
+ mov (8) mudMSGHDR_UVCOPY(0)<1> r27.0<8;8,1>:ud
+ send (8) udDNDI_UV_RESP(0)<1> r36 0x4 0x2390001:ud
+
+ //Update Header for Save
+ mov (1) mudMSGHDR_UVCOPY(0,2)<1> 0x3000F:ud // U/V block width and height (8x4)
+
+
+
+// FileName: DN_Save_Y_16x8.asm
+// Author: Vivek Kumar
+// Description: Save one 16x8 blocks of Y channel of DN output for reference
+
+
+mov (8) mudDN_Y_OUT(0,0)<1> r0<8;8,1>:ud // message header
+mov (2) mudDN_Y_OUT(0,0)<1> r7.0<2;2,1>:w { NoDDClr } // X origin
+mov (1) mudDN_Y_OUT(0,2)<1> 0x7000F:ud { NoDDChk } // block width and height (16x8)
+
+//send out data through data port
+send (8) null<1>:d mudDN_Y_OUT 0x5 0xA0A8018:ud
+
+
+
+// FileName: DN_Save_UV_NV12_16x8.asm
+// Author: Vivek Kumar
+// Description: Save UV for 16x8 block through DATAPORT
+
+
+
+// FileName: UVCopy_Save_16x8.asm
+// Author: Vivek Kumar
+// Description: Save UV for 16x8 block through DATAPORT
+
+
+//Reuse the header from Load component
+//Header is modified at the end of load - to be usable for save.
+
+ mov (8) mudMSGHDR_UVCOPY(1)<1> udDNDI_UV_RESP(0)<8;8,1>
+ mov (8) mudMSGHDR_UVCOPY(2)<1> udDNDI_UV_RESP(1)<8;8,1>
+ send (8) null<1>:d r36 0x5 0x60A8019: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
+/*
+ * 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.
+ */
// 9 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 7 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 7 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 7 // Total instruction count
// 1 // Total kernel count
--- /dev/null
+/*
+ * 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.
+ */
+// 87 // Total instruction count
+// 1 // Total kernel count
+
+.kernel PA_DI_422CP
+.code
+
+
+
+// FileName: DI.asm
+// Author: Vivek Kumar
+// Description: Tasks for DI only 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=====================================
+
+
+//=======================================================================
+
+
+//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 0x4AE8003: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(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> r7.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2
+mov (1) mudMSGHDR_STMM(0,1)<1> r7.1<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: 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> 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: 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 r7.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
+mov (1) r27.1<1>:ud r7.1<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 r7.26<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
+
+
+
+//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
--- /dev/null
+/*
+ * 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.
+ */
+// 57 // Total instruction count
+// 1 // Total kernel count
+
+.kernel PA_DI_PA
+.code
+
+
+
+// FileName: DI.asm
+// Author: Vivek Kumar
+// Description: Tasks for DI only 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=====================================
+
+
+//=======================================================================
+
+
+//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 0x4AE8003: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(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> r7.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2
+mov (1) mudMSGHDR_STMM(0,1)<1> r7.1<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: 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> 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: DI_Save_PA_16x4.asm
+// Author: Vivek Kumar
+// Description: Save two 16x4 blocks of DI output in Packed format
+
+
+add (4) a0.4<1>:uw r2.28<4;4,1>:ub 608:w // Initial Y,U,V offset in YUV422 block; it starts at m20
+
+mov (8) r27.0<1>:ud r0.0<8;8,1>:ud
+shl (1) r27.0<1>:d r7.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
+mov (1) r27.1<1>:d r7.1<0;1,0>:w { NoDDClr, NoDDChk } // Block origin
+mov (1) r27.2<1>:ud 0x3001F:ud { NoDDChk } // Block width and height (32x8)
+
+//prepare the message headers
+mov (8) r18.0<1>:ud r27<8;8,1>:ud
+mov (8) r23.0<1>:ud r27<8;8,1>:ud
+
+// Pack 2nd field Y
+ mov (16) r[a0.4, 0]<2> ubDNDI_RESP(0,0) { NoDDClr }
+ mov (16) r[a0.4, 32]<2> ubDNDI_RESP(0,16) { NoDDClr }
+ mov (16) r[a0.4, 64]<2> ubDNDI_RESP(0,32) { NoDDClr }
+ mov (16) r[a0.4, 96]<2> ubDNDI_RESP(0,48) { NoDDClr }
+// Pack 2nd field U
+ mov (8) r[a0.5, 0]<4> ubDNDI_RESP(2,1)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 32]<4> ubDNDI_RESP(2,17)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 64]<4> ubDNDI_RESP(2,33)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 96]<4> ubDNDI_RESP(2,49)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+// Pack 2nd field V
+ mov (8) r[a0.6, 0]<4> ubDNDI_RESP(2,0)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 32]<4> ubDNDI_RESP(2,16)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 64]<4> ubDNDI_RESP(2,32)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 96]<4> ubDNDI_RESP(2,48)<16;8,2> { NoDDChk } //Vpixels
+
+// Pack 1st field Y
+ mov (16) r[a0.4, 160]<2> ubDNDI_RESP(4,0) { NoDDClr }
+ mov (16) r[a0.4, 192]<2> ubDNDI_RESP(4,16) { NoDDClr }
+ mov (16) r[a0.4, 224]<2> ubDNDI_RESP(4,32) { NoDDClr }
+ mov (16) r[a0.4, 256]<2> ubDNDI_RESP(4,48) { NoDDClr }
+// Pack 1st field U
+ mov (8) r[a0.5, 160]<4> ubDNDI_RESP(6,1)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 192]<4> ubDNDI_RESP(6,17)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 224]<4> ubDNDI_RESP(6,33)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 256]<4> ubDNDI_RESP(6,49)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+// Pack 1st field V
+ mov (8) r[a0.6, 160]<4> ubDNDI_RESP(6,0)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 192]<4> ubDNDI_RESP(6,16)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 224]<4> ubDNDI_RESP(6,32)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 256]<4> ubDNDI_RESP(6,48)<16;8,2> { NoDDChk } //Vpixels
+
+//save the previous frame
+send (8) null<1>:d r18.0 0x5 0xA0A801B:ud
+
+//save the current frame
+send (8) null<1>:d r23.0 0x5 0xA0A801E: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
--- /dev/null
+/*
+ * 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.
+ */
+// 127 // Total instruction count
+// 1 // Total kernel count
+
+.kernel PA_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=====================================
+
+
+//=======================================================================
+
+
+//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 0x4CE8003: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(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> r7.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2
+mov (1) mudMSGHDR_STMM(0,1)<1> r7.1<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 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 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> 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> 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_Save_PA_16x4.asm
+// Author: Vivek Kumar
+// Description: Save one 16x4 blocks of DN output in Packed format for reference
+
+
+ // check top/bottom field first
+cmp.e.f0.0 (1) null<1>:w r1.28<0;1,0>:ub 1:w
+
+add (4) a0.4<1>:uw r4.0<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> 0x3001F:ud { NoDDChk } // block width and height (32x8)
+
+(f0.0) jmpi (1) TOP_FIELD_FIRST
+
+BOTTOM_FIELD_FIRST:
+ mov (16) r[a0.4, 0]<2>:ub ubDNDI_RESP(10,0) { NoDDClr } // 2nd field luma from current frame (line 0,2)
+ mov (16) r[a0.4, 32]<2>:ub ubDNDI_RESP(4,16) { NoDDClr } // 1st field luma from current frame (line 1,3)
+ mov (16) r[a0.4, 64]<2>:ub ubDNDI_RESP(10,16) { NoDDClr } // 2nd field luma from current frame (line 0,2)
+ mov (16) r[a0.4, 96]<2>:ub ubDNDI_RESP(5,16) { NoDDClr } // 1st field luma from current frame (line 1,3)
+ mov (8) r[a0.5, 0]<4>:ub ubDNDI_RESP(11,1)<16;8,2> { NoDDClr, NoDDChk } // 2nd field U from current frame (line 0,2)
+ mov (8) r[a0.5, 32]<4>:ub ubDNDI_RESP(6,17)<16;8,2> { NoDDClr, NoDDChk } // 1st field U from current frame (line 1,3)
+ mov (8) r[a0.5, 64]<4>:ub ubDNDI_RESP(11,17)<16;8,2> { NoDDClr, NoDDChk } // 2nd field U from current frame (line 0,2)
+ mov (8) r[a0.5, 96]<4>:ub ubDNDI_RESP(7,17)<16;8,2> { NoDDClr, NoDDChk } // 1st field U from current frame (line 1,3)
+ mov (8) r[a0.6, 0]<4>:ub ubDNDI_RESP(11,0)<16;8,2> { NoDDChk } // 2nd field V from current frame (line 0,2)
+ mov (8) r[a0.6, 32]<4>:ub ubDNDI_RESP(6,16)<16;8,2> { NoDDChk } // 1st field U from current frame (line 1,3)
+ mov (8) r[a0.6, 64]<4>:ub ubDNDI_RESP(11,16)<16;8,2> { NoDDChk } // 2nd field V from current frame (line 0,2)
+ mov (8) r[a0.6, 96]<4>:ub ubDNDI_RESP(7,16)<16;8,2> { NoDDChk } // 1st field U from current frame (line 1,3)
+jmpi (1) SAVE_DN_CURR
+
+TOP_FIELD_FIRST:
+ mov (16) r[a0.4, 0]<2>:ub ubDNDI_RESP(4,0) { NoDDClr } // 1st field luma from current frame (line 0,2)
+ mov (16) r[a0.4, 32]<2>:ub ubDNDI_RESP(10,0) { NoDDClr } // 2nd field luma from current frame (line 1,3)
+ mov (16) r[a0.4, 64]<2>:ub ubDNDI_RESP(5,0) { NoDDClr } // 1st field luma from current frame (line 0,2)
+ mov (16) r[a0.4, 96]<2>:ub ubDNDI_RESP(10,16) { NoDDClr } // 2nd field luma from current frame (line 1,3)
+ mov (8) r[a0.5, 0]<4>:ub ubDNDI_RESP(6,1)<16;8,2> { NoDDClr, NoDDChk } // 1st field U from current frame (line 0,2)
+ mov (8) r[a0.5, 32]<4>:ub ubDNDI_RESP(11,1)<16;8,2> { NoDDClr, NoDDChk } // 2nd field U from current frame (line 1,3)
+ mov (8) r[a0.5, 64]<4>:ub ubDNDI_RESP(7,1)<16;8,2> { NoDDClr, NoDDChk } // 1st field U from current frame (line 0,2)
+ mov (8) r[a0.5, 96]<4>:ub ubDNDI_RESP(11,17)<16;8,2> { NoDDClr, NoDDChk } // 2nd field U from current frame (line 1,3)
+ mov (8) r[a0.6, 0]<4>:ub ubDNDI_RESP(6,0)<16;8,2> { NoDDChk } // 1st field V from current frame (line 0,2)
+ mov (8) r[a0.6, 32]<4>:ub ubDNDI_RESP(11,0)<16;8,2> { NoDDChk } // 2nd field V from current frame (line 1,3)
+ mov (8) r[a0.6, 64]<4>:ub ubDNDI_RESP(7,0)<16;8,2> { NoDDChk } // 1st field V from current frame (line 0,2)
+ mov (8) r[a0.6, 96]<4>:ub ubDNDI_RESP(11,16)<16;8,2> { NoDDChk } // 2nd field V from current frame (line 1,3)
+
+SAVE_DN_CURR:
+//send out data through data port
+send (8) null<1>:d r31.0 0x5 0xA0A8018: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 r7.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
+mov (1) r27.1<1>:ud r7.1<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 r7.26<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
+
+
+
+//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
--- /dev/null
+/*
+ * 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.
+ */
+// 97 // Total instruction count
+// 1 // Total kernel count
+
+.kernel PA_DNDI_PA
+.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=====================================
+
+
+//=======================================================================
+
+
+//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 0x4CE8003: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(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> r7.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2
+mov (1) mudMSGHDR_STMM(0,1)<1> r7.1<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 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 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> 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> 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_Save_PA_16x4.asm
+// Author: Vivek Kumar
+// Description: Save one 16x4 blocks of DN output in Packed format for reference
+
+
+ // check top/bottom field first
+cmp.e.f0.0 (1) null<1>:w r1.28<0;1,0>:ub 1:w
+
+add (4) a0.4<1>:uw r4.0<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> 0x3001F:ud { NoDDChk } // block width and height (32x8)
+
+(f0.0) jmpi (1) TOP_FIELD_FIRST
+
+BOTTOM_FIELD_FIRST:
+ mov (16) r[a0.4, 0]<2>:ub ubDNDI_RESP(10,0) { NoDDClr } // 2nd field luma from current frame (line 0,2)
+ mov (16) r[a0.4, 32]<2>:ub ubDNDI_RESP(4,16) { NoDDClr } // 1st field luma from current frame (line 1,3)
+ mov (16) r[a0.4, 64]<2>:ub ubDNDI_RESP(10,16) { NoDDClr } // 2nd field luma from current frame (line 0,2)
+ mov (16) r[a0.4, 96]<2>:ub ubDNDI_RESP(5,16) { NoDDClr } // 1st field luma from current frame (line 1,3)
+ mov (8) r[a0.5, 0]<4>:ub ubDNDI_RESP(11,1)<16;8,2> { NoDDClr, NoDDChk } // 2nd field U from current frame (line 0,2)
+ mov (8) r[a0.5, 32]<4>:ub ubDNDI_RESP(6,17)<16;8,2> { NoDDClr, NoDDChk } // 1st field U from current frame (line 1,3)
+ mov (8) r[a0.5, 64]<4>:ub ubDNDI_RESP(11,17)<16;8,2> { NoDDClr, NoDDChk } // 2nd field U from current frame (line 0,2)
+ mov (8) r[a0.5, 96]<4>:ub ubDNDI_RESP(7,17)<16;8,2> { NoDDClr, NoDDChk } // 1st field U from current frame (line 1,3)
+ mov (8) r[a0.6, 0]<4>:ub ubDNDI_RESP(11,0)<16;8,2> { NoDDChk } // 2nd field V from current frame (line 0,2)
+ mov (8) r[a0.6, 32]<4>:ub ubDNDI_RESP(6,16)<16;8,2> { NoDDChk } // 1st field U from current frame (line 1,3)
+ mov (8) r[a0.6, 64]<4>:ub ubDNDI_RESP(11,16)<16;8,2> { NoDDChk } // 2nd field V from current frame (line 0,2)
+ mov (8) r[a0.6, 96]<4>:ub ubDNDI_RESP(7,16)<16;8,2> { NoDDChk } // 1st field U from current frame (line 1,3)
+jmpi (1) SAVE_DN_CURR
+
+TOP_FIELD_FIRST:
+ mov (16) r[a0.4, 0]<2>:ub ubDNDI_RESP(4,0) { NoDDClr } // 1st field luma from current frame (line 0,2)
+ mov (16) r[a0.4, 32]<2>:ub ubDNDI_RESP(10,0) { NoDDClr } // 2nd field luma from current frame (line 1,3)
+ mov (16) r[a0.4, 64]<2>:ub ubDNDI_RESP(5,0) { NoDDClr } // 1st field luma from current frame (line 0,2)
+ mov (16) r[a0.4, 96]<2>:ub ubDNDI_RESP(10,16) { NoDDClr } // 2nd field luma from current frame (line 1,3)
+ mov (8) r[a0.5, 0]<4>:ub ubDNDI_RESP(6,1)<16;8,2> { NoDDClr, NoDDChk } // 1st field U from current frame (line 0,2)
+ mov (8) r[a0.5, 32]<4>:ub ubDNDI_RESP(11,1)<16;8,2> { NoDDClr, NoDDChk } // 2nd field U from current frame (line 1,3)
+ mov (8) r[a0.5, 64]<4>:ub ubDNDI_RESP(7,1)<16;8,2> { NoDDClr, NoDDChk } // 1st field U from current frame (line 0,2)
+ mov (8) r[a0.5, 96]<4>:ub ubDNDI_RESP(11,17)<16;8,2> { NoDDClr, NoDDChk } // 2nd field U from current frame (line 1,3)
+ mov (8) r[a0.6, 0]<4>:ub ubDNDI_RESP(6,0)<16;8,2> { NoDDChk } // 1st field V from current frame (line 0,2)
+ mov (8) r[a0.6, 32]<4>:ub ubDNDI_RESP(11,0)<16;8,2> { NoDDChk } // 2nd field V from current frame (line 1,3)
+ mov (8) r[a0.6, 64]<4>:ub ubDNDI_RESP(7,0)<16;8,2> { NoDDChk } // 1st field V from current frame (line 0,2)
+ mov (8) r[a0.6, 96]<4>:ub ubDNDI_RESP(11,16)<16;8,2> { NoDDChk } // 2nd field V from current frame (line 1,3)
+
+SAVE_DN_CURR:
+//send out data through data port
+send (8) null<1>:d r31.0 0x5 0xA0A8018:ud
+
+
+
+// FileName: DI_Save_PA_16x4.asm
+// Author: Vivek Kumar
+// Description: Save two 16x4 blocks of DI output in Packed format
+
+
+add (4) a0.4<1>:uw r2.28<4;4,1>:ub 608:w // Initial Y,U,V offset in YUV422 block; it starts at m20
+
+mov (8) r27.0<1>:ud r0.0<8;8,1>:ud
+shl (1) r27.0<1>:d r7.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
+mov (1) r27.1<1>:d r7.1<0;1,0>:w { NoDDClr, NoDDChk } // Block origin
+mov (1) r27.2<1>:ud 0x3001F:ud { NoDDChk } // Block width and height (32x8)
+
+//prepare the message headers
+mov (8) r18.0<1>:ud r27<8;8,1>:ud
+mov (8) r23.0<1>:ud r27<8;8,1>:ud
+
+// Pack 2nd field Y
+ mov (16) r[a0.4, 0]<2> ubDNDI_RESP(0,0) { NoDDClr }
+ mov (16) r[a0.4, 32]<2> ubDNDI_RESP(0,16) { NoDDClr }
+ mov (16) r[a0.4, 64]<2> ubDNDI_RESP(0,32) { NoDDClr }
+ mov (16) r[a0.4, 96]<2> ubDNDI_RESP(0,48) { NoDDClr }
+// Pack 2nd field U
+ mov (8) r[a0.5, 0]<4> ubDNDI_RESP(2,1)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 32]<4> ubDNDI_RESP(2,17)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 64]<4> ubDNDI_RESP(2,33)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 96]<4> ubDNDI_RESP(2,49)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+// Pack 2nd field V
+ mov (8) r[a0.6, 0]<4> ubDNDI_RESP(2,0)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 32]<4> ubDNDI_RESP(2,16)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 64]<4> ubDNDI_RESP(2,32)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 96]<4> ubDNDI_RESP(2,48)<16;8,2> { NoDDChk } //Vpixels
+
+// Pack 1st field Y
+ mov (16) r[a0.4, 160]<2> ubDNDI_RESP(4,0) { NoDDClr }
+ mov (16) r[a0.4, 192]<2> ubDNDI_RESP(4,16) { NoDDClr }
+ mov (16) r[a0.4, 224]<2> ubDNDI_RESP(4,32) { NoDDClr }
+ mov (16) r[a0.4, 256]<2> ubDNDI_RESP(4,48) { NoDDClr }
+// Pack 1st field U
+ mov (8) r[a0.5, 160]<4> ubDNDI_RESP(6,1)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 192]<4> ubDNDI_RESP(6,17)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 224]<4> ubDNDI_RESP(6,33)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 256]<4> ubDNDI_RESP(6,49)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+// Pack 1st field V
+ mov (8) r[a0.6, 160]<4> ubDNDI_RESP(6,0)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 192]<4> ubDNDI_RESP(6,16)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 224]<4> ubDNDI_RESP(6,32)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 256]<4> ubDNDI_RESP(6,48)<16;8,2> { NoDDChk } //Vpixels
+
+//save the previous frame
+send (8) null<1>:d r18.0 0x5 0xA0A801B:ud
+
+//save the current frame
+send (8) null<1>:d r23.0 0x5 0xA0A801E: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
--- /dev/null
+/*
+ * 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.
+ */
+// 1319 // Total instruction count
+// 1 // Total kernel count
+
+
+.kernel YUY2_DNUV_YUY2
+.code
+
+
+
+//Module : DN_UV_Setup
+//Author : Tatiya, Rupesh
+//Description : Initial Set-up for DN_UV
+
+
+
+
+// Module name : ChromaDenoise.inc
+// Author : Tatiya, Rupesh
+
+
+
+
+// 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
+
+
+//Interface:
+//Static Parameters:
+//r1
+
+
+//======================================================
+//Interface for serpent mode Chroma Denoise, added by Le
+//======================================================
+//r1
+
+
+//noise history thresholds (low and high)
+
+
+//temporal difference thresholds (high and low)
+
+
+//noise history thresholds (low and high)
+//#define ubNoiseHistMaxHigh r1.22
+//#define ubNoiseHistMaxLow r1.23
+//#define ubNoiseHistDeltaHigh r1.24
+//#define ubNoiseHistDeltaLow r1.25
+
+//Gaussian thresholds
+
+
+//temporal difference thresholds (default)
+
+
+//r2
+//history thresholds (default)
+
+
+//denoise factor (0-63)
+
+
+//====================== Binding table (Explicit To DNUV)=========================================
+//Used by DN_UV kernels
+
+
+ //Pointer to Current Frame UV
+
+
+//r1-r6
+ //CURBE GRFs used as TEMP : Used for max computation and storing max temporarily. : r1-r6
+
+
+ .declare ubCURBE_TEMP Base=r1.0 ElementSize=1 Type=ub
+ .declare uwCURBE_TEMP Base=r1.0 ElementSize=2 Type=uw
+ .declare wCURBE_TEMP Base=r1.0 ElementSize=2 Type=w
+ .declare fCURBE_TEMP Base=r1.0 ElementSize=4 Type=f
+ .declare udCURBE_TEMP Base=r1.0 ElementSize=4 Type=ud
+ .declare uwMAX_ABS_DIFF Base=r5.0 ElementSize=2 Type=uw
+
+ //r1
+
+
+ //r3
+
+
+ //r4
+
+//r7
+ //All of the following has to defined in Same GRF for optimal performance.
+
+
+//r8-24
+ //Previous Frame UV
+
+ .declare udPREV_UV Base=r8.0 ElementSize=4 Type=ud
+ .declare ubPREV_UV Base=r8.0 ElementSize=1 Type=ub
+
+
+//r25-48
+ //TEMP Space for any Usage.
+
+
+//=========================================================================
+//Definations and declarations for serpent mode Chroma Denoise, added by Le
+//=========================================================================
+
+
+ .declare udGNE_UV Base=r24.0 ElementSize=4 Type=ud
+ .declare fGNE_UV Base=r24.0 ElementSize=4 Type=f
+ .declare ubGNE_UV Base=r24.0 ElementSize=1 Type=ub
+
+ .declare udMSGHDR_BNE_SERP Base=r25.0 ElementSize=4 Type=ud
+ .declare udMSGSRC_BNE_SERP Base=r26.0 ElementSize=4 Type=ud
+
+
+ .declare ubDN_UV_Thresholds Base=r26.0 ElementSize=1 Type=ub
+ .declare ubDN_UV_Thresholds_Temp Base=r27.0 ElementSize=1 Type=ub
+ .declare udDN_UV_Thresholds Base=r26.0 ElementSize=4 Type=ud
+ .declare udDN_UV_Thresholds_Temp Base=r27.0 ElementSize=4 Type=ud
+ .declare fDN_UV_Thresholds Base=r26.0 ElementSize=4 Type=f
+ .declare fDN_UV_Thresholds_Temp Base=r27.0 ElementSize=4 Type=f
+
+
+//====================================================================================
+
+
+ //TEMP23: To hold V data for PL3 surfaces
+ .declare udCURR_V_TEMP Base=r25.0 ElementSize=4 Type=ud
+ .declare ubCURR_V_TEMP Base=r25.0 ElementSize=1 Type=ub
+
+ //GRFs to calculate Median: r25-r42
+ .declare ubMEDIAN_TEMP Base=r25.0 ElementSize=1 Type=ub
+
+ //18 GRFs to hold difference : r25-r42
+ .declare wDIFF Base=r25.0 ElementSize=2 Type=w
+ .declare uwDIFF Base=r25.0 ElementSize=2 Type=uw
+
+ //Temporal Diff
+ .declare wDIFF_TEMPORAL Base=r25.0 ElementSize=2 Type=w
+ .declare ubDIFF_TEMPORAL Base=r25.0 ElementSize=1 Type=ub
+
+ //4 GRFs to hold Sobel Value : r43-46
+ .declare wSOBEL_X Base=r43.0 ElementSize=2 Type=w
+ .declare uwSOBEL Base=r43.0 ElementSize=2 Type=uw
+
+
+ //2 GRFs to hold SOAD temporarily: r47-48
+ .declare uwSOAD Base=r47.0 ElementSize=2 Type=uw
+
+ //Temp GRFs to hold extra YUYV pixels: r43-r48
+ .declare ubTEMP5 Base=r43.0 ElementSize=1 Type=ub
+
+ //Temp GRFs in Median Calculation: r47-r48
+ .declare ubTEMP1 Base=r47.0 ElementSize=1 Type=ub
+
+ .declare uwTEMP0 Base=r48.0 ElementSize=2 Type=uw
+ .declare ubTEMP0 Base=r48.0 ElementSize=1 Type=ub
+
+ //Temp Space to store Median : r49-50
+
+ .declare ubMEDIAN Base=r49.0 ElementSize=1 Type=ub
+
+//r49
+
+
+//r50
+ //Message Source
+
+
+//r51
+ //DN_UV History Surface
+
+ .declare udHIST_UV Base=r51.0 ElementSize=4 Type=ud
+ .declare ubHIST_UV Base=r51.0 ElementSize=1 Type=ub
+
+//r52 - r91
+ //r52
+ //Current Frame UV
+
+
+ .declare udCURR_UV Base=r52.0 ElementSize=4 Type=ud
+ .declare ubCURR_UV Base=r52.0 ElementSize=1 Type=ub
+
+ //r54
+ //CURBE COPY
+
+
+ //r55
+
+
+ .declare uwSOAD_MIN_8x4 Base=r56.0 ElementSize=2 Type=uw
+
+ //r61
+
+
+ //r62
+
+
+ //History Surface Temp Origin
+
+
+ //r63
+ //Current Frame Y Temp Origin
+
+
+ //BNE Surface Origin
+
+
+ //r70
+
+ .declare uwDIFF_TEMPORAL_SUM4x4 Base=r70.0 ElementSize=2 Type=uw //4 GRFs
+
+ //r74-91 : For Saving Dest UV (PL2/PL3)
+
+
+ .declare ubMSGPAYLOAD_UV0 Base=r75.0 ElementSize=1 Type=ub
+
+
+ .declare ubMSGPAYLOAD_U Base=r75.0 ElementSize=1 Type=ub
+
+
+ .declare ubMSGPAYLOAD_UV1 Base=r84.0 ElementSize=1 Type=ub
+
+
+ .declare ubMSGPAYLOAD_V Base=r84.0 ElementSize=1 Type=ub
+
+ //r90
+
+ .declare uwDIFF_TEMPORAL_SUM4x4_FINAL Base=r90.0 ElementSize=2 Type=uw //2 GRFs
+
+//r92-127
+ //Current Frame Y
+
+
+ //r92
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_0 Base=r92 ElementSize=2 Type=uw
+ //r101
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_1 Base=r101 ElementSize=2 Type=uw
+ //r110
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_2 Base=r110 ElementSize=2 Type=uw
+ //r119
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_3 Base=r119 ElementSize=2 Type=uw
+
+ .declare udCURR_Y0 Base=r93.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y0 Base=r93.0 ElementSize=1 Type=ub
+ .declare udCURR_Y1 Base=r102.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y1 Base=r102.0 ElementSize=1 Type=ub
+ .declare udCURR_Y2 Base=r111.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y2 Base=r111.0 ElementSize=1 Type=ub
+ .declare udCURR_Y3 Base=r120.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y3 Base=r120.0 ElementSize=1 Type=ub
+
+ //r92: To hold U data for PL3 surfaces
+ .declare udCURR_U_TEMP Base=r92.0 ElementSize=4 Type=ud
+ .declare ubCURR_U_TEMP Base=r92.0 ElementSize=1 Type=ub
+
+ //r112: To hold U data for PL3 surfaces
+ .declare udPREV_U_TEMP Base=r112.0 ElementSize=4 Type=ud
+ .declare ubPREV_U_TEMP Base=r112.0 ElementSize=1 Type=ub
+
+ //r120: To hold U data for PL3 surfaces
+ .declare udPREV_V_TEMP Base=r120.0 ElementSize=4 Type=ud
+ .declare ubPREV_V_TEMP Base=r120.0 ElementSize=1 Type=ub
+
+
+ // Initialize message source with r0.
+ mov (8) r50.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r92.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r101.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r110.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r119.0<1>:ud r0.0<8;8,1>:ud
+
+
+
+//Module Name : DN_UV_YUY2_Load_Curr_Frame_YUV
+//Author : Tatiya, Rupesh
+//Description : Loads Current Frame YUV data for YUY2 input.
+
+
+
+//Module name : DN_UV_Load_Curr_Frame_UV
+//Author : Tatiya, Rupesh
+//Description : Loads Current Frame (UV only).
+// We need 4 extra rows (2 per field) and 2 extra pixel (1 each side) for both U and V each.
+// The processing size is 16x16 U and V each. So we need : U size - 18x20, V size - 18x20, UV size - 36x20, YUYV size - 72x20.
+
+
+
+
+//72x20 interleaved YUYV block is partitioned as follows:
+// <------ 36 --------> <--------36 ------->
+// ------------------------------------------
+// | | 32x2 B1 | 32x2 B2 | |
+// | 4 |--------------------------------| 4 |
+// | x | | | x |
+// |20 | 32x8 A1 | 32x8 A3 | 20|
+// | |---------------|----------------| |
+// | C1| 32x8 A2 | 32x8 A4 | C2|
+// | | | | |
+// | |--------------------------------| |
+// | | 32x2 B3 | 32x2 B4 | |
+// ------------------------------------------
+//
+// Cordinates: (x, y), (x, y+8), (x+32, y), (x+32, y+8), (x-4, y-2), (x+64, y-2),(x, y-2), (x+32, y-2), (x, y+16), (x+32, y+16)
+
+ //UV surface origin: (2xORIX, ORIY)
+ add (2) r7.4<1>:w r7.0<2;2,1>:w r4.4<2;2,1>:w { AccWrEn } // Source Block origin
+ shl (1) r7.4<1>:w acc0.4<0;1,0>:w 1:w
+
+ //A1
+ mov (2) r92.0<1>:d r7.4<2;2,1>:w { AccWrEn } // Source Block origin
+ mov (1) r92.2<1>:ud 0x7001F:ud
+ send (8) udCURR_Y0(0)<1> r92 0x4 0x2890003:ud
+
+ //A2
+ mov (1) r101.0<1>:d acc0.0<0;1,0>:d
+ add (1) r101.1<1>:d acc0.1<0;1,0>:d 8:d
+ mov (1) r101.2<1>:ud 0x7001F:ud
+ send (8) udCURR_Y1(0)<1> r101 0x4 0x2890003:ud
+
+ //B1
+ mov (1) r50.0<1>:d acc0.0<0;1,0>:d
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d -2:d
+ mov (1) r50.2<1>:ud 0x1001F:ud
+ send (8) udCURR_UV(0)<1> r50 0x4 0x2290003:ud
+
+ //B3
+ mov (1) r50.0<1>:d acc0.0<0;1,0>:d
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d 16:d
+ send (8) udCURR_UV(18)<1> r50 0x4 0x2290003:ud
+
+ //C1
+ add (1) r50.0<1>:d acc0.0<0;1,0>:d -4:d
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d -2:d
+ mov (1) r50.2<1>:ud 0x130003:ud
+ send (8) ubTEMP5(0)<1> r50 0x4 0x2390003:ud
+
+ //A3
+ add (1) r110.0<1>:d acc0.0<0;1,0>:d 32:d
+ mov (1) r110.1<1>:d acc0.1<0;1,0>:d
+ mov (1) r110.2<1>:ud 0x7001F:ud
+ send (8) udCURR_Y2(0)<1> r110 0x4 0x2890003:ud
+
+ //A4
+ add (1) r119.0<1>:d acc0.0<0;1,0>:d 32:d
+ add (1) r119.1<1>:d acc0.1<0;1,0>:d 8:d
+ mov (1) r119.2<1>:ud 0x7001F:ud
+ send (8) udCURR_Y3(0)<1> r119 0x4 0x2890003:ud
+
+ //B2
+ add (1) r50.0<1>:d acc0.0<0;1,0>:d 32:d
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d -2:d
+ mov (1) r50.2<1>:ud 0x1001F:ud
+ send (8) udCURR_UV(20)<1> r50 0x4 0x2290003:ud
+
+ //B4
+ add (1) r50.0<1>:d acc0.0<0;1,0>:d 32:d
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d 16:d
+ send (8) udCURR_UV(38)<1> r50 0x4 0x2290003:ud
+
+ //C2
+ add (1) r50.0<1>:d acc0.0<0;1,0>:d 64:d
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d -2:d
+ mov (1) r50.2<1>:ud 0x130003:ud
+ send (8) ubTEMP5(3)<1> r50 0x4 0x2390003:ud
+
+ //History Origin, Current Y origin and BNE surface origin - all are in inline GRF. Use , . -rT.
+
+ //Calculate Origin For History Surface: (ORIX/4, ORIY/4)
+ shr (2) r7.2<1>:w r7.0<2;2,1>:w 2:w
+
+ //Calculate Origin For BNE Surface: (ORIX/8, ORIY/8)
+ shr (2) r7.6<1>:w r7.0<2;2,1>:w 3:w
+
+
+
+//Module Name : DN_UV_YUY2_Load_Prev_Frame_YUV.asm
+//Author : Tatiya, Rupesh
+//Description : Loads Pevious Frame YUV data for YUY2 input.
+
+
+
+//Module Name : DN_UV_Load_Prev_Frame_UV
+//Author : Tatiya, Rupesh
+//Description : Loads Prev Frame (UV only). U size - 16x16, V size - 16x16, UV size - 32x16, YUYV size - 64x16.
+
+
+
+
+ mov (2) r50.0<1>:d r7.4<2;2,1>:w { AccWrEn } // Source lock origin
+ mov (1) r50.2<1>:ud 0x7001F:ud // U/V block width and height (16x16)
+ send (8) udPREV_UV(0)<1> r50 0x4 0x2890000:ud
+
+ add (1) r50.1<1>:ud acc0.1<0;1,0>:d 8:w // Add 16 to X origin
+ send (8) udPREV_UV(8)<1> r50 0x4 0x2890000:ud
+
+ add (1) r50.0<1>:ud acc0.0<0;1,0>:d 32:w
+ mov (1) r50.1<1>:ud acc0.1<0;1,0>:d
+ send (8) udPREV_UV(16)<1> r50 0x4 0x2890000:ud
+
+ add (1) r50.1<1>:ud acc0.0<0;1,0>:d 8:w
+ send (8) udPREV_UV(24)<1> r50 0x4 0x2890000:ud
+
+
+ //TODO - See if History loading can be combined with Prev Frame Load. - rT
+
+
+//Module name : DN_UV_Load_Hist_UV
+//Author : Tatiya, Rupesh
+//Description : Load DN History for UV denoise. 4x4 for each U & V.
+
+
+
+
+ mov (2) r50.0<1>:d r7.2<2;2,1>:w
+ mov (1) r50.2<1>:ud 0x30007:ud
+ send (8) udHIST_UV(0)<1> r50 0x4 0x2190022:ud
+
+
+
+//Module Name: DN_UV_YUY2_Extract_Curr_Frame_UV
+//Author : Tatiya, Rupesh
+//Description: Extract UV data from current YUY2 frame.
+
+//72x20 interleaved YUYV block is partitioned as follows:
+// <------ 36 --------> <--------36 ------->
+// ------------------------------------------
+// | | 32x2 B1 | 32x2 B2 | |
+// | 4 |--------------------------------| 4 |
+// | x | | | x |
+// |20 | 32x8 A1 | 32x8 A3 | 20|
+// | |---------------|----------------| |
+// | C1| 32x8 A2 | 32x8 A4 | C2|
+// | | | | |
+// | |--------------------------------| |
+// | | 32x2 B3 | 32x2 B4 | |
+// ------------------------------------------
+
+ // Set SRC pointers according to Input packing i.e. YUYV, YVYU, UYVY, VYUY
+ add (1) a0.0<1>:uw r4.1<0;1,0>:ub 2976:w //A1
+ add (1) a0.1<1>:uw r4.1<0;1,0>:ub 3264:w //A2
+ add (1) a0.2<1>:uw r4.1<0;1,0>:ub 3552:w //A3
+ add (1) a0.3<1>:uw r4.1<0;1,0>:ub 3840:w //A4
+ add (1) a0.4<1>:uw r4.1<0;1,0>:ub 1664:w //B1
+ add (1) a0.5<1>:uw r4.1<0;1,0>:ub 2240:w //B3B2
+ add (1) a0.6<1>:uw r4.1<0;1,0>:ub 2880:w //B4
+ add (1) a0.7<1>:uw r4.1<0;1,0>:ub 1376:w //C1C2
+
+ //Left 20x20 UV : 16x16 UV (Original)+4 extra rows(2 per field on top/bottom)+4 extra pixels(2 on left/right)
+
+ //A1
+ mov (16) ubCURR_UV(2,2)<1> r[a0.0, 0]<32;16,2>
+ mov (16) ubCURR_UV(3,2)<1> r[a0.0, 32]<32;16,2>
+ mov (16) ubCURR_UV(4,2)<1> r[a0.0, 64]<32;16,2>
+ mov (16) ubCURR_UV(5,2)<1> r[a0.0, 96]<32;16,2>
+ mov (16) ubCURR_UV(6,2)<1> r[a0.0, 128]<32;16,2>
+ mov (16) ubCURR_UV(7,2)<1> r[a0.0, 160]<32;16,2>
+ mov (16) ubCURR_UV(8,2)<1> r[a0.0, 192]<32;16,2>
+ mov (16) ubCURR_UV(9,2)<1> r[a0.0, 224]<32;16,2>
+
+ //A2
+ mov (16) ubCURR_UV(10,2)<1> r[a0.1, 0]<32;16,2>
+ mov (16) ubCURR_UV(11,2)<1> r[a0.1, 32]<32;16,2>
+ mov (16) ubCURR_UV(12,2)<1> r[a0.1, 64]<32;16,2>
+ mov (16) ubCURR_UV(13,2)<1> r[a0.1, 96]<32;16,2>
+ mov (16) ubCURR_UV(14,2)<1> r[a0.1, 128]<32;16,2>
+ mov (16) ubCURR_UV(15,2)<1> r[a0.1, 160]<32;16,2>
+ mov (16) ubCURR_UV(16,2)<1> r[a0.1, 192]<32;16,2>
+ mov (16) ubCURR_UV(17,2)<1> r[a0.1, 224]<32;16,2>
+
+ //B1
+ mov (16) ubCURR_UV(0,2)<1> r[a0.4, 0]<32;16,2>
+ mov (16) ubCURR_UV(1,2)<1> r[a0.4, 32]<32;16,2>
+
+ //B3
+ mov (16) ubCURR_UV(18,2)<1> r[a0.5, 0]<32;16,2>
+ mov (16) ubCURR_UV(19,2)<1> r[a0.5, 32]<32;16,2>
+
+ //TODO - Find a way to reduce this 40 SIMD2 instructions - rT
+ //C1
+ mov (2) ubCURR_UV(0,0)<1> r[a0.7, 0]<4;2,2>
+ mov (2) ubCURR_UV(1,0)<1> r[a0.7, 4]<4;2,2>
+ mov (2) ubCURR_UV(2,0)<1> r[a0.7, 8]<4;2,2>
+ mov (2) ubCURR_UV(3,0)<1> r[a0.7, 12]<4;2,2>
+ mov (2) ubCURR_UV(4,0)<1> r[a0.7, 16]<4;2,2>
+ mov (2) ubCURR_UV(5,0)<1> r[a0.7, 20]<4;2,2>
+ mov (2) ubCURR_UV(6,0)<1> r[a0.7, 24]<4;2,2>
+ mov (2) ubCURR_UV(7,0)<1> r[a0.7, 28]<4;2,2>
+ mov (2) ubCURR_UV(8,0)<1> r[a0.7, 32]<4;2,2>
+ mov (2) ubCURR_UV(9,0)<1> r[a0.7, 36]<4;2,2>
+ mov (2) ubCURR_UV(10,0)<1> r[a0.7, 40]<4;2,2>
+ mov (2) ubCURR_UV(11,0)<1> r[a0.7, 44]<4;2,2>
+ mov (2) ubCURR_UV(12,0)<1> r[a0.7, 48]<4;2,2>
+ mov (2) ubCURR_UV(13,0)<1> r[a0.7, 52]<4;2,2>
+ mov (2) ubCURR_UV(14,0)<1> r[a0.7, 56]<4;2,2>
+ mov (2) ubCURR_UV(15,0)<1> r[a0.7, 60]<4;2,2>
+ mov (2) ubCURR_UV(16,0)<1> r[a0.7, 64]<4;2,2>
+ mov (2) ubCURR_UV(17,0)<1> r[a0.7, 68]<4;2,2>
+ mov (2) ubCURR_UV(18,0)<1> r[a0.7, 72]<4;2,2>
+ mov (2) ubCURR_UV(19,0)<1> r[a0.7, 76]<4;2,2>
+
+ //2 right bytes from B2 - 2 rows
+ mov (2) ubCURR_UV(0,18)<1> r[a0.5, 64]<4;2,2>
+ mov (2) ubCURR_UV(1,18)<1> r[a0.5, 96]<4;2,2>
+
+ //2 right bytes from A3 - 8 rows
+ mov (2) ubCURR_UV(2,18)<1> r[a0.2, 0]<4;2,2>
+ mov (2) ubCURR_UV(3,18)<1> r[a0.2, 32]<4;2,2>
+ mov (2) ubCURR_UV(4,18)<1> r[a0.2, 64]<4;2,2>
+ mov (2) ubCURR_UV(5,18)<1> r[a0.2, 96]<4;2,2>
+ mov (2) ubCURR_UV(6,18)<1> r[a0.2, 128]<4;2,2>
+ mov (2) ubCURR_UV(7,18)<1> r[a0.2, 160]<4;2,2>
+ mov (2) ubCURR_UV(8,18)<1> r[a0.2, 192]<4;2,2>
+ mov (2) ubCURR_UV(9,18)<1> r[a0.2, 224]<4;2,2>
+
+ //2 right bytes from A4 - 8 rows
+ mov (2) ubCURR_UV(10,18)<1> r[a0.3, 0]<4;2,2>
+ mov (2) ubCURR_UV(11,18)<1> r[a0.3, 32]<4;2,2>
+ mov (2) ubCURR_UV(12,18)<1> r[a0.3, 64]<4;2,2>
+ mov (2) ubCURR_UV(13,18)<1> r[a0.3, 96]<4;2,2>
+ mov (2) ubCURR_UV(14,18)<1> r[a0.3, 128]<4;2,2>
+ mov (2) ubCURR_UV(15,18)<1> r[a0.3, 160]<4;2,2>
+ mov (2) ubCURR_UV(16,18)<1> r[a0.3, 192]<4;2,2>
+ mov (2) ubCURR_UV(17,18)<1> r[a0.3, 224]<4;2,2>
+
+ //2 right bytes from B4 - 2 rows
+ mov (2) ubCURR_UV(18,18)<1> r[a0.6, 0]<4;2,2>
+ mov (2) ubCURR_UV(19,18)<1> r[a0.6, 32]<4;2,2>
+
+ //Right 20x20 UV : 16x16 UV (Original)+4 extra rows(2 per field on top/bottom)+4 extra pixels(2 on left/right)
+
+ //A3
+ mov (16) ubCURR_UV(22,2)<1> r[a0.2, 0]<32;16,2>
+ mov (16) ubCURR_UV(23,2)<1> r[a0.2, 32]<32;16,2>
+ mov (16) ubCURR_UV(24,2)<1> r[a0.2, 64]<32;16,2>
+ mov (16) ubCURR_UV(25,2)<1> r[a0.2, 96]<32;16,2>
+ mov (16) ubCURR_UV(26,2)<1> r[a0.2, 128]<32;16,2>
+ mov (16) ubCURR_UV(27,2)<1> r[a0.2, 160]<32;16,2>
+ mov (16) ubCURR_UV(28,2)<1> r[a0.2, 192]<32;16,2>
+ mov (16) ubCURR_UV(29,2)<1> r[a0.2, 224]<32;16,2>
+
+ //A4
+ mov (16) ubCURR_UV(30,2)<1> r[a0.3, 0]<32;16,2>
+ mov (16) ubCURR_UV(31,2)<1> r[a0.3, 32]<32;16,2>
+ mov (16) ubCURR_UV(32,2)<1> r[a0.3, 64]<32;16,2>
+ mov (16) ubCURR_UV(33,2)<1> r[a0.3, 96]<32;16,2>
+ mov (16) ubCURR_UV(34,2)<1> r[a0.3, 128]<32;16,2>
+ mov (16) ubCURR_UV(35,2)<1> r[a0.3, 160]<32;16,2>
+ mov (16) ubCURR_UV(36,2)<1> r[a0.3, 192]<32;16,2>
+ mov (16) ubCURR_UV(37,2)<1> r[a0.3, 224]<32;16,2>
+
+ //B2
+ mov (16) ubCURR_UV(20,2)<1> r[a0.5, 64]<32;16,2>
+ mov (16) ubCURR_UV(21,2)<1> r[a0.5, 96]<32;16,2>
+
+ //B4
+ mov (16) ubCURR_UV(38,2)<1> r[a0.6, 0]<32;16,2>
+ mov (16) ubCURR_UV(39,2)<1> r[a0.6, 32]<32;16,2>
+
+ //TODO - Find a way to reduce this 40 SIMD2 instructions - rT
+ //C2
+ mov (2) ubCURR_UV(20,18)<1> r[a0.7, 96]<4;2,2>
+ mov (2) ubCURR_UV(21,18)<1> r[a0.7, 100]<4;2,2>
+ mov (2) ubCURR_UV(22,18)<1> r[a0.7, 104]<4;2,2>
+ mov (2) ubCURR_UV(23,18)<1> r[a0.7, 108]<4;2,2>
+ mov (2) ubCURR_UV(24,18)<1> r[a0.7, 112]<4;2,2>
+ mov (2) ubCURR_UV(25,18)<1> r[a0.7, 116]<4;2,2>
+ mov (2) ubCURR_UV(26,18)<1> r[a0.7, 120]<4;2,2>
+ mov (2) ubCURR_UV(27,18)<1> r[a0.7, 124]<4;2,2>
+ mov (2) ubCURR_UV(28,18)<1> r[a0.7, 128]<4;2,2>
+ mov (2) ubCURR_UV(29,18)<1> r[a0.7, 132]<4;2,2>
+ mov (2) ubCURR_UV(30,18)<1> r[a0.7, 136]<4;2,2>
+ mov (2) ubCURR_UV(31,18)<1> r[a0.7, 140]<4;2,2>
+ mov (2) ubCURR_UV(32,18)<1> r[a0.7, 144]<4;2,2>
+ mov (2) ubCURR_UV(33,18)<1> r[a0.7, 148]<4;2,2>
+ mov (2) ubCURR_UV(34,18)<1> r[a0.7, 152]<4;2,2>
+ mov (2) ubCURR_UV(35,18)<1> r[a0.7, 156]<4;2,2>
+ mov (2) ubCURR_UV(36,18)<1> r[a0.7, 160]<4;2,2>
+ mov (2) ubCURR_UV(37,18)<1> r[a0.7, 164]<4;2,2>
+ mov (2) ubCURR_UV(38,18)<1> r[a0.7, 168]<4;2,2>
+ mov (2) ubCURR_UV(39,18)<1> r[a0.7, 172]<4;2,2>
+
+ //2 left bytes from B1 - 2 rows
+ mov (2) ubCURR_UV(20,0)<1> r[a0.4, 28]<4;2,2>
+ mov (2) ubCURR_UV(21,0)<1> r[a0.4, 60]<4;2,2>
+
+ //2 left bytes from A1 - 8 rows
+ mov (2) ubCURR_UV(22,0)<1> r[a0.0, 28]<4;2,2>
+ mov (2) ubCURR_UV(23,0)<1> r[a0.0, 60]<4;2,2>
+ mov (2) ubCURR_UV(24,0)<1> r[a0.0, 92]<4;2,2>
+ mov (2) ubCURR_UV(25,0)<1> r[a0.0, 124]<4;2,2>
+ mov (2) ubCURR_UV(26,0)<1> r[a0.0, 156]<4;2,2>
+ mov (2) ubCURR_UV(27,0)<1> r[a0.0, 188]<4;2,2>
+ mov (2) ubCURR_UV(28,0)<1> r[a0.0, 220]<4;2,2>
+ mov (2) ubCURR_UV(29,0)<1> r[a0.0, 252]<4;2,2>
+
+ //2 left bytes from A2 - 8 rows
+ mov (2) ubCURR_UV(30,0)<1> r[a0.1, 28]<4;2,2>
+ mov (2) ubCURR_UV(31,0)<1> r[a0.1, 60]<4;2,2>
+ mov (2) ubCURR_UV(32,0)<1> r[a0.1, 92]<4;2,2>
+ mov (2) ubCURR_UV(33,0)<1> r[a0.1, 124]<4;2,2>
+ mov (2) ubCURR_UV(34,0)<1> r[a0.1, 156]<4;2,2>
+ mov (2) ubCURR_UV(35,0)<1> r[a0.1, 188]<4;2,2>
+ mov (2) ubCURR_UV(36,0)<1> r[a0.1, 220]<4;2,2>
+ mov (2) ubCURR_UV(37,0)<1> r[a0.1, 252]<4;2,2>
+
+ //2 left bytes from B3 - 2 rows
+ mov (2) ubCURR_UV(38,0)<1> r[a0.5, 28]<4;2,2>
+ mov (2) ubCURR_UV(39,0)<1> r[a0.5, 60]<4;2,2>
+
+
+
+// Module Name : DN_UV_YUY2_Extract_Prev_Frame_UV
+// Author : Tatiya, Rupesh
+// Description : Extract UV from previous frame YUY2.
+
+ // Set SRC pointers according to Input packing i.e. YUYV, YVYU, UYVY, VYUY
+ add (1) a0.0<1>:uw r4.1<0;1,0>:ub 256:w
+ add (1) a0.1<1>:uw r4.1<0;1,0>:ub 768:w
+
+ mov (16) ubPREV_UV(0,0)<1> r[a0.0, 0]<32;16,2>:ub
+ mov (16) ubPREV_UV(0,16)<1> r[a0.0, 32]<32;16,2>:ub
+ mov (16) ubPREV_UV(1,0)<1> r[a0.0, 64]<32;16,2>:ub
+ mov (16) ubPREV_UV(1,16)<1> r[a0.0, 96]<32;16,2>:ub
+ mov (16) ubPREV_UV(2,0)<1> r[a0.0, 128]<32;16,2>:ub
+ mov (16) ubPREV_UV(2,16)<1> r[a0.0, 160]<32;16,2>:ub
+ mov (16) ubPREV_UV(3,0)<1> r[a0.0, 192]<32;16,2>:ub
+ mov (16) ubPREV_UV(3,16)<1> r[a0.0, 224]<32;16,2>:ub
+ mov (16) ubPREV_UV(4,0)<1> r[a0.0, 256]<32;16,2>:ub
+ mov (16) ubPREV_UV(4,16)<1> r[a0.0, 288]<32;16,2>:ub
+ mov (16) ubPREV_UV(5,0)<1> r[a0.0, 320]<32;16,2>:ub
+ mov (16) ubPREV_UV(5,16)<1> r[a0.0, 352]<32;16,2>:ub
+ mov (16) ubPREV_UV(6,0)<1> r[a0.0, 384]<32;16,2>:ub
+ mov (16) ubPREV_UV(6,16)<1> r[a0.0, 416]<32;16,2>:ub
+ mov (16) ubPREV_UV(7,0)<1> r[a0.0, 448]<32;16,2>:ub
+ mov (16) ubPREV_UV(7,16)<1> r[a0.0, 480]<32;16,2>:ub
+
+ mov (16) ubPREV_UV(8,0)<1> r[a0.1, 0]<32;16,2>:ub
+ mov (16) ubPREV_UV(8,16)<1> r[a0.1, 32]<32;16,2>:ub
+ mov (16) ubPREV_UV(9,0)<1> r[a0.1, 64]<32;16,2>:ub
+ mov (16) ubPREV_UV(9,16)<1> r[a0.1, 96]<32;16,2>:ub
+ mov (16) ubPREV_UV(10,0)<1> r[a0.1, 128]<32;16,2>:ub
+ mov (16) ubPREV_UV(10,16)<1> r[a0.1, 160]<32;16,2>:ub
+ mov (16) ubPREV_UV(11,0)<1> r[a0.1, 192]<32;16,2>:ub
+ mov (16) ubPREV_UV(11,16)<1> r[a0.1, 224]<32;16,2>:ub
+ mov (16) ubPREV_UV(12,0)<1> r[a0.1, 256]<32;16,2>:ub
+ mov (16) ubPREV_UV(12,16)<1> r[a0.1, 288]<32;16,2>:ub
+ mov (16) ubPREV_UV(13,0)<1> r[a0.1, 320]<32;16,2>:ub
+ mov (16) ubPREV_UV(13,16)<1> r[a0.1, 352]<32;16,2>:ub
+ mov (16) ubPREV_UV(14,0)<1> r[a0.1, 384]<32;16,2>:ub
+ mov (16) ubPREV_UV(14,16)<1> r[a0.1, 416]<32;16,2>:ub
+ mov (16) ubPREV_UV(15,0)<1> r[a0.1, 448]<32;16,2>:ub
+ mov (16) ubPREV_UV(15,16)<1> r[a0.1, 480]<32;16,2>:ub
+
+
+
+//Module Name : DN_UV_Noise_Detection_UV
+//Author : Tatiya, Rupesh
+//Description : Performs noise detection on 16x16 U and 16x16 V each.
+
+
+
+//Module Name : DN_UV_Move_CURBE_Inline_UV.asm
+//Author : Tatiya, Rupesh
+
+
+
+
+ //Mov CURBE data to another space - so that it can be used as Temp Space --> r1 - r6
+ mov (4) r54.28<1>:ub r2.28<4;4,1>:ub //Dest. YUY2 offset
+ mov (2) r54.5<1>:ud r4.0<4;2,2>:ud //Src YUY2 offset and Origin offset
+ mov (4) r55.28<1>:ub r1.0<4;4,1>:ub
+
+ mov (8) r61.20<1>:ub r1.4<8;8,1>:ub
+ mov (4) r61.28<1>:ub r1.12<4;4,1>:ub
+
+ //Move Inline Data to another space - so that it can be used as Temp Space --> r7
+ mov (4) r62.10<1>:w r7.0<4;4,1>:w
+ mov (4) r63.10<1>:w r7.4<4;4,1>:w
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ mov (1) a0.0:uw 1664:uw
+ mov (1) a0.1:uw 1816:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 1792:uw
+ mov (1) a0.1:uw 1820:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 1920:uw
+ mov (1) a0.1:uw 1848:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2048:uw
+ mov (1) a0.1:uw 1852:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ mov (1) a0.0:uw 2304:uw
+ mov (1) a0.1:uw 1880:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2432:uw
+ mov (1) a0.1:uw 1884:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2560:uw
+ mov (1) a0.1:uw 1912:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2688:uw
+ mov (1) a0.1:uw 1916:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+//Module : DN_UV_Noise_Reduction_UV
+//Author : Tatiya, Rupesh
+//Description : Performs Noise Reduction on 16x16 U and 16x16 V.
+//Tasks : 1. Update weight history
+// 2. Find if it block is motion block
+// 3. Compute Denoised Pixel.
+
+
+
+
+//History is 1+1 byte every 4x4 U and 4x4 V.
+
+ cmp.l.f0.0 (16) null<1>:w ubHIST_UV(0,0)<16;16,1> r61.20<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w ubHIST_UV(0,0)<16;16,1> r61.22<0;2,1>:ub
+
+ mov (16) uwCURBE_TEMP(0)<1> 0:w
+ mov (16) uwCURBE_TEMP(1)<1> 0:w
+
+ //Compute diff betn curr and prev. - First 16 lines
+ // 8 lines here
+ add (16) wDIFF_TEMPORAL(0)<1> ubCURR_UV(2,2)<16;16,1> -ubPREV_UV(0,0)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(1)<1> ubCURR_UV(3,2)<16;16,1> -ubPREV_UV(0,16)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(2)<1> ubCURR_UV(4,2)<16;16,1> -ubPREV_UV(0,32)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(3)<1> ubCURR_UV(5,2)<16;16,1> -ubPREV_UV(0,48)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(4)<1> ubCURR_UV(6,2)<16;16,1> -ubPREV_UV(0,64)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(5)<1> ubCURR_UV(7,2)<16;16,1> -ubPREV_UV(0,80)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(6)<1> ubCURR_UV(8,2)<16;16,1> -ubPREV_UV(0,96)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(7)<1> ubCURR_UV(9,2)<16;16,1> -ubPREV_UV(0,112)<16;16,1> //Diff UV interleaved
+
+ //Update WT HIST
+ (-f0.0) shr (16) uwCURBE_TEMP(0)<1> ubHIST_UV(0,0)<16;16,1> 1:w
+ (f1.0) add (16) uwCURBE_TEMP(2)<1> ubHIST_UV(0,0)<16;16,1> r61.24<0;2,1>:ub
+ (f0.0) mov (16) uwCURBE_TEMP(2)<1> r61.20<0;2,1>:ub
+ (-f0.0.anyv) mov (16) uwCURBE_TEMP(2)<1> ubHIST_UV(0,0)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w ubHIST_UV(0,16)<16;16,1> r61.20<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w ubHIST_UV(0,16)<16;16,1> r61.22<0;2,1>:ub
+
+ //Compute diff betn curr and prev. - First 16 lines
+ // 8 more lines here
+ add (16) wDIFF_TEMPORAL(8)<1> ubCURR_UV(10,2)<16;16,1> -ubPREV_UV(0,128)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(9)<1> ubCURR_UV(11,2)<16;16,1> -ubPREV_UV(0,144)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(10)<1> ubCURR_UV(12,2)<16;16,1> -ubPREV_UV(0,160)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(11)<1> ubCURR_UV(13,2)<16;16,1> -ubPREV_UV(0,176)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(12)<1> ubCURR_UV(14,2)<16;16,1> -ubPREV_UV(0,192)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(13)<1> ubCURR_UV(15,2)<16;16,1> -ubPREV_UV(0,208)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(14)<1> ubCURR_UV(16,2)<16;16,1> -ubPREV_UV(0,224)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(15)<1> ubCURR_UV(17,2)<16;16,1> -ubPREV_UV(0,240)<16;16,1> //Diff UV interleaved
+
+ (-f0.0) shr (16) uwCURBE_TEMP(1)<1> ubHIST_UV(0,16)<16;16,1> 1:w
+ (f1.0) add (16) uwCURBE_TEMP(3)<1> ubHIST_UV(0,16)<16;16,1> r61.24<0;2,1>:ub
+ (f0.0) mov (16) uwCURBE_TEMP(3)<1> r61.20<0;2,1>:ub
+ (-f0.0.anyv) mov(16) uwCURBE_TEMP(3)<1> ubHIST_UV(0,16)<16;16,1>
+
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(0)<16;16,1> (abs)wDIFF_TEMPORAL(1)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(2)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(3)<16;16,1>
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(4)<16;16,1> (abs)wDIFF_TEMPORAL(5)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(6)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(7)<16;16,1>
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(8)<16;16,1> (abs)wDIFF_TEMPORAL(9)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(10)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(2)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(11)<16;16,1>
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(12)<16;16,1> (abs)wDIFF_TEMPORAL(13)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(14)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(3)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(15)<16;16,1>
+
+//Compute diff betn curr and prev. - Second 16 lines
+//13 lines.
+ add (16) wDIFF_TEMPORAL(16)<1> ubCURR_UV(22,2)<16;16,1> -ubPREV_UV(8,0)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(17)<1> ubCURR_UV(23,2)<16;16,1> -ubPREV_UV(8,16)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(18)<1> ubCURR_UV(24,2)<16;16,1> -ubPREV_UV(8,32)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(19)<1> ubCURR_UV(25,2)<16;16,1> -ubPREV_UV(8,48)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(20)<1> ubCURR_UV(26,2)<16;16,1> -ubPREV_UV(8,64)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(21)<1> ubCURR_UV(27,2)<16;16,1> -ubPREV_UV(8,80)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(22)<1> ubCURR_UV(28,2)<16;16,1> -ubPREV_UV(8,96)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(23)<1> ubCURR_UV(29,2)<16;16,1> -ubPREV_UV(8,112)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(24)<1> ubCURR_UV(30,2)<16;16,1> -ubPREV_UV(8,128)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(25)<1> ubCURR_UV(31,2)<16;16,1> -ubPREV_UV(8,144)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(26)<1> ubCURR_UV(32,2)<16;16,1> -ubPREV_UV(8,160)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(27)<1> ubCURR_UV(33,2)<16;16,1> -ubPREV_UV(8,176)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(28)<1> ubCURR_UV(34,2)<16;16,1> -ubPREV_UV(8,192)<16;16,1> //Diff UV interleaved
+
+//3 more lines
+ add (16) wCURBE_TEMP(4)<1> ubCURR_UV(35,2)<16;16,1> -ubPREV_UV(8,208)<16;16,1> //Diff UV interleaved
+ add (16) wCURBE_TEMP(5)<1> ubCURR_UV(36,2)<16;16,1> -ubPREV_UV(8,224)<16;16,1> //Diff UV interleaved
+ add (16) wCURBE_TEMP(6)<1> ubCURR_UV(37,2)<16;16,1> -ubPREV_UV(8,240)<16;16,1> //Diff UV interleaved
+
+ //16x4 to 8x4 - First 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> uwDIFF_TEMPORAL_SUM4x4(2,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(2,2)<4;2,1>
+
+ //8x4 to 4x4 - First 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4_FINAL(0)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1> { AccWrEn }
+
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(16)<16;16,1> (abs)wDIFF_TEMPORAL(17)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(18)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(19)<16;16,1>
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(20)<16;16,1> (abs)wDIFF_TEMPORAL(21)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(22)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(23)<16;16,1>
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(24)<16;16,1> (abs)wDIFF_TEMPORAL(25)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(26)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(2)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(27)<16;16,1>
+
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(28)<16;16,1> (abs)wCURBE_TEMP(4)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wCURBE_TEMP(5)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(3)<1> acc0.0<16;16,1>:uw (abs)wCURBE_TEMP(6)<16;16,1>
+
+ //Find if block is motion block - First 16 lines
+ cmp.g.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_FINAL(0)<16;16,1> r61.26<0;2,1>:ub
+
+ //Move TEMPORAL_SUM4x4 for SIMD16 use later.
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,0)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,2)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,4)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,6)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,8)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,10)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,12)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,14)<0;2,1>
+
+ //Pick Appropriate Weight History Based on motion. - First 16 lines
+ (-f0.0) mov (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(2)<16;16,1>
+
+ //Actual DN - First 16 lines
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(0)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(0)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(2,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(2,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(2,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,0)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,8)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(0)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(0)<1> ubCURR_UV(2,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(0)<1> wDIFF_TEMPORAL(0)<16;16,1> ubCURR_UV(2,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(1)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(1)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(3,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(3,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(3,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,16)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,24)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(1)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(1)<1> ubCURR_UV(3,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(1)<1> wDIFF_TEMPORAL(1)<16;16,1> ubCURR_UV(3,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(2)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(2)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(4,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(4,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(4,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,32)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,40)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(2)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(2)<1> ubCURR_UV(4,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(2)<1> wDIFF_TEMPORAL(2)<16;16,1> ubCURR_UV(4,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(3)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(3)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(5,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(5,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(5,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,48)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,56)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(3)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(3)<1> ubCURR_UV(5,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(3)<1> wDIFF_TEMPORAL(3)<16;16,1> ubCURR_UV(5,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(4)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(4)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(6,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(6,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(6,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,64)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,72)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(4)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(4)<1> ubCURR_UV(6,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(4)<1> wDIFF_TEMPORAL(4)<16;16,1> ubCURR_UV(6,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(5)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(5)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(7,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(7,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(7,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,80)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,88)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(5)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(5)<1> ubCURR_UV(7,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(5)<1> wDIFF_TEMPORAL(5)<16;16,1> ubCURR_UV(7,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(6)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(6)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(8,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(8,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(8,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,96)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,104)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(6)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(6)<1> ubCURR_UV(8,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(6)<1> wDIFF_TEMPORAL(6)<16;16,1> ubCURR_UV(8,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(7)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(7)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(9,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(9,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(9,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,112)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,120)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(7)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(7)<1> ubCURR_UV(9,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(7)<1> wDIFF_TEMPORAL(7)<16;16,1> ubCURR_UV(9,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(8)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(8)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(10,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(10,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(10,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,128)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,136)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(8)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(8)<1> ubCURR_UV(10,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(8)<1> wDIFF_TEMPORAL(8)<16;16,1> ubCURR_UV(10,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(9)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(9)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(11,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(11,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(11,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,144)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,152)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(9)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(9)<1> ubCURR_UV(11,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(9)<1> wDIFF_TEMPORAL(9)<16;16,1> ubCURR_UV(11,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(10)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(10)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(12,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(12,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(12,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,160)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,168)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(10)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(10)<1> ubCURR_UV(12,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(10)<1> wDIFF_TEMPORAL(10)<16;16,1> ubCURR_UV(12,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(11)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(11)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(13,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(13,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(13,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,176)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,184)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(11)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(11)<1> ubCURR_UV(13,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(11)<1> wDIFF_TEMPORAL(11)<16;16,1> ubCURR_UV(13,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(12)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(12)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(14,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(14,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(14,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,192)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,200)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(12)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(12)<1> ubCURR_UV(14,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(12)<1> wDIFF_TEMPORAL(12)<16;16,1> ubCURR_UV(14,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(13)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(13)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(15,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(15,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(15,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,208)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,216)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(13)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(13)<1> ubCURR_UV(15,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(13)<1> wDIFF_TEMPORAL(13)<16;16,1> ubCURR_UV(15,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(14)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(14)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(16,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(16,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(16,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,224)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,232)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(14)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(14)<1> ubCURR_UV(16,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(14)<1> wDIFF_TEMPORAL(14)<16;16,1> ubCURR_UV(16,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(15)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(15)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(17,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(17,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(17,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,240)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,248)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(15)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(15)<1> ubCURR_UV(17,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(15)<1> wDIFF_TEMPORAL(15)<16;16,1> ubCURR_UV(17,2)<16;16,1>
+
+
+ //16x4 to 8x4 - Second 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> uwDIFF_TEMPORAL_SUM4x4(2,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(2,2)<4;2,1>
+
+ //8x4 to 4x4 - Second 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4_FINAL(1)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1> { AccWrEn }
+
+ //Find if block is motion block - Second 16 lines
+ cmp.g.f1.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_FINAL(1)<16;16,1> r61.26<0;2,1>:ub
+
+ //Move TEMPORAL_SUM4x4 for SIMD16 use later.
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,0)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,2)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,4)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,6)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,8)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,10)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,12)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,14)<0;2,1>
+
+ //Pick Appropriate Weight History Based on motion. - Second 16 lines
+ (-f1.0) mov (16) uwCURBE_TEMP(1)<1> uwCURBE_TEMP(3)<16;16,1>
+
+ //Actual DN - Second 16 lines
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(16)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(16)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(22,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(22,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(22,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,0)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,8)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(16)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(16)<1> ubCURR_UV(22,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(16)<1> wDIFF_TEMPORAL(16)<16;16,1> ubCURR_UV(22,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(17)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(17)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(23,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(23,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(23,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,16)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,24)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(17)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(17)<1> ubCURR_UV(23,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(17)<1> wDIFF_TEMPORAL(17)<16;16,1> ubCURR_UV(23,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(18)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(18)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(24,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(24,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(24,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,32)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,40)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(18)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(18)<1> ubCURR_UV(24,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(18)<1> wDIFF_TEMPORAL(18)<16;16,1> ubCURR_UV(24,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(19)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(19)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(25,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(25,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(25,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,48)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,56)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(19)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(19)<1> ubCURR_UV(25,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(19)<1> wDIFF_TEMPORAL(19)<16;16,1> ubCURR_UV(25,2)<16;16,1>
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(20)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(20)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(26,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(26,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(26,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,64)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,72)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(20)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(20)<1> ubCURR_UV(26,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(20)<1> wDIFF_TEMPORAL(20)<16;16,1> ubCURR_UV(26,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(21)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(21)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(27,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(27,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(27,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,80)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,88)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(21)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(21)<1> ubCURR_UV(27,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(21)<1> wDIFF_TEMPORAL(21)<16;16,1> ubCURR_UV(27,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(22)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(22)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(28,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(28,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(28,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,96)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,104)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(22)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(22)<1> ubCURR_UV(28,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(22)<1> wDIFF_TEMPORAL(22)<16;16,1> ubCURR_UV(28,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(23)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(23)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(29,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(29,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(29,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,112)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,120)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(23)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(23)<1> ubCURR_UV(29,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(23)<1> wDIFF_TEMPORAL(23)<16;16,1> ubCURR_UV(29,2)<16;16,1>
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(24)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(24)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(30,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(30,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(30,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,128)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,136)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(24)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(24)<1> ubCURR_UV(30,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(24)<1> wDIFF_TEMPORAL(24)<16;16,1> ubCURR_UV(30,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(25)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(25)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(31,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(31,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(31,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,144)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,152)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(25)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(25)<1> ubCURR_UV(31,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(25)<1> wDIFF_TEMPORAL(25)<16;16,1> ubCURR_UV(31,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(26)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(26)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(32,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(32,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(32,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,160)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,168)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(26)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(26)<1> ubCURR_UV(32,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(26)<1> wDIFF_TEMPORAL(26)<16;16,1> ubCURR_UV(32,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(27)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(27)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(33,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(33,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(33,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,176)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,184)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(27)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(27)<1> ubCURR_UV(33,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(27)<1> wDIFF_TEMPORAL(27)<16;16,1> ubCURR_UV(33,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(28)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(28)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(34,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(34,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(34,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,192)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,200)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(28)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(28)<1> ubCURR_UV(34,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(28)<1> wDIFF_TEMPORAL(28)<16;16,1> ubCURR_UV(34,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wCURBE_TEMP(4)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wCURBE_TEMP(4)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(35,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(35,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(35,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,208)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,216)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wCURBE_TEMP(4)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wCURBE_TEMP(4)<1> ubCURR_UV(35,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wCURBE_TEMP(4)<1> wCURBE_TEMP(4)<16;16,1> ubCURR_UV(35,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wCURBE_TEMP(5)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wCURBE_TEMP(5)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(36,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(36,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(36,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,224)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,232)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wCURBE_TEMP(5)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wCURBE_TEMP(5)<1> ubCURR_UV(36,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wCURBE_TEMP(5)<1> wCURBE_TEMP(5)<16;16,1> ubCURR_UV(36,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wCURBE_TEMP(6)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wCURBE_TEMP(6)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(37,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(37,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(37,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,240)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,248)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wCURBE_TEMP(6)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wCURBE_TEMP(6)<1> ubCURR_UV(37,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wCURBE_TEMP(6)<1> wCURBE_TEMP(6)<16;16,1> ubCURR_UV(37,2)<16;16,1>
+
+ //Pack Weight History WORD -> BYTE
+ mov (16) ubCURBE_TEMP(3,0)<1> ubCURBE_TEMP(0)<32;16,2>
+ mov (16) ubCURBE_TEMP(3,16)<1> ubCURBE_TEMP(1)<32;16,2>
+
+
+
+//Module Name : DN_UV_Compute_BNE_UV
+//Author : Tatiya, Rupesh
+//Description : Computes minimum SOAD for each 16x4 block.
+
+ cmp.l.f0.0 (8) null:w uwSOAD_MIN_8x4(0,12)<16;4,1> uwSOAD_MIN_8x4(2,12)<16;4,1>
+ (f0.0)sel (8) uwCURBE_TEMP(1,0)<1> uwSOAD_MIN_8x4(0,12)<16;4,1> uwSOAD_MIN_8x4(2,12)<16;4,1>
+
+ mov (8) ubCURBE_TEMP(1)<1> ubCURBE_TEMP(1)<16;8,2>
+
+
+
+//Module Name : DN_UV_YUY2_Pack_Denoised_UV
+//Name : Tatiya, Rupesh
+//Description : Pack UV denoised data based on YUY2 input.
+
+
+
+//Module Name : DN_UV_Pack_Denoised_UV
+//Name : Tatiya, Rupesh
+//Description : Pack UV denoised data based on PL2/PL3/PA.
+
+
+ add (1) a0.0<1>:uw r54.21<0;1,0>:ub 2976:w
+ add (1) a0.1<1>:uw r54.21<0;1,0>:ub 3264:w
+ add (1) a0.2<1>:uw r54.21<0;1,0>:ub 3552:w
+ add (1) a0.3<1>:uw r54.21<0;1,0>:ub 3840:w
+
+//First 8 lines.
+ mov (16) r[a0.0, 0]<2>:ub ubDIFF_TEMPORAL(0)<32;16,2>
+ mov (16) r[a0.0, 32]<2>:ub ubDIFF_TEMPORAL(1)<32;16,2>
+ mov (16) r[a0.0, 64]<2>:ub ubDIFF_TEMPORAL(2)<32;16,2>
+ mov (16) r[a0.0, 96]<2>:ub ubDIFF_TEMPORAL(3)<32;16,2>
+ mov (16) r[a0.0, 128]<2>:ub ubDIFF_TEMPORAL(4)<32;16,2>
+ mov (16) r[a0.0, 160]<2>:ub ubDIFF_TEMPORAL(5)<32;16,2>
+ mov (16) r[a0.0, 192]<2>:ub ubDIFF_TEMPORAL(6)<32;16,2>
+ mov (16) r[a0.0, 224]<2>:ub ubDIFF_TEMPORAL(7)<32;16,2>
+
+//Second 8 lines
+ mov (16) r[a0.1, 0]<2>:ub ubDIFF_TEMPORAL(8)<32;16,2>
+ mov (16) r[a0.1, 32]<2>:ub ubDIFF_TEMPORAL(9)<32;16,2>
+ mov (16) r[a0.1, 64]<2>:ub ubDIFF_TEMPORAL(10)<32;16,2>
+ mov (16) r[a0.1, 96]<2>:ub ubDIFF_TEMPORAL(11)<32;16,2>
+ mov (16) r[a0.1, 128]<2>:ub ubDIFF_TEMPORAL(12)<32;16,2>
+ mov (16) r[a0.1, 160]<2>:ub ubDIFF_TEMPORAL(13)<32;16,2>
+ mov (16) r[a0.1, 192]<2>:ub ubDIFF_TEMPORAL(14)<32;16,2>
+ mov (16) r[a0.1, 224]<2>:ub ubDIFF_TEMPORAL(15)<32;16,2>
+
+//Third 8 lines
+ mov (16) r[a0.2, 0]<2>:ub ubDIFF_TEMPORAL(16)<32;16,2>
+ mov (16) r[a0.2, 32]<2>:ub ubDIFF_TEMPORAL(17)<32;16,2>
+ mov (16) r[a0.2, 64]<2>:ub ubDIFF_TEMPORAL(18)<32;16,2>
+ mov (16) r[a0.2, 96]<2>:ub ubDIFF_TEMPORAL(19)<32;16,2>
+ mov (16) r[a0.2, 128]<2>:ub ubDIFF_TEMPORAL(20)<32;16,2>
+ mov (16) r[a0.2, 160]<2>:ub ubDIFF_TEMPORAL(21)<32;16,2>
+ mov (16) r[a0.2, 192]<2>:ub ubDIFF_TEMPORAL(22)<32;16,2>
+ mov (16) r[a0.2, 224]<2>:ub ubDIFF_TEMPORAL(23)<32;16,2>
+
+//Fourth 8 lines
+//5 lines first
+ mov (16) r[a0.3, 0]<2>:ub ubDIFF_TEMPORAL(24)<32;16,2>
+ mov (16) r[a0.3, 32]<2>:ub ubDIFF_TEMPORAL(25)<32;16,2>
+ mov (16) r[a0.3, 64]<2>:ub ubDIFF_TEMPORAL(26)<32;16,2>
+ mov (16) r[a0.3, 96]<2>:ub ubDIFF_TEMPORAL(27)<32;16,2>
+ mov (16) r[a0.3, 128]<2>:ub ubDIFF_TEMPORAL(28)<32;16,2>
+
+//3 more lines
+ mov (16) r[a0.3, 160]<2>:ub ubCURBE_TEMP(4)<32;16,2>
+ mov (16) r[a0.3, 192]<2>:ub ubCURBE_TEMP(5)<32;16,2>
+ mov (16) r[a0.3, 224]<2>:ub ubCURBE_TEMP(6)<32;16,2>
+
+
+ //TODO - See if History saving can be combined with Curr Frame Save. - rT
+
+
+//Module Name : DN_UV_Save_Hist_UV
+//Author : Tatiya, Rupesh
+//Description : Saves DN history for UV data.
+
+ mov (8) r3.0<1>:ud r0.0<8;8,1>:ud
+ mov (2) r3.0<1>:d r62.12<2;2,1>:w
+ mov (1) r3.2<1>:d 0x30007:ud
+
+ send (8) null<1>:d r3 0x5 0x40A8021:ud
+
+
+
+//Module Name : DN_UV_Save_BNE_UV
+//Author : Tatiya, Rupesh
+//Description : Saves BNE values for 16x16 U and 16x16 V.
+
+ mov (8) r1.0<1>:ud r0.0<8;8,1>:ud
+ mov (2) r1.0<1>:d r63.12<2;2,1>:w
+ mov (1) r1.2<1>:d 0x10003:ud
+
+ send (8) null<1>:d r1 0x5 0x40A8023:ud
+
+
+
+//Module Name : DN_UV_YUY2_Save_Curr_Frame_YUV
+//Author : Tatiya, Rupesh
+
+
+
+//Module Name : DN_UV_Load_Curr_Frame_Y
+//Author : Tatiya, Rupesh
+//Description : Saves Y or YUY2 of Current frame.
+
+
+
+
+ mov (8) acc0.0<1>:ud r0.0<8;8,1>:ud
+ shl (1) r62.10<1>:w r62.10<0;1,0>:w 1:w
+ mov (1) acc0.0<1>:d r62.10<0;1,0>:w
+ mov (1) acc0.1<1>:d r62.11<0;1,0>:w
+
+ mov (1) acc0.2<1>:d 0x7001F:ud
+
+ mov (8) r92.0<1>:ud acc0.0<8;8,1>:ud
+
+ mov (8) r101.0<1>:ud acc0.0<8;8,1>:ud
+ mov (8) r110.0<1>:ud acc0.0<8;8,1>:ud
+ mov (8) r119.0<1>:ud acc0.0<8;8,1>:ud
+
+ add (1) r101.1<1>:d acc0.1<0;1,0>:d 8:d
+
+ add (1) r110.0<1>:d acc0.0<0;1,0>:d 32:d
+
+ add (1) r119.0<1>:d acc0.0<0;1,0>:d 32:d
+ add (1) r119.1<1>:d acc0.1<0;1,0>:d 8:d
+
+ send (8) null<1>:d r92 0x5 0x120A8018:ud
+ send (8) null<1>:d r101 0x5 0x120A8018:ud
+ send (8) null<1>:d r110 0x5 0x120A8018:ud
+ send (8) null<1>:d r119 0x5 0x120A8018:ud
+
+
+
+//End of Thread message
+
+mov (8) r127<1>:ud r0.0<8;8,1>:ud
+ send (1) null<1>:d r127 0x27 0x02000010
+
+
+ //All sub-routines here
+
+
+// Module Name : Noise_Detection
+// Author : Tatiya, Rupesh
+// Description : Performs noise detection on 32 pixels of U (8x4) and 32 pixels of V (8x4).
+
+DN_UV_NOISE_DETECTION_UV:
+
+// Find Field Block Median
+//
+// Purpose : Find the median value of the nine pixels in the same field
+// which are centered at current pixel.
+//
+// Works on 9 pixels centered at the current pixel
+// NOTE: pixels are within same field.
+// v4 - current pixel
+//
+// v2 v1 v0
+// * * * <--- Different field - not used
+// v5 v4 v3
+// * * * <--- Different field - not used
+// v8 v7 v6
+
+// Algorithm to find median modifies the data.
+// Copy the data needed to calculate median so the original source data stays intact.
+//
+
+//TODO - Change Interleaved implementation to separated one if - , does not work on predication. - rT
+
+//Delete Later - rT
+//mov (1) pCUR_UV:uw 52*32:uw
+
+// v0
+mov (16) ubMEDIAN_TEMP(0,0)<1> r[a0.0,0]<16;16,1>
+// v0
+mov (16) ubMEDIAN_TEMP(0,16)<1> r[a0.0,32]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(1,0)<1> r[a0.0,2]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(1,16)<1> r[a0.0,34]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(2,0)<1> r[a0.0,4]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(2,16)<1> r[a0.0,36]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(3,0)<1> r[a0.0,64]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(3,16)<1> r[a0.0,96]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(4,0)<1> r[a0.0,66]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(4,16)<1> r[a0.0,98]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(5,0)<1> r[a0.0,68]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(5,16)<1> r[a0.0,100]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(6,0)<1> r[a0.0,128]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(6,16)<1> r[a0.0,160]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(7,0)<1> r[a0.0,130]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(7,16)<1> r[a0.0,162]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(8,0)<1> r[a0.0,132]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(8,16)<1> r[a0.0,164]<16;16,1>
+
+//TODO - Optimize one instruction here.
+add (1) a0.0:uw a0.0<0;1,0>:uw 64:uw
+// v0
+mov (16) ubMEDIAN_TEMP(9,0)<1> r[a0.0,0]<16;16,1>
+// v0
+mov (16) ubMEDIAN_TEMP(9,16)<1> r[a0.0,32]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(10,0)<1> r[a0.0,2]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(10,16)<1> r[a0.0,34]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(11,0)<1> r[a0.0,4]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(11,16)<1> r[a0.0,36]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(12,0)<1> r[a0.0,64]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(12,16)<1> r[a0.0,96]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(13,0)<1> r[a0.0,66]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(13,16)<1> r[a0.0,98]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(14,0)<1> r[a0.0,68]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(14,16)<1> r[a0.0,100]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(15,0)<1> r[a0.0,128]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(15,16)<1> r[a0.0,160]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(16,0)<1> r[a0.0,130]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(16,16)<1> r[a0.0,162]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(17,0)<1> r[a0.0,132]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(17,16)<1> r[a0.0,164]<16;16,1>
+
+//TODO - Optimize one instruction here.
+add (1) a0.0:uw a0.0<0;1,0>:uw 64:uw
+
+// MedianSwap
+//
+// MedianSwap(inOutLeft, inOutRight)
+// {
+// if (inOutLeft > inOutRight)
+// {
+// temp = inOutLeft
+// inOutLeft = inOutRight
+// inOutRight = temp
+// }
+// }
+
+// MedianSwap(v1, v2) - U
+// MedianSwap(v4, v5) - U
+// MedianSwap(v1, v2) - V
+// MedianSwap(v4, v5) - V
+
+cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(1,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(5,0)<32;16,2>
+cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(1,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(1,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(1,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(1,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+(f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(5,0)<32;16,2>
+(f1.0) mov (16) ubMEDIAN_TEMP(1,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+(f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubTEMP1(0,0)<16;16,1>
+(f0.1) mov (16) ubMEDIAN_TEMP(5,0)<2> ubTEMP1(0,16)<16;16,1>
+(f1.0) mov (16) ubMEDIAN_TEMP(2,1)<2> ubTEMP1(1,0)<16;16,1>
+(f1.1) mov (16) ubMEDIAN_TEMP(5,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// MedianSwap(v7, v8) - U
+// MedianSwap(v0, v1) - U
+// MedianSwap(v7, v8) - V
+// MedianSwap(v0, v1) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(7,0)<32;16,2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(0,0)<32;16,2> ubMEDIAN_TEMP(1,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(7,1)<32;16,2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(0,1)<32;16,2> ubMEDIAN_TEMP(1,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(7,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(0,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(7,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(0,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(7,0)<2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(0,0)<2> ubMEDIAN_TEMP(1,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(7,1)<2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(0,1)<2> ubMEDIAN_TEMP(1,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(8,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(1,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(8,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(1,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v3, v4) - U
+ // MedianSwap(v6, v7) - U
+ // MedianSwap(v3, v4) - V
+ // MedianSwap(v6, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(3,0)<32;16,2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(6,0)<32;16,2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(3,1)<32;16,2> ubMEDIAN_TEMP(4,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(6,1)<32;16,2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(3,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(6,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(3,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(6,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(3,0)<2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(6,0)<2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(3,1)<2> ubMEDIAN_TEMP(4,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(6,1)<2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(7,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(4,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(7,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v1, v2) - U
+ // MedianSwap(v4, v5) - U
+ // MedianSwap(v1, v2) - V
+ // MedianSwap(v4, v5) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(1,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(1,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(1,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(1,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(1,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(1,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(5,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(2,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(5,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v7, v8) - U
+ // MedianSwap(v0, v3) - U
+ // MedianSwap(v7, v8) - V
+ // MedianSwap(v0, v3) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(7,0)<32;16,2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(0,0)<32;16,2> ubMEDIAN_TEMP(3,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(7,1)<32;16,2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(0,1)<32;16,2> ubMEDIAN_TEMP(3,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(7,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(0,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(7,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(0,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(7,0)<2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(0,0)<2> ubMEDIAN_TEMP(3,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(7,1)<2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(0,1)<2> ubMEDIAN_TEMP(3,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(8,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(3,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(8,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(3,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // NOTE:
+ // Compare v0 to v6 to find the minimum.
+ // Store the minimum for future use.
+ //TODO - Find if MIN is needed.
+ //cmp.l.f0.0 (16) null:w ubMEDIAN_TEMP(%1+0,0)<32;16,2> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //cmp.l.f1.0 (16) null:w ubMEDIAN_TEMP(%1+0,1)<32;16,2> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+ //(f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,0)<32;16,2>
+ //(f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,1)<32;16,2>
+ //(-f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //(-f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+
+ // MedianSwap(v5, v8) - U
+ // MedianSwap(v4, v7) - U
+ // MedianSwap(v5, v8) - V
+ // MedianSwap(v4, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(5,0)<32;16,2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(5,1)<32;16,2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(5,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(5,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(5,0)<2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(5,1)<2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(8,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(7,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(8,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(7,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// NOTE:
+// Compare v2 to v8 to find the maximum.
+// Store the maximum for future use.
+ //TODO - Find if MAX is needed.
+// cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(%1+2,0)<32;16,2> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+// cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(%1+2,1)<32;16,2> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+//(f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,0)<32;16,2>
+//(f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,1)<32;16,2>
+//(-f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+//(-f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+
+// MedianSwap(v3, v6) - U
+// MedianSwap(v1, v4) - U
+// MedianSwap(v3, v6) - V
+// MedianSwap(v1, v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(3,0)<32;16,2> ubMEDIAN_TEMP(6,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(1,0)<32;16,2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(3,1)<32;16,2> ubMEDIAN_TEMP(6,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(1,1)<32;16,2> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(6,0)<2> ubMEDIAN_TEMP(3,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(1,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(6,1)<2> ubMEDIAN_TEMP(3,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(1,1)<32;16,2>
+
+ // MedianSwap(v2,v5) - U
+ // MedianSwap(v4,v7) - U
+ // MedianSwap(v2,v5) - V
+ // MedianSwap(v4,v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(2,0)<32;16,2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(2,1)<32;16,2> ubMEDIAN_TEMP(5,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(2,1)<2> ubMEDIAN_TEMP(5,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(2,1)<2> ubTEMP1(0,16)<16;16,1>
+
+ // MedianSwap(v6,v4) - U
+ // MedianSwap(v6,v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(6,0)<32;16,2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(6,1)<32;16,2> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(6,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(6,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(10,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(14,0)<32;16,2>
+cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(10,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(10,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(10,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(10,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+(f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(14,0)<32;16,2>
+(f1.0) mov (16) ubMEDIAN_TEMP(10,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+(f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubTEMP1(0,0)<16;16,1>
+(f0.1) mov (16) ubMEDIAN_TEMP(14,0)<2> ubTEMP1(0,16)<16;16,1>
+(f1.0) mov (16) ubMEDIAN_TEMP(11,1)<2> ubTEMP1(1,0)<16;16,1>
+(f1.1) mov (16) ubMEDIAN_TEMP(14,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// MedianSwap(v7, v8) - U
+// MedianSwap(v0, v1) - U
+// MedianSwap(v7, v8) - V
+// MedianSwap(v0, v1) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(16,0)<32;16,2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(9,0)<32;16,2> ubMEDIAN_TEMP(10,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(16,1)<32;16,2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(9,1)<32;16,2> ubMEDIAN_TEMP(10,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(16,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(9,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(16,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(9,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(16,0)<2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(9,0)<2> ubMEDIAN_TEMP(10,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(16,1)<2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(9,1)<2> ubMEDIAN_TEMP(10,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(17,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(10,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(17,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(10,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v3, v4) - U
+ // MedianSwap(v6, v7) - U
+ // MedianSwap(v3, v4) - V
+ // MedianSwap(v6, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(12,0)<32;16,2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(15,0)<32;16,2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(12,1)<32;16,2> ubMEDIAN_TEMP(13,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(15,1)<32;16,2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(12,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(15,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(12,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(15,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(12,0)<2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(15,0)<2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(12,1)<2> ubMEDIAN_TEMP(13,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(15,1)<2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(16,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(13,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(16,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v1, v2) - U
+ // MedianSwap(v4, v5) - U
+ // MedianSwap(v1, v2) - V
+ // MedianSwap(v4, v5) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(10,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(10,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(10,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(10,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(10,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(10,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(14,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(11,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(14,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v7, v8) - U
+ // MedianSwap(v0, v3) - U
+ // MedianSwap(v7, v8) - V
+ // MedianSwap(v0, v3) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(16,0)<32;16,2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(9,0)<32;16,2> ubMEDIAN_TEMP(12,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(16,1)<32;16,2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(9,1)<32;16,2> ubMEDIAN_TEMP(12,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(16,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(9,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(16,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(9,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(16,0)<2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(9,0)<2> ubMEDIAN_TEMP(12,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(16,1)<2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(9,1)<2> ubMEDIAN_TEMP(12,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(17,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(12,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(17,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(12,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // NOTE:
+ // Compare v0 to v6 to find the minimum.
+ // Store the minimum for future use.
+ //TODO - Find if MIN is needed.
+ //cmp.l.f0.0 (16) null:w ubMEDIAN_TEMP(%1+0,0)<32;16,2> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //cmp.l.f1.0 (16) null:w ubMEDIAN_TEMP(%1+0,1)<32;16,2> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+ //(f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,0)<32;16,2>
+ //(f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,1)<32;16,2>
+ //(-f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //(-f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+
+ // MedianSwap(v5, v8) - U
+ // MedianSwap(v4, v7) - U
+ // MedianSwap(v5, v8) - V
+ // MedianSwap(v4, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(14,0)<32;16,2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(14,1)<32;16,2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(14,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(14,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(14,0)<2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(14,1)<2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(17,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(16,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(17,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(16,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// NOTE:
+// Compare v2 to v8 to find the maximum.
+// Store the maximum for future use.
+ //TODO - Find if MAX is needed.
+// cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(%1+2,0)<32;16,2> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+// cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(%1+2,1)<32;16,2> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+//(f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,0)<32;16,2>
+//(f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,1)<32;16,2>
+//(-f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+//(-f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+
+// MedianSwap(v3, v6) - U
+// MedianSwap(v1, v4) - U
+// MedianSwap(v3, v6) - V
+// MedianSwap(v1, v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(12,0)<32;16,2> ubMEDIAN_TEMP(15,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(10,0)<32;16,2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(12,1)<32;16,2> ubMEDIAN_TEMP(15,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(10,1)<32;16,2> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(15,0)<2> ubMEDIAN_TEMP(12,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(10,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(15,1)<2> ubMEDIAN_TEMP(12,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(10,1)<32;16,2>
+
+ // MedianSwap(v2,v5) - U
+ // MedianSwap(v4,v7) - U
+ // MedianSwap(v2,v5) - V
+ // MedianSwap(v4,v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(11,0)<32;16,2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(11,1)<32;16,2> ubMEDIAN_TEMP(14,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(11,1)<2> ubMEDIAN_TEMP(14,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(11,1)<2> ubTEMP1(0,16)<16;16,1>
+
+ // MedianSwap(v6,v4) - U
+ // MedianSwap(v6,v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(15,0)<32;16,2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(15,1)<32;16,2> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(15,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(15,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+// Sobel Value calculation for the current pixel v4
+// v2 v1 v0
+// * * * <--- Different field - not used
+// v5 v4 v3
+// * * * <--- Different field - not used
+// v8 v7 v6
+//
+// Gx = -v0 - 2*v3 - v6 + v2 + 2*v5 + v8
+// Gy = v0 + 2*v1 + v2 - v6 - 2*v7 - v8
+//
+// Sobel = (|Gx| + |Gy|) >> 3
+
+//TODO - Change Later - rT
+add (1) a0.0:uw a0.0<0;1,0>:uw -128:uw
+
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,64]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,132]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,0]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,128]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,4]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(0)<1> r[a0.0,68]<16;16,1>:ub 2:w
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,96]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,164]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,32]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,160]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,36]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(1)<1> r[a0.0,100]<16;16,1>:ub 2:w
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,128]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,196]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,64]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,192]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,68]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(2)<1> r[a0.0,132]<16;16,1>:ub 2:w
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,160]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,228]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,96]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,224]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,100]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(3)<1> r[a0.0,164]<16;16,1>:ub 2:w
+
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,2]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,0]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,132]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,4]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,128]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,130]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(0)<16;16,1>
+
+shr (16) uwSOBEL(0)<1> acc0.0<16;16,1>:uw 3:uw
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,34]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,32]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,164]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,36]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,160]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,162]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(1)<16;16,1>
+
+shr (16) uwSOBEL(1)<1> acc0.0<16;16,1>:uw 3:uw
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,66]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,64]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,196]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,68]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,192]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,194]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(2)<16;16,1>
+
+shr (16) uwSOBEL(2)<1> acc0.0<16;16,1>:uw 3:uw
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,98]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,96]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,228]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,100]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,224]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,226]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(3)<16;16,1>
+
+shr (16) uwSOBEL(3)<1> acc0.0<16;16,1>:uw 3:uw
+
+//Mov Median in CURBE_TEMP to free up temp space.
+mov (16) ubMEDIAN(0,0)<1> ubMEDIAN_TEMP(4,0)<16;16,1>
+mov (16) ubMEDIAN(0,16)<1> ubMEDIAN_TEMP(4,16)<16;16,1>
+mov (16) ubMEDIAN(0,32)<1> ubMEDIAN_TEMP(13,0)<16;16,1>
+mov (16) ubMEDIAN(0,48)<1> ubMEDIAN_TEMP(13,16)<16;16,1>
+
+// Find:
+// absDiff = abs(ubCurY - ubMedian)
+// Find the difference between pixel and median value.
+
+//Median is interleaved. So difference is also interleaved.
+
+//------------------------------------------------------------------------------------------
+//Process 16 U and 16 V pixels here and rest later.
+// first row - v0,v1,v2
+add (16) wDIFF(0)<1> r[a0.0,0]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(1)<1> r[a0.0,2]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(2)<1> r[a0.0,4]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+
+// second row - v3,v4,v5
+add (16) wDIFF(3)<1> r[a0.0,64]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(4)<1> r[a0.0,66]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(5)<1> r[a0.0,68]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+
+// third row - v6,v7,v8
+add (16) wDIFF(6)<1> r[a0.0,128]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(7)<1> r[a0.0,130]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(8)<1> r[a0.0,132]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+// first row - v0,v1,v2
+add (16) wDIFF(9)<1> r[a0.0,32]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(10)<1> r[a0.0,34]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(11)<1> r[a0.0,36]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+
+// second row - v3,v4,v5
+add (16) wDIFF(12)<1> r[a0.0,96]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(13)<1> r[a0.0,98]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(14)<1> r[a0.0,100]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+
+// third row - v6,v7,v8
+add (16) wDIFF(15)<1> r[a0.0,160]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(16)<1> r[a0.0,162]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(17)<1> r[a0.0,164]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+
+//TODO - Change Later - rT
+add (1) a0.0:uw a0.0<0;1,0>:uw 64:uw
+
+// Find sum of all absolute differences AND
+// maximum absolute difference for 16 U and 16 V here.
+//First 2 rows of 8x4
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(2)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(3)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(4)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(5)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(6)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(7)<16;16,1>
+ add (16) uwSOAD(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF(8)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//------------
+ //DIFF(0-7) is not needed here. Populate it.
+ // first row - v0,v1,v2
+ add (16) wDIFF(0)<1> r[a0.0,0]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(1)<1> r[a0.0,2]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(2)<1> r[a0.0,4]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+
+ // second row - v3,v4,v5
+ add (16) wDIFF(3)<1> r[a0.0,64]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(4)<1> r[a0.0,66]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(5)<1> r[a0.0,68]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+
+ // third row - v6,v7
+ add (16) wDIFF(6)<1> r[a0.0,128]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(7)<1> r[a0.0,130]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+//------------
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(0)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+
+//------------
+ //Load v8 - DIFF(8)
+ add (16) wDIFF(8)<1> r[a0.0,132]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+//------------
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(11)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(12)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(13)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(14)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(15)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(16)<16;16,1>
+ add (16) uwSOAD(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF(17)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//------------
+ //DIFF(0-7) is not needed here. Populate it.
+ // first row - v0,v1,v2
+ add (16) wDIFF(9)<1> r[a0.0,32]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(10)<1> r[a0.0,34]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(11)<1> r[a0.0,36]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+
+ // second row - v3,v4,v5
+ add (16) wDIFF(12)<1> r[a0.0,96]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(13)<1> r[a0.0,98]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(14)<1> r[a0.0,100]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+
+ // third row - v6,v7
+ add (16) wDIFF(15)<1> r[a0.0,160]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(16)<1> r[a0.0,162]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+//------------
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(1)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+
+//------------
+ //Load v8 - DIFF(8)
+ add (16) wDIFF(17)<1> r[a0.0,164]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+//------------
+
+//if ((sobel_edge_measure < m_SobelEdgeThreshold) && ((block_max-block_min) < m_LocalDiffThreshold))
+// if (sigma_mb_min > sigma)
+// sigma_mb_min = sigma;
+
+//NOTE: block_min is always zero as median is one of the value in 3x3 block. So no need o calculate it.
+// So just do -
+//if ((sobel_edge_measure < m_SobelEdgeThreshold) && ((block_max) < m_LocalDiffThreshold) && ( sigma < sigma_mb_min))
+// sigma_mb_min = sigma;
+
+//We are processing 32 bytes of U and 32 bytes of V - each of size 8x4.
+//Compare first 8 bytes with max possible (255).
+//Start above condition from second 8 bytes.
+
+//TODO - Change Later - rT
+// mov (1) pCUR_MIN_SOAD_8x4:uw 1752:uw //r54.24:ub
+
+//First row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(0)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(0)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(0)<16;16,1> 255:uw
+(f0.0) sel (16) uwSOBEL(0)<1> uwSOAD(0)<16;16,1> 255:uw
+
+//Second row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(1)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(1)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(1)<16;16,1> uwSOBEL(0)<16;16,1>
+(f0.0) mov (16) uwSOBEL(0)<1> uwSOAD(1)<16;16,1>
+
+// Find sum of all absolute differences AND
+// maximum absolute difference for 16 U and 16 V here.
+//Second 2 rows of 8x4
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(2)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(3)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(4)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(5)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(6)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(7)<16;16,1>
+ add (16) uwSOAD(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF(8)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(0)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(11)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(12)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(13)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(14)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(15)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(16)<16;16,1>
+ add (16) uwSOAD(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF(17)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(1)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+
+//Third row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(2)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(0)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(0)<16;16,1> uwSOBEL(0)<16;16,1>
+(f0.0) mov (16) uwSOBEL(0)<1> uwSOAD(0)<16;16,1>
+
+//Fourth row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(3)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(1)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(1)<16;16,1> uwSOBEL(0)<16;16,1>
+(f0.0) mov (16) uwSOBEL(0)<1> uwSOAD(1)<16;16,1>
+
+ cmp.l.f0.0 (8) null:uw uwSOBEL(0,0)<8;8,1> uwSOBEL(0,8)<8;8,1>
+(f0.0) sel (8) uwSOBEL(0)<1> uwSOBEL(0,0)<8;8,1> uwSOBEL(0,8)<8;8,1>
+
+ cmp.l.f0.0 (4) null:uw uwSOBEL(0,0)<4;4,1> uwSOBEL(0,4)<4;4,1>
+(f0.0) sel (4) uwSOBEL(0)<1> uwSOBEL(0,0)<4;4,1> uwSOBEL(0,4)<4;4,1>
+
+ cmp.l.f0.0 (2) null:uw uwSOBEL(0,0)<2;2,1> uwSOBEL(0,2)<2;2,1>
+(f0.0) sel (2) r[a0.1,0]<1>:uw uwSOBEL(0,0)<2;2,1> uwSOBEL(0,2)<2;2,1>
+
+
+
+
+
+
+// End of common.inc
+
+mov (1) ip:ud r7.7<0;1,0>:d
+
+
+.end_code
+.end_kernel
--- /dev/null
+/*
+ * 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
--- /dev/null
+/*
+ * 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.
+ */
+// 57 // Total instruction count
+// 1 // Total kernel count
+
+.kernel PA_DN_PA
+.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
+
+
+
+//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
+/*
+ * 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.
+ */
// 12 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 10 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 10 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 10 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 15 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 13 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 13 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 13 // Total instruction count
// 1 // Total kernel count
--- /dev/null
+/*
+ * 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.
+ */
+// 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=====================================
+
+
+//=======================================================================
+
+
+//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 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) r7.0<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> r7.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2
+mov (1) mudMSGHDR_STMM(0,1)<1> r7.1<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 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 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> 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> 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 r7.0<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> r7.0<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 r7.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
+mov (1) r27.1<1>:ud r7.1<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 r7.26<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
--- /dev/null
+/*
+ * 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.
+ */
+// 90 // Total instruction count
+// 1 // Total kernel count
+
+.kernel PL3_DNDI_PA
+.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=====================================
+
+
+//=======================================================================
+
+
+//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 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) r7.0<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> r7.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2
+mov (1) mudMSGHDR_STMM(0,1)<1> r7.1<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 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 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> 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> 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 r7.0<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> r7.0<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_PA_16x4.asm
+// Author: Vivek Kumar
+// Description: Save two 16x4 blocks of DI output in Packed format
+
+
+add (4) a0.4<1>:uw r2.28<4;4,1>:ub 608:w // Initial Y,U,V offset in YUV422 block; it starts at m20
+
+mov (8) r27.0<1>:ud r0.0<8;8,1>:ud
+shl (1) r27.0<1>:d r7.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
+mov (1) r27.1<1>:d r7.1<0;1,0>:w { NoDDClr, NoDDChk } // Block origin
+mov (1) r27.2<1>:ud 0x3001F:ud { NoDDChk } // Block width and height (32x8)
+
+//prepare the message headers
+mov (8) r18.0<1>:ud r27<8;8,1>:ud
+mov (8) r23.0<1>:ud r27<8;8,1>:ud
+
+// Pack 2nd field Y
+ mov (16) r[a0.4, 0]<2> ubDNDI_RESP(0,0) { NoDDClr }
+ mov (16) r[a0.4, 32]<2> ubDNDI_RESP(0,16) { NoDDClr }
+ mov (16) r[a0.4, 64]<2> ubDNDI_RESP(0,32) { NoDDClr }
+ mov (16) r[a0.4, 96]<2> ubDNDI_RESP(0,48) { NoDDClr }
+// Pack 2nd field U
+ mov (8) r[a0.5, 0]<4> ubDNDI_RESP(2,1)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 32]<4> ubDNDI_RESP(2,17)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 64]<4> ubDNDI_RESP(2,33)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 96]<4> ubDNDI_RESP(2,49)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+// Pack 2nd field V
+ mov (8) r[a0.6, 0]<4> ubDNDI_RESP(2,0)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 32]<4> ubDNDI_RESP(2,16)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 64]<4> ubDNDI_RESP(2,32)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 96]<4> ubDNDI_RESP(2,48)<16;8,2> { NoDDChk } //Vpixels
+
+// Pack 1st field Y
+ mov (16) r[a0.4, 160]<2> ubDNDI_RESP(4,0) { NoDDClr }
+ mov (16) r[a0.4, 192]<2> ubDNDI_RESP(4,16) { NoDDClr }
+ mov (16) r[a0.4, 224]<2> ubDNDI_RESP(4,32) { NoDDClr }
+ mov (16) r[a0.4, 256]<2> ubDNDI_RESP(4,48) { NoDDClr }
+// Pack 1st field U
+ mov (8) r[a0.5, 160]<4> ubDNDI_RESP(6,1)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 192]<4> ubDNDI_RESP(6,17)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 224]<4> ubDNDI_RESP(6,33)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 256]<4> ubDNDI_RESP(6,49)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+// Pack 1st field V
+ mov (8) r[a0.6, 160]<4> ubDNDI_RESP(6,0)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 192]<4> ubDNDI_RESP(6,16)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 224]<4> ubDNDI_RESP(6,32)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 256]<4> ubDNDI_RESP(6,48)<16;8,2> { NoDDChk } //Vpixels
+
+//save the previous frame
+send (8) null<1>:d r18.0 0x5 0xA0A801B:ud
+
+//save the current frame
+send (8) null<1>:d r23.0 0x5 0xA0A801E: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
--- /dev/null
+/*
+ * 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.
+ */
+// 1295 // Total instruction count
+// 1 // Total kernel count
+
+
+.kernel PL3_DNUV_PL3
+.code
+
+
+
+//Module : DN_UV_Setup
+//Author : Tatiya, Rupesh
+//Description : Initial Set-up for DN_UV
+
+
+
+
+// Module name : ChromaDenoise.inc
+// Author : Tatiya, Rupesh
+
+
+
+
+// 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
+
+
+//Interface:
+//Static Parameters:
+//r1
+
+
+//======================================================
+//Interface for serpent mode Chroma Denoise, added by Le
+//======================================================
+//r1
+
+
+//noise history thresholds (low and high)
+
+
+//temporal difference thresholds (high and low)
+
+
+//noise history thresholds (low and high)
+//#define ubNoiseHistMaxHigh r1.22
+//#define ubNoiseHistMaxLow r1.23
+//#define ubNoiseHistDeltaHigh r1.24
+//#define ubNoiseHistDeltaLow r1.25
+
+//Gaussian thresholds
+
+
+//temporal difference thresholds (default)
+
+
+//r2
+//history thresholds (default)
+
+
+//denoise factor (0-63)
+
+
+//====================== Binding table (Explicit To DNUV)=========================================
+//Used by DN_UV kernels
+
+
+ //Pointer to Current Frame UV
+
+
+//r1-r6
+ //CURBE GRFs used as TEMP : Used for max computation and storing max temporarily. : r1-r6
+
+
+ .declare ubCURBE_TEMP Base=r1.0 ElementSize=1 Type=ub
+ .declare uwCURBE_TEMP Base=r1.0 ElementSize=2 Type=uw
+ .declare wCURBE_TEMP Base=r1.0 ElementSize=2 Type=w
+ .declare fCURBE_TEMP Base=r1.0 ElementSize=4 Type=f
+ .declare udCURBE_TEMP Base=r1.0 ElementSize=4 Type=ud
+ .declare uwMAX_ABS_DIFF Base=r5.0 ElementSize=2 Type=uw
+
+ //r1
+
+
+ //r3
+
+
+ //r4
+
+//r7
+ //All of the following has to defined in Same GRF for optimal performance.
+
+
+//r8-24
+ //Previous Frame UV
+
+ .declare udPREV_UV Base=r8.0 ElementSize=4 Type=ud
+ .declare ubPREV_UV Base=r8.0 ElementSize=1 Type=ub
+
+
+//r25-48
+ //TEMP Space for any Usage.
+
+
+//=========================================================================
+//Definations and declarations for serpent mode Chroma Denoise, added by Le
+//=========================================================================
+
+
+ .declare udGNE_UV Base=r24.0 ElementSize=4 Type=ud
+ .declare fGNE_UV Base=r24.0 ElementSize=4 Type=f
+ .declare ubGNE_UV Base=r24.0 ElementSize=1 Type=ub
+
+ .declare udMSGHDR_BNE_SERP Base=r25.0 ElementSize=4 Type=ud
+ .declare udMSGSRC_BNE_SERP Base=r26.0 ElementSize=4 Type=ud
+
+
+ .declare ubDN_UV_Thresholds Base=r26.0 ElementSize=1 Type=ub
+ .declare ubDN_UV_Thresholds_Temp Base=r27.0 ElementSize=1 Type=ub
+ .declare udDN_UV_Thresholds Base=r26.0 ElementSize=4 Type=ud
+ .declare udDN_UV_Thresholds_Temp Base=r27.0 ElementSize=4 Type=ud
+ .declare fDN_UV_Thresholds Base=r26.0 ElementSize=4 Type=f
+ .declare fDN_UV_Thresholds_Temp Base=r27.0 ElementSize=4 Type=f
+
+
+//====================================================================================
+
+
+ //TEMP23: To hold V data for PL3 surfaces
+ .declare udCURR_V_TEMP Base=r25.0 ElementSize=4 Type=ud
+ .declare ubCURR_V_TEMP Base=r25.0 ElementSize=1 Type=ub
+
+ //GRFs to calculate Median: r25-r42
+ .declare ubMEDIAN_TEMP Base=r25.0 ElementSize=1 Type=ub
+
+ //18 GRFs to hold difference : r25-r42
+ .declare wDIFF Base=r25.0 ElementSize=2 Type=w
+ .declare uwDIFF Base=r25.0 ElementSize=2 Type=uw
+
+ //Temporal Diff
+ .declare wDIFF_TEMPORAL Base=r25.0 ElementSize=2 Type=w
+ .declare ubDIFF_TEMPORAL Base=r25.0 ElementSize=1 Type=ub
+
+ //4 GRFs to hold Sobel Value : r43-46
+ .declare wSOBEL_X Base=r43.0 ElementSize=2 Type=w
+ .declare uwSOBEL Base=r43.0 ElementSize=2 Type=uw
+
+
+ //2 GRFs to hold SOAD temporarily: r47-48
+ .declare uwSOAD Base=r47.0 ElementSize=2 Type=uw
+
+ //Temp GRFs to hold extra YUYV pixels: r43-r48
+ .declare ubTEMP5 Base=r43.0 ElementSize=1 Type=ub
+
+ //Temp GRFs in Median Calculation: r47-r48
+ .declare ubTEMP1 Base=r47.0 ElementSize=1 Type=ub
+
+ .declare uwTEMP0 Base=r48.0 ElementSize=2 Type=uw
+ .declare ubTEMP0 Base=r48.0 ElementSize=1 Type=ub
+
+ //Temp Space to store Median : r49-50
+
+ .declare ubMEDIAN Base=r49.0 ElementSize=1 Type=ub
+
+//r49
+
+
+//r50
+ //Message Source
+
+
+//r51
+ //DN_UV History Surface
+
+ .declare udHIST_UV Base=r51.0 ElementSize=4 Type=ud
+ .declare ubHIST_UV Base=r51.0 ElementSize=1 Type=ub
+
+//r52 - r91
+ //r52
+ //Current Frame UV
+
+
+ .declare udCURR_UV Base=r52.0 ElementSize=4 Type=ud
+ .declare ubCURR_UV Base=r52.0 ElementSize=1 Type=ub
+
+ //r54
+ //CURBE COPY
+
+
+ //r55
+
+
+ .declare uwSOAD_MIN_8x4 Base=r56.0 ElementSize=2 Type=uw
+
+ //r61
+
+
+ //r62
+
+
+ //History Surface Temp Origin
+
+
+ //r63
+ //Current Frame Y Temp Origin
+
+
+ //BNE Surface Origin
+
+
+ //r70
+
+ .declare uwDIFF_TEMPORAL_SUM4x4 Base=r70.0 ElementSize=2 Type=uw //4 GRFs
+
+ //r74-91 : For Saving Dest UV (PL2/PL3)
+
+
+ .declare ubMSGPAYLOAD_UV0 Base=r75.0 ElementSize=1 Type=ub
+
+
+ .declare ubMSGPAYLOAD_U Base=r75.0 ElementSize=1 Type=ub
+
+
+ .declare ubMSGPAYLOAD_UV1 Base=r84.0 ElementSize=1 Type=ub
+
+
+ .declare ubMSGPAYLOAD_V Base=r84.0 ElementSize=1 Type=ub
+
+ //r90
+
+ .declare uwDIFF_TEMPORAL_SUM4x4_FINAL Base=r90.0 ElementSize=2 Type=uw //2 GRFs
+
+//r92-127
+ //Current Frame Y
+
+
+ //r92
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_0 Base=r92 ElementSize=2 Type=uw
+ //r101
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_1 Base=r101 ElementSize=2 Type=uw
+ //r110
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_2 Base=r110 ElementSize=2 Type=uw
+ //r119
+ .declare uwDIFF_TEMPORAL_SUM4x4_TEMP_3 Base=r119 ElementSize=2 Type=uw
+
+ .declare udCURR_Y0 Base=r93.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y0 Base=r93.0 ElementSize=1 Type=ub
+ .declare udCURR_Y1 Base=r102.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y1 Base=r102.0 ElementSize=1 Type=ub
+ .declare udCURR_Y2 Base=r111.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y2 Base=r111.0 ElementSize=1 Type=ub
+ .declare udCURR_Y3 Base=r120.0 ElementSize=4 Type=ud
+ .declare ubCURR_Y3 Base=r120.0 ElementSize=1 Type=ub
+
+ //r92: To hold U data for PL3 surfaces
+ .declare udCURR_U_TEMP Base=r92.0 ElementSize=4 Type=ud
+ .declare ubCURR_U_TEMP Base=r92.0 ElementSize=1 Type=ub
+
+ //r112: To hold U data for PL3 surfaces
+ .declare udPREV_U_TEMP Base=r112.0 ElementSize=4 Type=ud
+ .declare ubPREV_U_TEMP Base=r112.0 ElementSize=1 Type=ub
+
+ //r120: To hold U data for PL3 surfaces
+ .declare udPREV_V_TEMP Base=r120.0 ElementSize=4 Type=ud
+ .declare ubPREV_V_TEMP Base=r120.0 ElementSize=1 Type=ub
+
+
+ // Initialize message source with r0.
+ mov (8) r50.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r92.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r101.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r110.0<1>:ud r0.0<8;8,1>:ud
+ mov (8) r119.0<1>:ud r0.0<8;8,1>:ud
+
+
+
+//Module Name : DN_UV_PL3_Load_Curr_Frame_UV
+//Author : Tatiya, Rupesh
+//Description : Loads Current Frame U/V data for PL3 input.
+
+
+
+//Module name : DN_UV_Load_Curr_Frame_UV
+//Author : Tatiya, Rupesh
+//Description : Loads Current Frame (UV only).
+// We need 4 extra rows (2 per field) and 2 extra pixel (1 each side) for both U and V each.
+// The processing size is 16x16 U and V each. So we need : U size - 18x20, V size - 18x20, UV size - 36x20, YUYV size - 72x20.
+
+
+
+
+//18x20 U/V block is partitioned as follows:
+// <------ 18 ------>
+// ------------------
+// | 18x8 A1 |
+// | |
+// |----------------|
+// | 18x8 A2 |
+// | |
+// |----------------|
+// | 18x4 A2 |
+// |----------------|
+//
+// Cordinates: (x-1, y-2), (x-1, y+6), (x-1, y+14)
+
+//1. Load U data into starting at CURR_Y0 (r93-r122)
+//2. Load V data into TEMP space (r25-r44)
+
+ //U/V surface origin: (ORIX/2, ORIY/2)
+ add (2) r7.4<1>:w r7.0<2;2,1>:w r4.4<2;2,1>:w { AccWrEn } // Source Block origin
+ shr (2) r7.4<1>:w acc0.4<2;2,1>:w 1:w //U Data
+ mov (2) acc0.0<1>:d r7.4<2;2,1>:w
+
+ //A1
+ add (1) r50.0<1>:d acc0.0<0;1,0>:d -1:d
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d -2:d
+ mov (1) r50.2<1>:ud 0x70011:ud
+ send (8) udCURR_U_TEMP(0)<1> r50 0x4 0x2890004:ud
+
+ //A2
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d 6:d
+ send (8) udCURR_U_TEMP(8)<1> r50 0x4 0x2890004:ud
+
+ //A3
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d 14:d
+ mov (1) r50.2<1>:ud 0x30011:ud
+ send (8) udCURR_U_TEMP(16)<1> r50 0x4 0x2490004:ud
+
+ //V Data
+ //A1
+ add (1) r50.0<1>:d acc0.0<0;1,0>:d -1:d
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d -2:d
+ mov (1) r50.2<1>:ud 0x70011:ud
+ send (8) udCURR_V_TEMP(0)<1> r50 0x4 0x2890005:ud
+
+ //A2
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d 6:d
+ send (8) udCURR_V_TEMP(8)<1> r50 0x4 0x2890005:ud
+
+ //A3
+ add (1) r50.1<1>:d acc0.1<0;1,0>:d 14:d
+ mov (1) r50.2<1>:ud 0x30011:ud
+ send (8) udCURR_V_TEMP(16)<1> r50 0x4 0x2490005:ud
+
+ //History Origin, Current Y origin and BNE surface origin - all are in inline GRF. Use , . -rT.
+
+ //Calculate Origin For History Surface: (ORIX/4, ORIY/8)
+ mov (16) acc0.0<1>:w r7.0<0;2,1>:w { AccWrEn }
+ shr (1) r7.2<1>:w acc0.2<0;1,0>:w 2:w
+ shr (1) r7.3<1>:w acc0.3<0;1,0>:w 3:w
+
+ //Calculate Origin For BNE Surface: (ORIX/8, ORIY/16)
+ shr (1) r7.6<1>:w acc0.6<0;1,0>:w 3:w
+ shr (1) r7.7<1>:w acc0.7<0;1,0>:w 4:w
+
+
+
+//Module Name : DN_UV_PL3_Load_Prev_Frame_UV.asm
+//Author : Tatiya, Rupesh
+//Description : Loads Pevious Frame UV data for PL3 input.
+
+
+
+//Module Name : DN_UV_Load_Prev_Frame_UV
+//Author : Tatiya, Rupesh
+//Description : Loads Prev Frame (UV only). U size - 16x16, V size - 16x16, UV size - 32x16, YUYV size - 64x16.
+
+
+
+
+//1. Load U in bottom half of UV space for prev frame (r17-r24)
+//2. Load V in bottom quarter of Y space for curr frame (r120-r127)
+
+ mov (2) r50.0<1>:d r7.4<2;2,1>:w { AccWrEn } // Source lock origin
+ mov (1) r50.2<1>:ud 0xF000F:ud // U/V block width and height (16x16)
+
+ mov (8) r49.0<1>:ud r50<8;8,1>:ud
+
+ send (8) udPREV_U_TEMP(0)<1> r50 0x4 0x2890001:ud //U data
+ send (8) udPREV_V_TEMP(0)<1> r49 0x4 0x2890002:ud //V data
+
+
+ //TODO - See if History loading can be combined with Prev Frame Load. - rT
+
+
+//Module name : DN_UV_Load_Hist_UV
+//Author : Tatiya, Rupesh
+//Description : Load DN History for UV denoise. 4x4 for each U & V.
+
+
+
+
+ mov (2) r50.0<1>:d r7.2<2;2,1>:w
+ mov (1) r50.2<1>:ud 0x30007:ud
+ send (8) udHIST_UV(0)<1> r50 0x4 0x2190022:ud
+
+
+
+//File Name : DN_UV_PL3_Interleave_Curr_Frame_UV.asm
+//Author : Tatiya, Rupesh
+//Description : Interleave separately loaded U and V for PL3 format.
+// This is needed because Noise Detection and Noise Reduction works on interleaved UV data.
+
+//1. U data: Starting at CURR_Y0 (r93-r122)
+//2. V data: TEMP space (r25-r44)
+
+//In one GRF, we need 10 U (1+8+1) bytes, but there's no SIMD10. So use SIMD16 and discard last 6 bytes.
+
+
+//Move U data
+ mov (16) ubCURR_UV(0,0)<2> ubCURR_U_TEMP(0,0)<16;16,1>
+ mov (16) ubCURR_UV(20,0)<2> ubCURR_U_TEMP(0,8)<16;16,1>
+
+ mov (16) ubCURR_UV(1,0)<2> ubCURR_U_TEMP(1,0)<16;16,1>
+ mov (16) ubCURR_UV(21,0)<2> ubCURR_U_TEMP(1,8)<16;16,1>
+
+ mov (16) ubCURR_UV(2,0)<2> ubCURR_U_TEMP(2,0)<16;16,1>
+ mov (16) ubCURR_UV(22,0)<2> ubCURR_U_TEMP(2,8)<16;16,1>
+
+ mov (16) ubCURR_UV(3,0)<2> ubCURR_U_TEMP(3,0)<16;16,1>
+ mov (16) ubCURR_UV(23,0)<2> ubCURR_U_TEMP(3,8)<16;16,1>
+
+ mov (16) ubCURR_UV(4,0)<2> ubCURR_U_TEMP(4,0)<16;16,1>
+ mov (16) ubCURR_UV(24,0)<2> ubCURR_U_TEMP(4,8)<16;16,1>
+
+ mov (16) ubCURR_UV(5,0)<2> ubCURR_U_TEMP(5,0)<16;16,1>
+ mov (16) ubCURR_UV(25,0)<2> ubCURR_U_TEMP(5,8)<16;16,1>
+
+ mov (16) ubCURR_UV(6,0)<2> ubCURR_U_TEMP(6,0)<16;16,1>
+ mov (16) ubCURR_UV(26,0)<2> ubCURR_U_TEMP(6,8)<16;16,1>
+
+ mov (16) ubCURR_UV(7,0)<2> ubCURR_U_TEMP(7,0)<16;16,1>
+ mov (16) ubCURR_UV(27,0)<2> ubCURR_U_TEMP(7,8)<16;16,1>
+
+ mov (16) ubCURR_UV(8,0)<2> ubCURR_U_TEMP(8,0)<16;16,1>
+ mov (16) ubCURR_UV(28,0)<2> ubCURR_U_TEMP(8,8)<16;16,1>
+
+ mov (16) ubCURR_UV(9,0)<2> ubCURR_U_TEMP(9,0)<16;16,1>
+ mov (16) ubCURR_UV(29,0)<2> ubCURR_U_TEMP(9,8)<16;16,1>
+
+ mov (16) ubCURR_UV(10,0)<2> ubCURR_U_TEMP(10,0)<16;16,1>
+ mov (16) ubCURR_UV(30,0)<2> ubCURR_U_TEMP(10,8)<16;16,1>
+
+ mov (16) ubCURR_UV(11,0)<2> ubCURR_U_TEMP(11,0)<16;16,1>
+ mov (16) ubCURR_UV(31,0)<2> ubCURR_U_TEMP(11,8)<16;16,1>
+
+ mov (16) ubCURR_UV(12,0)<2> ubCURR_U_TEMP(12,0)<16;16,1>
+ mov (16) ubCURR_UV(32,0)<2> ubCURR_U_TEMP(12,8)<16;16,1>
+
+ mov (16) ubCURR_UV(13,0)<2> ubCURR_U_TEMP(13,0)<16;16,1>
+ mov (16) ubCURR_UV(33,0)<2> ubCURR_U_TEMP(13,8)<16;16,1>
+
+ mov (16) ubCURR_UV(14,0)<2> ubCURR_U_TEMP(14,0)<16;16,1>
+ mov (16) ubCURR_UV(34,0)<2> ubCURR_U_TEMP(14,8)<16;16,1>
+
+ mov (16) ubCURR_UV(15,0)<2> ubCURR_U_TEMP(15,0)<16;16,1>
+ mov (16) ubCURR_UV(35,0)<2> ubCURR_U_TEMP(15,8)<16;16,1>
+
+ mov (16) ubCURR_UV(16,0)<2> ubCURR_U_TEMP(16,0)<16;16,1>
+ mov (16) ubCURR_UV(36,0)<2> ubCURR_U_TEMP(16,8)<16;16,1>
+
+ mov (16) ubCURR_UV(17,0)<2> ubCURR_U_TEMP(17,0)<16;16,1>
+ mov (16) ubCURR_UV(37,0)<2> ubCURR_U_TEMP(17,8)<16;16,1>
+
+ mov (16) ubCURR_UV(18,0)<2> ubCURR_U_TEMP(18,0)<16;16,1>
+ mov (16) ubCURR_UV(38,0)<2> ubCURR_U_TEMP(18,8)<16;16,1>
+
+ mov (16) ubCURR_UV(19,0)<2> ubCURR_U_TEMP(19,0)<16;16,1>
+ mov (16) ubCURR_UV(39,0)<2> ubCURR_U_TEMP(19,8)<16;16,1>
+
+
+//Move V data
+ mov (16) ubCURR_UV(0,1)<2> ubCURR_V_TEMP(0,0)<16;16,1>
+ mov (16) ubCURR_UV(20,1)<2> ubCURR_V_TEMP(0,8)<16;16,1>
+ mov (16) ubCURR_UV(1,1)<2> ubCURR_V_TEMP(1,0)<16;16,1>
+ mov (16) ubCURR_UV(21,1)<2> ubCURR_V_TEMP(1,8)<16;16,1>
+ mov (16) ubCURR_UV(2,1)<2> ubCURR_V_TEMP(2,0)<16;16,1>
+ mov (16) ubCURR_UV(22,1)<2> ubCURR_V_TEMP(2,8)<16;16,1>
+ mov (16) ubCURR_UV(3,1)<2> ubCURR_V_TEMP(3,0)<16;16,1>
+ mov (16) ubCURR_UV(23,1)<2> ubCURR_V_TEMP(3,8)<16;16,1>
+ mov (16) ubCURR_UV(4,1)<2> ubCURR_V_TEMP(4,0)<16;16,1>
+ mov (16) ubCURR_UV(24,1)<2> ubCURR_V_TEMP(4,8)<16;16,1>
+ mov (16) ubCURR_UV(5,1)<2> ubCURR_V_TEMP(5,0)<16;16,1>
+ mov (16) ubCURR_UV(25,1)<2> ubCURR_V_TEMP(5,8)<16;16,1>
+ mov (16) ubCURR_UV(6,1)<2> ubCURR_V_TEMP(6,0)<16;16,1>
+ mov (16) ubCURR_UV(26,1)<2> ubCURR_V_TEMP(6,8)<16;16,1>
+ mov (16) ubCURR_UV(7,1)<2> ubCURR_V_TEMP(7,0)<16;16,1>
+ mov (16) ubCURR_UV(27,1)<2> ubCURR_V_TEMP(7,8)<16;16,1>
+ mov (16) ubCURR_UV(8,1)<2> ubCURR_V_TEMP(8,0)<16;16,1>
+ mov (16) ubCURR_UV(28,1)<2> ubCURR_V_TEMP(8,8)<16;16,1>
+ mov (16) ubCURR_UV(9,1)<2> ubCURR_V_TEMP(9,0)<16;16,1>
+ mov (16) ubCURR_UV(29,1)<2> ubCURR_V_TEMP(9,8)<16;16,1>
+ mov (16) ubCURR_UV(10,1)<2> ubCURR_V_TEMP(10,0)<16;16,1>
+ mov (16) ubCURR_UV(30,1)<2> ubCURR_V_TEMP(10,8)<16;16,1>
+ mov (16) ubCURR_UV(11,1)<2> ubCURR_V_TEMP(11,0)<16;16,1>
+ mov (16) ubCURR_UV(31,1)<2> ubCURR_V_TEMP(11,8)<16;16,1>
+ mov (16) ubCURR_UV(12,1)<2> ubCURR_V_TEMP(12,0)<16;16,1>
+ mov (16) ubCURR_UV(32,1)<2> ubCURR_V_TEMP(12,8)<16;16,1>
+ mov (16) ubCURR_UV(13,1)<2> ubCURR_V_TEMP(13,0)<16;16,1>
+ mov (16) ubCURR_UV(33,1)<2> ubCURR_V_TEMP(13,8)<16;16,1>
+ mov (16) ubCURR_UV(14,1)<2> ubCURR_V_TEMP(14,0)<16;16,1>
+ mov (16) ubCURR_UV(34,1)<2> ubCURR_V_TEMP(14,8)<16;16,1>
+ mov (16) ubCURR_UV(15,1)<2> ubCURR_V_TEMP(15,0)<16;16,1>
+ mov (16) ubCURR_UV(35,1)<2> ubCURR_V_TEMP(15,8)<16;16,1>
+ mov (16) ubCURR_UV(16,1)<2> ubCURR_V_TEMP(16,0)<16;16,1>
+ mov (16) ubCURR_UV(36,1)<2> ubCURR_V_TEMP(16,8)<16;16,1>
+ mov (16) ubCURR_UV(17,1)<2> ubCURR_V_TEMP(17,0)<16;16,1>
+ mov (16) ubCURR_UV(37,1)<2> ubCURR_V_TEMP(17,8)<16;16,1>
+ mov (16) ubCURR_UV(18,1)<2> ubCURR_V_TEMP(18,0)<16;16,1>
+ mov (16) ubCURR_UV(38,1)<2> ubCURR_V_TEMP(18,8)<16;16,1>
+ mov (16) ubCURR_UV(19,1)<2> ubCURR_V_TEMP(19,0)<16;16,1>
+ mov (16) ubCURR_UV(39,1)<2> ubCURR_V_TEMP(19,8)<16;16,1>
+
+
+
+//File Name : DN_UV_PL3_Interleave_Prev_Frame_UV.asm
+//Author : Tatiya, Rupesh
+//Description : Interleave separately loaded U and V for PL3 format.
+// This is needed because Noise Detection and Noise Reduction works on interleaved UV data.
+
+//1.U Data: bottom half of UV space for prev frame (r17-r24)
+//2.V Data: bottom quarter of Y space for curr frame (r120-r127)
+
+ mov (16) ubPREV_UV(0,0)<2> ubPREV_U_TEMP(0,0)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(8,0)<2> ubPREV_U_TEMP(0,8)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(1,0)<2> ubPREV_U_TEMP(1,0)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(9,0)<2> ubPREV_U_TEMP(1,8)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(2,0)<2> ubPREV_U_TEMP(2,0)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(10,0)<2> ubPREV_U_TEMP(2,8)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(3,0)<2> ubPREV_U_TEMP(3,0)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(11,0)<2> ubPREV_U_TEMP(3,8)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(4,0)<2> ubPREV_U_TEMP(4,0)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(12,0)<2> ubPREV_U_TEMP(4,8)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(5,0)<2> ubPREV_U_TEMP(5,0)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(13,0)<2> ubPREV_U_TEMP(5,8)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(6,0)<2> ubPREV_U_TEMP(6,0)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(14,0)<2> ubPREV_U_TEMP(6,8)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(7,0)<2> ubPREV_U_TEMP(7,0)<16;8,1> { NoDDClr }
+ mov (16) ubPREV_UV(15,0)<2> ubPREV_U_TEMP(7,8)<16;8,1> { NoDDClr }
+
+ mov (16) ubPREV_UV(0,1)<2> ubPREV_V_TEMP(0,0)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(8,1)<2> ubPREV_V_TEMP(0,8)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(1,1)<2> ubPREV_V_TEMP(1,0)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(9,1)<2> ubPREV_V_TEMP(1,8)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(2,1)<2> ubPREV_V_TEMP(2,0)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(10,1)<2> ubPREV_V_TEMP(2,8)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(3,1)<2> ubPREV_V_TEMP(3,0)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(11,1)<2> ubPREV_V_TEMP(3,8)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(4,1)<2> ubPREV_V_TEMP(4,0)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(12,1)<2> ubPREV_V_TEMP(4,8)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(5,1)<2> ubPREV_V_TEMP(5,0)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(13,1)<2> ubPREV_V_TEMP(5,8)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(6,1)<2> ubPREV_V_TEMP(6,0)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(14,1)<2> ubPREV_V_TEMP(6,8)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(7,1)<2> ubPREV_V_TEMP(7,0)<16;8,1> { NoDDChk }
+ mov (16) ubPREV_UV(15,1)<2> ubPREV_V_TEMP(7,8)<16;8,1> { NoDDChk }
+
+
+
+//Module Name : DN_UV_420_Load_Curr_Frame_Y
+//Author : Tatiya, Rupesh
+//Description : Load Curr Frame Y data for 420 Input
+
+
+
+//Module Name : DN_UV_Load_Curr_Frame_Y
+//Author : Tatiya, Rupesh
+//Description : Loads Y of Current frame.
+
+
+
+
+ //For 16x16 U and 16x16 V for 420, we need to read 32x32 Y.
+
+ mov (8) acc0.0<1>:ud r0.0<8;8,1>:ud
+ mov (1) acc0.2<1>:ud 0xF000F:ud
+ add (2) acc0.0<1>:ud r7.0<2;2,1>:w r4.4<2;2,1>:w
+
+ mov (8) r92.0<1>:ud acc0.0<8;8,1>:ud
+
+ mov (8) r101.0<1>:ud acc0.0<8;8,1>:ud
+ mov (8) r110.0<1>:ud acc0.0<8;8,1>:ud
+ mov (8) r119.0<1>:ud acc0.0<8;8,1>:ud
+
+ add (1) r101.1<1>:d acc0.1<0;1,0>:d 16:d
+
+ add (1) r110.0<1>:d acc0.0<0;1,0>:d 16:d
+
+ add (2) r119.0<1>:d acc0.0<2;2,1>:d 16:d
+
+ send (8) udCURR_Y0(0)<1> r92 0x4 0x2890003:ud
+ send (8) udCURR_Y1(0)<1> r101 0x4 0x2890003:ud
+ send (8) udCURR_Y2(0)<1> r110 0x4 0x2890003:ud
+ send (8) udCURR_Y3(0)<1> r119 0x4 0x2890003:ud
+
+
+
+//Module Name : DN_UV_Noise_Detection_UV
+//Author : Tatiya, Rupesh
+//Description : Performs noise detection on 16x16 U and 16x16 V each.
+
+
+
+//Module Name : DN_UV_Move_CURBE_Inline_UV.asm
+//Author : Tatiya, Rupesh
+
+
+
+
+ //Mov CURBE data to another space - so that it can be used as Temp Space --> r1 - r6
+ mov (4) r54.28<1>:ub r2.28<4;4,1>:ub //Dest. YUY2 offset
+ mov (2) r54.5<1>:ud r4.0<4;2,2>:ud //Src YUY2 offset and Origin offset
+ mov (4) r55.28<1>:ub r1.0<4;4,1>:ub
+
+ mov (8) r61.20<1>:ub r1.4<8;8,1>:ub
+ mov (4) r61.28<1>:ub r1.12<4;4,1>:ub
+
+ //Move Inline Data to another space - so that it can be used as Temp Space --> r7
+ mov (4) r62.10<1>:w r7.0<4;4,1>:w
+ mov (4) r63.10<1>:w r7.4<4;4,1>:w
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ mov (1) a0.0:uw 1664:uw
+ mov (1) a0.1:uw 1816:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 1792:uw
+ mov (1) a0.1:uw 1820:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 1920:uw
+ mov (1) a0.1:uw 1848:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2048:uw
+ mov (1) a0.1:uw 1852:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ mov (1) a0.0:uw 2304:uw
+ mov (1) a0.1:uw 1880:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2432:uw
+ mov (1) a0.1:uw 1884:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2560:uw
+ mov (1) a0.1:uw 1912:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+
+
+//Module Name : DN_UV_Noise_Detection_Set_Top_Region_N
+//Author : Tatiya, Rupesh
+//Description : Sets sub-region region N from Top region.
+
+
+ //TODO - remove one instruction here using arithmatic. -rT
+ mov (1) a0.0:uw 2688:uw
+ mov (1) a0.1:uw 1916:uw
+
+
+
+
+add (1) r7.7<1>:d ip:ud 32:d { NoCompact }
+ jmpi (1) DN_UV_NOISE_DETECTION_UV { NoCompact }
+
+
+
+//Module : DN_UV_Noise_Reduction_UV
+//Author : Tatiya, Rupesh
+//Description : Performs Noise Reduction on 16x16 U and 16x16 V.
+//Tasks : 1. Update weight history
+// 2. Find if it block is motion block
+// 3. Compute Denoised Pixel.
+
+
+
+
+//History is 1+1 byte every 4x4 U and 4x4 V.
+
+ cmp.l.f0.0 (16) null<1>:w ubHIST_UV(0,0)<16;16,1> r61.20<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w ubHIST_UV(0,0)<16;16,1> r61.22<0;2,1>:ub
+
+ mov (16) uwCURBE_TEMP(0)<1> 0:w
+ mov (16) uwCURBE_TEMP(1)<1> 0:w
+
+ //Compute diff betn curr and prev. - First 16 lines
+ // 8 lines here
+ add (16) wDIFF_TEMPORAL(0)<1> ubCURR_UV(2,2)<16;16,1> -ubPREV_UV(0,0)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(1)<1> ubCURR_UV(3,2)<16;16,1> -ubPREV_UV(0,16)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(2)<1> ubCURR_UV(4,2)<16;16,1> -ubPREV_UV(0,32)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(3)<1> ubCURR_UV(5,2)<16;16,1> -ubPREV_UV(0,48)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(4)<1> ubCURR_UV(6,2)<16;16,1> -ubPREV_UV(0,64)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(5)<1> ubCURR_UV(7,2)<16;16,1> -ubPREV_UV(0,80)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(6)<1> ubCURR_UV(8,2)<16;16,1> -ubPREV_UV(0,96)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(7)<1> ubCURR_UV(9,2)<16;16,1> -ubPREV_UV(0,112)<16;16,1> //Diff UV interleaved
+
+ //Update WT HIST
+ (-f0.0) shr (16) uwCURBE_TEMP(0)<1> ubHIST_UV(0,0)<16;16,1> 1:w
+ (f1.0) add (16) uwCURBE_TEMP(2)<1> ubHIST_UV(0,0)<16;16,1> r61.24<0;2,1>:ub
+ (f0.0) mov (16) uwCURBE_TEMP(2)<1> r61.20<0;2,1>:ub
+ (-f0.0.anyv) mov (16) uwCURBE_TEMP(2)<1> ubHIST_UV(0,0)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w ubHIST_UV(0,16)<16;16,1> r61.20<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w ubHIST_UV(0,16)<16;16,1> r61.22<0;2,1>:ub
+
+ //Compute diff betn curr and prev. - First 16 lines
+ // 8 more lines here
+ add (16) wDIFF_TEMPORAL(8)<1> ubCURR_UV(10,2)<16;16,1> -ubPREV_UV(0,128)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(9)<1> ubCURR_UV(11,2)<16;16,1> -ubPREV_UV(0,144)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(10)<1> ubCURR_UV(12,2)<16;16,1> -ubPREV_UV(0,160)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(11)<1> ubCURR_UV(13,2)<16;16,1> -ubPREV_UV(0,176)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(12)<1> ubCURR_UV(14,2)<16;16,1> -ubPREV_UV(0,192)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(13)<1> ubCURR_UV(15,2)<16;16,1> -ubPREV_UV(0,208)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(14)<1> ubCURR_UV(16,2)<16;16,1> -ubPREV_UV(0,224)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(15)<1> ubCURR_UV(17,2)<16;16,1> -ubPREV_UV(0,240)<16;16,1> //Diff UV interleaved
+
+ (-f0.0) shr (16) uwCURBE_TEMP(1)<1> ubHIST_UV(0,16)<16;16,1> 1:w
+ (f1.0) add (16) uwCURBE_TEMP(3)<1> ubHIST_UV(0,16)<16;16,1> r61.24<0;2,1>:ub
+ (f0.0) mov (16) uwCURBE_TEMP(3)<1> r61.20<0;2,1>:ub
+ (-f0.0.anyv) mov(16) uwCURBE_TEMP(3)<1> ubHIST_UV(0,16)<16;16,1>
+
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(0)<16;16,1> (abs)wDIFF_TEMPORAL(1)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(2)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(3)<16;16,1>
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(4)<16;16,1> (abs)wDIFF_TEMPORAL(5)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(6)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(7)<16;16,1>
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(8)<16;16,1> (abs)wDIFF_TEMPORAL(9)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(10)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(2)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(11)<16;16,1>
+ //16x16 to 16x4 - First 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(12)<16;16,1> (abs)wDIFF_TEMPORAL(13)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(14)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(3)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(15)<16;16,1>
+
+//Compute diff betn curr and prev. - Second 16 lines
+//13 lines.
+ add (16) wDIFF_TEMPORAL(16)<1> ubCURR_UV(22,2)<16;16,1> -ubPREV_UV(8,0)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(17)<1> ubCURR_UV(23,2)<16;16,1> -ubPREV_UV(8,16)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(18)<1> ubCURR_UV(24,2)<16;16,1> -ubPREV_UV(8,32)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(19)<1> ubCURR_UV(25,2)<16;16,1> -ubPREV_UV(8,48)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(20)<1> ubCURR_UV(26,2)<16;16,1> -ubPREV_UV(8,64)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(21)<1> ubCURR_UV(27,2)<16;16,1> -ubPREV_UV(8,80)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(22)<1> ubCURR_UV(28,2)<16;16,1> -ubPREV_UV(8,96)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(23)<1> ubCURR_UV(29,2)<16;16,1> -ubPREV_UV(8,112)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(24)<1> ubCURR_UV(30,2)<16;16,1> -ubPREV_UV(8,128)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(25)<1> ubCURR_UV(31,2)<16;16,1> -ubPREV_UV(8,144)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(26)<1> ubCURR_UV(32,2)<16;16,1> -ubPREV_UV(8,160)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(27)<1> ubCURR_UV(33,2)<16;16,1> -ubPREV_UV(8,176)<16;16,1> //Diff UV interleaved
+ add (16) wDIFF_TEMPORAL(28)<1> ubCURR_UV(34,2)<16;16,1> -ubPREV_UV(8,192)<16;16,1> //Diff UV interleaved
+
+//3 more lines
+ add (16) wCURBE_TEMP(4)<1> ubCURR_UV(35,2)<16;16,1> -ubPREV_UV(8,208)<16;16,1> //Diff UV interleaved
+ add (16) wCURBE_TEMP(5)<1> ubCURR_UV(36,2)<16;16,1> -ubPREV_UV(8,224)<16;16,1> //Diff UV interleaved
+ add (16) wCURBE_TEMP(6)<1> ubCURR_UV(37,2)<16;16,1> -ubPREV_UV(8,240)<16;16,1> //Diff UV interleaved
+
+ //16x4 to 8x4 - First 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> uwDIFF_TEMPORAL_SUM4x4(2,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(2,2)<4;2,1>
+
+ //8x4 to 4x4 - First 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4_FINAL(0)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1> { AccWrEn }
+
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(16)<16;16,1> (abs)wDIFF_TEMPORAL(17)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(18)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(19)<16;16,1>
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(20)<16;16,1> (abs)wDIFF_TEMPORAL(21)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(22)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(23)<16;16,1>
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(24)<16;16,1> (abs)wDIFF_TEMPORAL(25)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(26)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(2)<1> acc0.0<16;16,1>:uw (abs)wDIFF_TEMPORAL(27)<16;16,1>
+
+ //16x16 to 16x4 - Second 16 lines
+ add (16) acc0.0<1>:uw (abs)wDIFF_TEMPORAL(28)<16;16,1> (abs)wCURBE_TEMP(4)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wCURBE_TEMP(5)<16;16,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(3)<1> acc0.0<16;16,1>:uw (abs)wCURBE_TEMP(6)<16;16,1>
+
+ //Find if block is motion block - First 16 lines
+ cmp.g.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_FINAL(0)<16;16,1> r61.26<0;2,1>:ub
+
+ //Move TEMPORAL_SUM4x4 for SIMD16 use later.
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,0)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,2)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,4)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,6)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,8)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,10)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,12)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(0,14)<0;2,1>
+
+ //Pick Appropriate Weight History Based on motion. - First 16 lines
+ (-f0.0) mov (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(2)<16;16,1>
+
+ //Actual DN - First 16 lines
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(0)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(0)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(2,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(2,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(2,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,0)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,8)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(0)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(0)<1> ubCURR_UV(2,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(0)<1> wDIFF_TEMPORAL(0)<16;16,1> ubCURR_UV(2,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(1)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(1)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(3,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(3,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(3,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,16)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,24)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(1)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(1)<1> ubCURR_UV(3,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(1)<1> wDIFF_TEMPORAL(1)<16;16,1> ubCURR_UV(3,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(2)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(2)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(4,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(4,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(4,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,32)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,40)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(2)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(2)<1> ubCURR_UV(4,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(2)<1> wDIFF_TEMPORAL(2)<16;16,1> ubCURR_UV(4,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(3)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(3)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(5,2)<8;8,1> -uwCURBE_TEMP(0,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(5,10)<8;8,1> -uwCURBE_TEMP(0,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(5,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,48)<8;8,1> uwCURBE_TEMP(0,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,56)<8;8,1> uwCURBE_TEMP(0,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(3)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(3)<1> ubCURR_UV(5,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(3)<1> wDIFF_TEMPORAL(3)<16;16,1> ubCURR_UV(5,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(4)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(4)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(6,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(6,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(6,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,64)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,72)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(4)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(4)<1> ubCURR_UV(6,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(4)<1> wDIFF_TEMPORAL(4)<16;16,1> ubCURR_UV(6,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(5)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(5)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(7,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(7,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(7,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,80)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,88)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(5)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(5)<1> ubCURR_UV(7,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(5)<1> wDIFF_TEMPORAL(5)<16;16,1> ubCURR_UV(7,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(6)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(6)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(8,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(8,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(8,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,96)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,104)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(6)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(6)<1> ubCURR_UV(8,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(6)<1> wDIFF_TEMPORAL(6)<16;16,1> ubCURR_UV(8,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(7)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(7)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(9,2)<8;8,1> -uwCURBE_TEMP(0,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(9,10)<8;8,1> -uwCURBE_TEMP(0,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(9,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,112)<8;8,1> uwCURBE_TEMP(0,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,120)<8;8,1> uwCURBE_TEMP(0,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(7)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(7)<1> ubCURR_UV(9,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(7)<1> wDIFF_TEMPORAL(7)<16;16,1> ubCURR_UV(9,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(8)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(8)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(10,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(10,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(10,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,128)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,136)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(8)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(8)<1> ubCURR_UV(10,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(8)<1> wDIFF_TEMPORAL(8)<16;16,1> ubCURR_UV(10,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(9)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(9)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(11,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(11,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(11,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,144)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,152)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(9)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(9)<1> ubCURR_UV(11,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(9)<1> wDIFF_TEMPORAL(9)<16;16,1> ubCURR_UV(11,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(10)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(10)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(12,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(12,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(12,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,160)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,168)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(10)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(10)<1> ubCURR_UV(12,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(10)<1> wDIFF_TEMPORAL(10)<16;16,1> ubCURR_UV(12,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(11)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(11)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(13,2)<8;8,1> -uwCURBE_TEMP(0,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(13,10)<8;8,1> -uwCURBE_TEMP(0,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(13,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,176)<8;8,1> uwCURBE_TEMP(0,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,184)<8;8,1> uwCURBE_TEMP(0,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(11)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(11)<1> ubCURR_UV(13,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(11)<1> wDIFF_TEMPORAL(11)<16;16,1> ubCURR_UV(13,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(12)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(12)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(14,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(14,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(14,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,192)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,200)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(12)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(12)<1> ubCURR_UV(14,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(12)<1> wDIFF_TEMPORAL(12)<16;16,1> ubCURR_UV(14,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(13)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(13)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(15,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(15,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(15,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,208)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,216)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(13)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(13)<1> ubCURR_UV(15,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(13)<1> wDIFF_TEMPORAL(13)<16;16,1> ubCURR_UV(15,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(14)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(14)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(16,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(16,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(16,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,224)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,232)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(14)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(14)<1> ubCURR_UV(16,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(14)<1> wDIFF_TEMPORAL(14)<16;16,1> ubCURR_UV(16,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(15)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(15)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(17,2)<8;8,1> -uwCURBE_TEMP(0,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(17,10)<8;8,1> -uwCURBE_TEMP(0,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(17,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(0,240)<8;8,1> uwCURBE_TEMP(0,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(0,248)<8;8,1> uwCURBE_TEMP(0,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(15)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(15)<1> ubCURR_UV(17,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(15)<1> wDIFF_TEMPORAL(15)<16;16,1> ubCURR_UV(17,2)<16;16,1>
+
+
+ //16x4 to 8x4 - Second 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4(0)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1>
+ add (16) uwDIFF_TEMPORAL_SUM4x4(1)<1> uwDIFF_TEMPORAL_SUM4x4(2,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(2,2)<4;2,1>
+
+ //8x4 to 4x4 - Second 16 lines
+ add (16) uwDIFF_TEMPORAL_SUM4x4_FINAL(1)<1> uwDIFF_TEMPORAL_SUM4x4(0,0)<4;2,1> uwDIFF_TEMPORAL_SUM4x4(0,2)<4;2,1> { AccWrEn }
+
+ //Find if block is motion block - Second 16 lines
+ cmp.g.f1.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_FINAL(1)<16;16,1> r61.26<0;2,1>:ub
+
+ //Move TEMPORAL_SUM4x4 for SIMD16 use later.
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,0)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,2)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,4)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,6)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,8)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,10)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,0)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,12)<0;2,1>
+ mov (8) uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0,8)<1> uwDIFF_TEMPORAL_SUM4x4_FINAL(1,14)<0;2,1>
+
+ //Pick Appropriate Weight History Based on motion. - Second 16 lines
+ (-f1.0) mov (16) uwCURBE_TEMP(1)<1> uwCURBE_TEMP(3)<16;16,1>
+
+ //Actual DN - Second 16 lines
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(16)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(16)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(22,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(22,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(22,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,0)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,8)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(16)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(16)<1> ubCURR_UV(22,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(16)<1> wDIFF_TEMPORAL(16)<16;16,1> ubCURR_UV(22,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(17)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(17)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(23,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(23,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(23,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,16)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,24)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(17)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(17)<1> ubCURR_UV(23,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(17)<1> wDIFF_TEMPORAL(17)<16;16,1> ubCURR_UV(23,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(18)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(18)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(24,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(24,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(24,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,32)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,40)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(18)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(18)<1> ubCURR_UV(24,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(18)<1> wDIFF_TEMPORAL(18)<16;16,1> ubCURR_UV(24,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(19)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(19)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(25,2)<8;8,1> -uwCURBE_TEMP(1,0)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(25,10)<8;8,1> -uwCURBE_TEMP(1,2)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(25,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,48)<8;8,1> uwCURBE_TEMP(1,0)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,56)<8;8,1> uwCURBE_TEMP(1,2)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(19)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(19)<1> ubCURR_UV(25,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_0(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(19)<1> wDIFF_TEMPORAL(19)<16;16,1> ubCURR_UV(25,2)<16;16,1>
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(20)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(20)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(26,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(26,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(26,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,64)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,72)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(20)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(20)<1> ubCURR_UV(26,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(20)<1> wDIFF_TEMPORAL(20)<16;16,1> ubCURR_UV(26,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(21)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(21)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(27,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(27,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(27,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,80)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,88)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(21)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(21)<1> ubCURR_UV(27,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(21)<1> wDIFF_TEMPORAL(21)<16;16,1> ubCURR_UV(27,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(22)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(22)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(28,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(28,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(28,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,96)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,104)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(22)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(22)<1> ubCURR_UV(28,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(22)<1> wDIFF_TEMPORAL(22)<16;16,1> ubCURR_UV(28,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(23)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(23)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(29,2)<8;8,1> -uwCURBE_TEMP(1,4)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(29,10)<8;8,1> -uwCURBE_TEMP(1,6)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(29,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,112)<8;8,1> uwCURBE_TEMP(1,4)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,120)<8;8,1> uwCURBE_TEMP(1,6)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(23)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(23)<1> ubCURR_UV(29,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_1(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(23)<1> wDIFF_TEMPORAL(23)<16;16,1> ubCURR_UV(29,2)<16;16,1>
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(24)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(24)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(30,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(30,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(30,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,128)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,136)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(24)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(24)<1> ubCURR_UV(30,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(24)<1> wDIFF_TEMPORAL(24)<16;16,1> ubCURR_UV(30,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(25)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(25)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(31,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(31,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(31,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,144)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,152)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(25)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(25)<1> ubCURR_UV(31,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(25)<1> wDIFF_TEMPORAL(25)<16;16,1> ubCURR_UV(31,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(26)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(26)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(32,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(32,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(32,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,160)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,168)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(26)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(26)<1> ubCURR_UV(32,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(26)<1> wDIFF_TEMPORAL(26)<16;16,1> ubCURR_UV(32,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(27)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(27)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(33,2)<8;8,1> -uwCURBE_TEMP(1,8)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(33,10)<8;8,1> -uwCURBE_TEMP(1,10)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(33,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,176)<8;8,1> uwCURBE_TEMP(1,8)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,184)<8;8,1> uwCURBE_TEMP(1,10)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(27)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(27)<1> ubCURR_UV(33,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_2(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(27)<1> wDIFF_TEMPORAL(27)<16;16,1> ubCURR_UV(33,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(28)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wDIFF_TEMPORAL(28)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(34,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(34,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(34,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,192)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,200)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wDIFF_TEMPORAL(28)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wDIFF_TEMPORAL(28)<1> ubCURR_UV(34,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wDIFF_TEMPORAL(28)<1> wDIFF_TEMPORAL(28)<16;16,1> ubCURR_UV(34,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wCURBE_TEMP(4)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wCURBE_TEMP(4)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(35,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(35,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(35,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,208)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,216)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wCURBE_TEMP(4)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wCURBE_TEMP(4)<1> ubCURR_UV(35,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wCURBE_TEMP(4)<1> wCURBE_TEMP(4)<16;16,1> ubCURR_UV(35,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wCURBE_TEMP(5)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wCURBE_TEMP(5)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(36,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(36,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(36,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,224)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,232)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wCURBE_TEMP(5)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wCURBE_TEMP(5)<1> ubCURR_UV(36,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wCURBE_TEMP(5)<1> wCURBE_TEMP(5)<16;16,1> ubCURR_UV(36,2)<16;16,1>
+
+ cmp.l.f0.0 (16) null<1>:w (abs)wCURBE_TEMP(6)<16;16,1> r61.28<0;2,1>:ub
+ cmp.l.f1.0 (16) null<1>:w (abs)wCURBE_TEMP(6)<16;16,1> r61.30<0;2,1>:ub
+ mul (8) acc0.0<1>:w ubCURR_UV(37,2)<8;8,1> -uwCURBE_TEMP(1,12)<0;2,1>
+ mul (8) acc0.8<1>:w ubCURR_UV(37,10)<8;8,1> -uwCURBE_TEMP(1,14)<0;2,1>
+ mac (16) acc0<1>:w ubCURR_UV(37,2)<16;16,1> 256:w
+ mac (8) acc0.0<1>:w ubPREV_UV(8,240)<8;8,1> uwCURBE_TEMP(1,12)<0;2,1>
+ mac (8) acc0.8<1>:w ubPREV_UV(8,248)<8;8,1> uwCURBE_TEMP(1,14)<0;2,1>
+ add (16) acc0<1>:w acc0<16;16,1>:w 128:w
+ (f0.0) shr (16) wCURBE_TEMP(6)<1> acc0<16;16,1>:w 8:w
+ (-f0.0) mov (16) wCURBE_TEMP(6)<1> ubCURR_UV(37,2)<16;16,1>
+ cmp.le.f0.0 (16) null<1>:w uwDIFF_TEMPORAL_SUM4x4_TEMP_3(0)<16;16,1> r61.26<0;2,1>:ub
+ (-f0.0.allv) avg (16) wCURBE_TEMP(6)<1> wCURBE_TEMP(6)<16;16,1> ubCURR_UV(37,2)<16;16,1>
+
+ //Pack Weight History WORD -> BYTE
+ mov (16) ubCURBE_TEMP(3,0)<1> ubCURBE_TEMP(0)<32;16,2>
+ mov (16) ubCURBE_TEMP(3,16)<1> ubCURBE_TEMP(1)<32;16,2>
+
+
+
+//Module Name : DN_UV_Compute_BNE_UV
+//Author : Tatiya, Rupesh
+//Description : Computes minimum SOAD for each 16x4 block.
+
+ cmp.l.f0.0 (8) null:w uwSOAD_MIN_8x4(0,12)<16;4,1> uwSOAD_MIN_8x4(2,12)<16;4,1>
+ (f0.0)sel (8) uwCURBE_TEMP(1,0)<1> uwSOAD_MIN_8x4(0,12)<16;4,1> uwSOAD_MIN_8x4(2,12)<16;4,1>
+
+ mov (8) ubCURBE_TEMP(1)<1> ubCURBE_TEMP(1)<16;8,2>
+
+
+
+//File Name : DN_UV_PL3_Unpack_Denoised_UV.asm
+//Author : Tatiya, Rupesh
+//Description : Upack the interleaved UV data
+
+//First 16 lines.
+ mov (8) ubMSGPAYLOAD_U(0,0)<1> ubDIFF_TEMPORAL(0,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(0,16)<1> ubDIFF_TEMPORAL(1,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(0,0)<1> ubDIFF_TEMPORAL(0,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(0,16)<1> ubDIFF_TEMPORAL(1,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(1,0)<1> ubDIFF_TEMPORAL(2,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(1,16)<1> ubDIFF_TEMPORAL(3,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(1,0)<1> ubDIFF_TEMPORAL(2,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(1,16)<1> ubDIFF_TEMPORAL(3,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(2,0)<1> ubDIFF_TEMPORAL(4,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(2,16)<1> ubDIFF_TEMPORAL(5,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(2,0)<1> ubDIFF_TEMPORAL(4,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(2,16)<1> ubDIFF_TEMPORAL(5,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(3,0)<1> ubDIFF_TEMPORAL(6,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(3,16)<1> ubDIFF_TEMPORAL(7,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(3,0)<1> ubDIFF_TEMPORAL(6,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(3,16)<1> ubDIFF_TEMPORAL(7,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(4,0)<1> ubDIFF_TEMPORAL(8,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(4,16)<1> ubDIFF_TEMPORAL(9,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(4,0)<1> ubDIFF_TEMPORAL(8,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(4,16)<1> ubDIFF_TEMPORAL(9,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(5,0)<1> ubDIFF_TEMPORAL(10,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(5,16)<1> ubDIFF_TEMPORAL(11,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(5,0)<1> ubDIFF_TEMPORAL(10,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(5,16)<1> ubDIFF_TEMPORAL(11,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(6,0)<1> ubDIFF_TEMPORAL(12,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(6,16)<1> ubDIFF_TEMPORAL(13,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(6,0)<1> ubDIFF_TEMPORAL(12,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(6,16)<1> ubDIFF_TEMPORAL(13,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(7,0)<1> ubDIFF_TEMPORAL(14,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(7,16)<1> ubDIFF_TEMPORAL(15,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(7,0)<1> ubDIFF_TEMPORAL(14,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(7,16)<1> ubDIFF_TEMPORAL(15,2)<32;8,4>
+
+//Second 16 lines.
+//12 lines first
+ mov (8) ubMSGPAYLOAD_U(0,8)<1> ubDIFF_TEMPORAL(16,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(0,24)<1> ubDIFF_TEMPORAL(17,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(0,8)<1> ubDIFF_TEMPORAL(16,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(0,24)<1> ubDIFF_TEMPORAL(17,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(1,8)<1> ubDIFF_TEMPORAL(18,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(1,24)<1> ubDIFF_TEMPORAL(19,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(1,8)<1> ubDIFF_TEMPORAL(18,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(1,24)<1> ubDIFF_TEMPORAL(19,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(2,8)<1> ubDIFF_TEMPORAL(20,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(2,24)<1> ubDIFF_TEMPORAL(21,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(2,8)<1> ubDIFF_TEMPORAL(20,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(2,24)<1> ubDIFF_TEMPORAL(21,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(3,8)<1> ubDIFF_TEMPORAL(22,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(3,24)<1> ubDIFF_TEMPORAL(23,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(3,8)<1> ubDIFF_TEMPORAL(22,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(3,24)<1> ubDIFF_TEMPORAL(23,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(4,8)<1> ubDIFF_TEMPORAL(24,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(4,24)<1> ubDIFF_TEMPORAL(25,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(4,8)<1> ubDIFF_TEMPORAL(24,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(4,24)<1> ubDIFF_TEMPORAL(25,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(5,8)<1> ubDIFF_TEMPORAL(26,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(5,24)<1> ubDIFF_TEMPORAL(27,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(5,8)<1> ubDIFF_TEMPORAL(26,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(5,24)<1> ubDIFF_TEMPORAL(27,2)<32;8,4>
+
+ //3 lines next
+ mov (8) ubMSGPAYLOAD_U(6,8)<1> ubDIFF_TEMPORAL(28,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(6,24)<1> ubCURBE_TEMP(4,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(6,8)<1> ubDIFF_TEMPORAL(28,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(6,24)<1> ubCURBE_TEMP(4,2)<32;8,4>
+
+ mov (8) ubMSGPAYLOAD_U(7,8)<1> ubCURBE_TEMP(5,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_U(7,24)<1> ubCURBE_TEMP(6,0)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(7,8)<1> ubCURBE_TEMP(5,2)<32;8,4>
+ mov (8) ubMSGPAYLOAD_V(7,24)<1> ubCURBE_TEMP(6,2)<32;8,4>
+
+
+
+//Module Name : DN_UV_420_Save_Curr_Frame_Y
+//Author : Tatiya, Rupesh
+//Description : Save Curr Frame Y data for 420 Input
+
+
+
+//Module Name : DN_UV_Load_Curr_Frame_Y
+//Author : Tatiya, Rupesh
+//Description : Saves Y or YUY2 of Current frame.
+
+
+
+
+ mov (8) acc0.0<1>:ud r0.0<8;8,1>:ud
+ mov (2) acc0.0<1>:d r62.10<2;2,1>:w
+
+ mov (1) acc0.2<1>:d 0xF000F:ud
+
+ mov (8) r92.0<1>:ud acc0.0<8;8,1>:ud
+
+ mov (8) r101.0<1>:ud acc0.0<8;8,1>:ud
+ mov (8) r110.0<1>:ud acc0.0<8;8,1>:ud
+ mov (8) r119.0<1>:ud acc0.0<8;8,1>:ud
+
+ add (1) r101.1<1>:d acc0.1<0;1,0>:d 16:d
+
+ add (1) r110.0<1>:d acc0.0<0;1,0>:d 16:d
+
+ add (2) r119.0<1>:d acc0.0<2;2,1>:d 16:d
+
+ send (8) null<1>:d r92 0x5 0x120A8018:ud
+ send (8) null<1>:d r101 0x5 0x120A8018:ud
+ send (8) null<1>:d r110 0x5 0x120A8018:ud
+ send (8) null<1>:d r119 0x5 0x120A8018:ud
+
+
+ //TODO - See if History saving can be combined with Curr Frame Save. - rT
+
+
+//Module Name : DN_UV_Save_Hist_UV
+//Author : Tatiya, Rupesh
+//Description : Saves DN history for UV data.
+
+ mov (8) r3.0<1>:ud r0.0<8;8,1>:ud
+ mov (2) r3.0<1>:d r62.12<2;2,1>:w
+ mov (1) r3.2<1>:d 0x30007:ud
+
+ send (8) null<1>:d r3 0x5 0x40A8021:ud
+
+
+
+//Module Name : DN_UV_Save_BNE_UV
+//Author : Tatiya, Rupesh
+//Description : Saves BNE values for 16x16 U and 16x16 V.
+
+ mov (8) r1.0<1>:ud r0.0<8;8,1>:ud
+ mov (2) r1.0<1>:d r63.12<2;2,1>:w
+ mov (1) r1.2<1>:d 0x10003:ud
+
+ send (8) null<1>:d r1 0x5 0x40A8023:ud
+
+
+
+//File Name : DN_UV_PL3_Save_Curr_Frame_UV.asm
+//Author : Tatiya, Rupesh
+//Description : Save U and V data for PL3 surface
+
+
+
+//Module name : DN_UV_Save_Curr_Frame_UV
+//Author : Tatiya, Rupesh
+//Description : Saves Current Frame (UV only).
+
+
+
+
+ mov (8) r74<1>:ud r0.0<8;8,1>:ud
+ mov (8) r83<1>:ud r0.0<8;8,1>:ud
+
+ shr (2) r74.0<1>:d r62.10<2;2,1>:w 1:w
+ mov (1) r74.2<1>:d 0xF000F:ud
+
+ mov (8) r83.0<1>:ud r74.0<8;8,1>:ud
+
+ send (8) null<1>:d r74 0x5 0x120A8019:ud
+ send (8) null<1>:d r83 0x5 0x120A801A:ud
+
+
+
+//End of Thread message
+
+mov (8) r127<1>:ud r0.0<8;8,1>:ud
+ send (1) null<1>:d r127 0x27 0x02000010
+
+
+ //All sub-routines here
+
+
+// Module Name : Noise_Detection
+// Author : Tatiya, Rupesh
+// Description : Performs noise detection on 32 pixels of U (8x4) and 32 pixels of V (8x4).
+
+DN_UV_NOISE_DETECTION_UV:
+
+// Find Field Block Median
+//
+// Purpose : Find the median value of the nine pixels in the same field
+// which are centered at current pixel.
+//
+// Works on 9 pixels centered at the current pixel
+// NOTE: pixels are within same field.
+// v4 - current pixel
+//
+// v2 v1 v0
+// * * * <--- Different field - not used
+// v5 v4 v3
+// * * * <--- Different field - not used
+// v8 v7 v6
+
+// Algorithm to find median modifies the data.
+// Copy the data needed to calculate median so the original source data stays intact.
+//
+
+//TODO - Change Interleaved implementation to separated one if - , does not work on predication. - rT
+
+//Delete Later - rT
+//mov (1) pCUR_UV:uw 52*32:uw
+
+// v0
+mov (16) ubMEDIAN_TEMP(0,0)<1> r[a0.0,0]<16;16,1>
+// v0
+mov (16) ubMEDIAN_TEMP(0,16)<1> r[a0.0,32]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(1,0)<1> r[a0.0,2]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(1,16)<1> r[a0.0,34]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(2,0)<1> r[a0.0,4]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(2,16)<1> r[a0.0,36]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(3,0)<1> r[a0.0,64]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(3,16)<1> r[a0.0,96]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(4,0)<1> r[a0.0,66]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(4,16)<1> r[a0.0,98]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(5,0)<1> r[a0.0,68]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(5,16)<1> r[a0.0,100]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(6,0)<1> r[a0.0,128]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(6,16)<1> r[a0.0,160]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(7,0)<1> r[a0.0,130]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(7,16)<1> r[a0.0,162]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(8,0)<1> r[a0.0,132]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(8,16)<1> r[a0.0,164]<16;16,1>
+
+//TODO - Optimize one instruction here.
+add (1) a0.0:uw a0.0<0;1,0>:uw 64:uw
+// v0
+mov (16) ubMEDIAN_TEMP(9,0)<1> r[a0.0,0]<16;16,1>
+// v0
+mov (16) ubMEDIAN_TEMP(9,16)<1> r[a0.0,32]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(10,0)<1> r[a0.0,2]<16;16,1>
+// v1
+mov (16) ubMEDIAN_TEMP(10,16)<1> r[a0.0,34]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(11,0)<1> r[a0.0,4]<16;16,1>
+// v2
+mov (16) ubMEDIAN_TEMP(11,16)<1> r[a0.0,36]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(12,0)<1> r[a0.0,64]<16;16,1>
+// v3
+mov (16) ubMEDIAN_TEMP(12,16)<1> r[a0.0,96]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(13,0)<1> r[a0.0,66]<16;16,1>
+// v4
+mov (16) ubMEDIAN_TEMP(13,16)<1> r[a0.0,98]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(14,0)<1> r[a0.0,68]<16;16,1>
+// v5
+mov (16) ubMEDIAN_TEMP(14,16)<1> r[a0.0,100]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(15,0)<1> r[a0.0,128]<16;16,1>
+// v6
+mov (16) ubMEDIAN_TEMP(15,16)<1> r[a0.0,160]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(16,0)<1> r[a0.0,130]<16;16,1>
+// v7
+mov (16) ubMEDIAN_TEMP(16,16)<1> r[a0.0,162]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(17,0)<1> r[a0.0,132]<16;16,1>
+// v8
+mov (16) ubMEDIAN_TEMP(17,16)<1> r[a0.0,164]<16;16,1>
+
+//TODO - Optimize one instruction here.
+add (1) a0.0:uw a0.0<0;1,0>:uw 64:uw
+
+// MedianSwap
+//
+// MedianSwap(inOutLeft, inOutRight)
+// {
+// if (inOutLeft > inOutRight)
+// {
+// temp = inOutLeft
+// inOutLeft = inOutRight
+// inOutRight = temp
+// }
+// }
+
+// MedianSwap(v1, v2) - U
+// MedianSwap(v4, v5) - U
+// MedianSwap(v1, v2) - V
+// MedianSwap(v4, v5) - V
+
+cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(1,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(5,0)<32;16,2>
+cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(1,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(1,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(1,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(1,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+(f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(5,0)<32;16,2>
+(f1.0) mov (16) ubMEDIAN_TEMP(1,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+(f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubTEMP1(0,0)<16;16,1>
+(f0.1) mov (16) ubMEDIAN_TEMP(5,0)<2> ubTEMP1(0,16)<16;16,1>
+(f1.0) mov (16) ubMEDIAN_TEMP(2,1)<2> ubTEMP1(1,0)<16;16,1>
+(f1.1) mov (16) ubMEDIAN_TEMP(5,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// MedianSwap(v7, v8) - U
+// MedianSwap(v0, v1) - U
+// MedianSwap(v7, v8) - V
+// MedianSwap(v0, v1) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(7,0)<32;16,2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(0,0)<32;16,2> ubMEDIAN_TEMP(1,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(7,1)<32;16,2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(0,1)<32;16,2> ubMEDIAN_TEMP(1,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(7,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(0,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(7,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(0,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(7,0)<2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(0,0)<2> ubMEDIAN_TEMP(1,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(7,1)<2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(0,1)<2> ubMEDIAN_TEMP(1,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(8,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(1,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(8,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(1,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v3, v4) - U
+ // MedianSwap(v6, v7) - U
+ // MedianSwap(v3, v4) - V
+ // MedianSwap(v6, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(3,0)<32;16,2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(6,0)<32;16,2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(3,1)<32;16,2> ubMEDIAN_TEMP(4,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(6,1)<32;16,2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(3,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(6,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(3,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(6,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(3,0)<2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(6,0)<2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(3,1)<2> ubMEDIAN_TEMP(4,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(6,1)<2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(7,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(4,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(7,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v1, v2) - U
+ // MedianSwap(v4, v5) - U
+ // MedianSwap(v1, v2) - V
+ // MedianSwap(v4, v5) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(1,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(1,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(1,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(1,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(1,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(1,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(5,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(5,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(2,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(5,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v7, v8) - U
+ // MedianSwap(v0, v3) - U
+ // MedianSwap(v7, v8) - V
+ // MedianSwap(v0, v3) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(7,0)<32;16,2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(0,0)<32;16,2> ubMEDIAN_TEMP(3,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(7,1)<32;16,2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(0,1)<32;16,2> ubMEDIAN_TEMP(3,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(7,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(0,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(7,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(0,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(7,0)<2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(0,0)<2> ubMEDIAN_TEMP(3,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(7,1)<2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(0,1)<2> ubMEDIAN_TEMP(3,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(8,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(3,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(8,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(3,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // NOTE:
+ // Compare v0 to v6 to find the minimum.
+ // Store the minimum for future use.
+ //TODO - Find if MIN is needed.
+ //cmp.l.f0.0 (16) null:w ubMEDIAN_TEMP(%1+0,0)<32;16,2> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //cmp.l.f1.0 (16) null:w ubMEDIAN_TEMP(%1+0,1)<32;16,2> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+ //(f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,0)<32;16,2>
+ //(f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,1)<32;16,2>
+ //(-f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //(-f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+
+ // MedianSwap(v5, v8) - U
+ // MedianSwap(v4, v7) - U
+ // MedianSwap(v5, v8) - V
+ // MedianSwap(v4, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(5,0)<32;16,2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(5,1)<32;16,2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(5,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(5,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(5,0)<2> ubMEDIAN_TEMP(8,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(5,1)<2> ubMEDIAN_TEMP(8,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(8,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(7,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(8,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(7,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// NOTE:
+// Compare v2 to v8 to find the maximum.
+// Store the maximum for future use.
+ //TODO - Find if MAX is needed.
+// cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(%1+2,0)<32;16,2> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+// cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(%1+2,1)<32;16,2> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+//(f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,0)<32;16,2>
+//(f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,1)<32;16,2>
+//(-f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+//(-f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+
+// MedianSwap(v3, v6) - U
+// MedianSwap(v1, v4) - U
+// MedianSwap(v3, v6) - V
+// MedianSwap(v1, v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(3,0)<32;16,2> ubMEDIAN_TEMP(6,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(1,0)<32;16,2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(3,1)<32;16,2> ubMEDIAN_TEMP(6,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(1,1)<32;16,2> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(6,0)<2> ubMEDIAN_TEMP(3,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(1,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(6,1)<2> ubMEDIAN_TEMP(3,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(1,1)<32;16,2>
+
+ // MedianSwap(v2,v5) - U
+ // MedianSwap(v4,v7) - U
+ // MedianSwap(v2,v5) - V
+ // MedianSwap(v4,v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(2,0)<32;16,2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(2,1)<32;16,2> ubMEDIAN_TEMP(5,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubMEDIAN_TEMP(5,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(7,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(2,1)<2> ubMEDIAN_TEMP(5,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(7,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(4,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(2,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(2,1)<2> ubTEMP1(0,16)<16;16,1>
+
+ // MedianSwap(v6,v4) - U
+ // MedianSwap(v6,v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(6,0)<32;16,2> ubMEDIAN_TEMP(4,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(6,1)<32;16,2> ubMEDIAN_TEMP(4,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(6,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(6,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(4,0)<32;16,2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(4,1)<32;16,2> ubMEDIAN_TEMP(2,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(4,0)<2> ubMEDIAN_TEMP(2,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(4,1)<2> ubMEDIAN_TEMP(2,1)<32;16,2>
+cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(10,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(14,0)<32;16,2>
+cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(10,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(10,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(10,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(10,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+(f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(14,0)<32;16,2>
+(f1.0) mov (16) ubMEDIAN_TEMP(10,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+(f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+(f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubTEMP1(0,0)<16;16,1>
+(f0.1) mov (16) ubMEDIAN_TEMP(14,0)<2> ubTEMP1(0,16)<16;16,1>
+(f1.0) mov (16) ubMEDIAN_TEMP(11,1)<2> ubTEMP1(1,0)<16;16,1>
+(f1.1) mov (16) ubMEDIAN_TEMP(14,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// MedianSwap(v7, v8) - U
+// MedianSwap(v0, v1) - U
+// MedianSwap(v7, v8) - V
+// MedianSwap(v0, v1) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(16,0)<32;16,2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(9,0)<32;16,2> ubMEDIAN_TEMP(10,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(16,1)<32;16,2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(9,1)<32;16,2> ubMEDIAN_TEMP(10,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(16,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(9,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(16,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(9,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(16,0)<2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(9,0)<2> ubMEDIAN_TEMP(10,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(16,1)<2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(9,1)<2> ubMEDIAN_TEMP(10,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(17,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(10,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(17,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(10,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v3, v4) - U
+ // MedianSwap(v6, v7) - U
+ // MedianSwap(v3, v4) - V
+ // MedianSwap(v6, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(12,0)<32;16,2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(15,0)<32;16,2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(12,1)<32;16,2> ubMEDIAN_TEMP(13,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(15,1)<32;16,2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(12,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(15,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(12,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(15,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(12,0)<2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(15,0)<2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(12,1)<2> ubMEDIAN_TEMP(13,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(15,1)<2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(16,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(13,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(16,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v1, v2) - U
+ // MedianSwap(v4, v5) - U
+ // MedianSwap(v1, v2) - V
+ // MedianSwap(v4, v5) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(10,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(10,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(10,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(10,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(10,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(10,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(14,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(14,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(11,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(14,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // MedianSwap(v7, v8) - U
+ // MedianSwap(v0, v3) - U
+ // MedianSwap(v7, v8) - V
+ // MedianSwap(v0, v3) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(16,0)<32;16,2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(9,0)<32;16,2> ubMEDIAN_TEMP(12,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(16,1)<32;16,2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(9,1)<32;16,2> ubMEDIAN_TEMP(12,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(16,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(9,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(16,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(9,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(16,0)<2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(9,0)<2> ubMEDIAN_TEMP(12,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(16,1)<2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(9,1)<2> ubMEDIAN_TEMP(12,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(17,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(12,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(17,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(12,1)<2> ubTEMP1(1,16)<16;16,1>
+
+ // NOTE:
+ // Compare v0 to v6 to find the minimum.
+ // Store the minimum for future use.
+ //TODO - Find if MIN is needed.
+ //cmp.l.f0.0 (16) null:w ubMEDIAN_TEMP(%1+0,0)<32;16,2> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //cmp.l.f1.0 (16) null:w ubMEDIAN_TEMP(%1+0,1)<32;16,2> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+ //(f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,0)<32;16,2>
+ //(f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+0,1)<32;16,2>
+ //(-f0.0) mov (16) ubCURR_MIN(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,0)<32;16,2>
+ //(-f1.0) mov (16) ubCURR_MIN(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+6,1)<32;16,2>
+
+ // MedianSwap(v5, v8) - U
+ // MedianSwap(v4, v7) - U
+ // MedianSwap(v5, v8) - V
+ // MedianSwap(v4, v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(14,0)<32;16,2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(14,1)<32;16,2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(14,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(1,0)<1> ubMEDIAN_TEMP(14,1)<32;16,2>
+ mov (16) ubTEMP1(1,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(14,0)<2> ubMEDIAN_TEMP(17,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(14,1)<2> ubMEDIAN_TEMP(17,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(17,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(16,0)<2> ubTEMP1(0,16)<16;16,1>
+ (f1.0) mov (16) ubMEDIAN_TEMP(17,1)<2> ubTEMP1(1,0)<16;16,1>
+ (f1.1) mov (16) ubMEDIAN_TEMP(16,1)<2> ubTEMP1(1,16)<16;16,1>
+
+// NOTE:
+// Compare v2 to v8 to find the maximum.
+// Store the maximum for future use.
+ //TODO - Find if MAX is needed.
+// cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(%1+2,0)<32;16,2> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+// cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(%1+2,1)<32;16,2> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+//(f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,0)<32;16,2>
+//(f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+2,1)<32;16,2>
+//(-f0.0) mov (16) ubCURR_MAX(0,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,0)<32;16,2>
+//(-f1.0) mov (16) ubCURR_MAX(1,%2*16+0)<1> ubMEDIAN_TEMP(%1+8,1)<32;16,2>
+
+// MedianSwap(v3, v6) - U
+// MedianSwap(v1, v4) - U
+// MedianSwap(v3, v6) - V
+// MedianSwap(v1, v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(12,0)<32;16,2> ubMEDIAN_TEMP(15,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(10,0)<32;16,2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(12,1)<32;16,2> ubMEDIAN_TEMP(15,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(10,1)<32;16,2> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(15,0)<2> ubMEDIAN_TEMP(12,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(10,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(15,1)<2> ubMEDIAN_TEMP(12,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(10,1)<32;16,2>
+
+ // MedianSwap(v2,v5) - U
+ // MedianSwap(v4,v7) - U
+ // MedianSwap(v2,v5) - V
+ // MedianSwap(v4,v7) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(11,0)<32;16,2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ cmp.g.f1.0 (16) null:w ubMEDIAN_TEMP(11,1)<32;16,2> ubMEDIAN_TEMP(14,1)<32;16,2>
+ cmp.g.f1.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubMEDIAN_TEMP(14,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(16,0)<32;16,2>
+ (f1.0) mov (16) ubMEDIAN_TEMP(11,1)<2> ubMEDIAN_TEMP(14,1)<32;16,2>
+ (f1.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(16,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+ mov (16) ubTEMP1(0,0)<1> ubMEDIAN_TEMP(13,0)<32;16,2>
+ mov (16) ubTEMP1(0,16)<1> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(11,0)<2> ubTEMP1(0,0)<16;16,1>
+ (f0.1) mov (16) ubMEDIAN_TEMP(11,1)<2> ubTEMP1(0,16)<16;16,1>
+
+ // MedianSwap(v6,v4) - U
+ // MedianSwap(v6,v4) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(15,0)<32;16,2> ubMEDIAN_TEMP(13,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(15,1)<32;16,2> ubMEDIAN_TEMP(13,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(15,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(15,1)<32;16,2>
+
+ // MedianSwap(v4,v2) - U
+ // MedianSwap(v4,v2) - V
+ cmp.g.f0.0 (16) null:w ubMEDIAN_TEMP(13,0)<32;16,2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ cmp.g.f0.1 (16) null:w ubMEDIAN_TEMP(13,1)<32;16,2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+ (f0.0) mov (16) ubMEDIAN_TEMP(13,0)<2> ubMEDIAN_TEMP(11,0)<32;16,2>
+ (f0.1) mov (16) ubMEDIAN_TEMP(13,1)<2> ubMEDIAN_TEMP(11,1)<32;16,2>
+
+// Sobel Value calculation for the current pixel v4
+// v2 v1 v0
+// * * * <--- Different field - not used
+// v5 v4 v3
+// * * * <--- Different field - not used
+// v8 v7 v6
+//
+// Gx = -v0 - 2*v3 - v6 + v2 + 2*v5 + v8
+// Gy = v0 + 2*v1 + v2 - v6 - 2*v7 - v8
+//
+// Sobel = (|Gx| + |Gy|) >> 3
+
+//TODO - Change Later - rT
+add (1) a0.0:uw a0.0<0;1,0>:uw -128:uw
+
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,64]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,132]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,0]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,128]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,4]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(0)<1> r[a0.0,68]<16;16,1>:ub 2:w
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,96]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,164]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,32]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,160]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,36]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(1)<1> r[a0.0,100]<16;16,1>:ub 2:w
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,128]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,196]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,64]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,192]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,68]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(2)<1> r[a0.0,132]<16;16,1>:ub 2:w
+// - 2 * v3
+mul (16) acc0.0<1>:w r[a0.0,160]<16;16,1>:ub -2:w
+// + v8
+mac (16) acc0.0<1>:w r[a0.0,228]<16;16,1>:ub 1:w
+// - v0
+mac (16) acc0.0<1>:w r[a0.0,96]<16;16,1>:ub -1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,224]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,100]<16;16,1>:ub 1:w
+// + 2 * v5
+mac (16) wSOBEL_X(3)<1> r[a0.0,164]<16;16,1>:ub 2:w
+
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,2]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,0]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,132]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,4]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,128]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,130]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(0)<16;16,1>
+
+shr (16) uwSOBEL(0)<1> acc0.0<16;16,1>:uw 3:uw
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,34]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,32]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,164]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,36]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,160]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,162]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(1)<16;16,1>
+
+shr (16) uwSOBEL(1)<1> acc0.0<16;16,1>:uw 3:uw
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,66]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,64]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,196]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,68]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,192]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,194]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(2)<16;16,1>
+
+shr (16) uwSOBEL(2)<1> acc0.0<16;16,1>:uw 3:uw
+// + 2 * v1
+mul (16) acc0.0<1>:w r[a0.0,98]<16;16,1>:ub 2:w
+// + v0
+mac (16) acc0.0<1>:w r[a0.0,96]<16;16,1>:ub 1:w
+// - v8
+mac (16) acc0.0<1>:w r[a0.0,228]<16;16,1>:ub -1:w
+// + v2
+mac (16) acc0.0<1>:w r[a0.0,100]<16;16,1>:ub 1:w
+// - v6
+mac (16) acc0.0<1>:w r[a0.0,224]<16;16,1>:ub -1:w
+// - 2 * v7
+mac (16) acc0.0<1>:w r[a0.0,226]<16;16,1>:ub -2:w
+
+add (16) acc0.0<1>:uw (abs)acc0.0<16;16,1>:w (abs)wSOBEL_X(3)<16;16,1>
+
+shr (16) uwSOBEL(3)<1> acc0.0<16;16,1>:uw 3:uw
+
+//Mov Median in CURBE_TEMP to free up temp space.
+mov (16) ubMEDIAN(0,0)<1> ubMEDIAN_TEMP(4,0)<16;16,1>
+mov (16) ubMEDIAN(0,16)<1> ubMEDIAN_TEMP(4,16)<16;16,1>
+mov (16) ubMEDIAN(0,32)<1> ubMEDIAN_TEMP(13,0)<16;16,1>
+mov (16) ubMEDIAN(0,48)<1> ubMEDIAN_TEMP(13,16)<16;16,1>
+
+// Find:
+// absDiff = abs(ubCurY - ubMedian)
+// Find the difference between pixel and median value.
+
+//Median is interleaved. So difference is also interleaved.
+
+//------------------------------------------------------------------------------------------
+//Process 16 U and 16 V pixels here and rest later.
+// first row - v0,v1,v2
+add (16) wDIFF(0)<1> r[a0.0,0]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(1)<1> r[a0.0,2]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(2)<1> r[a0.0,4]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+
+// second row - v3,v4,v5
+add (16) wDIFF(3)<1> r[a0.0,64]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(4)<1> r[a0.0,66]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(5)<1> r[a0.0,68]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+
+// third row - v6,v7,v8
+add (16) wDIFF(6)<1> r[a0.0,128]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(7)<1> r[a0.0,130]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+add (16) wDIFF(8)<1> r[a0.0,132]<16;16,1>:ub -ubMEDIAN(0,0)<16;16,1>
+// first row - v0,v1,v2
+add (16) wDIFF(9)<1> r[a0.0,32]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(10)<1> r[a0.0,34]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(11)<1> r[a0.0,36]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+
+// second row - v3,v4,v5
+add (16) wDIFF(12)<1> r[a0.0,96]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(13)<1> r[a0.0,98]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(14)<1> r[a0.0,100]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+
+// third row - v6,v7,v8
+add (16) wDIFF(15)<1> r[a0.0,160]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(16)<1> r[a0.0,162]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+add (16) wDIFF(17)<1> r[a0.0,164]<16;16,1>:ub -ubMEDIAN(0,16)<16;16,1>
+
+//TODO - Change Later - rT
+add (1) a0.0:uw a0.0<0;1,0>:uw 64:uw
+
+// Find sum of all absolute differences AND
+// maximum absolute difference for 16 U and 16 V here.
+//First 2 rows of 8x4
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(2)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(3)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(4)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(5)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(6)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(7)<16;16,1>
+ add (16) uwSOAD(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF(8)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//------------
+ //DIFF(0-7) is not needed here. Populate it.
+ // first row - v0,v1,v2
+ add (16) wDIFF(0)<1> r[a0.0,0]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(1)<1> r[a0.0,2]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(2)<1> r[a0.0,4]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+
+ // second row - v3,v4,v5
+ add (16) wDIFF(3)<1> r[a0.0,64]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(4)<1> r[a0.0,66]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(5)<1> r[a0.0,68]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+
+ // third row - v6,v7
+ add (16) wDIFF(6)<1> r[a0.0,128]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+ add (16) wDIFF(7)<1> r[a0.0,130]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+//------------
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(0)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+
+//------------
+ //Load v8 - DIFF(8)
+ add (16) wDIFF(8)<1> r[a0.0,132]<16;16,1>:ub -ubMEDIAN(1,0)<16;16,1>
+//------------
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(11)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(12)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(13)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(14)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(15)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(16)<16;16,1>
+ add (16) uwSOAD(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF(17)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//------------
+ //DIFF(0-7) is not needed here. Populate it.
+ // first row - v0,v1,v2
+ add (16) wDIFF(9)<1> r[a0.0,32]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(10)<1> r[a0.0,34]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(11)<1> r[a0.0,36]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+
+ // second row - v3,v4,v5
+ add (16) wDIFF(12)<1> r[a0.0,96]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(13)<1> r[a0.0,98]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(14)<1> r[a0.0,100]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+
+ // third row - v6,v7
+ add (16) wDIFF(15)<1> r[a0.0,160]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+ add (16) wDIFF(16)<1> r[a0.0,162]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+//------------
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(1)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+
+//------------
+ //Load v8 - DIFF(8)
+ add (16) wDIFF(17)<1> r[a0.0,164]<16;16,1>:ub -ubMEDIAN(1,16)<16;16,1>
+//------------
+
+//if ((sobel_edge_measure < m_SobelEdgeThreshold) && ((block_max-block_min) < m_LocalDiffThreshold))
+// if (sigma_mb_min > sigma)
+// sigma_mb_min = sigma;
+
+//NOTE: block_min is always zero as median is one of the value in 3x3 block. So no need o calculate it.
+// So just do -
+//if ((sobel_edge_measure < m_SobelEdgeThreshold) && ((block_max) < m_LocalDiffThreshold) && ( sigma < sigma_mb_min))
+// sigma_mb_min = sigma;
+
+//We are processing 32 bytes of U and 32 bytes of V - each of size 8x4.
+//Compare first 8 bytes with max possible (255).
+//Start above condition from second 8 bytes.
+
+//TODO - Change Later - rT
+// mov (1) pCUR_MIN_SOAD_8x4:uw 1752:uw //r54.24:ub
+
+//First row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(0)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(0)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(0)<16;16,1> 255:uw
+(f0.0) sel (16) uwSOBEL(0)<1> uwSOAD(0)<16;16,1> 255:uw
+
+//Second row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(1)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(1)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(1)<16;16,1> uwSOBEL(0)<16;16,1>
+(f0.0) mov (16) uwSOBEL(0)<1> uwSOAD(1)<16;16,1>
+
+// Find sum of all absolute differences AND
+// maximum absolute difference for 16 U and 16 V here.
+//Second 2 rows of 8x4
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(2)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(3)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(4)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(5)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(6)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(7)<16;16,1>
+ add (16) uwSOAD(0)<1> acc0.0<16;16,1>:uw (abs)wDIFF(8)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(0)<16;16,1> (abs)wDIFF(1)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(2)<16;16,1> (abs)wDIFF(3)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(4)<16;16,1> (abs)wDIFF(5)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(6)<16;16,1> (abs)wDIFF(7)<16;16,1>
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(0)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(8)<16;16,1>
+//Compare 0-1, 2-3, 4-5, 6-7
+cmp.g.f0.0 (16) null:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+cmp.g.f0.1 (16) null:uw (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+cmp.g.f1.0 (16) null:uw (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+cmp.g.f1.1 (16) null:uw (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//Calculate SAD
+ add (16) acc0.0<1>:uw (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(11)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(12)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(13)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(14)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(15)<16;16,1>
+ add (16) acc0.0<1>:uw acc0.0<16;16,1>:uw (abs)wDIFF(16)<16;16,1>
+ add (16) uwSOAD(1)<1> acc0.0<16;16,1>:uw (abs)wDIFF(17)<16;16,1>
+
+(f0.0) sel (16) uwCURBE_TEMP(0)<1> (abs)wDIFF(9)<16;16,1> (abs)wDIFF(10)<16;16,1>
+(f0.1) sel (16) uwCURBE_TEMP(1)<1> (abs)wDIFF(11)<16;16,1> (abs)wDIFF(12)<16;16,1>
+(f1.0) sel (16) uwCURBE_TEMP(2)<1> (abs)wDIFF(13)<16;16,1> (abs)wDIFF(14)<16;16,1>
+(f1.1) sel (16) uwCURBE_TEMP(3)<1> (abs)wDIFF(15)<16;16,1> (abs)wDIFF(16)<16;16,1>
+
+//Compare Max(0,1) - Max(2,3), Max(4,5) - Max(6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+cmp.g.f0.1 (16) null:uw uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(1)<16;16,1>
+(f0.1)sel (16) uwCURBE_TEMP(2)<1> uwCURBE_TEMP(2)<16;16,1> uwCURBE_TEMP(3)<16;16,1>
+
+//Compare Max(0,1,2,3) - Max(4,5,6,7)
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+(f0.0)sel (16) uwCURBE_TEMP(0)<1> uwCURBE_TEMP(0)<16;16,1> uwCURBE_TEMP(2)<16;16,1>
+
+//Compare Max(0,1,2,3,4,5,6,7) - 8
+cmp.g.f0.0 (16) null:uw uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+(f0.0)sel (16) uwMAX_ABS_DIFF(1)<1> uwCURBE_TEMP(0)<16;16,1> (abs)wDIFF(17)<16;16,1>
+
+//Third row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(2)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(0)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(0)<16;16,1> uwSOBEL(0)<16;16,1>
+(f0.0) mov (16) uwSOBEL(0)<1> uwSOAD(0)<16;16,1>
+
+//Fourth row of 8x4
+ cmp.l.f0.0 (16) null:uw uwSOBEL(3)<16;16,1> r55.30<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwMAX_ABS_DIFF(1)<16;16,1> r55.28<0;2,1>:ub
+(f0.0) cmp.l.f0.0 (16) null:uw uwSOAD(1)<16;16,1> uwSOBEL(0)<16;16,1>
+(f0.0) mov (16) uwSOBEL(0)<1> uwSOAD(1)<16;16,1>
+
+ cmp.l.f0.0 (8) null:uw uwSOBEL(0,0)<8;8,1> uwSOBEL(0,8)<8;8,1>
+(f0.0) sel (8) uwSOBEL(0)<1> uwSOBEL(0,0)<8;8,1> uwSOBEL(0,8)<8;8,1>
+
+ cmp.l.f0.0 (4) null:uw uwSOBEL(0,0)<4;4,1> uwSOBEL(0,4)<4;4,1>
+(f0.0) sel (4) uwSOBEL(0)<1> uwSOBEL(0,0)<4;4,1> uwSOBEL(0,4)<4;4,1>
+
+ cmp.l.f0.0 (2) null:uw uwSOBEL(0,0)<2;2,1> uwSOBEL(0,2)<2;2,1>
+(f0.0) sel (2) r[a0.1,0]<1>:uw uwSOBEL(0,0)<2;2,1> uwSOBEL(0,2)<2;2,1>
+
+
+
+
+
+
+// End of common.inc
+
+mov (1) ip:ud r7.7<0;1,0>:d
+
+
+.end_code
+.end_kernel
--- /dev/null
+/*
+ * 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.
+ */
+// 117 // Total instruction count
+// 1 // Total kernel count
+
+.kernel PL3_DN_422CP
+.code
+
+
+
+// FileName: DN_PL_Core.asm
+// Author: Vivek Kumar
+// Description: Tasks for DN only case (16x8 block) for planar 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 0x45E8003: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_Load_UV_IMC3_16x8.asm
+// Author: Vivek Kumar
+// Description: Read UV for 16x8 block through DATAPORT
+
+
+
+// FileName: UVCopy_Load_16x8.asm
+// Author: Vivek Kumar
+// Description: Read UV for 16x8 block through DATAPORT
+
+//CHANGE : Read extra UV data to convert to 422. -rT
+//we are reading extra data in ALL cases irrespective of whether upsampling is reqd or not later on, to keep things simple.
+
+
+ add (2) r27.0<1>:d r7.0<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 0x40007:ud { NoDDChk } // U/V block width and height (8x5)
+ 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 0x2290001:ud
+ send (8) udDNDI_UV_RESP(2)<1> r38 0x4 0x2290002:ud
+
+ //Update Header for Save
+ mov (1) mudMSGHDR_UCOPY(0,2)<1> 0x30007:ud // U block width and height (8x4)
+ mov (1) mudMSGHDR_VCOPY(0,2)<1> 0x30007:ud // V block width and height (8x4)
+
+
+
+// FileName: DN_Save_Y_16x8.asm
+// Author: Vivek Kumar
+// Description: Save one 16x8 blocks of Y channel of DN output for reference
+
+
+mov (8) mudDN_Y_OUT(0,0)<1> r0<8;8,1>:ud // message header
+mov (2) mudDN_Y_OUT(0,0)<1> r7.0<2;2,1>:w { NoDDClr } // X origin
+mov (1) mudDN_Y_OUT(0,2)<1> 0x7000F:ud { NoDDChk } // block width and height (16x8)
+
+//send out data through data port
+send (8) null<1>:d mudDN_Y_OUT 0x5 0xA0A8018:ud
+
+
+
+// FileName: DN_Save_UV_IMC3_16x8.asm
+// Author: Vivek Kumar
+// Description: Save UV for 16x8 block through DATAPORT
+
+
+
+// FileName: UVCopy_Save_16x8.asm
+// Author: Vivek Kumar
+// Description: Save UV for 16x8 block through DATAPORT
+
+
+//Reuse the header from Load component
+//Header is modified at the end of load - to be usable for save.
+
+
+ mov (8) mudMSGHDR_UCOPY(1)<1> udDNDI_UV_RESP(0)<8;8,1>
+ mov (8) mudMSGHDR_VCOPY(1)<1> udDNDI_UV_RESP(2)<8;8,1>
+ send (4) null<1>:d r36 0x5 0x40A8019:ud
+ send (4) null<1>:d r38 0x5 0x40A801A:ud
+
+
+
+// FileName: DN_Upsample_UV_IMC3_16x8.asm
+// Author: Tatiya, Rupesh
+// Description: Upconvert 420 UV to 422
+
+
+
+// FileName: UVCopy_Upsample_UV_16x8.asm
+// Author: Tatiya, Rupesh
+// Description: Convert 42X UV to 422 - to be used for IECP.
+
+
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(0) ubDNDI_UV_RESP(0,0)<0;8,1> ubDNDI_UV_RESP(0,0)<8;8,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(4) ubDNDI_UV_RESP(2,0)<0;8,1> ubDNDI_UV_RESP(2,0)<8;8,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(1) ubDNDI_UV_RESP(0,8)<0;8,1> ubDNDI_UV_RESP(0,8)<8;8,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(5) ubDNDI_UV_RESP(2,8)<0;8,1> ubDNDI_UV_RESP(2,8)<8;8,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(2) ubDNDI_UV_RESP(0,16)<0;8,1> ubDNDI_UV_RESP(0,16)<8;8,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(6) ubDNDI_UV_RESP(2,16)<0;8,1> ubDNDI_UV_RESP(2,16)<8;8,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(3) ubDNDI_UV_RESP(0,24)<0;8,1> ubDNDI_UV_RESP(0,24)<8;8,1>
+ avg.sat (16) uwDNDI_UVCOPY_TEMP(7) ubDNDI_UV_RESP(2,24)<0;8,1> ubDNDI_UV_RESP(2,24)<8;8,1>
+
+ mov (16) ubDNDI_RESP(5, 1)<2> ubDNDI_UVCOPY_TEMP(0,0)<32;16,2> { NoDDClr }
+ mov (16) ubDNDI_RESP(5, 0)<2> ubDNDI_UVCOPY_TEMP(4,0)<32;16,2> { NoDDChk }
+ mov (16) ubDNDI_RESP(5, 33)<2> ubDNDI_UVCOPY_TEMP(1,0)<32;16,2> { NoDDClr }
+ mov (16) ubDNDI_RESP(5, 32)<2> ubDNDI_UVCOPY_TEMP(5,0)<32;16,2> { NoDDChk }
+ mov (16) ubDNDI_RESP(5, 65)<2> ubDNDI_UVCOPY_TEMP(2,0)<32;16,2> { NoDDClr }
+ mov (16) ubDNDI_RESP(5, 64)<2> ubDNDI_UVCOPY_TEMP(6,0)<32;16,2> { NoDDChk }
+ mov (16) ubDNDI_RESP(5, 97)<2> ubDNDI_UVCOPY_TEMP(3,0)<32;16,2> { NoDDClr }
+ mov (16) ubDNDI_RESP(5, 96)<2> ubDNDI_UVCOPY_TEMP(7,0)<32;16,2> { NoDDChk }
+
+
+
+// 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
--- /dev/null
+/*
+ * 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.
+ */
+// 44 // Total instruction count
+// 1 // Total kernel count
+
+.kernel PL3_DN_PL3
+.code
+
+
+
+// FileName: DN_PL_Core.asm
+// Author: Vivek Kumar
+// Description: Tasks for DN only case (16x8 block) for planar 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 0x45E8003: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_Load_UV_IMC3_16x8.asm
+// Author: Vivek Kumar
+// Description: Read UV for 16x8 block through DATAPORT
+
+
+
+// FileName: UVCopy_Load_16x8.asm
+// Author: Vivek Kumar
+// Description: Read UV for 16x8 block through DATAPORT
+
+//CHANGE : Read extra UV data to convert to 422. -rT
+//we are reading extra data in ALL cases irrespective of whether upsampling is reqd or not later on, to keep things simple.
+
+
+ add (2) r27.0<1>:d r7.0<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 0x40007:ud { NoDDChk } // U/V block width and height (8x5)
+ 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 0x2290001:ud
+ send (8) udDNDI_UV_RESP(2)<1> r38 0x4 0x2290002:ud
+
+ //Update Header for Save
+ mov (1) mudMSGHDR_UCOPY(0,2)<1> 0x30007:ud // U block width and height (8x4)
+ mov (1) mudMSGHDR_VCOPY(0,2)<1> 0x30007:ud // V block width and height (8x4)
+
+
+
+// FileName: DN_Save_Y_16x8.asm
+// Author: Vivek Kumar
+// Description: Save one 16x8 blocks of Y channel of DN output for reference
+
+
+mov (8) mudDN_Y_OUT(0,0)<1> r0<8;8,1>:ud // message header
+mov (2) mudDN_Y_OUT(0,0)<1> r7.0<2;2,1>:w { NoDDClr } // X origin
+mov (1) mudDN_Y_OUT(0,2)<1> 0x7000F:ud { NoDDChk } // block width and height (16x8)
+
+//send out data through data port
+send (8) null<1>:d mudDN_Y_OUT 0x5 0xA0A8018:ud
+
+
+
+// FileName: DN_Save_UV_IMC3_16x8.asm
+// Author: Vivek Kumar
+// Description: Save UV for 16x8 block through DATAPORT
+
+
+
+// FileName: UVCopy_Save_16x8.asm
+// Author: Vivek Kumar
+// Description: Save UV for 16x8 block through DATAPORT
+
+
+//Reuse the header from Load component
+//Header is modified at the end of load - to be usable for save.
+
+
+ mov (8) mudMSGHDR_UCOPY(1)<1> udDNDI_UV_RESP(0)<8;8,1>
+ mov (8) mudMSGHDR_VCOPY(1)<1> udDNDI_UV_RESP(2)<8;8,1>
+ 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
--- /dev/null
+/*
+ * 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.
+ */
+// 87 // Total instruction count
+// 1 // Total kernel count
+
+.kernel PL_DI_422CP
+.code
+
+
+
+// FileName: DI.asm
+// Author: Vivek Kumar
+// Description: Tasks for DI only 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=====================================
+
+
+//=======================================================================
+
+
+//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 0x4AE8003: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(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> r7.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2
+mov (1) mudMSGHDR_STMM(0,1)<1> r7.1<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: 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> 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: 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 r7.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
+mov (1) r27.1<1>:ud r7.1<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 r7.26<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
+
+
+
+//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
--- /dev/null
+/*
+ * 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.
+ */
+// 57 // Total instruction count
+// 1 // Total kernel count
+
+.kernel PL_DI_PA
+.code
+
+
+
+// FileName: DI.asm
+// Author: Vivek Kumar
+// Description: Tasks for DI only 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=====================================
+
+
+//=======================================================================
+
+
+//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 0x4AE8003: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(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> r7.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2
+mov (1) mudMSGHDR_STMM(0,1)<1> r7.1<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: 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> 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: DI_Save_PA_16x4.asm
+// Author: Vivek Kumar
+// Description: Save two 16x4 blocks of DI output in Packed format
+
+
+add (4) a0.4<1>:uw r2.28<4;4,1>:ub 608:w // Initial Y,U,V offset in YUV422 block; it starts at m20
+
+mov (8) r27.0<1>:ud r0.0<8;8,1>:ud
+shl (1) r27.0<1>:d r7.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
+mov (1) r27.1<1>:d r7.1<0;1,0>:w { NoDDClr, NoDDChk } // Block origin
+mov (1) r27.2<1>:ud 0x3001F:ud { NoDDChk } // Block width and height (32x8)
+
+//prepare the message headers
+mov (8) r18.0<1>:ud r27<8;8,1>:ud
+mov (8) r23.0<1>:ud r27<8;8,1>:ud
+
+// Pack 2nd field Y
+ mov (16) r[a0.4, 0]<2> ubDNDI_RESP(0,0) { NoDDClr }
+ mov (16) r[a0.4, 32]<2> ubDNDI_RESP(0,16) { NoDDClr }
+ mov (16) r[a0.4, 64]<2> ubDNDI_RESP(0,32) { NoDDClr }
+ mov (16) r[a0.4, 96]<2> ubDNDI_RESP(0,48) { NoDDClr }
+// Pack 2nd field U
+ mov (8) r[a0.5, 0]<4> ubDNDI_RESP(2,1)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 32]<4> ubDNDI_RESP(2,17)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 64]<4> ubDNDI_RESP(2,33)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 96]<4> ubDNDI_RESP(2,49)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+// Pack 2nd field V
+ mov (8) r[a0.6, 0]<4> ubDNDI_RESP(2,0)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 32]<4> ubDNDI_RESP(2,16)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 64]<4> ubDNDI_RESP(2,32)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 96]<4> ubDNDI_RESP(2,48)<16;8,2> { NoDDChk } //Vpixels
+
+// Pack 1st field Y
+ mov (16) r[a0.4, 160]<2> ubDNDI_RESP(4,0) { NoDDClr }
+ mov (16) r[a0.4, 192]<2> ubDNDI_RESP(4,16) { NoDDClr }
+ mov (16) r[a0.4, 224]<2> ubDNDI_RESP(4,32) { NoDDClr }
+ mov (16) r[a0.4, 256]<2> ubDNDI_RESP(4,48) { NoDDClr }
+// Pack 1st field U
+ mov (8) r[a0.5, 160]<4> ubDNDI_RESP(6,1)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 192]<4> ubDNDI_RESP(6,17)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 224]<4> ubDNDI_RESP(6,33)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+ mov (8) r[a0.5, 256]<4> ubDNDI_RESP(6,49)<16;8,2> { NoDDClr, NoDDChk } //U pixels
+// Pack 1st field V
+ mov (8) r[a0.6, 160]<4> ubDNDI_RESP(6,0)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 192]<4> ubDNDI_RESP(6,16)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 224]<4> ubDNDI_RESP(6,32)<16;8,2> { NoDDChk } //Vpixels
+ mov (8) r[a0.6, 256]<4> ubDNDI_RESP(6,48)<16;8,2> { NoDDChk } //Vpixels
+
+//save the previous frame
+send (8) null<1>:d r18.0 0x5 0xA0A801B:ud
+
+//save the current frame
+send (8) null<1>:d r23.0 0x5 0xA0A801E: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
+/*
+ * 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.
+ */
// 93 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 110 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 48 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 278 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 7 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 7 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 7 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 7 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 7 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 17 // Total instruction count
// 1 // Total kernel count
+/*
+ * 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.
+ */
// 10 // Total instruction count
// 1 // Total kernel count