| 1 |
/* |
| 2 |
* All Video Processing kernels |
| 3 |
* Copyright © <2010>, Intel Corporation. |
| 4 |
* |
| 5 |
* Permission is hereby granted, free of charge, to any person obtaining a |
| 6 |
* copy of this software and associated documentation files (the |
| 7 |
* "Software"), to deal in the Software without restriction, including |
| 8 |
* without limitation the rights to use, copy, modify, merge, publish, |
| 9 |
* distribute, sub license, and/or sell copies of the Software, and to |
| 10 |
* permit persons to whom the Software is furnished to do so, subject to |
| 11 |
* the following conditions: |
| 12 |
* |
| 13 |
* The above copyright notice and this permission notice (including the |
| 14 |
* next paragraph) shall be included in all copies or substantial portions |
| 15 |
* of the Software. |
| 16 |
* |
| 17 |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS |
| 18 |
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
| 19 |
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. |
| 20 |
* IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR |
| 21 |
* ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, |
| 22 |
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE |
| 23 |
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. |
| 24 |
* |
| 25 |
* This file was originally licensed under the following license |
| 26 |
* |
| 27 |
* Licensed under the Apache License, Version 2.0 (the "License"); |
| 28 |
* you may not use this file except in compliance with the License. |
| 29 |
* You may obtain a copy of the License at |
| 30 |
* |
| 31 |
* http://www.apache.org/licenses/LICENSE-2.0 |
| 32 |
* |
| 33 |
* Unless required by applicable law or agreed to in writing, software |
| 34 |
* distributed under the License is distributed on an "AS IS" BASIS, |
| 35 |
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 36 |
* See the License for the specific language governing permissions and |
| 37 |
* limitations under the License. |
| 38 |
* |
| 39 |
*/ |
| 40 |
|
| 41 |
#ifndef COMMON_INC |
| 42 |
#define COMMON_INC |
| 43 |
|
| 44 |
// Module name: common.inc |
| 45 |
// |
| 46 |
// Common header file for all Video-Processing kernels |
| 47 |
// |
| 48 |
|
| 49 |
.default_execution_size (16) |
| 50 |
.default_register_type :ub |
| 51 |
|
| 52 |
.reg_count_total 80 |
| 53 |
.reg_count_payload 4 |
| 54 |
|
| 55 |
|
| 56 |
//========== Common constants ========== |
| 57 |
|
| 58 |
// Bit position constants |
| 59 |
#define BIT0 0x01 |
| 60 |
#define BIT1 0x02 |
| 61 |
#define BIT2 0x04 |
| 62 |
#define BIT3 0x08 |
| 63 |
#define BIT4 0x10 |
| 64 |
#define BIT5 0x20 |
| 65 |
#define BIT6 0x40 |
| 66 |
#define BIT7 0x80 |
| 67 |
#define BIT8 0x0100 |
| 68 |
#define BIT9 0x0200 |
| 69 |
#define BIT10 0x0400 |
| 70 |
#define BIT11 0x0800 |
| 71 |
#define BIT12 0x1000 |
| 72 |
#define BIT13 0x2000 |
| 73 |
#define BIT14 0x4000 |
| 74 |
#define BIT15 0x8000 |
| 75 |
#define BIT16 0x00010000 |
| 76 |
#define BIT17 0x00020000 |
| 77 |
#define BIT18 0x00040000 |
| 78 |
#define BIT19 0x00080000 |
| 79 |
#define BIT20 0x00100000 |
| 80 |
#define BIT21 0x00200000 |
| 81 |
#define BIT22 0x00400000 |
| 82 |
#define BIT23 0x00800000 |
| 83 |
#define BIT24 0x01000000 |
| 84 |
#define BIT25 0x02000000 |
| 85 |
#define BIT26 0x04000000 |
| 86 |
#define BIT27 0x08000000 |
| 87 |
#define BIT28 0x10000000 |
| 88 |
#define BIT29 0x20000000 |
| 89 |
#define BIT30 0x40000000 |
| 90 |
#define BIT31 0x80000000 |
| 91 |
|
| 92 |
#define nGRFWIB 32 // GRF register width in byte |
| 93 |
#define nGRFWIW 16 // GRF register width in word |
| 94 |
#define nGRFWID 8 // GRF register width in dword |
| 95 |
|
| 96 |
#define nTOP_FIELD 0 |
| 97 |
#define nBOTTOM_FIELD 1 |
| 98 |
|
| 99 |
#define nPREVIOUS_FRAME 0 // Previous frame |
| 100 |
#define nCURRENT_FRAME 1 // Current frame |
| 101 |
#define nNEXT_FRAME 2 // Next frame |
| 102 |
|
| 103 |
#ifdef GT |
| 104 |
// GT DI Kernel |
| 105 |
#else // ILK |
| 106 |
// ILK DI Kernel |
| 107 |
#endif |
| 108 |
|
| 109 |
//=================================== |
| 110 |
|
| 111 |
//========== Macros ========== |
| 112 |
#define REGION(Width,HStride) <Width*HStride;Width,HStride> // Region definition when ExecSize = Width |
| 113 |
|
| 114 |
#define RegFile(a) a |
| 115 |
#define REG(r,n) _REG(RegFile(r),n) |
| 116 |
#define _REG(r,n) __REG(r,n) |
| 117 |
#define __REG(r,n) r##n.0 |
| 118 |
#define REG2(r,n,s) _REG2(RegFile(r),n,s) |
| 119 |
#define _REG2(r,n,s) __REG2(r,n,s) |
| 120 |
#define __REG2(r,n,s) r##n.##s |
| 121 |
|
| 122 |
#define dNULLREG null<1>:d |
| 123 |
#define wNULLREG null<1>:w |
| 124 |
|
| 125 |
#define KERNEL_ID(kernel_ID) mov NULLREG kernel_ID:ud |
| 126 |
|
| 127 |
|
| 128 |
#define NODDCLR |
| 129 |
#define NODDCLR_NODDCHK |
| 130 |
#define NODDCHK |
| 131 |
|
| 132 |
//#define NODDCLR { NoDDClr } |
| 133 |
//#define NODDCLR_NODDCHK { NoDDClr, NoDDChk } |
| 134 |
//#define NODDCHK { NoDDChk } |
| 135 |
|
| 136 |
|
| 137 |
//========== Defines ==================== |
| 138 |
|
| 139 |
|
| 140 |
//========== GRF partition ========== |
| 141 |
// r0 header : r0 (1 GRF) |
| 142 |
// Static parameters : r1 - r5 (5 GRFS) |
| 143 |
// Inline parameters : r6 - r7 (2 GRFs) |
| 144 |
// MSGSRC : r9 (1 GRF) |
| 145 |
// Top IO region : r10 - r33 (24 GRFS 8 for each component Y,U,V 16X8:w) |
| 146 |
// Free space : r34 - r55 (22 GRFS) |
| 147 |
// Bottom IO region : r56 - r79 (24 GRFS 8 for each component Y,U,V 16X8:w) |
| 148 |
//=================================== |
| 149 |
|
| 150 |
|
| 151 |
//========== Static Parameters ========== |
| 152 |
// r1 |
| 153 |
#define fPROCAMP_C0 r1.0 // DWORD 0, Procamp constant C0 in :f |
| 154 |
#define wPROCAMP_C0 r1.0 // DWORD 0, Procamp constant C0 in :w |
| 155 |
#define NUMBER_0002 r1.1 // DWORD 0, 0x0002 used in procamp for GT |
| 156 |
#define udCP_MessageFormat r1.0 // DWORD 0, bits 2:3 of DWORD. (CE) |
| 157 |
#define udCP_StatePointer r1.0 // DWORD 0, bits 31:5 of DWORD.(CE) |
| 158 |
|
| 159 |
#define ubSRC_CF_OFFSET r1.4 // DWORD 1, byte 0-2. SRC packed color format YUV offset in :ub |
| 160 |
|
| 161 |
#define ubDEST_RGB_FORMAT r1.8 // DWORD 2, byte 0. Dest RGB color format (0:ARGB FF:XRGB) |
| 162 |
#define ubDEST_CF_OFFSET r1.8 // DWORD 2, byte 0-2. Dest packed color format YUV offset in :ub |
| 163 |
|
| 164 |
#define fPROCAMP_C1 r1.3 // DWORD 3, Procamp constant C1 in :f |
| 165 |
#define wPROCAMP_C1 r1.6 // DWORD 3, Procamp constant C1 in :w |
| 166 |
#define NUMBER_0100 r1.7 // DWORD 3, 0x0100 used in procamp for GT |
| 167 |
|
| 168 |
#define fPROCAMP_C2 r1.4 // DWORD 4, Procamp constant C2 in :f |
| 169 |
#define wPROCAMP_C2 r1.8 // DWORD 4, Procamp constant C2 in :w |
| 170 |
|
| 171 |
#define uwSPITCH_DIV2 r1.10 // DWORD 5, byte 0-1. statistics surface pitch divided by 2 |
| 172 |
|
| 173 |
#define fVIDEO_STEP_Y r1.6 // DWORD 6, :f, AVS normalized reciprocal of Y Scaling factor |
| 174 |
#define ubSTMM_SHIFT r1.24 // DWORD 6, byte 0. Amount of right shift for the DI blending equation |
| 175 |
#define ubSTMM_MIN r1.25 // DWORD 6, byte 1. Min STMM for DI blending equation |
| 176 |
#define ubSTMM_MAX r1.26 // DWORD 6, byte 2. Max STMM for DI blending equation |
| 177 |
#define ubTFLD_FIRST r1.27 // DWORD 6, byte 3. Field parity order |
| 178 |
|
| 179 |
#define fPROCAMP_C5 r1.7 // DWORD 7, Procamp constant C3 in :f |
| 180 |
#define wPROCAMP_C5 r1.14 // DWORD 7, Procamp constant C3 in :w |
| 181 |
|
| 182 |
// r2 |
| 183 |
#define fPROCAMP_C3 r2.0 // DWORD 0, Procamp constant C4 in :f |
| 184 |
#define wPROCAMP_C3 r2.0 // DWORD 0, Procamp constant C4 in :w |
| 185 |
|
| 186 |
#define fCSC_C5 r2.2 // DWORD 2. WG+CSC constant C5 |
| 187 |
#define wCSC_C5 r2.4 // DWORD 2. WG+CSC constant C5 |
| 188 |
|
| 189 |
#define fPROCAMP_C4 r2.3 // DWORD 3, Procamp constant C5 in :f |
| 190 |
#define wPROCAMP_C4 r2.6 // DWORD 3, Procamp constant C5 in :w |
| 191 |
|
| 192 |
#define fCSC_C8 r2.4 // DWORD 4. WG+CSC constant C8 |
| 193 |
#define wCSC_C8 r2.8 // DWORD 4. WG+CSC constant C8 |
| 194 |
#define fCSC_C9 r2.7 // DWORD 7. WG+CSC constant C9 |
| 195 |
#define wCSC_C9 r2.14 // DWORD 7. WG+CSC constant C9 |
| 196 |
|
| 197 |
// r3 |
| 198 |
#define fCSC_C0 r3.0 // DWORD 0. WG+CSC constant C0 |
| 199 |
#define wCSC_C0 r3.0 // DWORD 0. WG+CSC constant C0 |
| 200 |
|
| 201 |
#define fSCALING_STEP_RATIO r3.1 // DWORD 1, = Alpha_X_Scaling_Step / Video_X_scaling_Step :f (blending) |
| 202 |
#define fALPHA_STEP_X r3.1 // DWORD 1, = 1/Scale X, 0.5 = 2x, in :f (blending) |
| 203 |
|
| 204 |
#define fALPHA_STEP_Y r3.2 // DWORD 2, = 1/Scale Y, in :f |
| 205 |
|
| 206 |
#define fCSC_C4 r3.3 // DWORD 3. WG+CSC constant C4 |
| 207 |
#define wCSC_C4 r3.6 // DWORD 3. WG+CSC constant C4 |
| 208 |
#define fCSC_C1 r3.4 // DWORD 4. WG+CSC constant C1 |
| 209 |
#define wCSC_C1 r3.8 // DWORD 4. WG+CSC constant C1 |
| 210 |
|
| 211 |
#define wSRC_H_ORI_OFFSET r3.10 // DWORD 5, bytes 0,1 :w |
| 212 |
#define wSRC_V_ORI_OFFSET r3.11 // DWORD 5, bytes 2,3 :w |
| 213 |
|
| 214 |
#define dCOLOR_PIXEL r3.6 // DWORD 6. Color pixel for Colorfill |
| 215 |
|
| 216 |
#define fCSC_C2 r3.6 // DWORD 6. WG+CSC constant C2 |
| 217 |
#define wCSC_C2 r3.12 // DWORD 6. WG+CSC constant C2 |
| 218 |
#define fCSC_C3 r3.7 // DWORD 7. WG+CSC constant C3 |
| 219 |
#define wCSC_C3 r3.14 // DWORD 7. WG+CSC constant C3 |
| 220 |
|
| 221 |
// r4 |
| 222 |
#define fCSC_C6 r4.0 // DWORD 0. WG+CSC constant C6 |
| 223 |
#define wCSC_C6 r4.0 // DWORD 0. WG+CSC constant C6 |
| 224 |
|
| 225 |
#define wFRAME_ENDX r4.2 // DWORD 1, word 0. Horizontal end = Origin+Width (in pixels)(for multiple blocks) |
| 226 |
#define wNUM_BLKS r4.3 // DWORD 1, word 1. Number of blocks to process (for multiple blocks) |
| 227 |
|
| 228 |
#define wCOPY_ORIX r4.5 // DWORD 2, word 1. A copy of X origin (for multiple blocks) |
| 229 |
#define uwNLAS_ENABLE r4.4 // DWORD 2, bit 15, NLAS enble bit |
| 230 |
|
| 231 |
#define fCSC_C7 r4.3 // DWORD 3. WG+CSC constant C7 |
| 232 |
#define wCSC_C7 r4.6 // DWORD 3. WG+CSC constant C7 |
| 233 |
#define fCSC_C10 r4.4 // DWORD 4. WG+CSC constant C10 |
| 234 |
#define wCSC_C10 r4.8 // DWORD 4. WG+CSC constant C10 |
| 235 |
|
| 236 |
#define fFRAME_VID_ORIX r4.5 // DWORD 5, Frame horizontal origin normalized for scale kernel |
| 237 |
|
| 238 |
#define fFRAME_ALPHA_ORIX r4.6 // DWORD 6. Normalized alpha horiz origin for the frame |
| 239 |
|
| 240 |
#define fCSC_C11 r4.7 // DWORD 7. WG+CSC constant C11 |
| 241 |
#define wCSC_C11 r4.14 // DWORD 7. WG+CSC constant C11 |
| 242 |
|
| 243 |
//======================================== |
| 244 |
|
| 245 |
//========== Inline parameters =========== |
| 246 |
// r5 |
| 247 |
#define wORIX r5.0 // DWORD 0, byte 0-1. :w, Destination Block Horizontal Origin in pel |
| 248 |
#define wORIY r5.1 // DWORD 0, byte 2-3. :w, Destination Block Vertical Origin in pel |
| 249 |
|
| 250 |
#define fSRC_VID_H_ORI r5.1 // DWORD 1, :f, SRC Y horizontal origin normalized for scale kernel |
| 251 |
|
| 252 |
#define fSRC_VID_V_ORI r5.2 // DWORD 2, :f, SRC Y vertical origin normalized for scale kernel |
| 253 |
|
| 254 |
#define fSRC_ALPHA_H_ORI r5.3 // DWORD 3, :f, Normalized alpha horizontal origin |
| 255 |
|
| 256 |
#define fSRC_ALPHA_V_ORI r5.4 // DWORD 4, :f, Normalized alpha vertical origin |
| 257 |
|
| 258 |
#define uwALPHA_MASK_X r5.10 // DWORD 5, byte 0-1 :w, H. alpha mask |
| 259 |
#define ubALPHA_MASK_Y r5.22 // DWORD 5, byte 2. :ub,V. alpha mask |
| 260 |
#define ubBLK_CNT_X r5.23 // DWORD 5, byte 3, :ub, Horizontal Block Count per thread |
| 261 |
|
| 262 |
// mask is used for each block. it will be reloaded from r6 below for the last block. |
| 263 |
#define udBLOCK_MASK r5.6 // DWORD 6 |
| 264 |
#define uwBLOCK_MASK_H r5.12 // DWORD 6, byte 0-1 :uw, Block horizontal mask used in non-DWord aligned kernels |
| 265 |
#define ubBLOCK_MASK_V r5.26 // DWORD 6, byte 2 :ub, Block vertical mask used in non-DWord aligned kernels |
| 266 |
#define ubNUM_BLKS r5.27 // DWORD 6, byte 3, :ub, Total Block Count per thread |
| 267 |
|
| 268 |
#define fVIDEO_STEP_X r5.7 // DWORD 7. :f, AVS normalized reciprocal of X Scaling factor |
| 269 |
|
| 270 |
// r6 |
| 271 |
#define fVIDEO_STEP_DELTA r6.0 // DWORD 0. :f, AVS normalized delta between 2 adjacent scaling steps (used for non-linear scaling) |
| 272 |
|
| 273 |
// mask is used for the last block (assume only M*1 and 1*N block partation aer supported) |
| 274 |
#define udBLOCK_MASK_2 r6.1 // DWORD 1 |
| 275 |
#define uwBLOCK_MASK_H_RIGHT r6.2 // DWORD 1, byte 0-1 :uw, Block horizontal mask used in non-DWord aligned kernels (right) |
| 276 |
#define ubBLOCK_MASK_V_BOTTOM r6.6 // DWORD 1, byte 2 :ub, Block vertical mask used in non-DWord aligned kernels |
| 277 |
#define uwBLOCK_MASK_H_MIDDLE r6.4 // DWORD 2, byte 0-1 :uw, Block horizontal mask used in non-DWord aligned kernels (left) |
| 278 |
|
| 279 |
|
| 280 |
//====================== Binding table ========================================= |
| 281 |
|
| 282 |
#if defined(DNDI) |
| 283 |
// DNDI Surface Binding Table |
| 284 |
//#define nBI_SRC_CURR 0 // Current input frame surface |
| 285 |
//#define nBI_SRC_PRIV 1 // Denoised previous input frame surface |
| 286 |
//#define nBI_SRC_STAT 2 // Statistics input surface (STMM / Noise motion history) |
| 287 |
//#define nBI_DEST_1ST 3 // 1st deinterlaced output frame surface |
| 288 |
// #define nBI_DEST_YUV 3 // Dest frame YUV (for DN only) |
| 289 |
//#define nBI_DEST_Y 3 // Dest frame Y (for DN only) |
| 290 |
//#define nBI_DEST_2ND 4 // 2nd deinterlaced output frame surface |
| 291 |
//#define nBI_DEST_DN_CURR 6 // Denoised current output frame surface |
| 292 |
//#define nBI_DEST_STAT 7 // Statistics output surface (STMM / Noise motion history) |
| 293 |
// #define nBI_DEST_U 8 // Dest frame U (for DN only) |
| 294 |
// #define nBI_DEST_V 9 // Dest frame V (for DN only) |
| 295 |
// #define nBI_SRC_U 10 // Src frame U (for DN only) |
| 296 |
// #define nBI_SRC_V 11 // Src frame V (for DN only) |
| 297 |
// #define nBI_SRC_UV 10 // Current src frame for UV |
| 298 |
|
| 299 |
#endif |
| 300 |
|
| 301 |
#if defined(INPUT_PL3) |
| 302 |
// PL3 Surface Binding Table |
| 303 |
// #define nBI_SRC_ALPHA 0 // Alpha |
| 304 |
// #define nBI_SRC_Y 1 // Current src frame |
| 305 |
// #define nBI_SRC_U 2 // Current src frame |
| 306 |
// #define nBI_SRC_V 3 // Current src frame |
| 307 |
// #define nBI_DEST_Y 10 // Dest frame |
| 308 |
// #define nBI_DEST_U 11 // Dest frame |
| 309 |
// #define nBI_DEST_V 12 // Dest frame |
| 310 |
// #define nBI_DEST_YUV 7 // Dest frame |
| 311 |
// #define nBI_DEST_RGB 7 // same num as BI_DEST_YUV, never used at the same time |
| 312 |
#endif |
| 313 |
|
| 314 |
#if defined(INPUT_PL2) |
| 315 |
// PL2 Surface Binding Table |
| 316 |
// #define nBI_SRC_ALPHA 0 // Alpha |
| 317 |
// #define nBI_SRC_Y 1 // Current src frame for Y + offseted UV |
| 318 |
// #define nBI_SRC_YUV 1 // Current src frame for YUV in case of NV12_AVS |
| 319 |
// #define nBI_SRC_UV 2 // Current src frame for UV |
| 320 |
// #define nBI_DEST_YUV 7 // Current dest frame for Y + offseted UV |
| 321 |
// #define nBI_DEST_RGB 7 // same num as BI_DEST_YUV, never used at the same time |
| 322 |
// #define nBI_DEST_Y 10 // Dest frame |
| 323 |
// #define nBI_DEST_U 11 // Dest frame |
| 324 |
// #define nBI_DEST_V 12 // Dest frame |
| 325 |
#endif |
| 326 |
|
| 327 |
#if defined(INPUT_PA) || defined(COLORFILL) |
| 328 |
// Packed Surface Binding Table |
| 329 |
// #define nBI_SRC_ALPHA 0 // Alpha |
| 330 |
// #define nBI_SRC_YUV 1 // Current src frame |
| 331 |
// #define nBI_DEST_YUV 3 // Dest frame |
| 332 |
// #define nBI_DEST_RGB 3 // same num as BI_DEST_YUV, never used at the same time |
| 333 |
#endif |
| 334 |
|
| 335 |
|
| 336 |
//supper binding table |
| 337 |
#define nBI_ALPHA_SRC 0 |
| 338 |
#define nBI_CURRENT_SRC_YUV 1 |
| 339 |
#define nBI_FIELD_COPY_SRC_1_YUV 1 |
| 340 |
#define nBI_CURRENT_SRC_Y 1 |
| 341 |
#define nBI_FIELD_COPY_SRC_1_Y 1 |
| 342 |
#define nBI_CURRENT_SRC_RGB 1 |
| 343 |
#define nBI_CURRENT_SRC_UV 2 |
| 344 |
#define nBI_FIELD_COPY_SRC_1_UV 2 |
| 345 |
#define nBI_CURRENT_SRC_U 2 |
| 346 |
#define nBI_FIELD_COPY_SRC_1_U 2 |
| 347 |
#define nBI_CURRENT_SRC_V 3 |
| 348 |
#define nBI_FIELD_COPY_SRC_1_V 3 |
| 349 |
#define nBI_TEMPORAL_REFERENCE_YUV 4 |
| 350 |
#define nBI_FIELD_COPY_SRC_2_YUV 4 |
| 351 |
#define nBI_TEMPORAL_REFERENCE_Y 4 |
| 352 |
#define nBI_FIELD_COPY_SRC_2_Y 4 |
| 353 |
#define nBI_CURRENT_SRC_YUV_HW_DI 4 |
| 354 |
#define nBI_TEMPORAL_REFERENCE_UV 5 |
| 355 |
#define nBI_FIELD_COPY_SRC_2_UV 5 |
| 356 |
#define nBI_TEMPORAL_REFERENCE_U 5 |
| 357 |
#define nBI_FIELD_COPY_SRC_2_U 5 |
| 358 |
#define nBI_DENOISED_PREV_HW_DI 5 |
| 359 |
#define nBI_TEMPORAL_REFERENCE_V 6 |
| 360 |
#define nBI_FIELD_COPY_SRC_2_V 6 |
| 361 |
#define nBI_STMM_HISTORY 6 |
| 362 |
#define nBI_DESTINATION_YUV 7 |
| 363 |
#define nBI_DESTINATION_RGB 7 |
| 364 |
#define nBI_DESTINATION_Y 7 |
| 365 |
#define nBI_DESTINATION_UV 8 |
| 366 |
#define nBI_DESTINATION_U 8 |
| 367 |
#define nBI_DESTINATION_V 9 |
| 368 |
#define nBI_DESTINATION_1_YUV 10 |
| 369 |
#define nBI_DESTINATION_1_Y 10 |
| 370 |
#define nBI_DESTINATION_1_UV 11 |
| 371 |
#define nBI_DESTINATION_1_U 11 |
| 372 |
#define nBI_DESTINATION_1_V 12 |
| 373 |
#define nBI_DESTINATION_2_YUV 13 |
| 374 |
#define nBI_DESTINATION_2_Y 13 |
| 375 |
#define nBI_DESTINATION_2_UV 14 |
| 376 |
#define nBI_DESTINATION_2_U 14 |
| 377 |
#define nBI_DESTINATION_2_V 15 |
| 378 |
#define nBI_STMM_HISTORY_OUTPUT 20 |
| 379 |
#define nBI_TEMPORAL_REFERENCE_YUV_PDI 21 |
| 380 |
#define nBI_TEMPORAL_REFERENCE_Y_PDI 21 |
| 381 |
#define nBI_TEMPORAL_REFERENCE_UV_PDI 22 |
| 382 |
#define nBI_TEMPORAL_REFERENCE_U_PDI 22 |
| 383 |
#define nBI_TEMPORAL_REFERENCE_V_PDI 23 |
| 384 |
#define nBI_SUBVIDEO_YUV 26 |
| 385 |
#define nBI_SUBVIDEO_Y 26 |
| 386 |
#define nBI_SUBVIDEO_UV 27 |
| 387 |
#define nBI_SUBVIDEO_U 27 |
| 388 |
#define nBI_SUBVIDEO_V 28 |
| 389 |
#define nBI_SUBPICTURE_YUV 29 |
| 390 |
#define nBI_SUBPICTURE_P8 29 |
| 391 |
#define nBI_SUBPICTURE_A8 30 |
| 392 |
#define nBI_GRAPHIC_YUV 31 |
| 393 |
#define nBI_GRAPHIC_P8 31 |
| 394 |
#define nBI_GRAPHIC_A8 32 |
| 395 |
|
| 396 |
|
| 397 |
|
| 398 |
//========== Planar Sampler State Table Index ========== |
| 399 |
#define nSI_SRC_ALPHA 0x000 // Sampler State for Alpha |
| 400 |
|
| 401 |
//Sampler Index for AVS/IEF messages |
| 402 |
#define nSI_SRC_Y 0x400 // Sampler State for Y |
| 403 |
#define nSI_SRC_U 0x800 // Sampler State for U |
| 404 |
#define nSI_SRC_V 0xC00 // Sampler State for V |
| 405 |
#define nSI_SRC_UV 0x800 // For NV12 surfaces |
| 406 |
#define nSI_SRC_YUV 0x400 // For Packed surfaces |
| 407 |
#define nSI_SRC_RGB 0x400 // For ARGB surfaces |
| 408 |
|
| 409 |
//Sampler Index for SIMD16 sampler messages |
| 410 |
#define nSI_SRC_SIMD16_Y 0x100 // Sampler State for Y |
| 411 |
#define nSI_SRC_SIMD16_U 0x200 // Sampler State for U |
| 412 |
#define nSI_SRC_SIMD16_V 0x300 // Sampler State for V |
| 413 |
#define nSI_SRC_SIMD16_UV 0x200 // For NV12 surfaces |
| 414 |
#define nSI_SRC_SIMD16_YUV 0x100 // For Packed surfaces |
| 415 |
#define nSI_SRC_SIMD16_RGB 0x100 // For ARGB surfaces |
| 416 |
|
| 417 |
|
| 418 |
|
| 419 |
// Common Registers |
| 420 |
#define pCF_Y_OFFSET a0.4 // Address register holding Y offset |
| 421 |
#define pCF_U_OFFSET a0.5 // Address register holding U offset |
| 422 |
#define pCF_V_OFFSET a0.6 // Address register holding V offset |
| 423 |
|
| 424 |
// #define YUV_ORI ORIX // Used by writing packed data to dport |
| 425 |
|
| 426 |
|
| 427 |
//================= Message Payload Header fields ============================== |
| 428 |
#define IDP r0.2:ud // Interface Descriptor Pointer |
| 429 |
|
| 430 |
//================= Common Message Descriptor TBD add common load and save ===== |
| 431 |
// Message descriptor for dataport media write |
| 432 |
#ifdef GT |
| 433 |
// Message Descriptors |
| 434 |
// = 000 0001 (min message len 1 - add later) 00000 (resp len 0) |
| 435 |
// 1 (header present 1) 0 0 1010 (media block write) 00000 |
| 436 |
// 00000000 (binding table index - set later) |
| 437 |
// = 0x02094000 |
| 438 |
#define nDPMW_MSGDSC 0x02094000 |
| 439 |
#define nDPMR_MSGDSC 0x02098000 // Data Port Media Block Read Message Descriptor |
| 440 |
// TBD |
| 441 |
#else // ILK |
| 442 |
// Message Descriptors |
| 443 |
// = 000 0001 (min message len 1 - add later) 00000 (resp len 0) |
| 444 |
// 1 (header present 1) 000 0 010 (media block write) 0000 |
| 445 |
// 00000000 (binding table index - set later) |
| 446 |
// = 0x02082000 |
| 447 |
#define nDPMW_MSGDSC 0x02082000 // Data Port Media Block Write Message Descriptor |
| 448 |
#define nDPMR_MSGDSC 0x0208A000 // Data Port Media Block Read Message Descriptor |
| 449 |
#endif |
| 450 |
|
| 451 |
// Message Length defines |
| 452 |
#define nMSGLEN_1 0x02000000 // Message Length of 1 GRF for Send |
| 453 |
#define nMSGLEN_2 0x04000000 // Message Length of 2 GRF for Send |
| 454 |
#define nMSGLEN_4 0x08000000 // Message Length of 4 GRF for Send |
| 455 |
#define nMSGLEN_8 0x10000000 // Message Length of 8 GRF for Send |
| 456 |
|
| 457 |
// Response Length defines |
| 458 |
#define nRESLEN_1 0x00100000 // Message Response Length of 1 GRF from Send |
| 459 |
#define nRESLEN_2 0x00200000 // Message Response Length of 2 GRF from Send |
| 460 |
#define nRESLEN_3 0x00300000 // Message Response Length of 3 GRF from Send |
| 461 |
#define nRESLEN_4 0x00400000 // Message Response Length of 4 GRF from Send |
| 462 |
#define nRESLEN_5 0x00500000 // Message Response Length of 5 GRF from Send |
| 463 |
#define nRESLEN_8 0x00800000 // Message Response Length of 8 GRF from Send |
| 464 |
#define nRESLEN_9 0x00900000 // Message Response Length of 9 GRF from Send |
| 465 |
#define nRESLEN_11 0x00B00000 // Message Response Length of 11 GRF from Send |
| 466 |
#define nRESLEN_12 0x00C00000 // Message Response Length of 12 GRF from Send |
| 467 |
#define nRESLEN_16 0x01000000 // Message Response Length of 16 GRF from Send |
| 468 |
|
| 469 |
// Block Width and Height Size defines |
| 470 |
#define nBLOCK_WIDTH_4 0x00000003 // Block Width 4 |
| 471 |
#define nBLOCK_WIDTH_5 0x00000004 // Block Width 5 |
| 472 |
#define nBLOCK_WIDTH_8 0x00000007 // Block Width 8 |
| 473 |
#define nBLOCK_WIDTH_9 0x00000008 // Block Width 9 |
| 474 |
#define nBLOCK_WIDTH_12 0x0000000B // Block Width 12 |
| 475 |
#define nBLOCK_WIDTH_16 0x0000000F // Block Width 16 |
| 476 |
#define nBLOCK_WIDTH_20 0x00000013 // Block Width 20 |
| 477 |
#define nBLOCK_WIDTH_32 0x0000001F // Block Width 32 |
| 478 |
#define nBLOCK_HEIGHT_1 0x00000000 // Block Height 1 |
| 479 |
#define nBLOCK_HEIGHT_2 0x00010000 // Block Height 2 |
| 480 |
#define nBLOCK_HEIGHT_4 0x00030000 // Block Height 4 |
| 481 |
#define nBLOCK_HEIGHT_5 0x00040000 // Block Height 5 |
| 482 |
#define nBLOCK_HEIGHT_8 0x00070000 // Block Height 8 |
| 483 |
|
| 484 |
// Extended Message Descriptors |
| 485 |
#define nEXTENDED_MATH 0x1 |
| 486 |
#define nSMPL_ENGINE 0x2 |
| 487 |
#define nMESSAGE_GATEWAY 0x3 |
| 488 |
#define nDATAPORT_READ 0x4 |
| 489 |
#define nDATAPORT_WRITE 0x5 |
| 490 |
#define nURB 0x6 |
| 491 |
#define nTS_EOT 0x27 // with End-Of-Thread bit ON |
| 492 |
|
| 493 |
// Common message descriptors: |
| 494 |
#ifdef GT |
| 495 |
#define nEOT_MSGDSC 0x02000010 // End of Thread Message Descriptor |
| 496 |
#define IF_NULL null:uw null:uw null:uw //for different if instructions on ILK and Gen6 |
| 497 |
#else //ILK |
| 498 |
#define nEOT_MSGDSC 0x02000000 // End of Thread Message Descriptor |
| 499 |
#define IF_NULL |
| 500 |
#endif |
| 501 |
|
| 502 |
|
| 503 |
//===================== Math Function Control =================================== |
| 504 |
#define mfcINV 0x1 // reciprocal |
| 505 |
#define mfcLOG 0x2 // log |
| 506 |
#define mfcEXP 0x3 // exponent |
| 507 |
#define mfcSQRT 0x4 // square root |
| 508 |
#define mfcRSQ 0x5 // reciprocal square root |
| 509 |
#define mfcSIN 0x6 // sine (in radians) |
| 510 |
#define mfcCOS 0x7 // cosine (in radians) |
| 511 |
#define mfcSINCOS 0x8 // dst0 = sin of src0, dst1 = cosine of src0 (in radians) - GT+ ONLY |
| 512 |
#define mfcPOW 0xA // abs(src0) raised to the src1 power |
| 513 |
#define mfcINT_DIV_QR 0xB // return quotient and remainder |
| 514 |
#define mfcINT_DIV_Q 0xC // return quotient |
| 515 |
#define mfcINT_DIV_R 0xD // return remainder |
| 516 |
|
| 517 |
|
| 518 |
//=================== Message related registers ================================= |
| 519 |
|
| 520 |
#ifdef GT |
| 521 |
#define udDUMMY_NULL |
| 522 |
#else // _ILK |
| 523 |
#define udDUMMY_NULL null:ud // Used in send inst as src0 |
| 524 |
#endif |
| 525 |
|
| 526 |
|
| 527 |
//----------- Message Registers ------------ |
| 528 |
#define mMSGHDR m1 // Message Payload Header |
| 529 |
#define mMSGHDRY m1 // Message Payload Header register for Y data |
| 530 |
#define mMSGHDRU m2 // Message Payload Header register for U data |
| 531 |
#define mMSGHDRV m3 // Message Payload Header register for V data |
| 532 |
#define mMSGHDRYA m4 // Second Message Payload Header register for Y data |
| 533 |
#define mMSGHDRH m5 // Message Payload Header register for motion history |
| 534 |
#define mMSGHDRY1 m1 // Message Payload Header register for first Y data |
| 535 |
#define mMSGHDRY2 m2 // Message Payload Header register for second Y data |
| 536 |
#define mMSGHDRY3 m3 // Message Payload Header register for third Y data |
| 537 |
#define mMSGHDRY4 m4 // Message Payload Header register for fourth Y data |
| 538 |
#define mMSGHDRY5 m5 // Message Payload Header register for fifth Y data |
| 539 |
#define mMSGHDRY6 m6 // Message Payload Header register for sixth Y data |
| 540 |
#define mMSGHDR_EOT m15 // Dummy Message Register for EOT |
| 541 |
|
| 542 |
#define rMSGSRC r8 // Message source register |
| 543 |
#define pMSGDSC a0.0:ud // Message Descriptor register (type DWORD) |
| 544 |
|
| 545 |
#define udMH_ORI rMSGSRC.0 // Data Port Media Block R/W message header block offset |
| 546 |
#define udMH_ORIX rMSGSRC.0 // Data Port Media Block R/W message header X offset |
| 547 |
#define udMH_ORIY rMSGSRC.1 // Data Port Media Block R/W message header Y offset |
| 548 |
#define udMH_SIZE rMSGSRC.2 // Data Port Media Block R/W message header block width & height |
| 549 |
|
| 550 |
// M2 - M9 for message data payload |
| 551 |
.declare mubMSGPAYLOAD Base=m2 ElementSize=1 SrcRegion=REGION(16,1) Type=ub |
| 552 |
.declare muwMSGPAYLOAD Base=m2 ElementSize=2 SrcRegion=REGION(16,1) Type=uw |
| 553 |
.declare mudMSGPAYLOAD Base=m2 ElementSize=4 SrcRegion=REGION(8,1) Type=ud |
| 554 |
.declare mfMSGPAYLOAD Base=m2 ElementSize=4 SrcRegion=REGION(8,1) Type=f |
| 555 |
|
| 556 |
//=================== End of thread instruction =========================== |
| 557 |
#ifdef GT |
| 558 |
#define END_THREAD mov (8) mMSGHDR_EOT<1>:ud r0.0<8;8,1>:ud \n\ |
| 559 |
send (1) null<1>:d mMSGHDR_EOT nTS_EOT nEOT_MSGDSC |
| 560 |
#else // ILK This should be changed to 1 instruction; I have tested it and it works - vK |
| 561 |
#define END_THREAD mov (8) mMSGHDR_EOT<1>:ud r0.0<8;8,1>:ud \n\ |
| 562 |
send (1) dNULLREG mMSGHDR_EOT udDUMMY_NULL nTS_EOT nEOT_MSGDSC:ud |
| 563 |
#endif |
| 564 |
|
| 565 |
|
| 566 |
//======================================================================= |
| 567 |
// Region declarations for SRC and DEST as TOP and BOT |
| 568 |
|
| 569 |
// Common I/O regions |
| 570 |
#define nREGION_1 1 |
| 571 |
#define nREGION_2 2 |
| 572 |
|
| 573 |
//*** These region base GRFs are fixed regardless planar/packed, and data alignment. |
| 574 |
//*** Each kernel is responsible to select the correct region declaration below. |
| 575 |
//*** YUV regions are not necessarily next to each other. |
| 576 |
#define nTOP_Y 10 // r10 - r17 (8 GRFs) |
| 577 |
#define nTOP_U 18 // r18 - r25 (8 GRFs) |
| 578 |
#define nTOP_V 26 // r26 - r33 (8 GRFs) |
| 579 |
|
| 580 |
#define nBOT_Y 56 // r56 - r63 (8 GRFs) |
| 581 |
#define nBOT_U 64 // r64 - r71 (8 GRFs) |
| 582 |
#define nBOT_V 72 // r72 - r79 (8 GRFs) |
| 583 |
|
| 584 |
// Define temp space for any usages |
| 585 |
#define nTEMP0 34 |
| 586 |
#define nTEMP1 35 |
| 587 |
#define nTEMP2 36 |
| 588 |
#define nTEMP3 37 |
| 589 |
#define nTEMP4 38 |
| 590 |
#define nTEMP5 39 |
| 591 |
#define nTEMP6 40 |
| 592 |
#define nTEMP7 41 |
| 593 |
#define nTEMP8 42 |
| 594 |
#define nTEMP10 44 |
| 595 |
#define nTEMP12 46 |
| 596 |
#define nTEMP14 48 |
| 597 |
#define nTEMP16 50 |
| 598 |
#define nTEMP17 51 |
| 599 |
#define nTEMP18 52 |
| 600 |
|
| 601 |
#define nTEMP24 58 |
| 602 |
|
| 603 |
// Common region 1 |
| 604 |
.declare ubTOP_Y Base=REG(r,nTOP_Y) ElementSize=1 SrcRegion=REGION(16,1) DstRegion=<1> Type=ub |
| 605 |
.declare ubTOP_U Base=REG(r,nTOP_U) ElementSize=1 SrcRegion=REGION(8,1) DstRegion=<1> Type=ub |
| 606 |
.declare ubTOP_V Base=REG(r,nTOP_V) ElementSize=1 SrcRegion=REGION(8,1) DstRegion=<1> Type=ub |
| 607 |
|
| 608 |
.declare uwTOP_Y Base=REG(r,nTOP_Y) ElementSize=2 SrcRegion=REGION(16,1) DstRegion=<1> Type=uw |
| 609 |
.declare uwTOP_U Base=REG(r,nTOP_U) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw |
| 610 |
.declare uwTOP_V Base=REG(r,nTOP_V) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw |
| 611 |
.declare ub2TOP_Y Base=REG(r,nTOP_Y) ElementSize=1 SrcRegion=REGION(16,2) DstRegion=<1> Type=ub |
| 612 |
.declare ub2TOP_U Base=REG(r,nTOP_U) ElementSize=1 SrcRegion=REGION(8,2) DstRegion=<1> Type=ub |
| 613 |
.declare ub2TOP_V Base=REG(r,nTOP_V) ElementSize=1 SrcRegion=REGION(8,2) DstRegion=<1> Type=ub |
| 614 |
|
| 615 |
.declare ub4TOP_Y Base=REG(r,nTOP_Y) ElementSize=1 SrcRegion=REGION(8,4) Type=ub |
| 616 |
.declare ub4TOP_U Base=REG(r,nTOP_U) ElementSize=1 SrcRegion=REGION(8,4) Type=ub |
| 617 |
.declare ub4TOP_V Base=REG(r,nTOP_V) ElementSize=1 SrcRegion=REGION(8,4) Type=ub |
| 618 |
|
| 619 |
.declare ubTOP_ARGB Base=REG(r,nTOP_Y) ElementSize=1 SrcRegion=REGION(8,4) Type=ub |
| 620 |
|
| 621 |
// Used by "send" instruction |
| 622 |
.declare udTOP_Y_IO Base=REG(r,nTOP_Y) ElementSize=4 SrcRegion=REGION(8,1) Type=ud |
| 623 |
.declare udTOP_U_IO Base=REG(r,nTOP_U) ElementSize=4 SrcRegion=REGION(8,1) Type=ud |
| 624 |
.declare udTOP_V_IO Base=REG(r,nTOP_V) ElementSize=4 SrcRegion=REGION(8,1) Type=ud |
| 625 |
|
| 626 |
// Common region 2 |
| 627 |
.declare ubBOT_Y Base=REG(r,nBOT_Y) ElementSize=1 SrcRegion=REGION(16,1) DstRegion=<1> Type=ub |
| 628 |
.declare ubBOT_U Base=REG(r,nBOT_U) ElementSize=1 SrcRegion=REGION(8,1) DstRegion=<1> Type=ub |
| 629 |
.declare ubBOT_V Base=REG(r,nBOT_V) ElementSize=1 SrcRegion=REGION(8,1) DstRegion=<1> Type=ub |
| 630 |
|
| 631 |
.declare uwBOT_Y Base=REG(r,nBOT_Y) ElementSize=2 SrcRegion=REGION(16,1) DstRegion=<1> Type=uw |
| 632 |
.declare uwBOT_U Base=REG(r,nBOT_U) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw |
| 633 |
.declare uwBOT_V Base=REG(r,nBOT_V) ElementSize=2 SrcRegion=REGION(8,1) DstRegion=<1> Type=uw |
| 634 |
.declare ub2BOT_Y Base=REG(r,nBOT_Y) ElementSize=1 SrcRegion=REGION(16,2) DstRegion=<1> Type=ub |
| 635 |
.declare ub2BOT_U Base=REG(r,nBOT_U) ElementSize=1 SrcRegion=REGION(8,2) DstRegion=<1> Type=ub |
| 636 |
.declare ub2BOT_V Base=REG(r,nBOT_V) ElementSize=1 SrcRegion=REGION(8,2) DstRegion=<1> Type=ub |
| 637 |
|
| 638 |
.declare ubBOT_ARGB Base=REG(r,nBOT_Y) ElementSize=1 SrcRegion=REGION(8,4) Type=ub |
| 639 |
|
| 640 |
// Used by "send" instruction |
| 641 |
.declare udBOT_Y_IO Base=REG(r,nBOT_Y) ElementSize=4 SrcRegion=REGION(8,1) Type=ud |
| 642 |
.declare udBOT_U_IO Base=REG(r,nBOT_U) ElementSize=4 SrcRegion=REGION(8,1) Type=ud |
| 643 |
.declare udBOT_V_IO Base=REG(r,nBOT_V) ElementSize=4 SrcRegion=REGION(8,1) Type=ud |
| 644 |
|
| 645 |
// End of common.inc |
| 646 |
|
| 647 |
#endif // COMMON_INC |