| 1 |
/* |
| 2 |
* All inter-prediction macroblock 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 |
// Kernel name: AVCMCInter.asm |
| 41 |
|
| 42 |
#ifdef INTERLABEL |
| 43 |
#undef INTERLABEL |
| 44 |
#endif |
| 45 |
|
| 46 |
#if defined(MBAFF) |
| 47 |
// < MBaff_Motion > |
| 48 |
#define INTERLABEL(x) x##_##MBF |
| 49 |
#elif defined(FIELD) |
| 50 |
// < FieldMB_Motion > |
| 51 |
#define INTERLABEL(x) x##_##FLD |
| 52 |
#else // FRAME |
| 53 |
// < FrameMB_Motion > |
| 54 |
#define INTERLABEL(x) x##_##FRM |
| 55 |
#endif |
| 56 |
// |
| 57 |
// Decoding an inter-prediction macroblock (conditional compile) |
| 58 |
// -DMBAFF : MBAff picture MB |
| 59 |
// -DFRAME : Frame picture MB |
| 60 |
// -DFIELD : Field picture MB |
| 61 |
// -DMBAFF -DMONO : MBAff mono picture MB |
| 62 |
// -DFRAME -DMONO : Frame mono picture MB |
| 63 |
// -DFIELD -DMONO : Field mono picture MB |
| 64 |
|
| 65 |
|
| 66 |
//#if !defined(__AVCMCInter__) // Make sure this is only included once |
| 67 |
//#define __AVCMCInter__ |
| 68 |
|
| 69 |
|
| 70 |
// TODO: header files need to be in sync with intra prediction |
| 71 |
#include "header.inc" |
| 72 |
#include "inter_Header.inc" |
| 73 |
|
| 74 |
// TODO: Kernel names for mono cases |
| 75 |
#if defined(MBAFF) |
| 76 |
.kernel MBAff_Motion |
| 77 |
MBAFF_MB: |
| 78 |
#elif defined(FIELD) |
| 79 |
.kernel FieldMB_Motion |
| 80 |
FIELD_MB: |
| 81 |
#else // Frame |
| 82 |
.kernel FrameMB_Motion |
| 83 |
FRAME_MB: |
| 84 |
#endif |
| 85 |
|
| 86 |
#ifdef _DEBUG |
| 87 |
// WA for FULSIM so we'll know which kernel is being debugged |
| 88 |
#if defined(MBAFF) |
| 89 |
mov (1) acc0:ud 0x0aaa55a5:ud |
| 90 |
#elif defined(FIELD) |
| 91 |
mov (1) acc0:ud 0x0baa55a5:ud |
| 92 |
#else // Frame |
| 93 |
mov (1) acc0:ud 0x0caa55a5:ud |
| 94 |
#endif |
| 95 |
#endif |
| 96 |
|
| 97 |
|
| 98 |
#ifdef SW_SCOREBOARD |
| 99 |
CALL(scoreboard_start_inter,1) |
| 100 |
#endif |
| 101 |
|
| 102 |
mov (8) gMSGSRC<1>:ud r0.0<8;8,1>:ud // Initialize message header payload with R0 |
| 103 |
|
| 104 |
and (1) gwMBTYPE<1> gMBTYPE:ub nMBTYPE_MASK:w // MB type |
| 105 |
shl (2) gX<1>:w gORIX<2;2,1>:ub 4:w // Convert MB origin to pixel unit |
| 106 |
|
| 107 |
// #include "process_inter16x16.asm" // Handle B_L0_16x16 case with zero MVs and weighted pred off. |
| 108 |
// In the case of B_L0_16x16 with zero MVs and weighted pred off, the kernel jumps to INTERLABEL(EXIT_LOOP). |
| 109 |
|
| 110 |
INTERLABEL(INIT_MBPARA): |
| 111 |
#include "initialize_MBPara.asm" |
| 112 |
|
| 113 |
|
| 114 |
//========================= BEGIN - LOOP_SUBMB =========================== |
| 115 |
mov (1) gLOOP_SUBMB:uw 0:uw // 0, 2, 4, 6 |
| 116 |
INTERLABEL(LOOP_SUBMB): |
| 117 |
|
| 118 |
//========================== BEGIN - LOOP_DIR ============================ |
| 119 |
// Prediction flag (gPREDFLAG - 0:Pred_L0, 1:Pred_L1, 2:BiPred) |
| 120 |
asr (1) gPREDFLAG:w gSUBMB_MODE:ub gLOOP_SUBMB:uw |
| 121 |
mov (1) gLOOP_DIR:uw 1:uw // 1, 0 |
| 122 |
and (1) gPREDFLAG:w gPREDFLAG:w 0x3:w |
| 123 |
INTERLABEL(LOOP_DIR): |
| 124 |
|
| 125 |
cmp.e.f0.0 (1) null:w gLOOP_DIR:w gPREDFLAG:w |
| 126 |
(f0.0) jmpi INTERLABEL(LOOP_DIR_CONTINUE) |
| 127 |
|
| 128 |
// Get binding table index |
| 129 |
// & reference picture parity (gREFPARITY - 0:top, 0x100:bottom, x:frame) |
| 130 |
// & address of interpolation result |
| 131 |
cmp.e.f0.1 (1) null:w gLOOP_DIR:w 1:w |
| 132 |
(f0.1) mov (1) gpINTP:ud nOFFSET_INTP0:ud {NoDDClr} // |
| 133 |
(f0.1) and (1) gBIDX:w r[pBIDX]:ub 0x7f:w {NoDDChk} // |
| 134 |
(-f0.1) mov (1) gpINTP:ud nOFFSET_INTP1:ud {NoDDClr} // |
| 135 |
(-f0.1) and (1) gBIDX:w r[pBIDX,4]:ub 0x7f:w {NoDDChk} // |
| 136 |
#if defined(MBAFF) || defined(FIELD) |
| 137 |
(f0.1) and (1) gREFPARITY:w r[pBIDX]:ub 0x80:w |
| 138 |
(-f0.1) and (1) gREFPARITY:w r[pBIDX,4]:ub 0x80:w |
| 139 |
shl (1) gREFPARITY:w gREFPARITY<0;1,0>:w 1:w |
| 140 |
#endif |
| 141 |
|
| 142 |
// Sub MB shape |
| 143 |
asr (1) gSHAPETEMP:w gSUBMB_SHAPE:ub gLOOP_SUBMB:w |
| 144 |
|
| 145 |
// Chroma MV adjustment & Set message descriptor for frame/field read |
| 146 |
#if defined(MBAFF) |
| 147 |
#include "chromaMVAdjust.asm" |
| 148 |
and.nz.f0.0 (1) null:uw gFIELDMBFLAG:ub nFIELDMB_MASK:uw |
| 149 |
(f0.0) add (1) gD0:ud gBIDX:uw nDWBRMSGDSC_SC_TF:ud |
| 150 |
(-f0.0) add (1) gMSGDSC_R:ud gBIDX:uw nDWBRMSGDSC_SC:ud |
| 151 |
(f0.0) add (1) gMSGDSC_R:ud gD0:ud gREFPARITY:uw |
| 152 |
#elif defined(FIELD) |
| 153 |
#include "chromaMVAdjust.asm" |
| 154 |
add (1) gMSGDSC_R:ud gBIDX:uw nDWBRMSGDSC_SC_TF:ud |
| 155 |
add (1) gMSGDSC_R:ud gMSGDSC_R:ud gREFPARITY:uw |
| 156 |
#else // FRAME |
| 157 |
add (1) gMSGDSC_R:ud gBIDX:uw nDWBRMSGDSC_SC:ud |
| 158 |
#endif |
| 159 |
|
| 160 |
and.nz.f0.1 (1) null:w gSHAPETEMP:w 3:w |
| 161 |
(f0.1) jmpi INTERLABEL(PROCESS4x4) |
| 162 |
|
| 163 |
//======================== BEGIN - PROCESS 8x8 =========================== |
| 164 |
|
| 165 |
// Reference block load |
| 166 |
#include "loadRef_Y_16x13.asm" |
| 167 |
#ifndef MONO |
| 168 |
#if defined(MBAFF) || defined(FIELD) |
| 169 |
add (1) r[pMV,2]:w r[pMV,2]:w gCHRMVADJ:w |
| 170 |
#endif |
| 171 |
#include "loadRef_C_10x5.asm" |
| 172 |
#endif |
| 173 |
|
| 174 |
// Interpolation |
| 175 |
//CALL_INTER(INTERLABEL(Interpolate_Y_8x8_Func), 1) |
| 176 |
#include "interpolate_Y_8x8.asm" |
| 177 |
#ifndef MONO |
| 178 |
//CALL_INTER(INTERLABEL(Interpolate_C_4x4_Func), 1) |
| 179 |
#include "interpolate_C_4x4.asm" |
| 180 |
#endif |
| 181 |
|
| 182 |
jmpi INTERLABEL(ROUND_SHIFT_C) |
| 183 |
//========================= END - PROCESS 8x8 ============================ |
| 184 |
|
| 185 |
//======================== BEGIN - LOOP_SUBMBPT ========================== |
| 186 |
INTERLABEL(PROCESS4x4): |
| 187 |
|
| 188 |
mov (1) gLOOP_SUBMBPT:uw 4:uw // 4, 3, 2, 1 |
| 189 |
INTERLABEL(LOOP_SUBMBPT): |
| 190 |
|
| 191 |
// Reference block load |
| 192 |
#include "loadRef_Y_16x9.asm" |
| 193 |
#ifndef MONO |
| 194 |
#if defined(MBAFF) || defined(FIELD) |
| 195 |
add (1) r[pMV,2]:w r[pMV,2]:w gCHRMVADJ:w |
| 196 |
#endif |
| 197 |
#include "loadRef_C_6x3.asm" |
| 198 |
#endif |
| 199 |
|
| 200 |
// Interpolation |
| 201 |
#include "interpolate_Y_4x4.asm" |
| 202 |
#ifndef MONO |
| 203 |
#include "interpolate_C_2x2.asm" |
| 204 |
#endif |
| 205 |
|
| 206 |
cmp.e.f0.0 (1) null:w gLOOP_SUBMBPT:uw 3:w |
| 207 |
add.z.f0.1 (1) gLOOP_SUBMBPT:uw gLOOP_SUBMBPT:uw -1:w |
| 208 |
add (1) pMV:w pMV:w 8:w |
| 209 |
(-f0.0) add (1) gpINTP:ud gpINTP:ud 0x00080008:ud // 8 & 8 |
| 210 |
(f0.0) add (1) gpINTP:ud gpINTP:ud 0x00180038:ud // 24 & 56 |
| 211 |
(-f0.1) jmpi INTERLABEL(LOOP_SUBMBPT) |
| 212 |
|
| 213 |
cmp.e.f0.1 null:w gLOOP_DIR:w 1:w |
| 214 |
add (1) pMV:w pMV:w -32:w |
| 215 |
(f0.1) mov (1) gpINTP:ud nOFFSET_INTP0:ud |
| 216 |
(-f0.1) mov (1) gpINTP:ud nOFFSET_INTP1:ud |
| 217 |
|
| 218 |
mov (1) pRESULT:uw gpINTPC:uw |
| 219 |
|
| 220 |
//========================= END - LOOP_SUBMBPT =========================== |
| 221 |
|
| 222 |
INTERLABEL(ROUND_SHIFT_C): |
| 223 |
|
| 224 |
#ifndef MONO |
| 225 |
#include "roundShift_C_4x4.asm" |
| 226 |
#endif |
| 227 |
|
| 228 |
INTERLABEL(LOOP_DIR_CONTINUE): |
| 229 |
|
| 230 |
add.nz.f0.1 (1) gLOOP_DIR:uw gLOOP_DIR:uw -1:w |
| 231 |
add (1) pMV:w pMV:w 4:w |
| 232 |
(-f0.1) jmpi INTERLABEL(LOOP_DIR) |
| 233 |
//=========================== END - LOOP_DIR ============================= |
| 234 |
|
| 235 |
INTERLABEL(Weighted_Prediction): |
| 236 |
#include "weightedPred.asm" |
| 237 |
|
| 238 |
and.z.f0.1 (16) null<1>:w gLOOP_SUBMB<0;1,0>:uw 2:w |
| 239 |
|
| 240 |
#include "recon_Y_8x8.asm" |
| 241 |
#ifndef MONO |
| 242 |
#include "recon_C_4x4.asm" |
| 243 |
|
| 244 |
(-f0.1) add (1) pERRORC:w pERRORC:w 48:w |
| 245 |
#endif |
| 246 |
|
| 247 |
cmp.e.f0.1 (1) null:w gLOOP_SUBMB:uw 6:w |
| 248 |
add (1) gLOOP_SUBMB:uw gLOOP_SUBMB:uw 2:w |
| 249 |
|
| 250 |
add (1) pWGT_BIDX:ud pWGT_BIDX:ud 0x00100001:ud // 12 & 1 |
| 251 |
add (1) pMV:w pMV:w gMVSTEP:w |
| 252 |
|
| 253 |
(-f0.1) jmpi INTERLABEL(LOOP_SUBMB) |
| 254 |
//========================== END - LOOP_SUBMB ============================ |
| 255 |
|
| 256 |
INTERLABEL(EXIT_LOOP): |
| 257 |
#include "writeRecon_YC.asm" |
| 258 |
|
| 259 |
#ifdef SW_SCOREBOARD |
| 260 |
wait n0:ud // Now wait for scoreboard to response |
| 261 |
#include "Soreboard_update.asm" // scorboard update function |
| 262 |
#else |
| 263 |
// Check for write commit first if SW scoreboard is disabled |
| 264 |
mov (1) gREG_WRITE_COMMIT_Y<1>:ud gREG_WRITE_COMMIT_Y<0;1,0>:ud // Make sure Y write is committed |
| 265 |
mov (1) gREG_WRITE_COMMIT_UV<1>:ud gREG_WRITE_COMMIT_UV<0;1,0>:ud // Make sure U/V write is committed |
| 266 |
#endif |
| 267 |
|
| 268 |
// Terminate the thread |
| 269 |
// |
| 270 |
END_THREAD |
| 271 |
|
| 272 |
|
| 273 |
//#include "Interpolate_Y_8x8_Func.asm" |
| 274 |
//#include "Interpolate_C_4x4_Func.asm" |
| 275 |
//#include "WeightedPred_Y_Func.asm" |
| 276 |
//#include "WeightedPred_C_Func.asm" |
| 277 |
|
| 278 |
|
| 279 |
.end_code |
| 280 |
|
| 281 |
.end_kernel |
| 282 |
|
| 283 |
|
| 284 |
//#endif // !defined(__AVCMCInter__) |