| 1 |
/* |
| 2 |
* Intra predict 8X8 luma block |
| 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 |
#if !defined(__INTRA_PRED_8X8_Y__) // Make sure this is only included once |
| 41 |
#define __INTRA_PRED_8X8_Y__ |
| 42 |
|
| 43 |
// Module name: intra_Pred_8X8_Y.asm |
| 44 |
// |
| 45 |
// Intra predict 8X8 luma block |
| 46 |
// |
| 47 |
//-------------------------------------------------------------------------- |
| 48 |
// Input data: |
| 49 |
// |
| 50 |
// REF_TOP: Top reference data stored in BYTE with p[-1,-1] at REF_TOP(0,-1), p[-1,-1] and [15,-1] adjusted |
| 51 |
// REF_LEFT: Left reference data stored in BYTE with p[-1,0] at REF_LEFT(0,2), REF_LEFT(0,1) (p[-1,-1]) adjusted |
| 52 |
// PRED_MODE: Intra prediction mode stored in 4 LSBs |
| 53 |
// INTRA_PRED_AVAIL: Top/Left available flag, (Bit0: Left, Bit1: Top) |
| 54 |
// |
| 55 |
// Output data: |
| 56 |
// |
| 57 |
// REG_INTRA_8X8_PRED: Predicted 8X8 block data |
| 58 |
//-------------------------------------------------------------------------- |
| 59 |
|
| 60 |
#define INTRA_REF REG_INTRA_TEMP_1 |
| 61 |
#define REF_TMP REG_INTRA_TEMP_2 |
| 62 |
|
| 63 |
intra_Pred_8x8_Y: |
| 64 |
|
| 65 |
// Reference sample filtering |
| 66 |
// |
| 67 |
// Set up boundary pixels for unified filtering |
| 68 |
mov (1) REF_TOP(0,16)<1> REF_TOP(0,15)REGION(1,0) // p[16,-1] = p[15,-1] |
| 69 |
mov (8) REF_LEFT(0,2+8)<1> REF_LEFT(0,2+7)REGION(1,0) // p[-1,8] = p[-1,7] |
| 70 |
|
| 71 |
// Top reference sample filtering (!!Consider instruction compression later) |
| 72 |
add (16) acc0<1>:w REF_TOP(0,-1)REGION(16,1) 2:w // p[x-1,-1]+2 |
| 73 |
mac (16) acc0<1>:w REF_TOP(0)REGION(16,1) 2:w // p[x-1,-1]+2*p[x,-1]+2 |
| 74 |
mac (16) acc0<1>:w REF_TOP(0,1)REGION(16,1) 1:w // p[x-1,-1]+2*p[x,-1]+p[x+1,-1]+2 |
| 75 |
shr (16) REF_TMP<1>:w acc0:w 2:w // (p[x-1,-1]+2*p[x,-1]+p[x+1,-1]+2)>>2 |
| 76 |
|
| 77 |
// Left reference sample filtering |
| 78 |
add (16) acc0<1>:w REF_LEFT(0)REGION(16,1) 2:w // p[-1,y-1]+2 |
| 79 |
mac (16) acc0<1>:w REF_LEFT(0,1)REGION(16,1) 2:w // p[-1,y-1]+2*p[-1,y]+2 |
| 80 |
mac (16) acc0<1>:w REF_LEFT(0,2)REGION(16,1) 1:w // p[-1,y-1]+2*p[-1,y]+p[-1,y+1]+2 |
| 81 |
shr (16) INTRA_REF<1>:w acc0:w 2:w // (p[-1,y-1]+2*p[-1,y]+p[-1,y+1]+2)>>2 |
| 82 |
|
| 83 |
// Re-assign filtered reference samples |
| 84 |
mov (16) REF_TOP(0)<1> REF_TMP<32;16,2>:ub // p'[x,-1], x=0...15 |
| 85 |
mov (8) REF_LEFT(0)<1> INTRA_REF.2<16;8,2>:ub // p'[-1,y], y=0...7 |
| 86 |
mov (1) REF_TOP(0,-1)<1> INTRA_REF<0;1,0>:ub // p'[-1,-1] |
| 87 |
|
| 88 |
// Select intra_8x8 prediction mode |
| 89 |
// |
| 90 |
and (1) PINTRAPRED_Y<1>:w PRED_MODE<0;1,0>:w 0x0F:w |
| 91 |
// WA for "jmpi" restriction |
| 92 |
mov (1) REG_INTRA_TEMP_1<1>:ud r[PINTRAPRED_Y, INTRA_8X8_OFFSET]:ub |
| 93 |
jmpi (1) REG_INTRA_TEMP_1<0;1,0>:d |
| 94 |
|
| 95 |
// Mode 0 |
| 96 |
#define PTMP a0.6 |
| 97 |
#define PTMP_D a0.3 |
| 98 |
INTRA_8X8_VERTICAL: |
| 99 |
$for(0,0; <4; 1,32) { |
| 100 |
add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PERROR,%2]<16;16,1>:w REF_TOP(0)<0;8,1> |
| 101 |
} |
| 102 |
RETURN |
| 103 |
|
| 104 |
// Mode 1 |
| 105 |
INTRA_8X8_HORIZONTAL: |
| 106 |
$for(0,0; <8; 2,32) { |
| 107 |
add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PERROR,%2]<16;16,1>:w REF_LEFT(0,%1)<1;8,0> |
| 108 |
} |
| 109 |
RETURN |
| 110 |
|
| 111 |
// Mode 2 |
| 112 |
INTRA_8X8_DC: |
| 113 |
// Rearrange reference samples for unified DC prediction code |
| 114 |
// |
| 115 |
and.nz.f0.0 (16) NULLREG INTRA_PRED_AVAIL<0;1,0>:w 2:w // Top macroblock available for intra prediction? |
| 116 |
and.nz.f0.1 (8) NULLREG INTRA_PRED_AVAIL<0;1,0>:w 1:w // Left macroblock available for intra prediction? |
| 117 |
(-f0.0.any16h) mov (16) REF_TOP_W(0)<1> 0x8080:uw |
| 118 |
(-f0.1.any8h) mov (8) REF_LEFT(0)<1> REF_TOP(0)REGION(8,1) |
| 119 |
(-f0.0.any8h) mov (8) REF_TOP(0)<1> REF_LEFT(0)REGION(8,1) |
| 120 |
|
| 121 |
// Perform DC prediction |
| 122 |
// |
| 123 |
add (8) PRED_YW(15)<1> REF_TOP(0)REGION(8,1) REF_LEFT(0)REGION(8,1) |
| 124 |
add (4) PRED_YW(15)<1> PRED_YW(15)REGION(4,1) PRED_YW(15,4)REGION(4,1) |
| 125 |
add (2) PRED_YW(15)<1> PRED_YW(15)REGION(2,1) PRED_YW(15,2)REGION(2,1) |
| 126 |
add (16) acc0<1>:w PRED_YW(15)REGION(1,0) PRED_YW(15,1)REGION(1,0) |
| 127 |
add (16) acc0<1>:w acc0:w 8:w |
| 128 |
shr (16) REG_INTRA_TEMP_0<1>:w acc0:w 4:w |
| 129 |
|
| 130 |
// Add error block |
| 131 |
$for(0,0; <4; 1,32) { |
| 132 |
add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PERROR,%2]<16;16,1>:w REG_INTRA_TEMP_0<16;16,1>:w |
| 133 |
} |
| 134 |
RETURN |
| 135 |
|
| 136 |
// Mode 3 |
| 137 |
INTRA_8X8_DIAG_DOWN_LEFT: |
| 138 |
mov (8) REF_TOP(0,16)<1> REF_TOP(0,15)REGION(8,1) // p[16,-1] = p[15,-1] |
| 139 |
add (16) acc0<1>:w REF_TOP(0,2)REGION(16,1) 2:w // p[x+2]+2 |
| 140 |
mac (16) acc0<1>:w REF_TOP(0,1)REGION(16,1) 2:w // 2*p[x+1]+p[x+2]+2 |
| 141 |
mac (16) acc0<1>:w REF_TOP(0)REGION(16,1) 1:w // p[x]+2*p[x+1]+p[x+2]+2 |
| 142 |
shr (16) REG_INTRA_TEMP_0<1>:w acc0<16;16,1>:w 2:w // (p[x]+2*p[x+1]+p[x+2]+2)>>2 |
| 143 |
|
| 144 |
// Add error block |
| 145 |
$for(0,0; <8; 2,32) { |
| 146 |
add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PERROR,%2]<16;16,1>:w REG_INTRA_TEMP_0.%1<1;8,1>:w |
| 147 |
} |
| 148 |
RETURN |
| 149 |
|
| 150 |
// Mode 4 |
| 151 |
INTRA_8X8_DIAG_DOWN_RIGHT: |
| 152 |
#define INTRA_REF REG_INTRA_TEMP_1 |
| 153 |
#define REF_TMP REG_INTRA_TEMP_2 |
| 154 |
|
| 155 |
// Set inverse shift count |
| 156 |
shl (4) REF_TMP<1>:ud REF_LEFT_D(0,1)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order bottom 4 pixels of left ref. |
| 157 |
shl (4) REF_TMP.4<1>:ud REF_LEFT_D(0)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order top 4 pixels of left ref. |
| 158 |
mov (8) INTRA_REF<1>:ub REF_TMP.3<32;8,4>:ub |
| 159 |
mov (16) INTRA_REF.8<1>:ub REF_TOP(0,-1)REGION(16,1) // INTRA_REF holds all reference data |
| 160 |
|
| 161 |
add (16) acc0<1>:w INTRA_REF.2<16;16,1>:ub 2:w // p[x+2]+2 |
| 162 |
mac (16) acc0<1>:w INTRA_REF.1<16;16,1>:ub 2:w // 2*p[x+1]+p[x+2]+2 |
| 163 |
mac (16) acc0<1>:w INTRA_REF<16;16,1>:ub 1:w // p[x]+2*p[x+1]+p[x+2]+2 |
| 164 |
shr (16) INTRA_REF<1>:w acc0<16;16,1>:w 2:w // (p[x]+2*p[x+1]+p[x+2]+2)>>2 |
| 165 |
|
| 166 |
// Store data in reversed order |
| 167 |
add (2) PBWDCOPY_8<1>:w INV_TRANS48<2;2,1>:b INTRA_TEMP_1*GRFWIB:w // Must match with INTRA_REF |
| 168 |
|
| 169 |
// Add error block |
| 170 |
$for(0,96; <8; 2,-32) { |
| 171 |
add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PBWDCOPY_8,%1*2]<8,1>:w r[PERROR,%2]<16;16,1>:w |
| 172 |
} |
| 173 |
RETURN |
| 174 |
|
| 175 |
// Mode 5 |
| 176 |
INTRA_8X8_VERT_RIGHT: |
| 177 |
#define INTRA_REF REG_INTRA_TEMP_1 |
| 178 |
#define REF_TMP REG_INTRA_TEMP_2 |
| 179 |
#define REF_TMP1 REG_INTRA_TEMP_3 |
| 180 |
|
| 181 |
// Set inverse shift count |
| 182 |
shl (4) REF_TMP<1>:ud REF_LEFT_D(0,1)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order bottom 4 pixels of left ref. |
| 183 |
shl (4) REF_TMP.4<1>:ud REF_LEFT_D(0)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order top 4 pixels of left ref. |
| 184 |
mov (8) INTRA_REF<1>:ub REF_TMP.3<32;8,4>:ub |
| 185 |
mov (16) INTRA_REF.8<1>:ub REF_TOP(0,-1)REGION(16,1) // INTRA_REF holds all reference data |
| 186 |
|
| 187 |
// Even rows |
| 188 |
avg (16) PRED_YW(14)<1> INTRA_REF.8<16;16,1> INTRA_REF.9<16;16,1> // avg(p[x-1],p[x]) |
| 189 |
// Odd rows |
| 190 |
add (16) acc0<1>:w INTRA_REF.3<16;16,1>:ub 2:w // p[x]+2 |
| 191 |
mac (16) acc0<1>:w INTRA_REF.2<16;16,1>:ub 2:w // 2*p[x-1]+p[x]+2 |
| 192 |
mac (16) acc0<1>:w INTRA_REF.1<16;16,1>:ub 1:w // p[x-2]+2*p[x-1]+p[x]+2 |
| 193 |
shr (16) REF_TMP<1>:w acc0:w 2:w // (p[x-2]+2*p[x-1]+p[x]+2)>>2 |
| 194 |
|
| 195 |
mov (8) INTRA_REF<1>:ub REF_TMP<16;8,2>:ub // Keep zVR = -1,-2,-3,-4,-5,-6,-7 sequencially |
| 196 |
mov (8) INTRA_REF.6<2>:ub REF_TMP.12<16;8,2>:ub // Keep zVR = -1,1,3,5,7,9,11,13 at even byte |
| 197 |
mov (8) INTRA_REF.7<2>:ub PRED_Y(14)REGION(8,2) // Combining zVR = 0,2,4,6,8,10,12,14 at odd byte |
| 198 |
|
| 199 |
add (2) PBWDCOPY_8<1>:w INV_TRANS8<2;2,1>:b INTRA_TEMP_1*GRFWIB:w // Must match with INTRA_REF |
| 200 |
|
| 201 |
// Add error block |
| 202 |
$for(0,96; <8; 2,-32) { |
| 203 |
add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PBWDCOPY_8,%1]<8,2>:ub r[PERROR,%2]<16;16,1>:w |
| 204 |
} |
| 205 |
RETURN |
| 206 |
|
| 207 |
// Mode 6 |
| 208 |
INTRA_8X8_HOR_DOWN: |
| 209 |
// Set inverse shift count |
| 210 |
shl (4) REF_TMP<1>:ud REF_LEFT_D(0,1)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order bottom 4 pixels of left ref. |
| 211 |
shl (4) REF_TMP.4<1>:ud REF_LEFT_D(0)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order top 4 pixels of left ref. |
| 212 |
mov (8) INTRA_REF<1>:ub REF_TMP.3<16;4,4>:ub |
| 213 |
mov (16) INTRA_REF.8<1>:ub REF_TOP(0,-1)REGION(16,1) // INTRA_REF holds all reference data |
| 214 |
|
| 215 |
// Odd pixels |
| 216 |
add (16) acc0<1>:w INTRA_REF.2<16;16,1>:ub 2:w // p[y]+2 |
| 217 |
mac (16) acc0<1>:w INTRA_REF.1<16;16,1>:ub 2:w // 2*p[y-1]+p[y]+2 |
| 218 |
mac (16) acc0<1>:w INTRA_REF.0<16;16,1>:ub 1:w // p[y-2]+2*p[y-1]+p[y]+2 |
| 219 |
shr (16) PRED_YW(14)<1> acc0:w 2:w // (p[y-2]+2*p[y-1]+p[y]+2)>>2 |
| 220 |
// Even pixels |
| 221 |
avg (16) INTRA_REF<1>:w INTRA_REF<16;16,1>:ub INTRA_REF.1<16;16,1>:ub // avg(p[y-1],p[y]) |
| 222 |
|
| 223 |
mov (8) INTRA_REF.1<2>:ub PRED_Y(14)REGION(8,2) // Combining odd pixels to form byte type |
| 224 |
mov (8) INTRA_REF.16<1>:ub PRED_Y(14,16)REGION(8,2) // Keep zVR = -2,-3,-4,-5,-6,-7 unchanged |
| 225 |
// Now INTRA_REF.0 - INTRA_REF.21 contain predicted data |
| 226 |
|
| 227 |
add (2) PBWDCOPY_8<1>:w INV_TRANS48<2;2,1>:b INTRA_TEMP_1*GRFWIB:w // Must match with INTRA_REF |
| 228 |
|
| 229 |
// Add error block |
| 230 |
$for(0,96; <13; 4,-32) { |
| 231 |
add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PBWDCOPY_8,%1]<8,1>:ub r[PERROR,%2]<16;16,1>:w |
| 232 |
} |
| 233 |
RETURN |
| 234 |
|
| 235 |
// Mode 7 |
| 236 |
INTRA_8X8_VERT_LEFT: |
| 237 |
// Even rows |
| 238 |
avg (16) PRED_YW(14)<1> REF_TOP(0)REGION(16,1) REF_TOP(0,1)REGION(16,1) // avg(p[x],p[x+1]) |
| 239 |
// Odd rows |
| 240 |
add (16) acc0<1>:w REF_TOP(0,2)REGION(16,1) 2:w // p[x+2]+2 |
| 241 |
mac (16) acc0<1>:w REF_TOP(0,1)REGION(16,1) 2:w // 2*p[x+1]+p[x+2]+2 |
| 242 |
mac (16) acc0<1>:w REF_TOP(0)REGION(16,1) 1:w // p[x]+2*p[x+1]+p[x+2]+2 |
| 243 |
shr (16) PRED_YW(15)<1> acc0<1;8,1>:w 2:w // (p[x]+2*p[x+1]+p[x+2]+2)>>2 |
| 244 |
|
| 245 |
// Add error block |
| 246 |
$for(0,0; <4; 1,32) { |
| 247 |
add.sat (16) r[PPREDBUF_Y,%2]<2>:ub PRED_YW(14,%1)<16;8,1> r[PERROR,%2]<16;16,1>:w |
| 248 |
} |
| 249 |
RETURN |
| 250 |
|
| 251 |
// Mode 8 |
| 252 |
INTRA_8X8_HOR_UP: |
| 253 |
// Set extra left reference pixels for unified prediction |
| 254 |
mov (8) REF_LEFT(0,8)<1> REF_LEFT(0,7)REGION(1,0) // Copy p[-1,7] to p[-1,y],y=8...15 |
| 255 |
|
| 256 |
// Even pixels |
| 257 |
avg (16) PRED_YW(14)<1> REF_LEFT(0)REGION(16,1) REF_LEFT(0,1)REGION(16,1) // avg(p[y],p[y+1]) |
| 258 |
// Odd pixels |
| 259 |
add (16) acc0<1>:w REF_LEFT(0,2)REGION(16,1) 2:w // p[y+2]+2 |
| 260 |
mac (16) acc0<1>:w REF_LEFT(0,1)REGION(16,1) 2:w // 2*p[y+1]+p[y+2]+2 |
| 261 |
mac (16) acc0<1>:w REF_LEFT(0)REGION(16,1) 1:w // p[y]+2*p[y+1]+p[y+2]+2 |
| 262 |
shr (16) PRED_YW(15)<1> acc0<1;8,1>:w 2:w // (p[y]+2*p[y+1]+p[y+2]+2)>>2 |
| 263 |
|
| 264 |
// Merge even/odd pixels |
| 265 |
// The predicted data need to be stored in byte type (22 bytes are required) |
| 266 |
mov (16) PRED_Y(14,1)<2> PRED_Y(15)REGION(16,2) |
| 267 |
|
| 268 |
// Add error block |
| 269 |
$for(0,0; <4; 1,32) { |
| 270 |
add.sat (16) r[PPREDBUF_Y,%2]<2>:ub PRED_Y(14,%1*4)<2;8,1> r[PERROR,%2]<16;16,1>:w |
| 271 |
} |
| 272 |
RETURN |
| 273 |
|
| 274 |
// End of intra_Pred_8X8_Y |
| 275 |
|
| 276 |
#endif // !defined(__INTRA_PRED_8X8_Y__) |