| 1 |
/* |
| 2 |
* Interpolation kernel for luminance motion compensation |
| 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: Interpolate_Y_8x8.asm |
| 41 |
// |
| 42 |
// Interpolation kernel for luminance motion compensation |
| 43 |
// |
| 44 |
// $Revision: 13 $ |
| 45 |
// $Date: 10/09/06 4:00p $ |
| 46 |
// |
| 47 |
|
| 48 |
|
| 49 |
//--------------------------------------------------------------- |
| 50 |
// In: pMV => Source address of MV |
| 51 |
// In: gMVX_FRAC<2;2,1>:w => MV fractional components |
| 52 |
// In: f0.1 (1) => If 1, vertical MV is integer |
| 53 |
// In: gpINTPY:uw => Destination address for interpolated result |
| 54 |
// In: Reference area staring from R43 |
| 55 |
// If horizontal/vertical MVs are all integer, 8x8 pixels are on R43~R44 (2 GRFs) |
| 56 |
// If only horz MV is integer, 8x13 pixels are on R43~R46 (4 GRFs) |
| 57 |
// If only vert MV is integer, 13x8 pixels are on R43~R46 (4 GRFs) |
| 58 |
// If no MVs are integer, 13x13 pixels are on R43~R49 (7 GRFs) |
| 59 |
//--------------------------------------------------------------- |
| 60 |
|
| 61 |
|
| 62 |
INTERLABEL(Interpolate_Y_8x8_Func): |
| 63 |
|
| 64 |
|
| 65 |
|
| 66 |
// Check whether MVX is integer MV |
| 67 |
and.z.f0.0 (1) null:w r[pMV,0]<0;1,0>:w 0x3:w |
| 68 |
(-f0.0) jmpi (1) INTERLABEL(Interpolate_Y_8x8_Func2) |
| 69 |
|
| 70 |
// TODO: remove this back-to-back read - huge latency.. |
| 71 |
mov (8) gubREF(6,2)<1> gubREF(3,0)<8;8,1> |
| 72 |
mov (8) gubREF(5,18)<1> gubREF(2,24)<8;8,1> {NoDDClr} |
| 73 |
mov (8) gubREF(5,2)<1> gubREF(2,16)<8;8,1> {NoDDChk} |
| 74 |
mov (8) gubREF(4,18)<1> gubREF(2,8)<8;8,1> {NoDDClr} |
| 75 |
mov (8) gubREF(4,2)<1> gubREF(2,0)<8;8,1> {NoDDChk} |
| 76 |
mov (8) gubREF(3,18)<1> gubREF(1,24)<8;8,1> {NoDDClr} |
| 77 |
mov (8) gubREF(3,2)<1> gubREF(1,16)<8;8,1> {NoDDChk} |
| 78 |
mov (8) gubREF(2,18)<1> gubREF(1,8)<8;8,1> {NoDDClr} |
| 79 |
mov (8) gubREF(2,2)<1> gubREF(1,0)<8;8,1> {NoDDChk} |
| 80 |
mov (8) gubREF(1,18)<1> gubREF(0,24)<8;8,1> {NoDDClr} |
| 81 |
mov (8) gubREF(1,2)<1> gubREF(0,16)<8;8,1> {NoDDChk} |
| 82 |
mov (8) gubREF(0,18)<1> gubREF(0,8)<8;8,1> |
| 83 |
mov (8) gubREF(0,2)<1> gubREF(0,0)<8;8,1> |
| 84 |
|
| 85 |
INTERLABEL(Interpolate_Y_8x8_Func2): |
| 86 |
|
| 87 |
// Compute the GRF address of the starting position of the reference area |
| 88 |
(-f0.1) mov (1) pREF:w nOFFSET_REF+2+nGRFWIB:w |
| 89 |
(f0.1) mov (1) pREF:w nOFFSET_REF+2:w |
| 90 |
mov (1) pRESULT:uw gpINTPY:uw |
| 91 |
|
| 92 |
/* |
| 93 |
* | | |
| 94 |
* - - 0 1 2 3 + - |
| 95 |
* 4 5 6 7 |
| 96 |
* 8 9 A B |
| 97 |
* C D E F |
| 98 |
* - - + - - - + - |
| 99 |
* | | |
| 100 |
*/ |
| 101 |
|
| 102 |
// Case 0 |
| 103 |
or.z.f0.1 (16) null:w gMVY_FRAC<0;1,0>:w gMVX_FRAC<0;1,0>:w |
| 104 |
(f0.1) mov (16) r[pRESULT]<1>:uw r[pREF]<16;8,1>:ub |
| 105 |
(f0.1) mov (16) r[pRESULT,nGRFWIB]<1>:uw r[pREF,nGRFWIB]<16;8,1>:ub |
| 106 |
(f0.1) mov (16) r[pRESULT,nGRFWIB*2]<1>:uw r[pREF,nGRFWIB*2]<16;8,1>:ub |
| 107 |
(f0.1) mov (16) r[pRESULT,nGRFWIB*3]<1>:uw r[pREF,nGRFWIB*3]<16;8,1>:ub |
| 108 |
(f0.1) jmpi INTERLABEL(Exit_Interpolate_Y_8x8) |
| 109 |
|
| 110 |
// Store all address registers |
| 111 |
mov (8) gpADDR<1>:w a0<8;8,1>:w |
| 112 |
|
| 113 |
mul.z.f0.0 (1) gW4:w gMVY_FRAC:w gMVX_FRAC:w |
| 114 |
add (1) pREF1:uw pREF0:uw nGRFWIB/2:uw |
| 115 |
and.nz.f0.1 (1) null gW4:w 1:w |
| 116 |
add (2) pREF2<1>:uw pREF0<2;2,1>:uw nGRFWIB:uw |
| 117 |
mov (4) gW0<1>:uw pREF0<4;4,1>:uw |
| 118 |
|
| 119 |
(f0.0) jmpi INTERLABEL(Interpolate_Y_H_8x8) |
| 120 |
(f0.1) jmpi INTERLABEL(Interpolate_Y_H_8x8) |
| 121 |
|
| 122 |
//----------------------------------------------------------------------- |
| 123 |
// CASE: A69BE (H/V interpolation) |
| 124 |
//----------------------------------------------------------------------- |
| 125 |
|
| 126 |
// Compute interim horizontal intepolation of 12 lines (not 9 lines) |
| 127 |
// add (1) pREF0<1>:ud pREF0<0;1,0>:ud 0xffeeffde:ud // (-18<<16)|(-34) |
| 128 |
add (1) pREF0<1>:uw pREF0<0;1,0>:uw -34:w |
| 129 |
add (1) pREF1<1>:uw pREF1<0;1,0>:uw -18:w {NoDDClr} |
| 130 |
mov (1) pRESD:ud nOFFSET_INTERIM3:ud {NoDDChk} |
| 131 |
|
| 132 |
// Check whether this position is 'A' |
| 133 |
cmp.e.f0.0 (1) null gW4:w 4:w |
| 134 |
|
| 135 |
$for(0;<6;2) { |
| 136 |
add (32) acc0<1>:w r[pREF,nGRFWIB*%1]<16;8,1>:ub r[pREF0,nGRFWIB*%1+5]<16;8,1>:ub {Compr} |
| 137 |
mac (32) acc0<1>:w r[pREF,nGRFWIB*%1+1]<16;8,1>:ub -5:w {Compr} |
| 138 |
mac (32) acc0<1>:w r[pREF,nGRFWIB*%1+2]<16;8,1>:ub 20:w {Compr} |
| 139 |
mac (32) acc0<1>:w r[pREF,nGRFWIB*%1+3]<16;8,1>:ub 20:w {Compr} |
| 140 |
mac (32) r[pRES,nGRFWIB*%1]<1>:w r[pREF,nGRFWIB*%1+4]<16;8,1>:ub -5:w {Compr} |
| 141 |
} |
| 142 |
// last line |
| 143 |
add (8) acc0<1>:w r[pREF,nGRFWIB*6]<8;8,1>:ub r[pREF,nGRFWIB*6+5]<8;8,1>:ub |
| 144 |
mac (8) acc0<1>:w r[pREF,nGRFWIB*6+1]<8;8,1>:ub -5:w |
| 145 |
mac (8) acc0<1>:w r[pREF,nGRFWIB*6+2]<8;8,1>:ub 20:w |
| 146 |
mac (8) acc0<1>:w r[pREF,nGRFWIB*6+3]<8;8,1>:ub 20:w |
| 147 |
mac (8) r[pRES,nGRFWIB*6]<1>:w r[pREF,nGRFWIB*6+4]<8;8,1>:ub -5:w |
| 148 |
|
| 149 |
// Compute interim/output vertical interpolation |
| 150 |
mov (1) pREF0:ud nOFFSET_INTERIM2:ud {NoDDClr} // set pREF0 and pREF1 at the same time |
| 151 |
mov (1) pREF2D:ud nOFFSET_INTERIM4:ud {NoDDChk,NoDDClr} // set pREF2 and pREF3 at the same time |
| 152 |
(f0.0) sel (1) pRES:uw gpINTPY:uw nOFFSET_INTERIM:uw {NoDDChk} // Case A vs. 69BE |
| 153 |
|
| 154 |
$for(0;<4;2) { |
| 155 |
add (32) acc0<1>:w r[pREF0,nGRFWIB*%1]<16;16,1>:w 512:w {Compr} |
| 156 |
mac (16) acc0<1>:w r[pREF2,nGRFWIB*%1]<8,1>:w -5:w |
| 157 |
mac (16) acc1<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:w -5:w |
| 158 |
mac (32) acc0<1>:w r[pREF0,nGRFWIB*%1+nGRFWIB]<16;16,1>:w 20:w {Compr} |
| 159 |
mac (16) acc0<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:w 20:w |
| 160 |
mac (16) acc1<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB+nGRFWIB]<8,1>:w 20:w |
| 161 |
mac (32) acc0<1>:w r[pREF0,(2+%1)*nGRFWIB]<16;16,1>:w -5:w {Compr} |
| 162 |
mac (16) acc0<1>:w r[pREF2,(2+%1)*nGRFWIB]<8,1>:w 1:w |
| 163 |
mac (16) acc1<1>:w r[pREF2,(2+%1)*nGRFWIB+nGRFWIB]<8,1>:w 1:w |
| 164 |
asr.sat (16) r[pRES,nGRFWIB*%1]<2>:ub acc0<16;16,1>:w 10:w |
| 165 |
asr.sat (16) r[pRES,nGRFWIB*%1+nGRFWIB]<2>:ub acc1<16;16,1>:w 10:w {SecHalf} |
| 166 |
} |
| 167 |
|
| 168 |
(f0.0) jmpi INTERLABEL(Return_Interpolate_Y_8x8) |
| 169 |
|
| 170 |
INTERLABEL(Interpolate_Y_H_8x8): |
| 171 |
|
| 172 |
cmp.e.f0.0 (1) null gMVX_FRAC:w 0:w |
| 173 |
cmp.e.f0.1 (1) null gMVY_FRAC:w 2:w |
| 174 |
(f0.0) jmpi INTERLABEL(Interpolate_Y_V_8x8) |
| 175 |
(f0.1) jmpi INTERLABEL(Interpolate_Y_V_8x8) |
| 176 |
|
| 177 |
//----------------------------------------------------------------------- |
| 178 |
// CASE: 123567DEF (H interpolation) |
| 179 |
//----------------------------------------------------------------------- |
| 180 |
|
| 181 |
add (4) pREF0<1>:uw gW0<4;4,1>:uw -2:w |
| 182 |
cmp.g.f0.0 (4) null:w gMVY_FRAC<0;1,0>:w 2:w |
| 183 |
cmp.e.f0.1 (1) null gMVX_FRAC:w 2:w |
| 184 |
(f0.0) add (4) pREF0<1>:uw pREF0<4;4,1>:uw nGRFWIB/2:uw |
| 185 |
|
| 186 |
cmp.e.f0.0 (1) null:w gMVY_FRAC<0;1,0>:w 0:w |
| 187 |
|
| 188 |
(f0.1) sel (1) pRES:uw gpINTPY:uw nOFFSET_INTERIM:uw // Case 26E vs. 1357DF |
| 189 |
|
| 190 |
// Compute interim/output horizontal interpolation |
| 191 |
$for(0;<4;2) { |
| 192 |
add (16) acc0<1>:w r[pREF0,nGRFWIB*%1]<8,1>:ub 16:w |
| 193 |
add (16) acc1<1>:w r[pREF0,nGRFWIB*%1+nGRFWIB]<8,1>:ub 16:w |
| 194 |
mac (16) acc0<1>:w r[pREF0,nGRFWIB*%1+1]<8,1>:ub -5:w |
| 195 |
mac (16) acc1<1>:w r[pREF0,nGRFWIB*%1+1+nGRFWIB]<8,1>:ub -5:w |
| 196 |
mac (16) acc0<1>:w r[pREF0,nGRFWIB*%1+2]<8,1>:ub 20:w |
| 197 |
mac (16) acc1<1>:w r[pREF0,nGRFWIB*%1+2+nGRFWIB]<8,1>:ub 20:w |
| 198 |
mac (16) acc0<1>:w r[pREF0,nGRFWIB*%1+3]<8,1>:ub 20:w |
| 199 |
mac (16) acc1<1>:w r[pREF0,nGRFWIB*%1+3+nGRFWIB]<8,1>:ub 20:w |
| 200 |
mac (16) acc0<1>:w r[pREF0,nGRFWIB*%1+4]<8,1>:ub -5:w |
| 201 |
mac (16) acc1<1>:w r[pREF0,nGRFWIB*%1+4+nGRFWIB]<8,1>:ub -5:w |
| 202 |
mac (16) acc0<1>:w r[pREF0,nGRFWIB*%1+5]<8,1>:ub 1:w |
| 203 |
mac (16) acc1<1>:w r[pREF0,nGRFWIB*%1+5+nGRFWIB]<8,1>:ub 1:w |
| 204 |
asr.sat (16) r[pRES,nGRFWIB*%1]<2>:ub acc0<16;16,1>:w 5:w |
| 205 |
asr.sat (16) r[pRES,nGRFWIB*%1+nGRFWIB]<2>:ub acc1<16;16,1>:w 5:w {SecHalf} |
| 206 |
} |
| 207 |
|
| 208 |
(-f0.1) jmpi INTERLABEL(Interpolate_Y_V_8x8) |
| 209 |
(-f0.0) jmpi INTERLABEL(Average_8x8) |
| 210 |
|
| 211 |
jmpi INTERLABEL(Return_Interpolate_Y_8x8) |
| 212 |
|
| 213 |
INTERLABEL(Interpolate_Y_V_8x8): |
| 214 |
|
| 215 |
cmp.e.f0.0 (1) null gMVY_FRAC:w 0:w |
| 216 |
(f0.0) jmpi INTERLABEL(Interpolate_Y_I_8x8) |
| 217 |
|
| 218 |
//----------------------------------------------------------------------- |
| 219 |
// CASE: 48C59D7BF (V interpolation) |
| 220 |
//----------------------------------------------------------------------- |
| 221 |
|
| 222 |
mov (2) pREF0<1>:uw gW0<4;2,2>:uw {NoDDClr} |
| 223 |
mov (2) pREF2<1>:uw gW1<2;2,1>:uw {NoDDChk,NoDDClr} |
| 224 |
mov (1) pRES:uw gpINTPY:uw {NoDDChk} |
| 225 |
|
| 226 |
cmp.g.f0.1 (4) null:w gMVX_FRAC<0;1,0>:w 2:w |
| 227 |
cmp.e.f0.0 (1) null:w gMVX_FRAC<0;1,0>:w 0:w |
| 228 |
(f0.1) add (4) pREF0<1>:uw pREF0<4;4,1>:uw 1:uw |
| 229 |
|
| 230 |
cmp.e.f0.1 (1) null gMVY_FRAC:w 2:w |
| 231 |
(-f0.0) jmpi INTERLABEL(VFILTER_8x8) |
| 232 |
(-f0.1) mov (1) pRES:uw nOFFSET_INTERIM:uw |
| 233 |
|
| 234 |
INTERLABEL(VFILTER_8x8): |
| 235 |
|
| 236 |
// Compute interim/output vertical interpolation |
| 237 |
$for(0;<4;2) { |
| 238 |
add (32) acc0<1>:w r[pREF0,nGRFWIB*%1-nGRFWIB]<16;8,1>:ub 16:w {Compr} |
| 239 |
mac (16) acc0<1>:w r[pREF2,nGRFWIB*%1-nGRFWIB]<8,1>:ub -5:w |
| 240 |
mac (16) acc1<1>:w r[pREF2,nGRFWIB*%1]<8,1>:ub -5:w |
| 241 |
mac (32) acc0<1>:w r[pREF0,nGRFWIB*%1]<16;8,1>:ub 20:w {Compr} |
| 242 |
mac (16) acc0<1>:w r[pREF2,nGRFWIB*%1]<8,1>:ub 20:w |
| 243 |
mac (16) acc1<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:ub 20:w |
| 244 |
mac (32) acc0<1>:w r[pREF0,nGRFWIB*%1+nGRFWIB]<16;8,1>:ub -5:w {Compr} |
| 245 |
mac (16) acc0<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:ub 1:w |
| 246 |
mac (16) acc1<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB+nGRFWIB]<8,1>:ub 1:w |
| 247 |
asr.sat (16) r[pRES,nGRFWIB*%1]<2>:ub acc0<16;16,1>:w 5:w |
| 248 |
asr.sat (16) r[pRES,nGRFWIB*%1+nGRFWIB]<2>:ub acc1<16;16,1>:w 5:w {SecHalf} |
| 249 |
} |
| 250 |
|
| 251 |
(-f0.0) jmpi INTERLABEL(Average_8x8) |
| 252 |
(f0.1) jmpi INTERLABEL(Return_Interpolate_Y_8x8) |
| 253 |
|
| 254 |
INTERLABEL(Interpolate_Y_I_8x8): |
| 255 |
|
| 256 |
//----------------------------------------------------------------------- |
| 257 |
// CASE: 134C (Integer position) |
| 258 |
//----------------------------------------------------------------------- |
| 259 |
|
| 260 |
mov (2) pREF0<1>:uw gW0<2;2,1>:uw {NoDDClr} |
| 261 |
|
| 262 |
mov (1) pRES:uw gpINTPY:uw {NoDDChk} |
| 263 |
|
| 264 |
cmp.e.f0.0 (2) null:w gMVX_FRAC<0;1,0>:w 3:w |
| 265 |
cmp.e.f0.1 (2) null:w gMVY_FRAC<0;1,0>:w 3:w |
| 266 |
(f0.0) add (2) pREF0<1>:uw pREF0<2;2,1>:uw 1:uw |
| 267 |
(f0.1) add (2) pREF0<1>:uw pREF0<2;2,1>:uw nGRFWIB/2:uw |
| 268 |
|
| 269 |
mov (16) r[pRES]<1>:uw r[pREF0]<8,1>:ub |
| 270 |
mov (16) r[pRES,nGRFWIB]<1>:uw r[pREF0,nGRFWIB]<8,1>:ub |
| 271 |
mov (16) r[pRES,nGRFWIB*2]<1>:uw r[pREF0,nGRFWIB*2]<8,1>:ub |
| 272 |
mov (16) r[pRES,nGRFWIB*3]<1>:uw r[pREF0,nGRFWIB*3]<8,1>:ub |
| 273 |
|
| 274 |
INTERLABEL(Average_8x8): |
| 275 |
|
| 276 |
//----------------------------------------------------------------------- |
| 277 |
// CASE: 13456789BCDEF (Average) |
| 278 |
//----------------------------------------------------------------------- |
| 279 |
|
| 280 |
// Average two interim results |
| 281 |
avg.sat (16) r[pRES,0]<2>:ub r[pRES,0]<32;16,2>:ub gubINTERIM_BUF(0) |
| 282 |
avg.sat (16) r[pRES,nGRFWIB]<2>:ub r[pRES,nGRFWIB]<32;16,2>:ub gubINTERIM_BUF(1) |
| 283 |
avg.sat (16) r[pRES,nGRFWIB*2]<2>:ub r[pRES,nGRFWIB*2]<32;16,2>:ub gubINTERIM_BUF(2) |
| 284 |
avg.sat (16) r[pRES,nGRFWIB*3]<2>:ub r[pRES,nGRFWIB*3]<32;16,2>:ub gubINTERIM_BUF(3) |
| 285 |
|
| 286 |
INTERLABEL(Return_Interpolate_Y_8x8): |
| 287 |
// Restore all address registers |
| 288 |
mov (8) a0<1>:w gpADDR<8;8,1>:w |
| 289 |
|
| 290 |
INTERLABEL(Exit_Interpolate_Y_8x8): |
| 291 |
|
| 292 |
// end of file |