Debian Codegraph @

src/shaders/post_processing/gen7/PA_DI_422CP.g4a

raw
1 /*
2 * Copyright 2000-2011 Intel Corporation All Rights Reserved
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the
6 * "Software"), to deal in the Software without restriction, including
7 * without limitation the rights to use, copy, modify, merge, publish,
8 * distribute, sub license, and/or sell copies of the Software, and to
9 * permit persons to whom the Software is furnished to do so, subject to
10 * the following conditions:
11 *
12 * The above copyright notice and this permission notice (including the
13 * next paragraph) shall be included in all copies or substantial portions
14 * of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
17 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
18 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
19 * IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
20 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
21 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
22 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
23 *
24 * This file was originally licensed under the following license
25 *
26 * Licensed under the Apache License, Version 2.0 (the "License");
27 * you may not use this file except in compliance with the License.
28 * You may obtain a copy of the License at
29 *
30 * http://www.apache.org/licenses/LICENSE-2.0
31 *
32 * Unless required by applicable law or agreed to in writing, software
33 * distributed under the License is distributed on an "AS IS" BASIS,
34 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
35 * See the License for the specific language governing permissions and
36 * limitations under the License.
37 */
38 // 87 // Total instruction count
39 // 1 // Total kernel count
40
41 .kernel PA_DI_422CP
42 .code
43
44
45
46 // FileName: DI.asm
47 // Author: Vivek Kumar
48 // Description: Tasks for DI only case (16x4 block)
49
50
51
52
53 // Module name: common.inc
54 //
55 // Common header file for all Video-Processing kernels
56 //
57
58 .default_execution_size (16)
59 .default_register_type :ub
60
61 .reg_count_total 128
62 .reg_count_payload 7
63
64 //========== Common constants ==========
65
66
67 //========== Macros ==========
68
69
70 //Fast Jump, For more details see "Set_Layer_N.asm"
71
72
73 //========== Defines ====================
74
75 //========== Static Parameters (Common To All) ==========
76 //r1
77
78
79 //r2
80
81 // e.g. byte0 byte1 byte2
82 // YUYV 0 1 3
83 // YVYU 0 3 1
84
85 //Color Pipe (IECP) parameters
86
87
88 //ByteCopy
89
90
91 //r4
92
93 // e.g. byte0 byte1 byte2
94 // YUYV 0 1 3
95 // YVYU 0 3 1
96
97
98 //========== Inline parameters (Common To All) ===========
99
100
101 //============== Binding Index Table===========
102 //Common between DNDI and DNUV
103
104
105 //================= Common Message Descriptor =====
106 // Message descriptor for thread spawning
107 // Message Descriptors
108 // = 000 0001 (min message len 1 ) 0,0000 (resp len 0 -add later)
109 // 0000,0000,0000
110 // 0001(Spawn a root thread),0001 (Root thread spawn thread)
111 // = 0x02000011
112 // Thread Spawner Message Descriptor
113
114
115 // Message descriptor for atomic operation add
116 // Message Descriptors
117 // = 000 0110 (min message len 6 ) 0,0000 (resp len 0 -add later)
118 // 1(header present)001,10(typed atomic operation)0(return enabled)0(slot group, low 8 bits),0111 (AOP_Add)
119 // 0000,0000 (Binding table index, added later)
120 // = 0x02000011
121
122 // Atomic Operation Add Message Descriptor
123
124
125 // Message descriptor for dataport media write
126 // Message Descriptors
127 // = 000 0001 (min message len 1 - add later) 00000 (resp len 0)
128 // 1 (header present 1) 0 1010 (media block write) 000000
129 // 00000000 (binding table index - set later)
130 // = 0x020A8000
131
132
133 // Message Length defines
134
135
136 // Response Length defines
137
138
139 // Block Width and Height Size defines
140
141
142 // Extended Message Descriptors
143
144
145 // Common message descriptors:
146
147
148 //===================== Math Function Control ===================================
149
150
151 //============ Message Registers ===============
152 // buf4 starts from r28
153
154
155 //#define mMSGHDR_EOT r43 // Dummy Message Register for EOT
156
157
158 .declare mubMSGPAYLOAD Base=r30 ElementSize=1 SrcRegion=<16;16,1> Type=ub
159 .declare muwMSGPAYLOAD Base=r30 ElementSize=2 SrcRegion=<16;16,1> Type=uw
160 .declare mudMSGPAYLOAD Base=r30 ElementSize=4 SrcRegion=<8;8,1> Type=ud
161 .declare mfMSGPAYLOAD Base=r30 ElementSize=4 SrcRegion=<8;8,1> Type=f
162
163 //=================== End of thread instruction ===========================
164
165
166 //=====================Pointers Used=====================================
167
168
169 //=======================================================================
170
171
172 //r11-r17
173 // Define temp space for any usages
174
175
176 // Common Buffers
177
178
179 // temp space for rotation
180
181 .declare fROBUF Base=r11.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=f
182
183 .declare udROBUF Base=r11.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=ud
184
185 .declare uwROBUF Base=r11.0 ElementSize=2 SrcRegion=<16;16,1> DstRegion=<1> Type=uw
186
187 .declare ubROBUF Base=r11.0 ElementSize=1 SrcRegion=<16;16,1> DstRegion=<1> Type=ub
188
189 .declare ub4ROBUF Base=r11.0 ElementSize=1 SrcRegion=<32;8,4> DstRegion=<4> Type=ub
190
191
192 // End of common.inc
193
194
195 // FileName: DNDI.inc
196 // Author: Vivek Kumar
197 // Description: Include file for DN, DI and DNDI
198 // Inputs: DI_ENABLE, DN_ENABLE, DN_PLANAR, DN_PACKED
199
200
201
202
203 // End of common.inc
204
205
206 //Interface:
207 //Static Parameters:
208 //r1
209
210
211 //====================== Binding table (Explicit To DNDI)=========================================
212
213
214 .declare mudMSGHDR_DNDI Base=r18 ElementSize=4 Type=ud
215 .declare mdMSGHDR_DNDI Base=r18 ElementSize=4 Type=d
216 .declare mwMSGHDR_DNDI Base=r18 ElementSize=2 Type=w
217
218
219 .declare mudMSGHDR_STMM Base=r20 ElementSize=4 Type=ud
220
221
222 .declare mudMSGHDR_HIST Base=r22 ElementSize=4 Type=ud
223
224
225 .declare mudMSGHDR_ENC_STATS Base=r24 ElementSize=4 Type=ud
226 .declare muwMSGHDR_ENC_STATS Base=r24 ElementSize=2 Type=uw
227 .declare mubMSGHDR_ENC_STATS Base=r24 ElementSize=1 Type=ub
228
229
230 .declare mudMSGHDR_DN_OUT Base=r31.0 ElementSize=4 Type=ud
231 .declare mdMSGHDR_DN_OUT Base=r31.0 ElementSize=4 Type=d
232 .declare mubMSGHDR_DN_OUT Base=r31.0 ElementSize=1 Type=ub
233
234
235 .declare mudMSGHDR_UVCOPY Base=r36 ElementSize=4 Type=ud
236 .declare mdMSGHDR_UVCOPY Base=r36 ElementSize=4 Type=d
237 .declare mudMSGHDR_UCOPY Base=r36 ElementSize=4 Type=ud
238 .declare mudMSGHDR_VCOPY Base=r38 ElementSize=4 Type=ud
239
240
241 .declare mudMSGHDR_DI_OUT1 Base=r18.0 ElementSize=4 Type=ud
242 .declare mubMSGHDR_DI_OUT1 Base=r18.0 ElementSize=1 Type=ub
243
244
245 .declare mudMSGHDR_DI_OUT2 Base=r23.0 ElementSize=4 Type=ud
246 .declare mubMSGHDR_DI_OUT2 Base=r23.0 ElementSize=1 Type=ub
247
248 //r45
249 //Use r45 as message header, so no need to "mov" the data.
250
251 .declare mudDN_Y_OUT Base=r45.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=ud
252
253 // Message response (Denoised & DI-ed pixels & statistics); Use buffer 5
254 .declare udDNDI_RESP Base=r46.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=ud
255 .declare uwDNDI_RESP Base=r46.0 ElementSize=2 SrcRegion=<16;16,1> DstRegion=<1> Type=uw
256 .declare ubDNDI_RESP Base=r46.0 ElementSize=1 SrcRegion=<16;16,1> DstRegion=<1> Type=ub
257
258 // Message response (UV Copy); Use buffer 5
259 .declare udDNDI_UV_RESP Base=r58.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=ud
260 .declare ubDNDI_UV_RESP Base=r58.0 ElementSize=1 SrcRegion=<16;16,1> DstRegion=<1> Type=ub
261
262 //Temp GRFs: For 42X to 422 Conversion
263 .declare uwDNDI_UVCOPY_TEMP Base=r10.0 ElementSize=2 SrcRegion=<16;16,1> DstRegion=<1> Type=uw //8 GRFs
264 .declare ubDNDI_UVCOPY_TEMP Base=r10.0 ElementSize=1 SrcRegion=<16;16,1> DstRegion=<1> Type=ub //8 GRFs
265 //---------------------------------------------------------------------------
266 // Message descriptors
267 //---------------------------------------------------------------------------
268 // Extended message descriptor
269 // Message descriptor for sampler read
270 // = 000 0010 (message len 2) 00000 (resp len - set later, 12 or 5 or 11)
271 // 1 (header present 1) 0 11 (SIMD32/64 mode)
272 // 1000 (message type) 0000 (DI state index)
273 // 00000000 (binding table index - set later)
274 // = 0x040b8000
275
276
277 // Attention: The Message Length is The Number of GRFs with Data Only, without the Header
278
279
280 //---------------------------------------------------------------------------
281 // VDI Return Data format
282 //---------------------------------------------------------------------------
283 // Defines for DI enabled
284
285
286 // Defines for DI disabled
287
288
289
290 // FileName: DNDI_Command.asm
291 // Author: Vivek Kumar
292 // Description: Sends a message to the VDI to process one DN (16x8) or DNDI (16x4) block
293
294 // Prepare the DNDI send command
295 mov (8) mudMSGHDR_DNDI(0)<1> r0.0<8;8,1>:ud // message header
296 mov (1) mwMSGHDR_DNDI(1,4)<1> r9.0<0;1,0>:w { NoDDClr } // horizontal origin // Do we need to add offset here? -vK
297 mov (1) mwMSGHDR_DNDI(1,12)<1> r9.1<0;1,0>:w { NoDDChk } // vertical origin // Can these 2 be combined? - vK
298
299 send (8) udDNDI_RESP(0)<1> r18 0x2 0x4AE8003:ud
300
301 // On Gen6, with VDI walker, use the XY pair returned rather than programmed above
302 // VDI_RETURNED_XY is ordered XY in case of walker enables and the same as programmed in case of walker disabled
303 mov (2) r9.0<1>:w uwDNDI_RESP(9,14)<2;2,1> // horizontal/Vertial origin in W.14 and W.15
304
305
306
307 // FileName: DI_STMM_Save.asm
308 // Author: Vivek Kumar
309 // Description: Saves DI STMM Data to statistics surface in case of DI enabled (for 16x4 block)
310
311 // Write STMM to memory
312 mov (8) mudMSGHDR_STMM(0)<1> r0.0<8;8,1>:ud // message header
313 mov (8) mudMSGHDR_STMM(1)<1> udDNDI_RESP(8,0) // Move STMM to MRF
314
315 shr (1) mudMSGHDR_STMM(0,0)<1> r9.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2
316 mov (1) mudMSGHDR_STMM(0,1)<1> r9.1<0;1,0>:w { NoDDClr, NoDDChk } // Y origin
317 mov (1) mudMSGHDR_STMM(0,2)<1> 0x30007:ud { NoDDChk } // block width and height (8x4)
318
319 send (8) null<1>:d r20 0x5 0x40A8021:ud
320
321
322
323 // FileName: DNDI_Enc_Stats_Save.asm
324 // Author: Vivek Kumar
325 // Description: Saves Encoder Statistics data to statistics surface in case of DI enabled (for 16x4 block)
326
327 // Write encoder statistics to memory
328 //Currently enable this only on Gen6 validation
329 mov (8) mudMSGHDR_ENC_STATS(1)<1> 0x0:ud // Init payload MRF
330 mov (8) mudMSGHDR_ENC_STATS(0)<1> r0.0<8;8,1>:ud // message header
331
332 shr (1) mudMSGHDR_ENC_STATS(0,0)<1> r9.0<0;1,0>:w 1:w { NoDDClr } //enable the flag after testing on si { NoDDClr } // X origin / 2
333 mul (1) acc0.1<1>:ud r9.1<0;1,0>:w 3:w // Y origin * 3
334 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
335 mov (1) mudMSGHDR_ENC_STATS(0,2)<1> 0x20007:ud { NoDDChk } //enable the flag after testing on si { NoDDChk } // block width and height (8x3)
336 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
337
338
339 //Data block for Encoder Statistics
340 //----------------------------------------------------
341 //| 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | Bytes
342 //----------------------------------------------------
343 //| BNE | MCNT | FCNT | TCNT | X | X | X | X |
344 //----------------------------------------------------
345 //| DcTpT | SVCM | DcBpT | DcTpB |
346 //----------------------------------------------------
347 //| SHCM | STAD | DcTcB | DcBpB |
348 //----------------------------------------------------
349 mov (1) mudMSGHDR_ENC_STATS(1,0)<1> udDNDI_RESP(9,1)<0;1,0> { NoDDClr } // Move encoder statistics to MRF
350 mov (2) mudMSGHDR_ENC_STATS(1,3)<2> udDNDI_RESP(9,3)<2;2,1> { NoDDClr, NoDDChk } // Move encoder statistics to MRF
351 mov (2) mudMSGHDR_ENC_STATS(1,2)<2> udDNDI_RESP(9,5)<2;2,1> { NoDDChk } // Move encoder statistics to MRF
352
353
354 send (8) null<1>:d r24 0x5 0x40A8021:ud
355
356
357
358 // FileName: DI_Save_422CP_16x4.asm
359 // Author: Vivek Kumar
360 // Description: Save two 16x4 blocks of DI output in 422 format to Color Pipe (IECP)
361
362
363 .declare mubMSGHDR_DI_OUT1_1 Base=r18.0 ElementSize=1 Type=ub
364
365
366 .declare mubMSGHDR_DI_OUT1_2 Base=r21.0 ElementSize=1 Type=ub
367
368
369 .declare mubMSGHDR_DI_OUT2_1 Base=r24.0 ElementSize=1 Type=ub
370
371
372 .declare mubMSGHDR_DI_OUT2_2 Base=r27.0 ElementSize=1 Type=ub
373
374
375 mov (8) r27.0<1>:ud r0.0<8;8,1>:ud
376 shl (1) r27.0<1>:ud r9.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled
377 mov (1) r27.1<1>:ud r9.1<0;1,0>:w { NoDDClr, NoDDChk } // Block origin
378 mov (1) r27.2<1>:ud 0x3000F:ud { NoDDClr, NoDDChk } // Block width and height (16x8)
379
380 //M0.3 - 0 - CP Enable, 1 - Area of Interest, 3:2 Message Format(TBD), 4:3 - Ignored, 31:5 CP state pointer
381 //Compose area-of-interest bit + color pipe state pointer
382 or (1) r27.3<1>:ud r2.4<0;1,0>:ud r9.26<0;1,0>:b { NoDDChk }
383
384 //prepare the message headers
385 mov (8) r18.0<1>:ud r27<8;8,1>:ud
386 mov (8) r24.0<1>:ud r27<8;8,1>:ud
387
388
389 // Pack 2nd field Y; First 8x4 block
390 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
391 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
392 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
393 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
394
395 // Pack 2nd field U, V; First 8x4 block
396 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
397 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
398
399 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
400 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
401 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
402 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
403
404 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
405 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
406
407
408 // Pack 2nd field Y; Second 8x4 block
409 mov (8) r21.0<1>:ud r18.0<8;8,1>:ud
410 add (1) r21.0<1>:ud r21.0<0;1,0>:w 0x10:w
411
412 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
413 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
414 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
415 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
416
417 // Pack 2nd field U, V; Second 8x4 block
418 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
419 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
420
421 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
422 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
423 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
424 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
425
426 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
427 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
428
429 send (8) null<1>:d r18.0 0x5 0x60A801B:ud
430 send (8) null<1>:d r21.0 0x5 0x60A801B:ud
431
432 // Pack 1st field Y; 1st 8x4 block
433 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
434 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
435 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
436 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
437
438 // Pack 1st field U,V; 1st 8x4 block
439 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
440 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
441
442 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
443 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
444 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
445 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
446
447 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
448 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
449
450 // Pack 1st field Y; 2nd 8x4 block
451 mov (8) r27.0<1>:ud r24.0<8;8,1>:ud
452 add (1) r27.0<1>:ud r27.0<0;1,0>:w 0x10:w
453
454 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
455 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
456 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
457 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
458
459 // Pack 1st field U, V; 2nd 8x4 block
460 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
461 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
462
463 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
464 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
465 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
466 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
467
468 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
469 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
470
471 send (8) null<1>:d r24.0 0x5 0x60A801E:ud
472 send (8) null<1>:d r27.0 0x5 0x60A801E:ud
473
474
475
476 //End of Thread message
477
478 mov (8) r127<1>:ud r0.0<8;8,1>:ud
479 send (1) null<1>:d r127 0x27 0x02000010
480
481
482 .end_code
483 .end_kernel