Details | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
5361 | serge | 1 | /* |
2 | * Copyright 2000-2011 Intel Corporation All Rights Reserved |
||
3 | * |
||
4 | * Licensed under the Apache License, Version 2.0 (the "License"); |
||
5 | * you may not use this file except in compliance with the License. |
||
6 | * You may obtain a copy of the License at |
||
7 | * |
||
8 | * http://www.apache.org/licenses/LICENSE-2.0 |
||
9 | * |
||
10 | * Unless required by applicable law or agreed to in writing, software |
||
11 | * distributed under the License is distributed on an "AS IS" BASIS, |
||
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
||
13 | * See the License for the specific language governing permissions and |
||
14 | * limitations under the License. |
||
15 | */ |
||
16 | // 87 // Total instruction count |
||
17 | // 1 // Total kernel count |
||
18 | |||
19 | .kernel PL_DI_422CP |
||
20 | .code |
||
21 | |||
22 | |||
23 | |||
24 | // FileName: DI.asm |
||
25 | // Author: Vivek Kumar |
||
26 | // Description: Tasks for DI only case (16x4 block) |
||
27 | |||
28 | |||
29 | |||
30 | |||
31 | // Module name: common.inc |
||
32 | // |
||
33 | // Common header file for all Video-Processing kernels |
||
34 | // |
||
35 | |||
36 | .default_execution_size (16) |
||
37 | .default_register_type :ub |
||
38 | |||
39 | .reg_count_total 128 |
||
40 | .reg_count_payload 7 |
||
41 | |||
42 | //========== Common constants ========== |
||
43 | |||
44 | |||
45 | //========== Macros ========== |
||
46 | |||
47 | |||
48 | //Fast Jump, For more details see "Set_Layer_N.asm" |
||
49 | |||
50 | |||
51 | //========== Defines ==================== |
||
52 | |||
53 | //========== Static Parameters (Common To All) ========== |
||
54 | //r1 |
||
55 | |||
56 | |||
57 | //r2 |
||
58 | |||
59 | // e.g. byte0 byte1 byte2 |
||
60 | // YUYV 0 1 3 |
||
61 | // YVYU 0 3 1 |
||
62 | |||
63 | //Color Pipe (IECP) parameters |
||
64 | |||
65 | |||
66 | //ByteCopy |
||
67 | |||
68 | |||
69 | //r4 |
||
70 | |||
71 | // e.g. byte0 byte1 byte2 |
||
72 | // YUYV 0 1 3 |
||
73 | // YVYU 0 3 1 |
||
74 | |||
75 | |||
76 | //========== Inline parameters (Common To All) =========== |
||
77 | |||
78 | |||
79 | //============== Binding Index Table=========== |
||
80 | //Common between DNDI and DNUV |
||
81 | |||
82 | |||
83 | //================= Common Message Descriptor ===== |
||
84 | // Message descriptor for thread spawning |
||
85 | // Message Descriptors |
||
86 | // = 000 0001 (min message len 1 ) 0,0000 (resp len 0 -add later) |
||
87 | // 0000,0000,0000 |
||
88 | // 0001(Spawn a root thread),0001 (Root thread spawn thread) |
||
89 | // = 0x02000011 |
||
90 | // Thread Spawner Message Descriptor |
||
91 | |||
92 | |||
93 | // Message descriptor for atomic operation add |
||
94 | // Message Descriptors |
||
95 | // = 000 0110 (min message len 6 ) 0,0000 (resp len 0 -add later) |
||
96 | // 1(header present)001,10(typed atomic operation)0(return enabled)0(slot group, low 8 bits),0111 (AOP_Add) |
||
97 | // 0000,0000 (Binding table index, added later) |
||
98 | // = 0x02000011 |
||
99 | |||
100 | // Atomic Operation Add Message Descriptor |
||
101 | |||
102 | |||
103 | // Message descriptor for dataport media write |
||
104 | // Message Descriptors |
||
105 | // = 000 0001 (min message len 1 - add later) 00000 (resp len 0) |
||
106 | // 1 (header present 1) 0 1010 (media block write) 000000 |
||
107 | // 00000000 (binding table index - set later) |
||
108 | // = 0x020A8000 |
||
109 | |||
110 | |||
111 | // Message Length defines |
||
112 | |||
113 | |||
114 | // Response Length defines |
||
115 | |||
116 | |||
117 | // Block Width and Height Size defines |
||
118 | |||
119 | |||
120 | // Extended Message Descriptors |
||
121 | |||
122 | |||
123 | // Common message descriptors: |
||
124 | |||
125 | |||
126 | //===================== Math Function Control =================================== |
||
127 | |||
128 | |||
129 | //============ Message Registers =============== |
||
130 | // buf4 starts from r28 |
||
131 | |||
132 | |||
133 | //#define mMSGHDR_EOT r43 // Dummy Message Register for EOT |
||
134 | |||
135 | |||
136 | .declare mubMSGPAYLOAD Base=r30 ElementSize=1 SrcRegion=<16;16,1> Type=ub |
||
137 | .declare muwMSGPAYLOAD Base=r30 ElementSize=2 SrcRegion=<16;16,1> Type=uw |
||
138 | .declare mudMSGPAYLOAD Base=r30 ElementSize=4 SrcRegion=<8;8,1> Type=ud |
||
139 | .declare mfMSGPAYLOAD Base=r30 ElementSize=4 SrcRegion=<8;8,1> Type=f |
||
140 | |||
141 | //=================== End of thread instruction =========================== |
||
142 | |||
143 | |||
144 | //=====================Pointers Used===================================== |
||
145 | |||
146 | |||
147 | //======================================================================= |
||
148 | |||
149 | |||
150 | //r9-r17 |
||
151 | // Define temp space for any usages |
||
152 | |||
153 | |||
154 | // Common Buffers |
||
155 | |||
156 | |||
157 | // temp space for rotation |
||
158 | |||
159 | .declare fROBUF Base=r9.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=f |
||
160 | |||
161 | .declare udROBUF Base=r9.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=ud |
||
162 | |||
163 | .declare uwROBUF Base=r9.0 ElementSize=2 SrcRegion=<16;16,1> DstRegion=<1> Type=uw |
||
164 | |||
165 | .declare ubROBUF Base=r9.0 ElementSize=1 SrcRegion=<16;16,1> DstRegion=<1> Type=ub |
||
166 | |||
167 | .declare ub4ROBUF Base=r9.0 ElementSize=1 SrcRegion=<32;8,4> DstRegion=<4> Type=ub |
||
168 | |||
169 | |||
170 | // End of common.inc |
||
171 | |||
172 | |||
173 | // FileName: DNDI.inc |
||
174 | // Author: Vivek Kumar |
||
175 | // Description: Include file for DN, DI and DNDI |
||
176 | // Inputs: DI_ENABLE, DN_ENABLE, DN_PLANAR, DN_PACKED |
||
177 | |||
178 | |||
179 | |||
180 | |||
181 | // End of common.inc |
||
182 | |||
183 | |||
184 | //Interface: |
||
185 | //Static Parameters: |
||
186 | //r1 |
||
187 | |||
188 | |||
189 | //====================== Binding table (Explicit To DNDI)========================================= |
||
190 | |||
191 | |||
192 | .declare mudMSGHDR_DNDI Base=r18 ElementSize=4 Type=ud |
||
193 | .declare mdMSGHDR_DNDI Base=r18 ElementSize=4 Type=d |
||
194 | .declare mwMSGHDR_DNDI Base=r18 ElementSize=2 Type=w |
||
195 | |||
196 | |||
197 | .declare mudMSGHDR_STMM Base=r20 ElementSize=4 Type=ud |
||
198 | |||
199 | |||
200 | .declare mudMSGHDR_HIST Base=r22 ElementSize=4 Type=ud |
||
201 | |||
202 | |||
203 | .declare mudMSGHDR_ENC_STATS Base=r24 ElementSize=4 Type=ud |
||
204 | .declare muwMSGHDR_ENC_STATS Base=r24 ElementSize=2 Type=uw |
||
205 | .declare mubMSGHDR_ENC_STATS Base=r24 ElementSize=1 Type=ub |
||
206 | |||
207 | |||
208 | .declare mudMSGHDR_DN_OUT Base=r31.0 ElementSize=4 Type=ud |
||
209 | .declare mdMSGHDR_DN_OUT Base=r31.0 ElementSize=4 Type=d |
||
210 | .declare mubMSGHDR_DN_OUT Base=r31.0 ElementSize=1 Type=ub |
||
211 | |||
212 | |||
213 | .declare mudMSGHDR_UVCOPY Base=r36 ElementSize=4 Type=ud |
||
214 | .declare mdMSGHDR_UVCOPY Base=r36 ElementSize=4 Type=d |
||
215 | .declare mudMSGHDR_UCOPY Base=r36 ElementSize=4 Type=ud |
||
216 | .declare mudMSGHDR_VCOPY Base=r38 ElementSize=4 Type=ud |
||
217 | |||
218 | |||
219 | .declare mudMSGHDR_DI_OUT1 Base=r18.0 ElementSize=4 Type=ud |
||
220 | .declare mubMSGHDR_DI_OUT1 Base=r18.0 ElementSize=1 Type=ub |
||
221 | |||
222 | |||
223 | .declare mudMSGHDR_DI_OUT2 Base=r23.0 ElementSize=4 Type=ud |
||
224 | .declare mubMSGHDR_DI_OUT2 Base=r23.0 ElementSize=1 Type=ub |
||
225 | |||
226 | //r45 |
||
227 | //Use r45 as message header, so no need to "mov" the data. |
||
228 | |||
229 | .declare mudDN_Y_OUT Base=r45.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=ud |
||
230 | |||
231 | // Message response (Denoised & DI-ed pixels & statistics); Use buffer 5 |
||
232 | .declare udDNDI_RESP Base=r46.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=ud |
||
233 | .declare uwDNDI_RESP Base=r46.0 ElementSize=2 SrcRegion=<16;16,1> DstRegion=<1> Type=uw |
||
234 | .declare ubDNDI_RESP Base=r46.0 ElementSize=1 SrcRegion=<16;16,1> DstRegion=<1> Type=ub |
||
235 | |||
236 | // Message response (UV Copy); Use buffer 5 |
||
237 | .declare udDNDI_UV_RESP Base=r58.0 ElementSize=4 SrcRegion=<8;8,1> DstRegion=<1> Type=ud |
||
238 | .declare ubDNDI_UV_RESP Base=r58.0 ElementSize=1 SrcRegion=<16;16,1> DstRegion=<1> Type=ub |
||
239 | |||
240 | //Temp GRFs: For 42X to 422 Conversion |
||
241 | .declare uwDNDI_UVCOPY_TEMP Base=r10.0 ElementSize=2 SrcRegion=<16;16,1> DstRegion=<1> Type=uw //8 GRFs |
||
242 | .declare ubDNDI_UVCOPY_TEMP Base=r10.0 ElementSize=1 SrcRegion=<16;16,1> DstRegion=<1> Type=ub //8 GRFs |
||
243 | //--------------------------------------------------------------------------- |
||
244 | // Message descriptors |
||
245 | //--------------------------------------------------------------------------- |
||
246 | // Extended message descriptor |
||
247 | // Message descriptor for sampler read |
||
248 | // = 000 0010 (message len 2) 00000 (resp len - set later, 12 or 5 or 11) |
||
249 | // 1 (header present 1) 0 11 (SIMD32/64 mode) |
||
250 | // 1000 (message type) 0000 (DI state index) |
||
251 | // 00000000 (binding table index - set later) |
||
252 | // = 0x040b8000 |
||
253 | |||
254 | |||
255 | // Attention: The Message Length is The Number of GRFs with Data Only, without the Header |
||
256 | |||
257 | |||
258 | //--------------------------------------------------------------------------- |
||
259 | // VDI Return Data format |
||
260 | //--------------------------------------------------------------------------- |
||
261 | // Defines for DI enabled |
||
262 | |||
263 | |||
264 | // Defines for DI disabled |
||
265 | |||
266 | |||
267 | |||
268 | // FileName: DNDI_Command.asm |
||
269 | // Author: Vivek Kumar |
||
270 | // Description: Sends a message to the VDI to process one DN (16x8) or DNDI (16x4) block |
||
271 | |||
272 | // Prepare the DNDI send command |
||
273 | mov (8) mudMSGHDR_DNDI(0)<1> r0.0<8;8,1>:ud // message header |
||
274 | mov (1) mwMSGHDR_DNDI(1,4)<1> r7.0<0;1,0>:w { NoDDClr } // horizontal origin // Do we need to add offset here? -vK |
||
275 | mov (1) mwMSGHDR_DNDI(1,12)<1> r7.1<0;1,0>:w { NoDDChk } // vertical origin // Can these 2 be combined? - vK |
||
276 | |||
277 | send (8) udDNDI_RESP(0)<1> r18 0x2 0x4AE8003:ud |
||
278 | |||
279 | // On Gen6, with VDI walker, use the XY pair returned rather than programmed above |
||
280 | // VDI_RETURNED_XY is ordered XY in case of walker enables and the same as programmed in case of walker disabled |
||
281 | mov (2) r7.0<1>:w uwDNDI_RESP(9,14)<2;2,1> // horizontal/Vertial origin in W.14 and W.15 |
||
282 | |||
283 | |||
284 | |||
285 | // FileName: DI_STMM_Save.asm |
||
286 | // Author: Vivek Kumar |
||
287 | // Description: Saves DI STMM Data to statistics surface in case of DI enabled (for 16x4 block) |
||
288 | |||
289 | // Write STMM to memory |
||
290 | mov (8) mudMSGHDR_STMM(0)<1> r0.0<8;8,1>:ud // message header |
||
291 | mov (8) mudMSGHDR_STMM(1)<1> udDNDI_RESP(8,0) // Move STMM to MRF |
||
292 | |||
293 | shr (1) mudMSGHDR_STMM(0,0)<1> r7.0<0;1,0>:w 1:w { NoDDClr } // X origin / 2 |
||
294 | mov (1) mudMSGHDR_STMM(0,1)<1> r7.1<0;1,0>:w { NoDDClr, NoDDChk } // Y origin |
||
295 | mov (1) mudMSGHDR_STMM(0,2)<1> 0x30007:ud { NoDDChk } // block width and height (8x4) |
||
296 | |||
297 | send (8) null<1>:d r20 0x5 0x40A8021:ud |
||
298 | |||
299 | |||
300 | |||
301 | // FileName: DNDI_Enc_Stats_Save.asm |
||
302 | // Author: Vivek Kumar |
||
303 | // Description: Saves Encoder Statistics data to statistics surface in case of DI enabled (for 16x4 block) |
||
304 | |||
305 | // Write encoder statistics to memory |
||
306 | //Currently enable this only on Gen6 validation |
||
307 | mov (8) mudMSGHDR_ENC_STATS(1)<1> 0x0:ud // Init payload MRF |
||
308 | mov (8) mudMSGHDR_ENC_STATS(0)<1> r0.0<8;8,1>:ud // message header |
||
309 | |||
310 | shr (1) mudMSGHDR_ENC_STATS(0,0)<1> r7.0<0;1,0>:w 1:w { NoDDClr } //enable the flag after testing on si { NoDDClr } // X origin / 2 |
||
311 | mul (1) acc0.1<1>:ud r7.1<0;1,0>:w 3:w // Y origin * 3 |
||
312 | 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 |
||
313 | mov (1) mudMSGHDR_ENC_STATS(0,2)<1> 0x20007:ud { NoDDChk } //enable the flag after testing on si { NoDDChk } // block width and height (8x3) |
||
314 | 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 |
||
315 | |||
316 | |||
317 | //Data block for Encoder Statistics |
||
318 | //---------------------------------------------------- |
||
319 | //| 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | Bytes |
||
320 | //---------------------------------------------------- |
||
321 | //| BNE | MCNT | FCNT | TCNT | X | X | X | X | |
||
322 | //---------------------------------------------------- |
||
323 | //| DcTpT | SVCM | DcBpT | DcTpB | |
||
324 | //---------------------------------------------------- |
||
325 | //| SHCM | STAD | DcTcB | DcBpB | |
||
326 | //---------------------------------------------------- |
||
327 | mov (1) mudMSGHDR_ENC_STATS(1,0)<1> udDNDI_RESP(9,1)<0;1,0> { NoDDClr } // Move encoder statistics to MRF |
||
328 | mov (2) mudMSGHDR_ENC_STATS(1,3)<2> udDNDI_RESP(9,3)<2;2,1> { NoDDClr, NoDDChk } // Move encoder statistics to MRF |
||
329 | mov (2) mudMSGHDR_ENC_STATS(1,2)<2> udDNDI_RESP(9,5)<2;2,1> { NoDDChk } // Move encoder statistics to MRF |
||
330 | |||
331 | |||
332 | send (8) null<1>:d r24 0x5 0x40A8021:ud |
||
333 | |||
334 | |||
335 | |||
336 | // FileName: DI_Save_422CP_16x4.asm |
||
337 | // Author: Vivek Kumar |
||
338 | // Description: Save two 16x4 blocks of DI output in 422 format to Color Pipe (IECP) |
||
339 | |||
340 | |||
341 | .declare mubMSGHDR_DI_OUT1_1 Base=r18.0 ElementSize=1 Type=ub |
||
342 | |||
343 | |||
344 | .declare mubMSGHDR_DI_OUT1_2 Base=r21.0 ElementSize=1 Type=ub |
||
345 | |||
346 | |||
347 | .declare mubMSGHDR_DI_OUT2_1 Base=r24.0 ElementSize=1 Type=ub |
||
348 | |||
349 | |||
350 | .declare mubMSGHDR_DI_OUT2_2 Base=r27.0 ElementSize=1 Type=ub |
||
351 | |||
352 | |||
353 | mov (8) r27.0<1>:ud r0.0<8;8,1>:ud |
||
354 | shl (1) r27.0<1>:ud r7.0<0;1,0>:w 1:w { NoDDClr } // H. block origin need to be doubled |
||
355 | mov (1) r27.1<1>:ud r7.1<0;1,0>:w { NoDDClr, NoDDChk } // Block origin |
||
356 | mov (1) r27.2<1>:ud 0x3000F:ud { NoDDClr, NoDDChk } // Block width and height (16x8) |
||
357 | |||
358 | //M0.3 - 0 - CP Enable, 1 - Area of Interest, 3:2 Message Format(TBD), 4:3 - Ignored, 31:5 CP state pointer |
||
359 | //Compose area-of-interest bit + color pipe state pointer |
||
360 | or (1) r27.3<1>:ud r2.4<0;1,0>:ud r7.26<0;1,0>:b { NoDDChk } |
||
361 | |||
362 | //prepare the message headers |
||
363 | mov (8) r18.0<1>:ud r27<8;8,1>:ud |
||
364 | mov (8) r24.0<1>:ud r27<8;8,1>:ud |
||
365 | |||
366 | |||
367 | // Pack 2nd field Y; First 8x4 block |
||
368 | 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 |
||
369 | 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 |
||
370 | 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 |
||
371 | 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 |
||
372 | |||
373 | // Pack 2nd field U, V; First 8x4 block |
||
374 | 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 |
||
375 | 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 |
||
376 | |||
377 | 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 |
||
378 | 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 |
||
379 | 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 |
||
380 | 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 |
||
381 | |||
382 | 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 |
||
383 | 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 |
||
384 | |||
385 | |||
386 | // Pack 2nd field Y; Second 8x4 block |
||
387 | mov (8) r21.0<1>:ud r18.0<8;8,1>:ud |
||
388 | add (1) r21.0<1>:ud r21.0<0;1,0>:w 0x10:w |
||
389 | |||
390 | 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 |
||
391 | 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 |
||
392 | 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 |
||
393 | 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 |
||
394 | |||
395 | // Pack 2nd field U, V; Second 8x4 block |
||
396 | 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 |
||
397 | 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 |
||
398 | |||
399 | 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 |
||
400 | 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 |
||
401 | 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 |
||
402 | 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 |
||
403 | |||
404 | 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 |
||
405 | 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 |
||
406 | |||
407 | send (8) null<1>:d r18.0 0x5 0x60A801B:ud |
||
408 | send (8) null<1>:d r21.0 0x5 0x60A801B:ud |
||
409 | |||
410 | // Pack 1st field Y; 1st 8x4 block |
||
411 | 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 |
||
412 | 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 |
||
413 | 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 |
||
414 | 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 |
||
415 | |||
416 | // Pack 1st field U,V; 1st 8x4 block |
||
417 | 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 |
||
418 | 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 |
||
419 | |||
420 | 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 |
||
421 | 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 |
||
422 | 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 |
||
423 | 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 |
||
424 | |||
425 | 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 |
||
426 | 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 |
||
427 | |||
428 | // Pack 1st field Y; 2nd 8x4 block |
||
429 | mov (8) r27.0<1>:ud r24.0<8;8,1>:ud |
||
430 | add (1) r27.0<1>:ud r27.0<0;1,0>:w 0x10:w |
||
431 | |||
432 | 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 |
||
433 | 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 |
||
434 | 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 |
||
435 | 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 |
||
436 | |||
437 | // Pack 1st field U, V; 2nd 8x4 block |
||
438 | 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 |
||
439 | 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 |
||
440 | |||
441 | 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 |
||
442 | 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 |
||
443 | 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 |
||
444 | 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 |
||
445 | |||
446 | 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 |
||
447 | 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 |
||
448 | |||
449 | send (8) null<1>:d r24.0 0x5 0x60A801E:ud |
||
450 | send (8) null<1>:d r27.0 0x5 0x60A801E:ud |
||
451 | |||
452 | |||
453 | |||
454 | //End of Thread message |
||
455 | |||
456 | mov (8) r127<1>:ud r0.0<8;8,1>:ud |
||
457 | send (1) null<1>:d r127 0x27 0x02000010 |
||
458 | |||
459 | |||
460 | .end_code |
||
461 | .end_kernel1>8;8,1>1>1>1>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;8,1>2>8;8,1>2>8;8,1>2>8;8,1>2>0;1,0>1>8;8,1>1>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;8,1>2>8;8,1>2>8;8,1>2>8;8,1>2>1>1>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;8,1>2>8;8,1>2>8;8,1>2>8;8,1>2>0;1,0>1>8;8,1>1>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;4,2>4>8;8,1>2>8;8,1>2>8;8,1>2>8;8,1>2>8;8,1>1>8;8,1>1>0;1,0>0;1,0>1>1>0;1,0>1>0;1,0>1>8;8,1>1>1>2;2,1>2>2;2,1>2>0;1,0>1>2;2,1>2;2,1>1>1>0;1,0>1>0;1,0>1>0;1,0>1>8;8,1>1>1>1>1>0;1,0>1>0;1,0>1>1>8;8,1>1>2;2,1>1>1>0;1,0>1>0;1,0>1>8;8,1>1>1>16;16,1>1>16;16,1>1>16;16,1>1>8;8,1>1>16;16,1>1>16;16,1>1>8;8,1>1>8;8,1>4>32;8,4>1>16;16,1>1>16;16,1>1>8;8,1>1>8;8,1>8;8,1>8;8,1>16;16,1>16;16,1> |