summaryrefslogtreecommitdiff
path: root/src/shaders/h264/mc/interpolate_Y_8x8.asm
blob: 4506de79442b21fec96b0dd3bcc4a8572c603121 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
/*
 * Interpolation kernel for luminance motion compensation
 * Copyright © <2010>, Intel Corporation.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the
 * "Software"), to deal in the Software without restriction, including
 * without limitation the rights to use, copy, modify, merge, publish,
 * distribute, sub license, and/or sell copies of the Software, and to
 * permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 *
 * The above copyright notice and this permission notice (including the
 * next paragraph) shall be included in all copies or substantial portions
 * of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
 * IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 *
 * This file was originally licensed under the following license
 *
 *  Licensed under the Apache License, Version 2.0 (the "License");
 *  you may not use this file except in compliance with the License.
 *  You may obtain a copy of the License at
 *
 *      http://www.apache.org/licenses/LICENSE-2.0
 *
 *  Unless required by applicable law or agreed to in writing, software
 *  distributed under the License is distributed on an "AS IS" BASIS,
 *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 *  See the License for the specific language governing permissions and
 *  limitations under the License.
 *
 */
// Kernel name: Interpolate_Y_8x8.asm
//
// Interpolation kernel for luminance motion compensation
//
//  $Revision: 13 $
//  $Date: 10/09/06 4:00p $
//


//---------------------------------------------------------------
// In: pMV => Source address of MV
// In: gMVX_FRAC<2;2,1>:w => MV fractional components
// In: f0.1 (1) => If 1, vertical MV is integer
// In: gpINTPY:uw => Destination address for interpolated result
// In: Reference area staring from R43
//		If horizontal/vertical MVs are all integer, 8x8 pixels are on R43~R44 (2 GRFs)
//		If only horz MV is integer, 8x13 pixels are on R43~R46 (4 GRFs)
//		If only vert MV is integer, 13x8 pixels are on R43~R46 (4 GRFs)
//		If no MVs are integer, 13x13 pixels are on R43~R49 (7 GRFs)
//---------------------------------------------------------------


INTERLABEL(Interpolate_Y_8x8_Func):



	// Check whether MVX is integer MV
	and.z.f0.0 (1) null:w			r[pMV,0]<0;1,0>:w				0x3:w
	(-f0.0)	jmpi (1) INTERLABEL(Interpolate_Y_8x8_Func2)
		
	// TODO: remove this back-to-back read - huge latency..
	mov (8)	gubREF(6,2)<1>	gubREF(3,0)<8;8,1>
    mov (8)	gubREF(5,18)<1>	gubREF(2,24)<8;8,1>		{NoDDClr}
	mov (8)	gubREF(5,2)<1>	gubREF(2,16)<8;8,1>		{NoDDChk}
	mov (8)	gubREF(4,18)<1>	gubREF(2,8)<8;8,1>		{NoDDClr}
	mov (8)	gubREF(4,2)<1>	gubREF(2,0)<8;8,1>		{NoDDChk}
	mov (8)	gubREF(3,18)<1>	gubREF(1,24)<8;8,1>		{NoDDClr}
	mov (8)	gubREF(3,2)<1>	gubREF(1,16)<8;8,1>		{NoDDChk}
	mov (8)	gubREF(2,18)<1>	gubREF(1,8)<8;8,1>		{NoDDClr}
	mov (8)	gubREF(2,2)<1>	gubREF(1,0)<8;8,1>		{NoDDChk}
	mov (8)	gubREF(1,18)<1>	gubREF(0,24)<8;8,1>		{NoDDClr}
	mov (8)	gubREF(1,2)<1>	gubREF(0,16)<8;8,1>		{NoDDChk}
	mov (8)	gubREF(0,18)<1>	gubREF(0,8)<8;8,1>	
    mov (8)	gubREF(0,2)<1>	gubREF(0,0)<8;8,1>

INTERLABEL(Interpolate_Y_8x8_Func2):

	// Compute the GRF address of the starting position of the reference area
    (-f0.1) mov (1)	pREF:w			nOFFSET_REF+2+nGRFWIB:w	
    (f0.1) mov (1)	pREF:w			nOFFSET_REF+2:w			
	mov (1)		pRESULT:uw			gpINTPY:uw	
	
	/*
	 *			 |		 |
	 *		 - - 0 1 2 3 + - 
	 *			 4 5 6 7
	 *			 8 9 A B
	 *			 C D E F
	 *		 - - + - - - + -
     *			 |		 |
	 */
	
	// Case 0
	or.z.f0.1 (16) null:w			gMVY_FRAC<0;1,0>:w				gMVX_FRAC<0;1,0>:w	
	(f0.1) mov (16)	r[pRESULT]<1>:uw				r[pREF]<16;8,1>:ub
	(f0.1) mov (16)	r[pRESULT,nGRFWIB]<1>:uw		r[pREF,nGRFWIB]<16;8,1>:ub
	(f0.1) mov (16)	r[pRESULT,nGRFWIB*2]<1>:uw		r[pREF,nGRFWIB*2]<16;8,1>:ub
	(f0.1) mov (16)	r[pRESULT,nGRFWIB*3]<1>:uw		r[pREF,nGRFWIB*3]<16;8,1>:ub
	(f0.1) jmpi INTERLABEL(Exit_Interpolate_Y_8x8)
	
	// Store all address registers
	mov (8)		gpADDR<1>:w			a0<8;8,1>:w
	
	mul.z.f0.0 (1) gW4:w			gMVY_FRAC:w						gMVX_FRAC:w
	add (1)		pREF1:uw			pREF0:uw						nGRFWIB/2:uw
	and.nz.f0.1 (1) null			gW4:w							1:w
	add (2)		pREF2<1>:uw			pREF0<2;2,1>:uw					nGRFWIB:uw
	mov (4)		gW0<1>:uw			pREF0<4;4,1>:uw

	(f0.0) jmpi INTERLABEL(Interpolate_Y_H_8x8)
	(f0.1) jmpi INTERLABEL(Interpolate_Y_H_8x8)
	
	//-----------------------------------------------------------------------
	// CASE: A69BE (H/V interpolation)
	//-----------------------------------------------------------------------
	
	// Compute interim horizontal intepolation of 12 lines (not 9 lines)
//	add (1)		pREF0<1>:ud			pREF0<0;1,0>:ud					0xffeeffde:ud	// (-18<<16)|(-34)
	add (1)		pREF0<1>:uw			pREF0<0;1,0>:uw					-34:w	
	add (1)		pREF1<1>:uw			pREF1<0;1,0>:uw					-18:w {NoDDClr}	
	mov (1)		pRESD:ud			nOFFSET_INTERIM3:ud					{NoDDChk}			
	
	// Check whether this position is 'A'    
	cmp.e.f0.0 (1) null				gW4:w							4:w
	
	$for(0;<6;2) {
	add (32)	acc0<1>:w			r[pREF,nGRFWIB*%1]<16;8,1>:ub			r[pREF0,nGRFWIB*%1+5]<16;8,1>:ub		{Compr}
	mac (32)	acc0<1>:w			r[pREF,nGRFWIB*%1+1]<16;8,1>:ub			-5:w	{Compr}
	mac (32)	acc0<1>:w			r[pREF,nGRFWIB*%1+2]<16;8,1>:ub			20:w	{Compr}
	mac (32)	acc0<1>:w			r[pREF,nGRFWIB*%1+3]<16;8,1>:ub			20:w	{Compr}
	mac (32)	r[pRES,nGRFWIB*%1]<1>:w		r[pREF,nGRFWIB*%1+4]<16;8,1>:ub	-5:w	{Compr}
	}
	// last line
	add (8)		acc0<1>:w			r[pREF,nGRFWIB*6]<8;8,1>:ub				r[pREF,nGRFWIB*6+5]<8;8,1>:ub
	mac (8)		acc0<1>:w			r[pREF,nGRFWIB*6+1]<8;8,1>:ub			-5:w
	mac (8)		acc0<1>:w			r[pREF,nGRFWIB*6+2]<8;8,1>:ub			20:w
	mac (8)		acc0<1>:w			r[pREF,nGRFWIB*6+3]<8;8,1>:ub			20:w
	mac (8)		r[pRES,nGRFWIB*6]<1>:w		r[pREF,nGRFWIB*6+4]<8;8,1>:ub	-5:w

    // Compute interim/output vertical interpolation 
    mov (1)		pREF0:ud			nOFFSET_INTERIM2:ud	{NoDDClr}			// set pREF0 and pREF1 at the same time
	mov (1)		pREF2D:ud			nOFFSET_INTERIM4:ud	{NoDDChk,NoDDClr}	// set pREF2 and pREF3 at the same time
	(f0.0) sel (1) pRES:uw			gpINTPY:uw	nOFFSET_INTERIM:uw {NoDDChk} // Case A vs. 69BE
    
	$for(0;<4;2) {
	add (32)	acc0<1>:w			r[pREF0,nGRFWIB*%1]<16;16,1>:w				512:w	{Compr}
	mac (16)	acc0<1>:w			r[pREF2,nGRFWIB*%1]<8,1>:w					-5:w
	mac (16)	acc1<1>:w			r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:w			-5:w
	mac (32)	acc0<1>:w			r[pREF0,nGRFWIB*%1+nGRFWIB]<16;16,1>:w		20:w	{Compr}
	mac (16)	acc0<1>:w			r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:w			20:w	
	mac (16)	acc1<1>:w			r[pREF2,nGRFWIB*%1+nGRFWIB+nGRFWIB]<8,1>:w	20:w	
	mac (32)	acc0<1>:w			r[pREF0,(2+%1)*nGRFWIB]<16;16,1>:w			-5:w	{Compr}
	mac (16)	acc0<1>:w			r[pREF2,(2+%1)*nGRFWIB]<8,1>:w				1:w
	mac (16)	acc1<1>:w			r[pREF2,(2+%1)*nGRFWIB+nGRFWIB]<8,1>:w		1:w
	asr.sat (16) r[pRES,nGRFWIB*%1]<2>:ub			acc0<16;16,1>:w				10:w
	asr.sat (16) r[pRES,nGRFWIB*%1+nGRFWIB]<2>:ub	acc1<16;16,1>:w				10:w {SecHalf}
	}
	
	(f0.0) jmpi INTERLABEL(Return_Interpolate_Y_8x8)
	
INTERLABEL(Interpolate_Y_H_8x8):
	
	cmp.e.f0.0 (1) null				gMVX_FRAC:w						0:w
	cmp.e.f0.1 (1) null				gMVY_FRAC:w						2:w
	(f0.0) jmpi INTERLABEL(Interpolate_Y_V_8x8)
	(f0.1) jmpi INTERLABEL(Interpolate_Y_V_8x8)
	
	//-----------------------------------------------------------------------
	// CASE: 123567DEF (H interpolation)
	//-----------------------------------------------------------------------

	add (4)		pREF0<1>:uw			gW0<4;4,1>:uw					-2:w		
	cmp.g.f0.0 (4) null:w			gMVY_FRAC<0;1,0>:w				2:w
	cmp.e.f0.1 (1) null				gMVX_FRAC:w						2:w
	(f0.0) add (4) pREF0<1>:uw		pREF0<4;4,1>:uw					nGRFWIB/2:uw

	cmp.e.f0.0 (1) null:w			gMVY_FRAC<0;1,0>:w				0:w

	(f0.1) sel (1) pRES:uw			gpINTPY:uw						nOFFSET_INTERIM:uw // Case 26E vs. 1357DF
	
	// Compute interim/output horizontal interpolation
	$for(0;<4;2) {
	add (16)	acc0<1>:w			r[pREF0,nGRFWIB*%1]<8,1>:ub				16:w
	add (16)	acc1<1>:w			r[pREF0,nGRFWIB*%1+nGRFWIB]<8,1>:ub		16:w
	mac (16)	acc0<1>:w			r[pREF0,nGRFWIB*%1+1]<8,1>:ub			-5:w
	mac (16)	acc1<1>:w			r[pREF0,nGRFWIB*%1+1+nGRFWIB]<8,1>:ub	-5:w
	mac (16)	acc0<1>:w			r[pREF0,nGRFWIB*%1+2]<8,1>:ub			20:w
	mac (16)	acc1<1>:w			r[pREF0,nGRFWIB*%1+2+nGRFWIB]<8,1>:ub	20:w
	mac (16)	acc0<1>:w			r[pREF0,nGRFWIB*%1+3]<8,1>:ub			20:w
	mac (16)	acc1<1>:w			r[pREF0,nGRFWIB*%1+3+nGRFWIB]<8,1>:ub	20:w
	mac (16)	acc0<1>:w			r[pREF0,nGRFWIB*%1+4]<8,1>:ub			-5:w
	mac (16)	acc1<1>:w			r[pREF0,nGRFWIB*%1+4+nGRFWIB]<8,1>:ub	-5:w
	mac (16)	acc0<1>:w			r[pREF0,nGRFWIB*%1+5]<8,1>:ub			1:w
	mac (16)	acc1<1>:w			r[pREF0,nGRFWIB*%1+5+nGRFWIB]<8,1>:ub	1:w
	asr.sat (16) r[pRES,nGRFWIB*%1]<2>:ub			acc0<16;16,1>:w		5:w
	asr.sat (16) r[pRES,nGRFWIB*%1+nGRFWIB]<2>:ub	acc1<16;16,1>:w		5:w {SecHalf}
    }
    
    (-f0.1) jmpi INTERLABEL(Interpolate_Y_V_8x8)
	(-f0.0) jmpi INTERLABEL(Average_8x8)
	
	jmpi INTERLABEL(Return_Interpolate_Y_8x8)

INTERLABEL(Interpolate_Y_V_8x8):

	cmp.e.f0.0 (1) null				gMVY_FRAC:w						0:w
	(f0.0) jmpi INTERLABEL(Interpolate_Y_I_8x8)
	
	//-----------------------------------------------------------------------
	// CASE: 48C59D7BF (V interpolation)
	//-----------------------------------------------------------------------

	mov (2)		pREF0<1>:uw			gW0<4;2,2>:uw	{NoDDClr}
	mov (2)		pREF2<1>:uw			gW1<2;2,1>:uw	{NoDDChk,NoDDClr}
	mov (1)		pRES:uw				gpINTPY:uw		{NoDDChk}

	cmp.g.f0.1 (4) null:w			gMVX_FRAC<0;1,0>:w				2:w
	cmp.e.f0.0 (1) null:w			gMVX_FRAC<0;1,0>:w				0:w
	(f0.1) add (4) pREF0<1>:uw		pREF0<4;4,1>:uw					1:uw

	cmp.e.f0.1 (1) null				gMVY_FRAC:w						2:w
	(-f0.0) jmpi INTERLABEL(VFILTER_8x8)
	(-f0.1) mov (1) pRES:uw		nOFFSET_INTERIM:uw
	
  INTERLABEL(VFILTER_8x8): 

	// Compute interim/output vertical interpolation
	$for(0;<4;2) {
	add (32)	acc0<1>:w			r[pREF0,nGRFWIB*%1-nGRFWIB]<16;8,1>:ub			16:w {Compr}
	mac (16)	acc0<1>:w			r[pREF2,nGRFWIB*%1-nGRFWIB]<8,1>:ub				-5:w
	mac (16)	acc1<1>:w			r[pREF2,nGRFWIB*%1]<8,1>:ub						-5:w
	mac (32)	acc0<1>:w			r[pREF0,nGRFWIB*%1]<16;8,1>:ub					20:w {Compr}
	mac (16)	acc0<1>:w			r[pREF2,nGRFWIB*%1]<8,1>:ub						20:w	
	mac (16)	acc1<1>:w			r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:ub				20:w	
	mac (32)	acc0<1>:w			r[pREF0,nGRFWIB*%1+nGRFWIB]<16;8,1>:ub			-5:w {Compr}
	mac (16)	acc0<1>:w			r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:ub				1:w
	mac (16)	acc1<1>:w			r[pREF2,nGRFWIB*%1+nGRFWIB+nGRFWIB]<8,1>:ub		1:w
	asr.sat (16) r[pRES,nGRFWIB*%1]<2>:ub			acc0<16;16,1>:w					5:w
	asr.sat (16) r[pRES,nGRFWIB*%1+nGRFWIB]<2>:ub	acc1<16;16,1>:w					5:w	{SecHalf}
	}

	(-f0.0) jmpi INTERLABEL(Average_8x8)
	(f0.1) jmpi INTERLABEL(Return_Interpolate_Y_8x8)

INTERLABEL(Interpolate_Y_I_8x8):

	//-----------------------------------------------------------------------
	// CASE: 134C (Integer position)
	//-----------------------------------------------------------------------
	
	mov (2)		pREF0<1>:uw			gW0<2;2,1>:uw		{NoDDClr}
			
	mov (1)		pRES:uw				gpINTPY:uw			{NoDDChk}

	cmp.e.f0.0 (2) null:w			gMVX_FRAC<0;1,0>:w				3:w
	cmp.e.f0.1 (2) null:w			gMVY_FRAC<0;1,0>:w				3:w
	(f0.0) add (2) pREF0<1>:uw		pREF0<2;2,1>:uw					1:uw 
	(f0.1) add (2) pREF0<1>:uw		pREF0<2;2,1>:uw					nGRFWIB/2:uw
	
	mov (16)	r[pRES]<1>:uw			r[pREF0]<8,1>:ub
	mov (16)	r[pRES,nGRFWIB]<1>:uw	r[pREF0,nGRFWIB]<8,1>:ub
	mov (16)	r[pRES,nGRFWIB*2]<1>:uw	r[pREF0,nGRFWIB*2]<8,1>:ub
	mov (16)	r[pRES,nGRFWIB*3]<1>:uw	r[pREF0,nGRFWIB*3]<8,1>:ub
	
INTERLABEL(Average_8x8):

	//-----------------------------------------------------------------------
	// CASE: 13456789BCDEF (Average)
	//-----------------------------------------------------------------------

	// Average two interim results
	avg.sat (16) r[pRES,0]<2>:ub			r[pRES,0]<32;16,2>:ub			gubINTERIM_BUF(0)	
	avg.sat (16) r[pRES,nGRFWIB]<2>:ub		r[pRES,nGRFWIB]<32;16,2>:ub		gubINTERIM_BUF(1)	
	avg.sat (16) r[pRES,nGRFWIB*2]<2>:ub	r[pRES,nGRFWIB*2]<32;16,2>:ub	gubINTERIM_BUF(2)	
	avg.sat (16) r[pRES,nGRFWIB*3]<2>:ub	r[pRES,nGRFWIB*3]<32;16,2>:ub	gubINTERIM_BUF(3)	

INTERLABEL(Return_Interpolate_Y_8x8):
	// Restore all address registers
	mov (8)		a0<1>:w					gpADDR<8;8,1>:w
	
INTERLABEL(Exit_Interpolate_Y_8x8):
	        
// end of file