Debian Codegraph @

src/shaders/vme/vp8_intra_frame_gen8.asm

raw
1 /*
2 * Copyright © 2014 Intel Corporation
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 * Authors:
25 * Zhao Yakui <[email protected]>
26 * Xiang Haihao <[email protected]>
27 * Li Zhong <[email protected]>
28 *
29 */
30
31 /*
32 * __START
33 */
34 __INTRA_START:
35 mov (16) tmp_reg0.0<1>:UD 0x0:UD {align1};
36 mov (16) tmp_reg2.0<1>:UD 0x0:UD {align1};
37 mov (16) tmp_reg4.0<1>:UD 0x0:UD {align1} ;
38 mov (16) tmp_reg6.0<1>:UD 0x0:UD {align1} ;
39
40 shl (2) read0_header.0<1>:D orig_xy_ub<2,2,1>:UB 4:UW {align1}; /* (x, y) * 16 */
41 add (1) read0_header.0<1>:D read0_header.0<0,1,0>:D -8:W {align1}; /* X offset */
42 add (1) read0_header.4<1>:D read0_header.4<0,1,0>:D -1:W {align1}; /* Y offset */
43 mov (1) read0_header.8<1>:UD BLOCK_32X1 {align1};
44 mov (1) read0_header.20<1>:UB thread_id_ub {align1}; /* dispatch id */
45
46 shl (2) read1_header.0<1>:D orig_xy_ub<2,2,1>:UB 4:UW {align1}; /* (x, y) * 16 */
47 add (1) read1_header.0<1>:D read1_header.0<0,1,0>:D -4:W {align1}; /* X offset */
48 mov (1) read1_header.8<1>:UD BLOCK_4X16 {align1};
49 mov (1) read1_header.20<1>:UB thread_id_ub {align1}; /* dispatch id */
50
51 shl (2) vme_m0.8<1>:UW orig_xy_ub<2,2,1>:UB 4:UW {align1}; /* (x, y) * 16 */
52 mov (1) vme_m0.20<1>:UB thread_id_ub {align1}; /* dispatch id */
53
54 mul (1) obw_m0.8<1>:UD w_in_mb_uw<0,1,0>:UW orig_y_ub<0,1,0>:UB {align1};
55 add (1) obw_m0.8<1>:UD obw_m0.8<0,1,0>:UD orig_x_ub<0,1,0>:UB {align1};
56 mul (1) obw_m0.8<1>:UD obw_m0.8<0,1,0>:UD 0x02:UD {align1};
57 mov (1) obw_m0.20<1>:UB thread_id_ub {align1}; /* dispatch id */
58
59 /*
60 * Media Read Message -- fetch Luma neighbor edge pixels
61 */
62 /* ROW */
63 mov (8) msg_reg0.0<1>:UD read0_header.0<8,8,1>:UD {align1};
64 send (8) msg_ind INEP_ROW<1>:UB null read(BIND_IDX_INEP, 0, 0, 4) mlen 1 rlen 1 {align1};
65
66 /* COL */
67 mov (8) msg_reg0.0<1>:UD read1_header.0<8,8,1>:UD {align1};
68 send (8) msg_ind INEP_COL0<1>:UB null read(BIND_IDX_INEP, 0, 0, 4) mlen 1 rlen 2 {align1};
69
70 /*
71 * Media Read Message -- fetch Chroma neighbor edge pixels
72 */
73 /* ROW */
74 shl (2) read0_header.0<1>:D orig_xy_ub<2,2,1>:UB 3:UW {align1}; /* x * 16 , y * 8 */
75 mul (1) read0_header.0<1>:D read0_header.0<0,1,0>:D 2:W {align1};
76 add (1) read0_header.0<1>:D read0_header.0<0,1,0>:D -8:W {align1}; /* X offset */
77 add (1) read0_header.4<1>:D read0_header.4<0,1,0>:D -1:W {align1}; /* Y offset */
78 mov (8) msg_reg0.0<1>:UD read0_header.0<8,8,1>:UD {align1};
79 send (8) msg_ind CHROMA_ROW<1>:UB null read(BIND_IDX_CBCR, 0, 0, 4) mlen 1 rlen 1 {align1};
80
81 /* COL */
82 shl (2) read1_header.0<1>:D orig_xy_ub<2,2,1>:UB 3:UW {align1}; /* x * 16, y * 8 */
83 mul (1) read1_header.0<1>:D read1_header.0<0,1,0>:D 2:W {align1};
84 add (1) read1_header.0<1>:D read1_header.0<0,1,0>:D -4:W {align1}; /* X offset */
85 mov (1) read1_header.8<1>:UD BLOCK_8X4 {align1};
86 mov (8) msg_reg0.0<1>:UD read1_header.0<8,8,1>:UD {align1};
87 send (8) msg_ind CHROMA_COL<1>:UB null read(BIND_IDX_CBCR, 0, 0, 4) mlen 1 rlen 1 {align1};
88
89 /* m2, get the MV/Mb cost passed by constant buffer
90 when creating EU thread by MEDIA_OBJECT */
91 mov (8) vme_msg_2<1>:UD r1.0<8,8,1>:UD {align1};
92
93 /* m3. This is changed for FWD/BWD cost center */
94 mov (8) vme_msg_3<1>:UD 0x0:UD {align1};
95
96 /* m4.*/
97 mov (8) vme_msg_4<1>:ud 0x0:ud {align1};
98
99 /* m5 */
100 mov (1) INEP_ROW.0<1>:UD 0x0:UD {align1};
101 and (1) INEP_ROW.4<1>:UD INEP_ROW.4<0,1,0>:UD 0xFF000000:UD {align1};
102 mov (8) vme_msg_5<1>:UD INEP_ROW.0<8,8,1>:UD {align1};
103
104 mov (1) tmp_reg0.0<1>:UB INTRA_PLANAR_MODE_MASK {align1}; /* vp8 don't support planar intra mode */
105 mov (1) tmp_reg0.1<1>:UB LUMA_CHROMA_MODE {align1}; /* Intra type: Luma + Chroma */
106
107 /* Intra mode mask && Intra compute type */
108 mov (1) vme_msg_5.4<1>:UW tmp_reg0.0<0,1,0>:UW {align1};
109
110 /* m6 */
111 mov (8) vme_msg_6<1>:UD 0x0:UD {align1};
112 mov (16) vme_msg_6.0<1>:UB INEP_COL0.3<32,8,4>:UB {align1};
113 mov (1) vme_msg_6.16<1>:UD INTRA_PREDICTORE_MODE {align1};
114
115 /* the penalty for Intra mode */
116 mov (1) vme_msg_6.28<1>:UD 0x010101:UD {align1};
117 mov (1) vme_msg_6.20<1>:UW CHROMA_ROW.6<0,1,0>:UW {align1};
118
119
120 /* m7 */
121
122 mov (4) vme_msg_7.16<1>:UD CHROMA_ROW.8<4,4,1>:UD {align1};
123 mov (8) vme_msg_7.0<1>:UW CHROMA_COL.2<16,8,2>:UW {align1};
124
125 /*
126 * VME message
127 */
128
129 /* m1 */
130 mov (1) intra_flag<1>:UW 0x0:UW {align1};
131 mov (1) intra_part_mask_ub<1>:UB LUMA_INTRA_8x8_DISABLE {align1}; /* vp8 don't support intra_8x8 mode*/
132
133 /* assign MB intra struct from the thread payload*/
134 mov (1) mb_intra_struct_ub<1>:UB input_mb_intra_ub<0,1,0>:UB {align1};
135
136 /* Disable DC HAAR component when calculating HARR SATD block */
137 mov (1) tmp_reg0.0<1>:UW DC_HARR_DISABLE:UW {align1};
138 mov (1) vme_m1.30<1>:UB tmp_reg0.0<0,1,0>:UB {align1};
139
140 mov (8) vme_msg_1<1>:UD vme_m1.0<8,8,1>:UD {align1};
141
142 /* m0 */
143 add (1) vme_m0.12<1>:UD vme_m0.12<0,1,0>:ud INTRA_SAD_HAAR:UD {align1};/* 16x16 Source, Intra_harr */
144 mov (1) vme_m0.15<1>:UB SUB_PART_8x4_DISABLE + SUB_PART_4x8_DISABLE {align1}; /* vp8 don't support 8x4 and 4x8 partion */
145 mov (8) vme_msg_0<1>:UD vme_m0.0<8,8,1>:UD {align1};
146
147 /* after verification it will be passed by using payload */
148 send (8)
149 vme_msg_ind
150 vme_wb<1>:UD
151 null
152 cre(
153 BIND_IDX_VME,
154 VME_SIC_MESSAGE_TYPE
155 )
156 mlen sic_vme_msg_length
157 rlen vme_wb_length
158 {align1};
159 /*
160 * Oword Block Write message
161 */
162 mov (8) msg_reg0.0<1>:UD obw_m0<8,8,1>:UD {align1};
163
164 mov (1) msg_reg1.0<1>:UD vme_wb.0<0,1,0>:UD {align1};
165 mov (1) msg_reg1.4<1>:UD vme_wb.16<0,1,0>:UD {align1};
166 mov (1) msg_reg1.8<1>:UD vme_wb.20<0,1,0>:UD {align1};
167 mov (1) msg_reg1.12<1>:UD vme_wb.24<0,1,0>:UD {align1};
168
169 /* Distortion, Intra (17-16), */
170 mov (1) msg_reg1.16<1>:UW vme_wb.12<0,1,0>:UW {align1};
171
172 mov (1) msg_reg1.20<1>:UD vme_wb.8<0,1,0>:UD {align1};
173 /* VME clock counts */
174 mov (1) msg_reg1.24<1>:UD vme_wb.28<0,1,0>:UD {align1};
175
176 mov (1) msg_reg1.28<1>:UD obw_m0.8<0,1,0>:UD {align1};
177
178 /* bind index 3, write 2 oword (32bytes), msg type: 8(OWord Block Write) */
179 send (16)
180 msg_ind
181 obw_wb
182 null
183 data_port(
184 OBW_CACHE_TYPE,
185 OBW_MESSAGE_TYPE,
186 OBW_CONTROL_2,
187 OBW_BIND_IDX,
188 OBW_WRITE_COMMIT_CATEGORY,
189 OBW_HEADER_PRESENT
190 )
191 mlen 2
192 rlen obw_wb_length
193 {align1};
194
195 __EXIT:
196 /*
197 * kill thread
198 */
199 mov (8) ts_msg_reg0<1>:UD r0<8,8,1>:UD {align1};
200 send (16) ts_msg_ind acc0<1>UW null thread_spawner(0, 0, 1) mlen 1 rlen 0 {align1 EOT};