2 * Copyright © <2010>, Intel Corporation.
4 * This program is licensed under the terms and conditions of the
5 * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
6 * http://www.opensource.org/licenses/eclipse-1.0.php.
9 // Modual name: IntraFrame.asm
11 // Make intra predition estimation for Intra frame
15 // Now, begin source code....
22 mov (16) tmp_reg0.0<1>:UD 0x0:UD {align1};
23 mov (16) tmp_reg2.0<1>:UD 0x0:UD {align1};
24 mov (16) tmp_reg4.0<1>:UD 0x0:UD {align1} ;
26 shl (2) read0_header.0<1>:D orig_xy_ub<2,2,1>:UB 4:UW {align1}; /* (x, y) * 16 */
27 add (1) read0_header.0<1>:D read0_header.0<0,1,0>:D -8:W {align1}; /* X offset */
28 add (1) read0_header.4<1>:D read0_header.4<0,1,0>:D -1:W {align1}; /* Y offset */
29 mov (1) read0_header.8<1>:UD BLOCK_32X1 {align1};
30 mov (1) read0_header.20<1>:UB thread_id_ub {align1}; /* dispatch id */
32 shl (2) read1_header.0<1>:D orig_xy_ub<2,2,1>:UB 4:UW {align1}; /* (x, y) * 16 */
33 add (1) read1_header.0<1>:D read1_header.0<0,1,0>:D -4:W {align1}; /* X offset */
34 mov (1) read1_header.8<1>:UD BLOCK_4X16 {align1};
35 mov (1) read1_header.20<1>:UB thread_id_ub {align1}; /* dispatch id */
37 shl (2) vme_m0.8<1>:UW orig_xy_ub<2,2,1>:UB 4:UW {align1}; /* (x, y) * 16 */
38 mov (1) vme_m0.20<1>:UB thread_id_ub {align1}; /* dispatch id */
40 mul (1) obw_m0.8<1>:UD w_in_mb_uw<0,1,0>:UW orig_y_ub<0,1,0>:UB {align1};
41 add (1) obw_m0.8<1>:UD obw_m0.8<0,1,0>:UD orig_x_ub<0,1,0>:UB {align1};
42 mov (1) obw_m0.20<1>:UB thread_id_ub {align1}; /* dispatch id */
45 * Media Read Message -- fetch neighbor edge pixels
49 mov (8) msg_reg0.0<1>:UD read0_header.0<8,8,1>:UD {align1};
50 send (8) msg_ind INEP_ROW<1>:UB null read(BIND_IDX_INEP, 0, 0, 4) mlen 1 rlen 1 {align1};
53 mov (8) msg_reg0.0<1>:UD read1_header.0<8,8,1>:UD {align1};
54 send (8) msg_ind INEP_COL0<1>:UB null read(BIND_IDX_INEP, 0, 0, 4) mlen 1 rlen 2 {align1};
60 mov (8) vme_msg_0.0<1>:UD vme_m0.0<8,8,1>:UD {align1};
63 mov (1) intra_flag<1>:UW 0x0:UW {align1} ;
64 cmp.nz.f0.0 (1) null<1>:UW transform_8x8_ub<0,1,0>:UB 1:UW {align1};
65 (f0.0) mov (1) intra_part_mask_ub<1>:UB LUMA_INTRA_8x8_DISABLE + LUMA_INTRA_4x4_DISABLE {align1};
67 cmp.nz.f0.0 (1) null<1>:UW orig_x_ub<0,1,0>:UB 0:UW {align1}; /* X != 0 */
68 (f0.0) add (1) mb_intra_struct_ub<1>:UB mb_intra_struct_ub<0,1,0>:UB INTRA_PRED_AVAIL_FLAG_AE {align1}; /* A */
70 cmp.nz.f0.0 (1) null<1>:UW orig_y_ub<0,1,0>:UB 0:UW {align1}; /* Y != 0 */
71 (f0.0) add (1) mb_intra_struct_ub<1>:UB mb_intra_struct_ub<0,1,0>:UB INTRA_PRED_AVAIL_FLAG_B {align1}; /* B */
73 mul.nz.f0.0 (1) null<1>:UW orig_x_ub<0,1,0>:UB orig_y_ub<0,1,0>:UB {align1}; /* X * Y != 0 */
74 (f0.0) add (1) mb_intra_struct_ub<1>:UB mb_intra_struct_ub<0,1,0>:UB INTRA_PRED_AVAIL_FLAG_D {align1}; /* D */
76 add (1) tmp_x_w<1>:W orig_x_ub<0,1,0>:UB 1:UW {align1}; /* X + 1 */
77 add (1) tmp_x_w<1>:W w_in_mb_uw<0,1,0>:UW -tmp_x_w<0,1,0>:W {align1}; /* width - (X + 1) */
78 mul.nz.f0.0 (1) null<1>:UD tmp_x_w<0,1,0>:W orig_y_ub<0,1,0>:UB {align1}; /* (width - (X + 1)) * Y != 0 */
79 (f0.0) add (1) mb_intra_struct_ub<1>:UB mb_intra_struct_ub<0,1,0>:UB INTRA_PRED_AVAIL_FLAG_C {align1}; /* C */
81 mov (8) vme_msg_1<1>:UD vme_m1.0<8,8,1>:UD {align1};
84 mov (8) vme_msg_2<1>:UD 0x0:UD {align1};
87 mov (8) vme_msg_3<1>:UD INEP_ROW.0<8,8,1>:UD {align1};
90 mov (8) vme_msg_4<1>:UD 0x0 {align1};
91 mov (16) vme_msg_4.0<1>:UB INEP_COL0.3<32,8,4>:UB {align1};
92 mov (1) vme_msg_4.16<1>:UD INTRA_PREDICTORE_MODE {align1};
102 VME_MESSAGE_TYPE_INTRA
105 rlen vme_intra_wb_length
109 * Oword Block Write message
111 mov (8) msg_reg0.0<1>:UD obw_m0<8,8,1>:UD {align1};
113 mov (1) msg_reg1.0<1>:UD vme_wb.0<0,1,0>:UD {align1};
114 mov (1) msg_reg1.4<1>:UD vme_wb.16<0,1,0>:UD {align1};
115 mov (1) msg_reg1.8<1>:UD vme_wb.20<0,1,0>:UD {align1};
116 mov (1) msg_reg1.12<1>:UD vme_wb.24<0,1,0>:UD {align1};
117 /* bind index 3, write 1 oword, msg type: 8(OWord Block Write) */
127 OBW_WRITE_COMMIT_CATEGORY,
134 add (1) read0_header.0<1>:D read0_header.0<0,1,0>:D 16:W {align1}; /* X offset: X += 16 */
135 add (1) read1_header.0<1>:D read1_header.0<0,1,0>:D 16:W {align1}; /* X offset: X += 16 */
136 add (1) vme_m0.8<1>:UW vme_m0.8<0,1,0>:UW 16:UW {align1}; /* Y = Y, X += 16 */
138 add (1) orig_x_ub<1>:ub orig_x_ub<0,1,0>:ub 1:uw {align1} ;
139 cmp.e.f0.0 (1) null<1>:uw w_in_mb_uw<0,1,0>:uw orig_x_ub<0,1,0>:ub {align1};
140 (f0.0)mov (1) orig_x_ub<1>:ub 0:uw {align1} ;
141 (f0.0)add (1) orig_y_ub<1>:ub orig_y_ub<0,1,0>:ub 1:uw {align1} ;
142 (f0.0)mov (1) read0_header.0<1>:D -8:W {align1}; /* X offset */
143 (f0.0)add (1) read0_header.4<1>:D read0_header.4<0,1,0>:D 16:W {align1}; /* Y offset */
144 (f0.0)mov (1) read1_header.0<1>:D -4:W {align1}; /* X offset */
145 (f0.0)add (1) read1_header.4<1>:D read1_header.4<0,1,0>:D 16:W {align1}; /* Y offset */
147 (f0.0)mov (1) vme_m0.8<1>:UW 0:UW {align1};
148 (f0.0)add (1) vme_m0.10<1>:UW vme_m0.10<0,1,0>:UW 16:UW {align1};
150 add (1) obw_m0.8<1>:UD obw_m0.8<0,1,0>:UD 1:uw {align1}; /* the new offset */
152 add.z.f0.0 (1) num_macroblocks<1>:w num_macroblocks<0,1,0>:w -1:w {align1} ;
153 (-f0.0)jmpi (1) __INTRA_LOOP ;
159 mov (8) msg_reg0<1>:UD r0<8,8,1>:UD {align1};
160 send (16) msg_ind acc0<1>UW null thread_spawner(0, 0, 1) mlen 1 rlen 0 {align1 EOT};