1/* GRF allocation:
2   g1~g30: constant buffer
3           g1~g2:intra IQ matrix
4           g3~g4:non intra IQ matrix
5           g5~g20:IDCT table
6   g31:    thread payload 
7   g32:    message descriptor for reading reference data
8   g58~g81:reference data
9   g82:    thread payload backup
10   g83~g106:IDCT data                           
11*/
12mov (1) g32.8<1>UD 0x007001FUD {align1};
13and.nz (1) null g32.16<1,1,1>UW 1UW {align1};
14(f0) jmpi read_backward;
15
16send (16) 0 g38.0<1>UW g32<8,8,1>UW read(4, 2, 0, 2) mlen 1 rlen 8 {align1};
17add (1) g32.4<1>UD g32.4<1,1,1>UD 8UD {align1};
18send (16) 0 g46.0<1>UW g32<8,8,1>UW read(4, 2, 0, 2) mlen 1 rlen 8 {align1};
19jmpi put_data;
20
21read_backward:
22send (16) 0 g38.0<1>UW g32<8,8,1>UW read(7, 2, 0, 2) mlen 1 rlen 8 {align1};
23add (1) g32.4<1>UD g32.4<1,1,1>UD 8UD {align1};
24send (16) 0 g46.0<1>UW g32<8,8,1>UW read(7, 2, 0, 2) mlen 1 rlen 8 {align1};
25
26put_data:
27mov (32) g58.0<1>UW g38.0<16,16,1>UB {align1 compr};
28mov (32) g60.0<1>UW g40.0<16,16,1>UB {align1 compr};
29mov (32) g62.0<1>UW g42.0<16,16,1>UB {align1 compr};
30mov (32) g64.0<1>UW g44.0<16,16,1>UB {align1 compr};
31mov (32) g66.0<1>UW g46.0<16,16,1>UB {align1 compr};
32mov (32) g68.0<1>UW g48.0<16,16,1>UB {align1 compr};
33mov (32) g70.0<1>UW g50.0<16,16,1>UB {align1 compr};
34mov (32) g72.0<1>UW g52.0<16,16,1>UB {align1 compr};
35
36