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   g58~g81:reference data
8   g82:    thread payload backup
9   g84~g107:IDCT data
10   g115:   message descriptor for reading reference data   
11*/
12mov (1) g115.8<1>UD 0x07001FUD {align1};
13and.nz (1) null g115.16<1,1,1>UW 1UW {align1};
14(f0) jmpi read_backward;
15
16send (16) 0 g40.0<1>UW g115<8,8,1>UW read(4,2,0,2) mlen 1 rlen 8 {align1};
17add (1) g115.4<1>UD g115.4<1,1,1>UD 8UD {align1};
18send (16) 0 g48.0<1>UW g115<8,8,1>UW read(4,2,0,2) mlen 1 rlen 8 {align1};
19jmpi put_data;
20
21read_backward:
22send (16) 0 g40.0<1>UW g115<8,8,1>UW read(7,2,0,2) mlen 1 rlen 8 {align1};
23add (1) g115.4<1>UD g115.4<1,1,1>UD 8UD {align1};
24send (16) 0 g48.0<1>UW g115<8,8,1>UW read(7,2,0,2) mlen 1 rlen 8 {align1};
25
26put_data:
27mov (16) g32.0<1>UW g40.0<16,16,1>UB {align1};
28mov (16) g33.0<1>UW g42.0<16,16,1>UB {align1};
29mov (16) g34.0<1>UW g44.0<16,16,1>UB {align1};
30mov (16) g35.0<1>UW g46.0<16,16,1>UB {align1};
31mov (16) g36.0<1>UW g48.0<16,16,1>UB {align1};
32mov (16) g37.0<1>UW g50.0<16,16,1>UB {align1};
33mov (16) g38.0<1>UW g52.0<16,16,1>UB {align1};
34mov (16) g39.0<1>UW g54.0<16,16,1>UB {align1};
35
36