2
g1~g30: constant buffer
4
g3~g4:non intra IQ matrix
7
g32: message descriptor for reading reference data
9
g82: thread payload backup
11
mov (8) g82.0<1>UD g31.0<8,8,1>UD {align1};
12
mov (1) g126.8<1>UD ip {align1};
13
add (1) ip g21.0<1,1,1>UD 0x50UD {align1}; //jump to the lib to do IDCT
15
//Y, (x', y') = (x, y) + (motion_vector.x >> 1, motion_vector.y >> 1)
16
asr (2) g31.14<1>W g82.14<2,2,1>W 1W {align1};
17
add (2) g32.0<1>UD g31.0<2,2,1>UD g31.14<2,2,1>W {align1};
18
mov (1) g32.16<1>UW 0UW {align1}; //0:forward 1:backward
19
mov (1) a0.0<1>UD 0x0A4EUD {align1}; //g82.14,motion vector
20
mov (1) g126.8<1>UD ip {align1};
21
add (1) ip g21.0<1,1,1>UD 0x00UD {align1}; //jump to the lib to read reference data
23
//UV, (x', y') = (x >> 1, y >> 1) + (motion_vector.x >> 2, motion_vector.y >> 2)
24
shr (2) g31.0<1>UD g31.0<2,2,1>UD 1UD {align1};
25
asr (2) g31.14<1>W g82.14<2,2,1>W 2W {align1};
26
add (2) g32.0<1>UD g31.0<2,2,1>UD g31.14<2,2,1>W {align1};
27
mov (1) g126.8<1>UD ip {align1};
28
add (1) ip g21.0<1,1,1>UD 0x10UD {align1}; //jump to the lib to read reference data
30
add (1) ip g21.0<1,1,1>UD 0x40UD {align1}; //jump to the lib to add the reference and idct data