1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
|
/*
* Weighted prediction of luminance and chrominance
* Copyright © <2010>, Intel Corporation.
*
* This program is licensed under the terms and conditions of the
* Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
* http://www.opensource.org/licenses/eclipse-1.0.php.
*
*/
// Kernel name: WeightedPred.asm
//
// Weighted prediction of luminance and chrominance
//
//#if !defined(__WeightedPred__) // Make sure this is only included once
//#define __WeightedPred__
and.z.f0.0 (1) gWEIGHTFLAG:w gWPREDFLAG:ub nWBIDIR_MASK:w
cmp.e.f0.1 (1) null:w gPREDFLAG:w 2:w
(-f0.0) jmpi INTERLABEL(WeightedPred)
(f0.1) jmpi INTERLABEL(DefaultWeightedPred_BiPred)
INTERLABEL(DefaultWeightedPred_UniPred):
cmp.e.f0.0 (1) null:w gPREDFLAG:w 0:w
(f0.0) jmpi INTERLABEL(Return_WeightedPred)
// luma
mov (32) gubYPRED(0)<2> gubINTPY1(0) {Compr}
mov (32) gubYPRED(2)<2> gubINTPY1(2) {Compr}
#ifndef MONO
// chroma
mov (32) gubCPRED(0)<2> gubINTPC1(0) {Compr}
#endif
jmpi INTERLABEL(Return_WeightedPred)
INTERLABEL(DefaultWeightedPred_BiPred):
// luma
avg.sat (32) gubYPRED(0)<2> gubINTPY0(0) gubINTPY1(0) {Compr}
avg.sat (32) gubYPRED(2)<2> gubINTPY0(2) gubINTPY1(2) {Compr}
#ifndef MONO
// chroma
avg.sat (32) gubCPRED(0)<2> gubINTPC0(0) gubINTPC1(0) {Compr}
#endif
jmpi INTERLABEL(Return_WeightedPred)
INTERLABEL(WeightedPred):
cmp.e.f0.1 (1) null:w gWEIGHTFLAG:w 0x80:w
(-f0.1) jmpi INTERLABEL(WeightedPred_Explicit)
cmp.e.f0.0 (1) null:w gPREDFLAG:w 2:w
(-f0.0) jmpi INTERLABEL(DefaultWeightedPred_UniPred)
mov (2) gYADD<1>:w 32:w {NoDDClr}
mov (2) gYSHIFT<1>:w 6:w {NoDDChk}
mov (4) gOFFSET<1>:w 0:w
mov (8) gWT0<2>:w r[pWGT,0]<0;2,1>:w
jmpi INTERLABEL(WeightedPred_LOOP)
// Explicit Prediction
INTERLABEL(WeightedPred_Explicit):
// WA for weighted prediction - 2007/09/06
#ifdef SW_W_128 // CTG SW WA
cmp.e.f0.1 (8) null:ud r[pWGT,0]<8;8,1>:uw gudW128(0)<0;1,0>
#else // ILK HW solution
and.ne.f0.1 (8) null:uw r[pWGT,12]<0;1,0>:ub 0x88848421:v // Expand W=128 flag to all components. 2 MSB are don't care
#endif
asr.nz.f0.0 (2) gBIPRED<1>:w gPREDFLAG<0;1,0>:w 1:w
asr (1) gWEIGHTFLAG:w gWEIGHTFLAG:w 6:w
(-f0.0) mov (2) gPREDFLAG1<1>:w gPREDFLAG<0;1,0>:w
(f0.0) mov (2) gPREDFLAG0<1>:ud 0x00010001:ud
(-f0.0) add (2) gPREDFLAG0<1>:w -gPREDFLAG1<2;2,1>:w 1:w
// WA for weighted prediction - 2007/09/06
(f0.1) mov (8) gWT0<1>:ud 0x00000080:ud
(-f0.1) mov (8) gWT0<2>:w r[pWGT,0]<16;8,2>:b
(-f0.1) mov (8) gO0<2>:w r[pWGT,1]<16;8,2>:b
mul (16) gWT0<1>:w gWT0<16;16,1>:w gPREDFLAG0<0;4,1>:w
// Compute addition
cmp.e.f0.1 (2) null<1>:w gYWDENOM<2;2,1>:ub 0:w
(-f0.1) shl (2) gW0<1>:w gWEIGHTFLAG<0;1,0>:w gYWDENOM<2;2,1>:ub
(f0.1) mov (2) gW0<1>:w 0:w
(-f0.1) asr (2) gW0<1>:w gW0<2;2,1>:w 1:w
shl (2) gYADD<1>:w gW0<2;2,1>:w gBIPRED<0;1,0>:w
(f0.1) add (2) gYADD<1>:w gYADD<2;2,1>:w gBIPRED<0;1,0>:w
// Compute shift
add (2) gYSHIFT<1>:w gYWDENOM<2;2,1>:ub gBIPRED<0;1,0>:w
// Compute offset
add (4) acc0<1>:w gO0<16;4,4>:w gO1<16;4,4>:w
add (4) acc0<1>:w acc0<4;4,1>:w gBIPRED<0;1,0>:w
asr (4) gOFFSET<1>:w acc0<4;4,1>:w gBIPRED<0;1,0>:w
INTERLABEL(WeightedPred_LOOP):
// luma
$for(0;<4;2) {
mul (16) acc0<1>:w gubINTPY0(%1) gWT0<0;1,0>:w
mul (16) acc1<1>:w gubINTPY0(%1+1) gWT0<0;1,0>:w
mac (16) acc0<1>:w gubINTPY1(%1) gWT1<0;1,0>:w
mac (16) acc1<1>:w gubINTPY1(%1+1) gWT1<0;1,0>:w
add (16) acc0<1>:w acc0<16;16,1>:w gYADD:w
add (16) acc1<1>:w acc1<16;16,1>:w gYADD:w
// Accumulator cannot be used as destination for ASR
asr (16) gwINTERIM_BUF3(0)<1> acc0<16;16,1>:w gYSHIFT:w
asr (16) gwINTERIM_BUF3(1)<1> acc1<16;16,1>:w gYSHIFT:w
add.sat (16) gubYPRED(%1)<2> gwINTERIM_BUF3(0) gOFFSET:w
add.sat (16) gubYPRED(%1+1)<2> gwINTERIM_BUF3(1) gOFFSET:w
}
#ifndef MONO
// chroma
mul (16) acc0<1>:w gubINTPC0(0) gUW0<0;2,4>:w
mul (16) acc1<1>:w gubINTPC0(1) gUW0<0;2,4>:w
mac (16) acc0<1>:w gubINTPC1(0) gUW1<0;2,4>:w
mac (16) acc1<1>:w gubINTPC1(1) gUW1<0;2,4>:w
add (16) acc0<1>:w acc0<16;16,1>:w gCADD:w
add (16) acc1<1>:w acc1<16;16,1>:w gCADD:w
// Accumulator cannot be used as destination for ASR
asr (16) gwINTERIM_BUF3(0)<1> acc0<16;16,1>:w gCSHIFT:w
asr (16) gwINTERIM_BUF3(1)<1> acc1<16;16,1>:w gCSHIFT:w
add.sat (16) gubCPRED(0)<2> gwINTERIM_BUF3(0) gUOFFSET<0;2,1>:w
add.sat (16) gubCPRED(1)<2> gwINTERIM_BUF3(1) gUOFFSET<0;2,1>:w
#endif
INTERLABEL(Return_WeightedPred):
//#endif // !defined(__WeightedPred__)
|
