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
|
//TEST:COMPARE_HLSL: -target dxbc-assembly -profile cs_4_0 -entry CSScanInBucket -entry CSScanBucketResult -entry CSScanAddBucketResult
//--------------------------------------------------------------------------------------
// File: ScanCS.hlsl
//
// A simple inclusive prefix sum(scan) implemented in CS4.0,
// using a typical up sweep and down sweep scheme
//
// Copyright (c) Microsoft Corporation. All rights reserved.
//--------------------------------------------------------------------------------------
StructuredBuffer<uint2> Input : register( t0 ); // Change uint2 here if scan other types, and
RWStructuredBuffer<uint2> Result : register( u0 ); // also here
#define groupthreads 128
groupshared uint4 bucket[groupthreads]; // Change uint4 to the "type x2" if scan other types, e.g.
// if scan uint2, then put uint4 here,
// if scan float, then put float2 here
void CSScan( uint3 DTid, uint GI, uint2 x ) // Change the type of x here if scan other types
{
// since CS40 can only support one shared memory for one shader, we use .xy and .zw as ping-ponging buffers
// if scan a single element type like int, search and replace all .xy to .x and .zw to .y below
bucket[GI].xy = x;
bucket[GI].zw = 0;
// Up sweep
[unroll]
for ( uint stride = 2; stride <= groupthreads; stride <<= 1 )
{
GroupMemoryBarrierWithGroupSync();
if ( (GI & (stride - 1)) == (stride - 1) )
{
bucket[GI].xy += bucket[GI - stride/2].xy;
}
}
if ( GI == (groupthreads - 1) )
{
bucket[GI].xy = 0;
}
// Down sweep
bool n = true;
[unroll]
for ( stride = groupthreads / 2; stride >= 1; stride >>= 1 )
{
GroupMemoryBarrierWithGroupSync();
uint a = stride - 1;
uint b = stride | a;
if ( n ) // ping-pong between passes
{
if ( ( GI & b) == b )
{
bucket[GI].zw = bucket[GI-stride].xy + bucket[GI].xy;
} else
if ( (GI & a) == a )
{
bucket[GI].zw = bucket[GI+stride].xy;
} else
{
bucket[GI].zw = bucket[GI].xy;
}
} else
{
if ( ( GI & b) == b )
{
bucket[GI].xy = bucket[GI-stride].zw + bucket[GI].zw;
} else
if ( (GI & a) == a )
{
bucket[GI].xy = bucket[GI+stride].zw;
} else
{
bucket[GI].xy = bucket[GI].zw;
}
}
n = !n;
}
Result[DTid.x] = bucket[GI].zw + x;
}
// scan in each bucket
[numthreads( groupthreads, 1, 1 )]
void CSScanInBucket( uint3 DTid : SV_DispatchThreadID, uint3 GTid : SV_GroupThreadID, uint GI: SV_GroupIndex )
{
uint2 x = Input[DTid.x]; // Change the type of x here if scan other types
CSScan( DTid, GI, x );
}
// record and scan the sum of each bucket
[numthreads( groupthreads, 1, 1 )]
void CSScanBucketResult( uint3 DTid : SV_DispatchThreadID, uint3 GTid : SV_GroupThreadID, uint GI: SV_GroupIndex )
{
uint2 x = Input[DTid.x*groupthreads - 1]; // Change the type of x here if scan other types
CSScan( DTid, GI, x );
}
StructuredBuffer<uint2> Input1 : register( t1 );
// add the bucket scanned result to each bucket to get the final result
[numthreads( groupthreads, 1, 1 )]
void CSScanAddBucketResult( uint3 Gid : SV_GroupID, uint3 DTid : SV_DispatchThreadID, uint3 GTid : SV_GroupThreadID, uint GI: SV_GroupIndex )
{
Result[DTid.x] = Input[DTid.x] + Input1[Gid.x];
}
|
