mirror of
https://mirror.ghproxy.com/https://github.com/dexyfex/CodeWalker
synced 2024-11-26 00:43:00 +08:00
73 lines
2.5 KiB
HLSL
73 lines
2.5 KiB
HLSL
|
//VERTICAL FILTER ONLY
|
||
|
//--------------------------------------------------------------------------------------
|
||
|
// File: FilterCS.hlsl
|
||
|
//
|
||
|
// The CSs for doing vertical and horizontal blur, used in CS path of
|
||
|
// HDRToneMappingCS11 sample
|
||
|
//
|
||
|
// Copyright (c) Microsoft Corporation. All rights reserved.
|
||
|
//--------------------------------------------------------------------------------------
|
||
|
StructuredBuffer<float4> InputBuf : register( t0 );
|
||
|
Texture2D InputTex : register( t1 );
|
||
|
RWStructuredBuffer<float4> Result : register( u0 );
|
||
|
|
||
|
cbuffer cb0
|
||
|
{
|
||
|
float4 g_avSampleWeights[15];
|
||
|
int2 g_outputsize;
|
||
|
int2 g_inputsize;
|
||
|
}
|
||
|
|
||
|
#define kernelhalf 7
|
||
|
#define groupthreads 128
|
||
|
groupshared float4 temp[groupthreads];
|
||
|
|
||
|
[numthreads( groupthreads, 1, 1 )]
|
||
|
void main( uint3 Gid : SV_GroupID, uint GI : SV_GroupIndex )
|
||
|
{
|
||
|
int offsety = GI - kernelhalf + (groupthreads - kernelhalf * 2) * Gid.y;
|
||
|
offsety = clamp( offsety, 0, g_inputsize.y-1 );
|
||
|
int offset = Gid.x + offsety * g_inputsize.x;
|
||
|
temp[GI] = InputBuf[offset];
|
||
|
|
||
|
GroupMemoryBarrierWithGroupSync();
|
||
|
|
||
|
// Vertical blur
|
||
|
if ( GI >= kernelhalf &&
|
||
|
GI < (groupthreads - kernelhalf) &&
|
||
|
( (GI - kernelhalf + (groupthreads - kernelhalf * 2) * Gid.y) < (uint)g_outputsize.y) )
|
||
|
{
|
||
|
float4 vOut = 0;
|
||
|
|
||
|
[unroll]
|
||
|
for ( int i = -kernelhalf; i <= kernelhalf; ++i )
|
||
|
vOut += temp[GI + i] * g_avSampleWeights[i + kernelhalf];
|
||
|
|
||
|
Result[Gid.x + (GI - kernelhalf + (groupthreads - kernelhalf * 2) * Gid.y) * g_outputsize.x] = float4(vOut.rgb, 1.0f);
|
||
|
}
|
||
|
}
|
||
|
|
||
|
//[numthreads( groupthreads, 1, 1 )]
|
||
|
//void CSHorizFilter( uint3 Gid : SV_GroupID, uint GI : SV_GroupIndex )
|
||
|
//{
|
||
|
// int2 coord = int2( GI - kernelhalf + (groupthreads - kernelhalf * 2) * Gid.x, Gid.y );
|
||
|
// coord = clamp( coord, int2(0, 0), int2(g_inputsize.x-1, g_inputsize.y-1) );
|
||
|
// temp[GI] = InputTex.Load( int3(coord, 0) );
|
||
|
//
|
||
|
// GroupMemoryBarrierWithGroupSync();
|
||
|
//
|
||
|
// // Horizontal blur
|
||
|
// if ( GI >= kernelhalf &&
|
||
|
// GI < (groupthreads - kernelhalf) &&
|
||
|
// ( (Gid.x * (groupthreads - 2 * kernelhalf) + GI - kernelhalf) < g_outputsize.x) )
|
||
|
// {
|
||
|
// float4 vOut = 0;
|
||
|
//
|
||
|
// [unroll]
|
||
|
// for ( int i = -kernelhalf; i <= kernelhalf; ++i )
|
||
|
// vOut += temp[GI + i] * g_avSampleWeights[i + kernelhalf];
|
||
|
//
|
||
|
// Result[GI - kernelhalf + (groupthreads - kernelhalf * 2) * Gid.x + Gid.y * g_outputsize.x] = float4(vOut.rgb, 1.0f);
|
||
|
// }
|
||
|
//}
|