Bart/bullet3
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Add the GPU rigid body pipeline from https://github.com/erwincoumans/experiments as a Bullet 3.x preview for Bullet 2.80

Browse Source
This commit is contained in:
erwin.coumans
2012-03-05 00:54:32 +00:00
parent 73c4646b40
commit 571af41cf6
257 changed files with 55106 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

73
Extras/RigidBodyGpuPipeline/opencl/primitives/AdlPrimitives/Copy/Copy.h Normal file
View File

@@ -0,0 +1,73 @@
/*
Copyright (c) 2012 Advanced Micro Devices, Inc.
This software is provided 'as-is', without any express or implied warranty.
In no event will the authors be held liable for any damages arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it freely,
subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
//Originally written by Takahiro Harada
#pragma once
#include <Adl/Adl.h>
#include <AdlPrimitives/Math/Math.h>
namespace adl
{
class CopyBase
{
public:
enum Option
{
PER_WI_1,
PER_WI_2,
PER_WI_4,
};
};
template<DeviceType TYPE>
class Copy : public CopyBase
{
public:
typedef Launcher::BufferInfo BufferInfo;
struct Data
{
const Device* m_device;
Kernel* m_copy1F4Kernel;
Kernel* m_copy2F4Kernel;
Kernel* m_copy4F4Kernel;
Kernel* m_copyF1Kernel;
Kernel* m_copyF2Kernel;
Buffer<int4>* m_constBuffer;
};
static
Data* allocate(const Device* deviceData);
static
void deallocate(Data* data);
static
void execute( Data* data, Buffer<float4>& dst, Buffer<float4>& src, int n, Option option = PER_WI_1);
static
void execute( Data* data, Buffer<float2>& dst, Buffer<float2>& src, int n);
static
void execute( Data* data, Buffer<float>& dst, Buffer<float>& src, int n);
};
#include <AdlPrimitives/Copy/CopyHost.inl>
#include <AdlPrimitives/Copy/Copy.inl>
};

151
Extras/RigidBodyGpuPipeline/opencl/primitives/AdlPrimitives/Copy/Copy.inl Normal file
View File

@@ -0,0 +1,151 @@
/*
Copyright (c) 2012 Advanced Micro Devices, Inc.
This software is provided 'as-is', without any express or implied warranty.
In no event will the authors be held liable for any damages arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it freely,
subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
//Originally written by Takahiro Harada
#define PATH "..\\..\\opencl\\primitives\\AdlPrimitives\\Copy\\CopyKernels"
#define KERNEL0 "Copy1F4Kernel"
#define KERNEL1 "Copy2F4Kernel"
#define KERNEL2 "Copy4F4Kernel"
#define KERNEL3 "CopyF1Kernel"
#define KERNEL4 "CopyF2Kernel"
#include <AdlPrimitives/Copy/CopyKernelsCL.h>
#include <AdlPrimitives/Copy/CopyKernelsDX11.h>
template<DeviceType TYPE>
typename Copy<TYPE>::Data* Copy<TYPE>::allocate( const Device* device )
{
ADLASSERT( TYPE == device->m_type );
const char* src[] =
#if defined(ADL_LOAD_KERNEL_FROM_STRING)
{copyKernelsCL, copyKernelsDX11};
// ADLASSERT(0);
#else
{0,0};
#endif
Data* data = new Data;
data->m_device = device;
data->m_copy1F4Kernel = device->getKernel( PATH, KERNEL0, 0, src[TYPE] );
data->m_copy2F4Kernel = device->getKernel( PATH, KERNEL1, 0, src[TYPE] );
data->m_copy4F4Kernel = device->getKernel( PATH, KERNEL2, 0, src[TYPE] );
data->m_copyF1Kernel = device->getKernel( PATH, KERNEL3, 0, src[TYPE] );
data->m_copyF2Kernel = device->getKernel( PATH, KERNEL4, 0, src[TYPE] );
data->m_constBuffer = new Buffer<int4>( device, 1, BufferBase::BUFFER_CONST );
return data;
}
template<DeviceType TYPE>
void Copy<TYPE>::deallocate( Data* data )
{
delete data->m_constBuffer;
delete data;
}
template<DeviceType TYPE>
void Copy<TYPE>::execute( Data* data, Buffer<float4>& dst, Buffer<float4>& src, int n, Option option )
{
ADLASSERT( TYPE == dst.getType() );
ADLASSERT( TYPE == src.getType() );
int4 constBuffer;
constBuffer.x = n;
switch (option)
{
case PER_WI_1:
{
BufferInfo bInfo[] = { BufferInfo( &dst ), BufferInfo( &src, true ) };
Launcher launcher( data->m_device, data->m_copy1F4Kernel );
launcher.setBuffers( bInfo, sizeof(bInfo)/sizeof(Launcher::BufferInfo) );
launcher.setConst( *data->m_constBuffer, constBuffer );
launcher.launch1D( n/1 );
}
break;
case PER_WI_2:
{
ADLASSERT( n%2 == 0 );
BufferInfo bInfo[] = { BufferInfo( &dst ), BufferInfo( &src, true ) };
Launcher launcher( data->m_device, data->m_copy2F4Kernel );
launcher.setBuffers( bInfo, sizeof(bInfo)/sizeof(Launcher::BufferInfo) );
launcher.setConst( *data->m_constBuffer, constBuffer );
launcher.launch1D( n/2 );
}
break;
case PER_WI_4:
{
ADLASSERT( n%4 == 0 );
BufferInfo bInfo[] = { BufferInfo( &dst ), BufferInfo( &src, true ) };
Launcher launcher( data->m_device, data->m_copy4F4Kernel );
launcher.setBuffers( bInfo, sizeof(bInfo)/sizeof(Launcher::BufferInfo) );
launcher.setConst( *data->m_constBuffer, constBuffer );
launcher.launch1D( n/4 );
}
break;
default:
ADLASSERT(0);
break;
};
}
template<DeviceType TYPE>
void Copy<TYPE>::execute( Data* data, Buffer<float2>& dst, Buffer<float2>& src, int n )
{
ADLASSERT( TYPE == dst.getType() );
ADLASSERT( TYPE == src.getType() );
int4 constBuffer;
constBuffer.x = n;
BufferInfo bInfo[] = { BufferInfo( &dst ), BufferInfo( &src, true ) };
Launcher launcher( data->m_device, data->m_copyF2Kernel );
launcher.setBuffers( bInfo, sizeof(bInfo)/sizeof(Launcher::BufferInfo) );
launcher.setConst( *data->m_constBuffer, constBuffer );
launcher.launch1D( n/1 );
}
template<DeviceType TYPE>
void Copy<TYPE>::execute( Data* data, Buffer<float>& dst, Buffer<float>& src, int n )
{
ADLASSERT( TYPE == dst.getType() );
ADLASSERT( TYPE == src.getType() );
int4 constBuffer;
constBuffer.x = n;
BufferInfo bInfo[] = { BufferInfo( &dst ), BufferInfo( &src, true ) };
Launcher launcher( data->m_device, data->m_copyF1Kernel );
launcher.setBuffers( bInfo, sizeof(bInfo)/sizeof(Launcher::BufferInfo) );
launcher.setConst( *data->m_constBuffer, constBuffer );
launcher.launch1D( n/1 );
}
#undef PATH
#undef KERNEL0
#undef KERNEL1
#undef KERNEL2
#undef KERNEL3
#undef KERNEL4

85
Extras/RigidBodyGpuPipeline/opencl/primitives/AdlPrimitives/Copy/CopyHost.inl Normal file
View File

@@ -0,0 +1,85 @@
/*
Copyright (c) 2012 Advanced Micro Devices, Inc.
This software is provided 'as-is', without any express or implied warranty.
In no event will the authors be held liable for any damages arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it freely,
subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
//Originally written by Takahiro Harada
template<>
class Copy<TYPE_HOST> : public CopyBase
{
public:
typedef Launcher::BufferInfo BufferInfo;
struct Data
{
};
static
Data* allocate(const Device* deviceData)
{
ADLASSERT( TYPE_HOST == deviceData->m_type );
return 0;
}
static
void deallocate(Data* data)
{
return;
}
static
void execute( Data* data, Buffer<float4>& dst, Buffer<float4>& src, int n, Option option = PER_WI_1)
{
ADLASSERT( TYPE_HOST == dst.getType() );
ADLASSERT( TYPE_HOST == src.getType() );
HostBuffer<float4>& dstH = (HostBuffer<float4>&)dst;
HostBuffer<float4>& srcH = (HostBuffer<float4>&)src;
for(int i=0; i<n; i++)
{
dstH[i] = srcH[i];
}
}
static
void execute( Data* data, Buffer<float2>& dst, Buffer<float2>& src, int n)
{
ADLASSERT( TYPE_HOST == dst.getType() );
ADLASSERT( TYPE_HOST == src.getType() );
HostBuffer<float2>& dstH = (HostBuffer<float2>&)dst;
HostBuffer<float2>& srcH = (HostBuffer<float2>&)src;
for(int i=0; i<n; i++)
{
dstH[i] = srcH[i];
}
}
static
void execute( Data* data, Buffer<float>& dst, Buffer<float>& src, int n)
{
ADLASSERT( TYPE_HOST == dst.getType() );
ADLASSERT( TYPE_HOST == src.getType() );
HostBuffer<float>& dstH = (HostBuffer<float>&)dst;
HostBuffer<float>& srcH = (HostBuffer<float>&)src;
for(int i=0; i<n; i++)
{
dstH[i] = srcH[i];
}
}
};

128
Extras/RigidBodyGpuPipeline/opencl/primitives/AdlPrimitives/Copy/CopyKernels.cl Normal file
View File

@@ -0,0 +1,128 @@
/*
Copyright (c) 2012 Advanced Micro Devices, Inc.
This software is provided 'as-is', without any express or implied warranty.
In no event will the authors be held liable for any damages arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it freely,
subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
//Originally written by Takahiro Harada
#pragma OPENCL EXTENSION cl_amd_printf : enable
#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable
typedef unsigned int u32;
#define GET_GROUP_IDX get_group_id(0)
#define GET_LOCAL_IDX get_local_id(0)
#define GET_GLOBAL_IDX get_global_id(0)
#define GET_GROUP_SIZE get_local_size(0)
#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)
#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)
#define AtomInc(x) atom_inc(&(x))
#define AtomInc1(x, out) out = atom_inc(&(x))
#define make_uint4 (uint4)
#define make_uint2 (uint2)
#define make_int2 (int2)
typedef struct
{
int m_n;
int m_padding[3];
} ConstBuffer;
__kernel
__attribute__((reqd_work_group_size(64,1,1)))
void Copy1F4Kernel(__global float4* dst, __global float4* src,
ConstBuffer cb)
{
int gIdx = GET_GLOBAL_IDX;
if( gIdx < cb.m_n )
{
float4 a0 = src[gIdx];
dst[ gIdx ] = a0;
}
}
__kernel
__attribute__((reqd_work_group_size(64,1,1)))
void Copy2F4Kernel(__global float4* dst, __global float4* src,
ConstBuffer cb)
{
int gIdx = GET_GLOBAL_IDX;
if( 2*gIdx <= cb.m_n )
{
float4 a0 = src[gIdx*2+0];
float4 a1 = src[gIdx*2+1];
dst[ gIdx*2+0 ] = a0;
dst[ gIdx*2+1 ] = a1;
}
}
__kernel
__attribute__((reqd_work_group_size(64,1,1)))
void Copy4F4Kernel(__global float4* dst, __global float4* src,
ConstBuffer cb)
{
int gIdx = GET_GLOBAL_IDX;
if( 4*gIdx <= cb.m_n )
{
int idx0 = gIdx*4+0;
int idx1 = gIdx*4+1;
int idx2 = gIdx*4+2;
int idx3 = gIdx*4+3;
float4 a0 = src[idx0];
float4 a1 = src[idx1];
float4 a2 = src[idx2];
float4 a3 = src[idx3];
dst[ idx0 ] = a0;
dst[ idx1 ] = a1;
dst[ idx2 ] = a2;
dst[ idx3 ] = a3;
}
}
__kernel
__attribute__((reqd_work_group_size(64,1,1)))
void CopyF1Kernel(__global float* dstF1, __global float* srcF1,
ConstBuffer cb)
{
int gIdx = GET_GLOBAL_IDX;
if( gIdx < cb.m_n )
{
float a0 = srcF1[gIdx];
dstF1[ gIdx ] = a0;
}
}
__kernel
__attribute__((reqd_work_group_size(64,1,1)))
void CopyF2Kernel(__global float2* dstF2, __global float2* srcF2,
ConstBuffer cb)
{
int gIdx = GET_GLOBAL_IDX;
if( gIdx < cb.m_n )
{
float2 a0 = srcF2[gIdx];
dstF2[ gIdx ] = a0;
}
}

130
Extras/RigidBodyGpuPipeline/opencl/primitives/AdlPrimitives/Copy/CopyKernels.hlsl Normal file
View File

@@ -0,0 +1,130 @@
/*
Copyright (c) 2012 Advanced Micro Devices, Inc.
This software is provided 'as-is', without any express or implied warranty.
In no event will the authors be held liable for any damages arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it freely,
subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not claim that you wrote the original software. If you use this software in a product, an acknowledgment in the product documentation would be appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
*/
//Originally written by Takahiro Harada
typedef uint u32;
#define GET_GROUP_IDX groupIdx.x
#define GET_LOCAL_IDX localIdx.x
#define GET_GLOBAL_IDX globalIdx.x
#define GROUP_LDS_BARRIER GroupMemoryBarrierWithGroupSync()
#define GROUP_MEM_FENCE
#define DEFAULT_ARGS uint3 globalIdx : SV_DispatchThreadID, uint3 localIdx : SV_GroupThreadID, uint3 groupIdx : SV_GroupID
#define AtomInc(x) InterlockedAdd(x, 1)
#define AtomInc1(x, out) InterlockedAdd(x, 1, out)
#define make_uint4 uint4
#define make_uint2 uint2
#define make_int2 int2
#define WG_SIZE 64
#define GET_GROUP_SIZE WG_SIZE
cbuffer CB : register( b0 )
{
int m_n;
int m_padding[3];
};
RWStructuredBuffer<float4> dst : register( u0 );
StructuredBuffer<float4> src : register( t0 );
[numthreads(WG_SIZE, 1, 1)]
void Copy1F4Kernel( DEFAULT_ARGS )
{
int gIdx = GET_GLOBAL_IDX;
if( gIdx < m_n )
{
float4 a0 = src[gIdx];
dst[ gIdx ] = a0;
}
}
[numthreads(WG_SIZE, 1, 1)]
void Copy2F4Kernel( DEFAULT_ARGS )
{
int gIdx = GET_GLOBAL_IDX;
if( 2*gIdx <= m_n )
{
float4 a0 = src[gIdx*2+0];
float4 a1 = src[gIdx*2+1];
dst[ gIdx*2+0 ] = a0;
dst[ gIdx*2+1 ] = a1;
}
}
[numthreads(WG_SIZE, 1, 1)]
void Copy4F4Kernel( DEFAULT_ARGS )
{
int gIdx = GET_GLOBAL_IDX;
if( 4*gIdx <= m_n )
{
int idx0 = gIdx*4+0;
int idx1 = gIdx*4+1;
int idx2 = gIdx*4+2;
int idx3 = gIdx*4+3;
float4 a0 = src[idx0];
float4 a1 = src[idx1];
float4 a2 = src[idx2];
float4 a3 = src[idx3];
dst[ idx0 ] = a0;
dst[ idx1 ] = a1;
dst[ idx2 ] = a2;
dst[ idx3 ] = a3;
}
}
RWStructuredBuffer<float> dstF1 : register( u0 );
StructuredBuffer<float> srcF1 : register( t0 );
[numthreads(WG_SIZE, 1, 1)]
void CopyF1Kernel( DEFAULT_ARGS )
{
int gIdx = GET_GLOBAL_IDX;
if( gIdx < m_n )
{
float a0 = srcF1[gIdx];
dstF1[ gIdx ] = a0;
}
}
RWStructuredBuffer<float2> dstF2 : register( u0 );
StructuredBuffer<float2> srcF2 : register( t0 );
[numthreads(WG_SIZE, 1, 1)]
void CopyF2Kernel( DEFAULT_ARGS )
{
int gIdx = GET_GLOBAL_IDX;
if( gIdx < m_n )
{
float2 a0 = srcF2[gIdx];
dstF2[ gIdx ] = a0;
}
}

119
Extras/RigidBodyGpuPipeline/opencl/primitives/AdlPrimitives/Copy/CopyKernelsCL.h Normal file
View File

@@ -0,0 +1,119 @@
static const char* copyKernelsCL= \
"/*\n"
" 2011 Takahiro Harada\n"
"*/\n"
"\n"
"#pragma OPENCL EXTENSION cl_amd_printf : enable\n"
"#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable\n"
"\n"
"typedef unsigned int u32;\n"
"#define GET_GROUP_IDX get_group_id(0)\n"
"#define GET_LOCAL_IDX get_local_id(0)\n"
"#define GET_GLOBAL_IDX get_global_id(0)\n"
"#define GET_GROUP_SIZE get_local_size(0)\n"
"#define GROUP_LDS_BARRIER barrier(CLK_LOCAL_MEM_FENCE)\n"
"#define GROUP_MEM_FENCE mem_fence(CLK_LOCAL_MEM_FENCE)\n"
"#define AtomInc(x) atom_inc(&(x))\n"
"#define AtomInc1(x, out) out = atom_inc(&(x))\n"
"\n"
"#define make_uint4 (uint4)\n"
"#define make_uint2 (uint2)\n"
"#define make_int2 (int2)\n"
"\n"
"typedef struct\n"
"{\n"
" int m_n;\n"
" int m_padding[3];\n"
"} ConstBuffer;\n"
"\n"
"\n"
"\n"
"__kernel\n"
"__attribute__((reqd_work_group_size(64,1,1)))\n"
"void Copy1F4Kernel(__global float4* dst, __global float4* src, \n"
" ConstBuffer cb)\n"
"{\n"
" int gIdx = GET_GLOBAL_IDX;\n"
"\n"
" if( gIdx < cb.m_n )\n"
" {\n"
" float4 a0 = src[gIdx];\n"
"\n"
" dst[ gIdx ] = a0;\n"
" }\n"
"}\n"
"\n"
"__kernel\n"
"__attribute__((reqd_work_group_size(64,1,1)))\n"
"void Copy2F4Kernel(__global float4* dst, __global float4* src, \n"
" ConstBuffer cb)\n"
"{\n"
" int gIdx = GET_GLOBAL_IDX;\n"
"\n"
" if( 2*gIdx <= cb.m_n )\n"
" {\n"
" float4 a0 = src[gIdx*2+0];\n"
" float4 a1 = src[gIdx*2+1];\n"
"\n"
" dst[ gIdx*2+0 ] = a0;\n"
" dst[ gIdx*2+1 ] = a1;\n"
" }\n"
"}\n"
"\n"
"__kernel\n"
"__attribute__((reqd_work_group_size(64,1,1)))\n"
"void Copy4F4Kernel(__global float4* dst, __global float4* src, \n"
" ConstBuffer cb)\n"
"{\n"
" int gIdx = GET_GLOBAL_IDX;\n"
"\n"
" if( 4*gIdx <= cb.m_n )\n"
" {\n"
" int idx0 = gIdx*4+0;\n"
" int idx1 = gIdx*4+1;\n"
" int idx2 = gIdx*4+2;\n"
" int idx3 = gIdx*4+3;\n"
"\n"
" float4 a0 = src[idx0];\n"
" float4 a1 = src[idx1];\n"
" float4 a2 = src[idx2];\n"
" float4 a3 = src[idx3];\n"
"\n"
" dst[ idx0 ] = a0;\n"
" dst[ idx1 ] = a1;\n"
" dst[ idx2 ] = a2;\n"
" dst[ idx3 ] = a3;\n"
" }\n"
"}\n"
"\n"
"__kernel\n"
"__attribute__((reqd_work_group_size(64,1,1)))\n"
"void CopyF1Kernel(__global float* dstF1, __global float* srcF1, \n"
" ConstBuffer cb)\n"
"{\n"
" int gIdx = GET_GLOBAL_IDX;\n"
"\n"
" if( gIdx < cb.m_n )\n"
" {\n"
" float a0 = srcF1[gIdx];\n"
"\n"
" dstF1[ gIdx ] = a0;\n"
" }\n"
"}\n"
"\n"
"__kernel\n"
"__attribute__((reqd_work_group_size(64,1,1)))\n"
"void CopyF2Kernel(__global float2* dstF2, __global float2* srcF2, \n"
" ConstBuffer cb)\n"
"{\n"
" int gIdx = GET_GLOBAL_IDX;\n"
"\n"
" if( gIdx < cb.m_n )\n"
" {\n"
" float2 a0 = srcF2[gIdx];\n"
"\n"
" dstF2[ gIdx ] = a0;\n"
" }\n"
"}\n"
"\n"
;

120
Extras/RigidBodyGpuPipeline/opencl/primitives/AdlPrimitives/Copy/CopyKernelsDX11.h Normal file
View File

@@ -0,0 +1,120 @@
static const char* copyKernelsDX11= \
"/*\n"
" 2011 Takahiro Harada\n"
"*/\n"
"\n"
"typedef uint u32;\n"
"\n"
"#define GET_GROUP_IDX groupIdx.x\n"
"#define GET_LOCAL_IDX localIdx.x\n"
"#define GET_GLOBAL_IDX globalIdx.x\n"
"#define GROUP_LDS_BARRIER GroupMemoryBarrierWithGroupSync()\n"
"#define GROUP_MEM_FENCE\n"
"#define DEFAULT_ARGS uint3 globalIdx : SV_DispatchThreadID, uint3 localIdx : SV_GroupThreadID, uint3 groupIdx : SV_GroupID\n"
"#define AtomInc(x) InterlockedAdd(x, 1)\n"
"#define AtomInc1(x, out) InterlockedAdd(x, 1, out)\n"
"\n"
"#define make_uint4 uint4\n"
"#define make_uint2 uint2\n"
"#define make_int2 int2\n"
"\n"
"#define WG_SIZE 64\n"
"\n"
"#define GET_GROUP_SIZE WG_SIZE\n"
"\n"
"\n"
"\n"
"cbuffer CB : register( b0 )\n"
"{\n"
" int m_n;\n"
" int m_padding[3];\n"
"};\n"
"\n"
"RWStructuredBuffer<float4> dst : register( u0 );\n"
"StructuredBuffer<float4> src : register( t0 );\n"
"\n"
"[numthreads(WG_SIZE, 1, 1)]\n"
"void Copy1F4Kernel( DEFAULT_ARGS )\n"
"{\n"
" int gIdx = GET_GLOBAL_IDX;\n"
"\n"
" if( gIdx < m_n )\n"
" {\n"
" float4 a0 = src[gIdx];\n"
"\n"
" dst[ gIdx ] = a0;\n"
" }\n"
"}\n"
"\n"
"[numthreads(WG_SIZE, 1, 1)]\n"
"void Copy2F4Kernel( DEFAULT_ARGS )\n"
"{\n"
" int gIdx = GET_GLOBAL_IDX;\n"
"\n"
" if( 2*gIdx <= m_n )\n"
" {\n"
" float4 a0 = src[gIdx*2+0];\n"
" float4 a1 = src[gIdx*2+1];\n"
"\n"
" dst[ gIdx*2+0 ] = a0;\n"
" dst[ gIdx*2+1 ] = a1;\n"
" }\n"
"}\n"
"\n"
"[numthreads(WG_SIZE, 1, 1)]\n"
"void Copy4F4Kernel( DEFAULT_ARGS )\n"
"{\n"
" int gIdx = GET_GLOBAL_IDX;\n"
"\n"
" if( 4*gIdx <= m_n )\n"
" {\n"
" int idx0 = gIdx*4+0;\n"
" int idx1 = gIdx*4+1;\n"
" int idx2 = gIdx*4+2;\n"
" int idx3 = gIdx*4+3;\n"
"\n"
" float4 a0 = src[idx0];\n"
" float4 a1 = src[idx1];\n"
" float4 a2 = src[idx2];\n"
" float4 a3 = src[idx3];\n"
"\n"
" dst[ idx0 ] = a0;\n"
" dst[ idx1 ] = a1;\n"
" dst[ idx2 ] = a2;\n"
" dst[ idx3 ] = a3;\n"
" }\n"
"}\n"
"\n"
"RWStructuredBuffer<float> dstF1 : register( u0 );\n"
"StructuredBuffer<float> srcF1 : register( t0 );\n"
"\n"
"[numthreads(WG_SIZE, 1, 1)]\n"
"void CopyF1Kernel( DEFAULT_ARGS )\n"
"{\n"
" int gIdx = GET_GLOBAL_IDX;\n"
"\n"
" if( gIdx < m_n )\n"
" {\n"
" float a0 = srcF1[gIdx];\n"
"\n"
" dstF1[ gIdx ] = a0;\n"
" }\n"
"\n"
"}\n"
"\n"
"RWStructuredBuffer<float2> dstF2 : register( u0 );\n"
"StructuredBuffer<float2> srcF2 : register( t0 );\n"
"\n"
"[numthreads(WG_SIZE, 1, 1)]\n"
"void CopyF2Kernel( DEFAULT_ARGS )\n"
"{\n"
" int gIdx = GET_GLOBAL_IDX;\n"
"\n"
" if( gIdx < m_n )\n"
" {\n"
" float2 a0 = srcF2[gIdx];\n"
"\n"
" dstF2[ gIdx ] = a0;\n"
" }\n"
"}\n"
;