318 lines
13 KiB
C++
318 lines
13 KiB
C++
// Redistribution and use in source and binary forms, with or without
|
|
// modification, are permitted provided that the following conditions
|
|
// are met:
|
|
// * Redistributions of source code must retain the above copyright
|
|
// notice, this list of conditions and the following disclaimer.
|
|
// * Redistributions in binary form must reproduce the above copyright
|
|
// notice, this list of conditions and the following disclaimer in the
|
|
// documentation and/or other materials provided with the distribution.
|
|
// * Neither the name of NVIDIA CORPORATION nor the names of its
|
|
// contributors may be used to endorse or promote products derived
|
|
// from this software without specific prior written permission.
|
|
//
|
|
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ''AS IS'' AND ANY
|
|
// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
|
// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
|
// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
|
// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
|
// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
|
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
|
// OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
|
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
|
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
//
|
|
// Copyright (c) 2008-2025 NVIDIA Corporation. All rights reserved.
|
|
// Copyright (c) 2004-2008 AGEIA Technologies, Inc. All rights reserved.
|
|
// Copyright (c) 2001-2004 NovodeX AG. All rights reserved.
|
|
|
|
#ifndef PX_CUDA_HELPERS_EXT_H
|
|
#define PX_CUDA_HELPERS_EXT_H
|
|
|
|
|
|
#include "foundation/PxPreprocessor.h"
|
|
|
|
#if PX_SUPPORT_GPU_PHYSX
|
|
#include "foundation/PxSimpleTypes.h"
|
|
#include "foundation/PxAssert.h"
|
|
#include "foundation/PxFoundation.h"
|
|
#include "cudamanager/PxCudaContextManager.h"
|
|
#include "cudamanager/PxCudaContext.h"
|
|
#include "cudamanager/PxCudaTypes.h"
|
|
|
|
|
|
namespace physx
|
|
{
|
|
namespace Ext
|
|
{
|
|
|
|
class PxCudaHelpersExt
|
|
{
|
|
public:
|
|
|
|
/**
|
|
* \brief Allocates a device buffer and returns the pointer to the memory
|
|
*
|
|
* The cuda context will get acquired automatically
|
|
*
|
|
* \param[in] cudaContextManager the PxCudaContextManager managing the CUDA context the allocation should be attributed to.
|
|
* \param[in] numElements the number of elements of type T to allocate.
|
|
*
|
|
* \return a pointer to the allocated memory.
|
|
*/
|
|
template<typename T>
|
|
static T* allocDeviceBuffer(PxCudaContextManager& cudaContextManager, PxU64 numElements)
|
|
{
|
|
PxScopedCudaLock lock(cudaContextManager);
|
|
|
|
CUdeviceptr ptr = 0;
|
|
PxCUresult result = cudaContextManager.getCudaContext()->memAlloc(&ptr, numElements * sizeof(T));
|
|
if (result != 0)
|
|
PxGetFoundation().error(PxErrorCode::eINTERNAL_ERROR, PX_FL, "allocDeviceBuffer failed with error code %i!\n", PxI32(result));
|
|
|
|
return reinterpret_cast<T*>(ptr);
|
|
}
|
|
|
|
/**
|
|
* \brief Frees a device buffer
|
|
*
|
|
* The cuda context will get acquired automatically
|
|
*
|
|
* \param[in] cudaContextManager the PxCudaContextManager managing the CUDA context the device buffer is attributed to.
|
|
* \param[in] deviceBuffer reference to a pointer pointing to the device memory buffer.
|
|
*/
|
|
template<typename T>
|
|
static void freeDeviceBuffer(PxCudaContextManager& cudaContextManager, T*& deviceBuffer)
|
|
{
|
|
if (deviceBuffer)
|
|
{
|
|
PxScopedCudaLock lock(cudaContextManager);
|
|
PxCUresult result = cudaContextManager.getCudaContext()->memFree(reinterpret_cast<CUdeviceptr>(deviceBuffer));
|
|
if (result != 0)
|
|
PxGetFoundation().error(PxErrorCode::eINTERNAL_ERROR, PX_FL, "freeDeviceBuffer failed with error code %i!\n", PxI32(result));
|
|
|
|
deviceBuffer = NULL;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* \brief Allocates a pinned host buffer and returns the pointer to the memory
|
|
*
|
|
* A pinned host buffer can be used on the gpu after getting a mapped device pointer from the pinned host buffer pointer, see getMappedDevicePtr
|
|
* The cuda context will get acquired automatically
|
|
*
|
|
* \param[in] cudaContextManager the PxCudaContextManager managing the CUDA context the allocation should be attributed to.
|
|
* \param[in] numElements the number of elements of type T to allocate.
|
|
*
|
|
* \return a pointer to the allocated memory.
|
|
*/
|
|
template<typename T>
|
|
static T* allocPinnedHostBuffer(PxCudaContextManager& cudaContextManager, PxU64 numElements)
|
|
{
|
|
PxScopedCudaLock lock(cudaContextManager);
|
|
|
|
void* ptr = NULL;
|
|
const unsigned int cuMemhostallocDevicemap = 0x02;
|
|
const unsigned int cuMemhostallocPortable = 0x01;
|
|
PxCUresult result = cudaContextManager.getCudaContext()->memHostAlloc(&ptr, numElements * sizeof(T), cuMemhostallocDevicemap | cuMemhostallocPortable);
|
|
if (result != 0)
|
|
PxGetFoundation().error(PxErrorCode::eINTERNAL_ERROR, PX_FL, "allocPinnedHostBuffer failed with error code %i!\n", PxI32(result));
|
|
|
|
return reinterpret_cast<T*>(ptr);
|
|
}
|
|
|
|
/**
|
|
* \brief Frees a pinned host buffer
|
|
*
|
|
* The cuda context will get acquired automatically
|
|
*
|
|
* \param[in] cudaContextManager the PxCudaContextManager managing the CUDA context the pinned memory buffer is attributed to.
|
|
* \param[in] pinnedHostBuffer reference to a pointer pointing to the pinned memory buffer.
|
|
*/
|
|
template<typename T>
|
|
static void freePinnedHostBuffer(PxCudaContextManager& cudaContextManager, T*& pinnedHostBuffer)
|
|
{
|
|
if (pinnedHostBuffer)
|
|
{
|
|
PxScopedCudaLock lock(cudaContextManager);
|
|
PxCUresult result = cudaContextManager.getCudaContext()->memFreeHost(pinnedHostBuffer);
|
|
if (result != 0)
|
|
PxGetFoundation().error(PxErrorCode::eINTERNAL_ERROR, PX_FL, "freePinnedHostBuffer failed with error code %i!\n", PxI32(result));
|
|
|
|
pinnedHostBuffer = NULL;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* \brief Copies a device buffer to the host
|
|
*
|
|
* The cuda context will get acquired automatically
|
|
*
|
|
* \param[in] cudaContextManager the PxCudaContextManager managing the CUDA context executing the operation.
|
|
* \param[in] hostBuffer pointer to the destination (pinned) host memory buffer.
|
|
* \param[in] deviceBuffer pointer to the source device buffer.
|
|
* \param[in] numElements the number of elements of type T to copy.
|
|
*/
|
|
template<typename T>
|
|
static void copyDToH(PxCudaContextManager& cudaContextManager, T* hostBuffer, const T* deviceBuffer, PxU64 numElements)
|
|
{
|
|
if (!deviceBuffer || !hostBuffer)
|
|
return;
|
|
|
|
PxScopedCudaLock lock(cudaContextManager);
|
|
PxU64 numBytes = numElements * sizeof(T);
|
|
|
|
PxCUresult result = cudaContextManager.getCudaContext()->memcpyDtoH(hostBuffer, CUdeviceptr(deviceBuffer), numBytes);
|
|
if (result != 0)
|
|
PxGetFoundation().error(PxErrorCode::eINTERNAL_ERROR, PX_FL, "copyDToH failed with error code %i!\n", PxI32(result));
|
|
}
|
|
|
|
/**
|
|
* \brief Copies a host buffer to the device
|
|
*
|
|
* The cuda context will get acquired automatically
|
|
*
|
|
* \param[in] cudaContextManager the PxCudaContextManager managing the CUDA context executing the operation.
|
|
* \param[in] deviceBuffer pointer to the destination device buffer.
|
|
* \param[in] hostBuffer pointer to the source (pinned) host memory buffer.
|
|
* \param[in] numElements the number of elements of type T to copy.
|
|
*/
|
|
template<typename T>
|
|
static void copyHToD(PxCudaContextManager& cudaContextManager, T* deviceBuffer, const T* hostBuffer, PxU64 numElements)
|
|
{
|
|
if (!deviceBuffer || !hostBuffer)
|
|
return;
|
|
|
|
PxScopedCudaLock lock(cudaContextManager);
|
|
PxU64 numBytes = numElements * sizeof(T);
|
|
|
|
PxCUresult result = cudaContextManager.getCudaContext()->memcpyHtoD(CUdeviceptr(deviceBuffer), hostBuffer, numBytes);
|
|
if (result != 0)
|
|
PxGetFoundation().error(PxErrorCode::eINTERNAL_ERROR, PX_FL, "copyHtoD failed with error code %i!\n", PxI32(result));
|
|
}
|
|
|
|
/**
|
|
* \brief Schedules device to host copy operation on the specified stream
|
|
*
|
|
* The cuda context will get acquired automatically
|
|
*
|
|
* \param[in] cudaContextManager the PxCudaContextManager managing the CUDA context executing the operation.
|
|
* \param[in] hostBuffer pointer to the destination (pinned) host memory buffer.
|
|
* \param[in] deviceBuffer pointer to the source device buffer.
|
|
* \param[in] numElements the number of elements of type T to copy.
|
|
* \param[in] stream the CUDA stream to perform the operation on.
|
|
*/
|
|
template<typename T>
|
|
static void copyDToHAsync(PxCudaContextManager& cudaContextManager, T* hostBuffer, const T* deviceBuffer, PxU64 numElements, CUstream stream)
|
|
{
|
|
if (!deviceBuffer || !hostBuffer)
|
|
return;
|
|
|
|
PxScopedCudaLock lock(cudaContextManager);
|
|
PxU64 numBytes = numElements * sizeof(T);
|
|
|
|
PxCUresult result = cudaContextManager.getCudaContext()->memcpyDtoHAsync(hostBuffer, CUdeviceptr(deviceBuffer), numBytes, stream);
|
|
if (result != 0)
|
|
PxGetFoundation().error(PxErrorCode::eINTERNAL_ERROR, PX_FL, "copyDtoHAsync failed with error code %i!\n", PxI32(result));
|
|
}
|
|
|
|
/**
|
|
* \brief Schedules host to device copy operation on the specified stream
|
|
*
|
|
* The cuda context will get acquired automatically
|
|
*
|
|
* \param[in] cudaContextManager the PxCudaContextManager managing the CUDA context executing the operation.
|
|
* \param[in] deviceBuffer pointer to the destination device buffer.
|
|
* \param[in] hostBuffer pointer to the source (pinned) host memory buffer.
|
|
* \param[in] numElements the number of elements of type T to copy.
|
|
* \param[in] stream the CUDA stream to perform the operation on.
|
|
*/
|
|
template<typename T>
|
|
static void copyHToDAsync(PxCudaContextManager& cudaContextManager, T* deviceBuffer, const T* hostBuffer, PxU64 numElements, CUstream stream)
|
|
{
|
|
if (!deviceBuffer || !hostBuffer)
|
|
return;
|
|
|
|
PxScopedCudaLock lock(cudaContextManager);
|
|
PxU64 numBytes = numElements * sizeof(T);
|
|
|
|
PxCUresult result = cudaContextManager.getCudaContext()->memcpyHtoDAsync(CUdeviceptr(deviceBuffer), hostBuffer, numBytes, stream);
|
|
if (result != 0)
|
|
PxGetFoundation().error(PxErrorCode::eINTERNAL_ERROR, PX_FL, "copyHtoDAsync failed with error code %i!\n", PxI32(result));
|
|
}
|
|
|
|
/**
|
|
* \brief Schedules device to device copy operation on the specified stream
|
|
*
|
|
* The cuda context will get acquired automatically
|
|
*
|
|
* \param[in] cudaContextManager the PxCudaContextManager managing the CUDA context executing the operation.
|
|
* \param[in] dstDeviceBuffer pointer to the destination device buffer.
|
|
* \param[in] srcDeviceBuffer pointer to the source device buffer.
|
|
* \param[in] numElements the number of elements of type T to copy.
|
|
* \param[in] stream the CUDA stream to perform the operation on.
|
|
*/
|
|
template<typename T>
|
|
static void copyDToDAsync(PxCudaContextManager& cudaContextManager, T* dstDeviceBuffer, const T* srcDeviceBuffer, PxU64 numElements, CUstream stream)
|
|
{
|
|
if (!srcDeviceBuffer || !dstDeviceBuffer)
|
|
return;
|
|
|
|
PxScopedCudaLock lock(cudaContextManager);
|
|
PxU64 numBytes = numElements * sizeof(T);
|
|
|
|
PxCUresult result = cudaContextManager.getCudaContext()->memcpyDtoDAsync(CUdeviceptr(dstDeviceBuffer), CUdeviceptr(srcDeviceBuffer), numBytes, stream);
|
|
if (result != 0)
|
|
PxGetFoundation().error(PxErrorCode::eINTERNAL_ERROR, PX_FL, "copyDtoDAsync failed with error code %i!\n", PxI32(result));
|
|
}
|
|
|
|
/**
|
|
* \brief Schedules a memset operation on the device on the specified stream. Only supported for 1 bytes or 4 byte data types.
|
|
*
|
|
* The cuda context will get acquired automatically
|
|
*
|
|
* \param[in] cudaContextManager the PxCudaContextManager managing the CUDA context executing the operation.
|
|
* \param[in] dstDeviceBuffer pointer to the destination device buffer.
|
|
* \param[in] value the value to set the memory to.
|
|
* \param[in] numElements the number of elements of type T to set.
|
|
* \param[in] stream the CUDA stream to perform the operation on.
|
|
*/
|
|
template<typename T>
|
|
static void memsetAsync(PxCudaContextManager& cudaContextManager, T* dstDeviceBuffer, const T& value, PxU64 numElements, CUstream stream)
|
|
{
|
|
PX_COMPILE_TIME_ASSERT(sizeof(T) == sizeof(PxU32) || sizeof(T) == sizeof(PxU8));
|
|
|
|
if (!dstDeviceBuffer)
|
|
return;
|
|
|
|
PxScopedCudaLock lock(cudaContextManager);
|
|
PxU64 numBytes = numElements * sizeof(T);
|
|
|
|
if (sizeof(T) == sizeof(PxU32))
|
|
{
|
|
PxCUresult result = cudaContextManager.getCudaContext()->memsetD32Async(CUdeviceptr(dstDeviceBuffer), reinterpret_cast<const PxU32&>(value), numBytes >> 2, stream);
|
|
if (result != 0)
|
|
PxGetFoundation().error(PxErrorCode::eINTERNAL_ERROR, PX_FL, "memsetAsync failed with error code %i!\n", PxI32(result));
|
|
}
|
|
else
|
|
{
|
|
PxCUresult result = cudaContextManager.getCudaContext()->memsetD8Async(CUdeviceptr(dstDeviceBuffer), reinterpret_cast<const PxU8&>(value), numBytes, stream);
|
|
if (result != 0)
|
|
PxGetFoundation().error(PxErrorCode::eINTERNAL_ERROR, PX_FL, "memsetAsync failed with error code %i!\n", PxI32(result));
|
|
}
|
|
}
|
|
};
|
|
|
|
#define PX_EXT_DEVICE_MEMORY_ALLOC(T, cudaContextManager, numElements) physx::Ext::PxCudaHelpersExt::allocDeviceBuffer<T>(cudaContextManager, numElements)
|
|
#define PX_EXT_DEVICE_MEMORY_FREE(cudaContextManager, deviceBuffer) physx::Ext::PxCudaHelpersExt::freeDeviceBuffer(cudaContextManager, deviceBuffer);
|
|
|
|
#define PX_EXT_PINNED_MEMORY_ALLOC(T, cudaContextManager, numElements) physx::Ext::PxCudaHelpersExt::allocPinnedHostBuffer<T>(cudaContextManager, numElements)
|
|
#define PX_EXT_PINNED_MEMORY_FREE(cudaContextManager, pinnedHostBuffer) physx::Ext::PxCudaHelpersExt::freePinnedHostBuffer(cudaContextManager, pinnedHostBuffer);
|
|
|
|
}
|
|
}
|
|
|
|
#endif
|
|
|
|
#endif
|