CaloRecGPU::CUDA_Helpers Namespace Reference

Classes
struct	CUDAStreamPtrHolder

Functions
void *	allocate (const size_t num)
	Allocates and returns the address of num bytes from GPU memory. More...
void	deallocate (void *address)
	Deallocates address in GPU memory. More...
void *	allocate_pinned (const size_t num)
	Allocates and returns the address of num bytes from CPU pinned memory. More...
void	deallocate_pinned (void *address)
	Deallocates address in CPU pinned memory. More...
void	GPU_to_CPU (void *dest, const void *const source, const size_t num)
	Copies num bytse from source in GPU memory to dest in CPU memory. More...
void	CPU_to_GPU (void *dest, const void *const source, const size_t num)
	Copies num bytes from source in CPU memory to dest in GPU memory. More...
void	GPU_to_GPU (void *dest, const void *const source, const size_t num)
	Copies num bytes from source to dest, both in GPU memory. More...
void	GPU_to_CPU_async (void *dest, const void *const source, const size_t num, CUDAStreamPtrHolder stream={})
	Copies num bytes from source in GPU memory to dest in CPU memory, asynchronously. More...
void	CPU_to_GPU_async (void *dest, const void *const source, const size_t num, CUDAStreamPtrHolder stream={})
	Copies num bytes from source in CPU memory to dest in GPU memory, asynchronously. More...
void	GPU_to_GPU_async (void *dest, const void *const source, const size_t num, CUDAStreamPtrHolder stream={})
	Copies num bytes from source to dest, both in GPU memory, asynchronously. More...
void	GPU_synchronize (CUDAStreamPtrHolder stream={})
	Synchronizes the stream. More...
void	optimize_block_and_grid_size (void *func, int &block_size, int &grid_size, const int dynamic_memory=0, const int block_size_limit=0)
	Optimizes block and grid size according to cudaOccupancyMaxPotentialBlockSize. More...
void	optimize_block_and_grid_size_for_cooperative_launch (void *func, int &block_size, int &grid_size, const int dynamic_memory=0, const int block_size_limit=0)
	Optimizes block and grid size for a cooperative launch. More...
bool	supports_cooperative_launches ()
bool	supports_dynamic_parallelism ()
std::string	GPU_name ()

Function Documentation

allocate()

void* CaloRecGPU::CUDA_Helpers::allocate

(

const size_t

num

)

Allocates and returns the address of num bytes from GPU memory.

allocate_pinned()

void* CaloRecGPU::CUDA_Helpers::allocate_pinned

(

const size_t

num

)

Allocates and returns the address of num bytes from CPU pinned memory.

CPU_to_GPU()

void CaloRecGPU::CUDA_Helpers::CPU_to_GPU	(	void *	dest,
		const void *const	source,
		const size_t	num
	)

Copies num bytes from source in CPU memory to dest in GPU memory.

CPU_to_GPU_async()

void CaloRecGPU::CUDA_Helpers::CPU_to_GPU_async	(	void *	dest,
		const void *const	source,
		const size_t	num,
		CUDAStreamPtrHolder	stream = {}
	)

Copies num bytes from source in CPU memory to dest in GPU memory, asynchronously.

deallocate()

void CaloRecGPU::CUDA_Helpers::deallocate

(

void *

address

)

Deallocates address in GPU memory.

deallocate_pinned()

void CaloRecGPU::CUDA_Helpers::deallocate_pinned

(

void *

address

)

Deallocates address in CPU pinned memory.

GPU_name()

std::string CaloRecGPU::CUDA_Helpers::GPU_name

(

)

GPU_synchronize()

void CaloRecGPU::CUDA_Helpers::GPU_synchronize

(

CUDAStreamPtrHolder

stream = {}

)

Synchronizes the stream.

If called with no value, synchronizes with cudaStreamPerThread.

GPU_to_CPU()

void CaloRecGPU::CUDA_Helpers::GPU_to_CPU	(	void *	dest,
		const void *const	source,
		const size_t	num
	)

Copies num bytse from source in GPU memory to dest in CPU memory.

GPU_to_CPU_async()

void CaloRecGPU::CUDA_Helpers::GPU_to_CPU_async	(	void *	dest,
		const void *const	source,
		const size_t	num,
		CUDAStreamPtrHolder	stream = {}
	)

Copies num bytes from source in GPU memory to dest in CPU memory, asynchronously.

GPU_to_GPU()

void CaloRecGPU::CUDA_Helpers::GPU_to_GPU	(	void *	dest,
		const void *const	source,
		const size_t	num
	)

Copies num bytes from source to dest, both in GPU memory.

GPU_to_GPU_async()

void CaloRecGPU::CUDA_Helpers::GPU_to_GPU_async	(	void *	dest,
		const void *const	source,
		const size_t	num,
		CUDAStreamPtrHolder	stream = {}
	)

Copies num bytes from source to dest, both in GPU memory, asynchronously.

optimize_block_and_grid_size()

void CaloRecGPU::CUDA_Helpers::optimize_block_and_grid_size	(	void *	func,
		int &	block_size,
		int &	grid_size,
		const int	dynamic_memory = 0,
		const int	block_size_limit = 0
	)

Optimizes block and grid size according to cudaOccupancyMaxPotentialBlockSize.

optimize_block_and_grid_size_for_cooperative_launch()

void CaloRecGPU::CUDA_Helpers::optimize_block_and_grid_size_for_cooperative_launch	(	void *	func,
		int &	block_size,
		int &	grid_size,
		const int	dynamic_memory = 0,
		const int	block_size_limit = 0
	)

Optimizes block and grid size for a cooperative launch.

supports_cooperative_launches()

bool CaloRecGPU::CUDA_Helpers::supports_cooperative_launches

(

)

supports_dynamic_parallelism()

bool CaloRecGPU::CUDA_Helpers::supports_dynamic_parallelism

(

)