CUDA

Detailed Description

This module implements CUDA specific data-structures and functions

Modules
	CUDA Allocators
	Hash Maps
	Parallel Primitives
	Sort Enactors
	CUDA Atomics

Classes
struct	cugar::cuda::arch
struct	cugar::cuda::cuda_devices
struct	cugar::cuda::load_pointer< T, MOD >
struct	cugar::cuda::store_reference< T, STORE_MOD, LOAD_MOD >
struct	cugar::cuda::store_pointer< T, STORE_MOD, LOAD_MOD >
struct	cugar::cuda::Timer
struct	cugar::cuda::ScopedTimer< T >

Enumerations
enum	cugar::cuda::CacheLoadModifier {   cugar::cuda::LOAD_DEFAULT, cugar::cuda::LOAD_CA, cugar::cuda::LOAD_CG, cugar::cuda::LOAD_CS,   cugar::cuda::LOAD_CV, cugar::cuda::LOAD_LDG, cugar::cuda::LOAD_VOLATILE }
	Enumeration of cache modifiers for memory load operations. More...
enum	cugar::cuda::CacheStoreModifier {   cugar::cuda::STORE_DEFAULT, cugar::cuda::STORE_WB, cugar::cuda::STORE_CG, cugar::cuda::STORE_CS,   cugar::cuda::STORE_WT, cugar::cuda::STORE_VOLATILE }
	Enumeration of cache modifiers for memory load operations. More...

Functions
cudaDeviceProp	cugar::cuda::get_device_properties ()
void	cugar::cuda::device_arch (uint32 &major, uint32 &minor)
uint32	cugar::cuda::max_grid_size ()
size_t	cugar::cuda::multiprocessor_count ()
size_t	cugar::cuda::smem_allocation_unit (const cudaDeviceProp &properties)
size_t	cugar::cuda::reg_allocation_unit (const cudaDeviceProp &properties, const size_t regsPerThread)
size_t	cugar::cuda::warp_allocation_multiple (const cudaDeviceProp &properties)
size_t	cugar::cuda::num_sides_per_multiprocessor (const cudaDeviceProp &properties)
size_t	cugar::cuda::max_blocks_per_multiprocessor (const cudaDeviceProp &properties)
size_t	cugar::cuda::num_regs_per_block (const cudaDeviceProp &properties, const cudaFuncAttributes &attributes, const size_t CTA_SIZE)
template<typename KernelFunction >
cudaFuncAttributes	cugar::cuda::function_attributes (KernelFunction kernel)
template<typename KernelFunction >
size_t	cugar::cuda::max_active_blocks_per_multiprocessor (KernelFunction kernel, const size_t CTA_SIZE, const size_t dynamic_smem_bytes)
template<typename KernelFunction >
size_t	cugar::cuda::max_active_blocks (KernelFunction kernel, const size_t CTA_SIZE, const size_t dynamic_smem_bytes)
template<typename KernelFunction >
size_t	cugar::cuda::num_registers (KernelFunction kernel)
template<typename KernelFunction >
size_t	cugar::cuda::max_blocksize_with_highest_occupancy (KernelFunction kernel, size_t dynamic_smem_bytes_per_thread)
bool	cugar::cuda::is_tcc_enabled ()
void	cugar::cuda::check_error (const char *message)
void	cugar::cuda::check_error (cudaError_t, const char *message)
template<uint32 N>
CUGAR_FORCEINLINE CUGAR_HOST_DEVICE void	cugar::cuda::syncthreads ()
template<CacheLoadModifier LOAD_MOD, typename T >
CUGAR_FORCEINLINE CUGAR_HOST_DEVICE T	cugar::cuda::load (const T *ptr)
template<CacheStoreModifier STORE_MOD, typename T >
CUGAR_FORCEINLINE CUGAR_HOST_DEVICE void	cugar::cuda::store (T *ptr, const T &value)
template<CacheLoadModifier MOD, typename T >
CUGAR_FORCEINLINE CUGAR_HOST_DEVICE load_pointer< T, MOD >	cugar::cuda::make_load_pointer (const T *it)
template<CacheStoreModifier STORE_MOD, CacheLoadModifier LOAD_MOD, typename T >
CUGAR_FORCEINLINE CUGAR_HOST_DEVICE store_pointer< T, STORE_MOD, LOAD_MOD >	cugar::cuda::make_store_pointer (const T *it)
	cugar::cuda::Timer::Timer ()
	cugar::cuda::Timer::~Timer ()
void	cugar::cuda::Timer::start ()
void	cugar::cuda::Timer::stop ()
float	cugar::cuda::Timer::seconds () const

Enumeration Type Documentation

CacheLoadModifier

enum cugar::cuda::CacheLoadModifier

Enumeration of cache modifiers for memory load operations.

Enumerator
LOAD_DEFAULT	Default (no modifier)
LOAD_CA	Cache at all levels.
LOAD_CG	Cache at global level.
LOAD_CS	Cache streaming (likely to be accessed once)
LOAD_CV	Cache as volatile (including cached system lines)
LOAD_LDG	Cache as texture.
LOAD_VOLATILE	Volatile (any memory space)

CacheStoreModifier

enum cugar::cuda::CacheStoreModifier

Enumeration of cache modifiers for memory load operations.

Enumerator
STORE_DEFAULT	Default (no modifier)
STORE_WB	Cache write-back all coherent levels.
STORE_CG	Cache at global level.
STORE_CS	Cache streaming (likely to be accessed once)
STORE_WT	Cache write-through (to system memory)
STORE_VOLATILE	Volatile shared (any memory space)

Function Documentation

load()

template<CacheLoadModifier LOAD_MOD, typename T >

CUGAR_FORCEINLINE CUGAR_HOST_DEVICE T cugar::cuda::load

(

const T *

ptr

)

issue a load

make_load_pointer()

template<CacheLoadModifier MOD, typename T >

CUGAR_FORCEINLINE CUGAR_HOST_DEVICE load_pointer<T,MOD> cugar::cuda::make_load_pointer

(

const T *

it

)

make a load_pointer

make_store_pointer()

template<CacheStoreModifier STORE_MOD, CacheLoadModifier LOAD_MOD, typename T >

CUGAR_FORCEINLINE CUGAR_HOST_DEVICE store_pointer<T,STORE_MOD,LOAD_MOD> cugar::cuda::make_store_pointer

(

const T *

it

)

make a store_pointer

seconds()

float cugar::cuda::Timer::seconds ( ) const

inline

elapsed seconds

start()

void cugar::cuda::Timer::start ( )

inline

start timing

stop()

void cugar::cuda::Timer::stop ( )

inline

stop timing

store()

template<CacheStoreModifier STORE_MOD, typename T >

CUGAR_FORCEINLINE CUGAR_HOST_DEVICE void cugar::cuda::store	(	T *	ptr,
		const T &	value
	)

issue a store

syncthreads()

template<uint32 N>

CUGAR_FORCEINLINE CUGAR_HOST_DEVICE void cugar::cuda::syncthreads

(

)

a generic syncthreads() implementation to synchronize contiguous blocks of N threads at a time

Timer()

cugar::cuda::Timer::Timer ( )

inline

constructor

~Timer()

cugar::cuda::Timer::~Timer ( )

inline

destructor