Realm::Cuda Namespace Reference

Classes
class	AddressInfoCudaArray
struct	AffineCopyInfo
struct	AffineCopyPair
struct	AffineFillInfo
struct	AffineFillRect
struct	AffineLargeFillInfo
struct	AffineSubRect
class	AutoGPUContext
class	ContextSynchronizer
class	CudaDeviceMemoryInfo
struct	CudaIpcImportRequest
class	CudaModule
class	CudaModuleConfig
struct	CudaRedOpDesc
struct	FatBin
class	GlobalRegistrations
class	GPU
class	GPUAllocation
	Class for managing the lifetime of a given gpu allocation. As instances of this class own an underlying resource they are not copyable and must be std::move'd (thus invalidating the original variable) or references made. More...
class	GPUChannel
class	GPUCompletionEvent
class	GPUCompletionNotification
class	GPUContextManager
class	GPUDynamicFBMemory
class	GPUEventPool
class	GPUFBIBMemory
class	GPUFBMemory
class	GPUfillChannel
class	GPUfillXferDes
class	GPUIndirectChannel
class	GPUIndirectRemoteChannel
class	GPUIndirectRemoteChannelInfo
class	GPUIndirectTransferCompletion
class	GPUIndirectXferDes
struct	GPUInfo
class	GPUProcessor
class	GPUreduceChannel
class	GPUreduceRemoteChannel
class	GPUreduceRemoteChannelInfo
class	GPUreduceXferDes
class	GPUReplHeapListener
class	GPURequest
class	GPUStream
class	GPUTransferCompletion
class	GPUWorker
class	GPUWorkFence
class	GPUWorkStart
class	GPUXferDes
class	GPUZCMemory
struct	MemcpyIndirectInfo
struct	MemcpyTransposeInfo
class	MemSpecificCudaArray
struct	RegisteredFunction
struct	RegisteredModule
struct	RegisteredVariable

Typedefs
typedef enum Realm::Cuda::nvmlIntNvLinkDeviceType_enum	nvmlIntNvLinkDeviceType_t
typedef void(*	StreamAwareTaskFuncPtr) (const void *args, size_t arglen, const void *user_data, size_t user_data_len, Processor proc, CUstream_st *stream)
typedef char	Uuid[UUID_SIZE]

Enumerations
enum	GPUMemcpyKind { GPU_MEMCPY_HOST_TO_DEVICE , GPU_MEMCPY_DEVICE_TO_HOST , GPU_MEMCPY_DEVICE_TO_DEVICE , GPU_MEMCPY_PEER_TO_PEER }
enum	nvmlIntNvLinkDeviceType_enum { NVML_NVLINK_DEVICE_TYPE_GPU = 0x00 , NVML_NVLINK_DEVICE_TYPE_IBMNPU = 0x01 , NVML_NVLINK_DEVICE_TYPE_SWITCH = 0x02 , NVML_NVLINK_DEVICE_TYPE_UNKNOWN = 0xFF }

Functions
CUresult	cuGetProcAddress (const char *, void **, int, int)
CUresult	cuCtxRecordEvent (CUcontext hctx, CUevent event)
	CUDA_DRIVER_APIS (DECL_FNPTR_EXTERN)
nvmlReturn_t	nvmlDeviceGetNvLinkRemoteDeviceType (nvmlDevice_t device, unsigned int link, nvmlIntNvLinkDeviceType_t *pNvLinkDeviceType)
REALM_PUBLIC_API CUstream_st *	get_task_cuda_stream ()
REALM_PUBLIC_API void	set_task_ctxsync_required (bool is_required)
REALM_PUBLIC_API bool	get_cuda_device_uuid (Processor p, Uuid *uuid)
REALM_PUBLIC_API bool	get_cuda_device_id (Processor p, int *device)

Variables
CudaModule *	cuda_module_singleton
bool	cudart_hijack_active
int	cudart_hijack_nongpu_sync

Typedef Documentation

nvmlIntNvLinkDeviceType_t

typedef enum Realm::Cuda::nvmlIntNvLinkDeviceType_enum Realm::Cuda::nvmlIntNvLinkDeviceType_t

StreamAwareTaskFuncPtr

typedef void(* Realm::Cuda::StreamAwareTaskFuncPtr) (const void *args, size_t arglen, const void *user_data, size_t user_data_len, Processor proc, CUstream_st *stream)

Uuid

typedef char Realm::Cuda::Uuid[UUID_SIZE]

Enumeration Type Documentation

GPUMemcpyKind

enum Realm::Cuda::GPUMemcpyKind

Enumerator
GPU_MEMCPY_HOST_TO_DEVICE
GPU_MEMCPY_DEVICE_TO_HOST
GPU_MEMCPY_DEVICE_TO_DEVICE
GPU_MEMCPY_PEER_TO_PEER

nvmlIntNvLinkDeviceType_enum

enum Realm::Cuda::nvmlIntNvLinkDeviceType_enum

Enumerator
NVML_NVLINK_DEVICE_TYPE_GPU
NVML_NVLINK_DEVICE_TYPE_IBMNPU
NVML_NVLINK_DEVICE_TYPE_SWITCH
NVML_NVLINK_DEVICE_TYPE_UNKNOWN

Function Documentation

cuCtxRecordEvent()

CUresult Realm::Cuda::cuCtxRecordEvent	(	CUcontext	hctx,
		CUevent	event
	)

CUDA_DRIVER_APIS()

Realm::Cuda::CUDA_DRIVER_APIS

(

DECL_FNPTR_EXTERN

)

cuGetProcAddress()

CUresult Realm::Cuda::cuGetProcAddress	(	const char *	,
		void **	,
		int	,
		int
	)

get_cuda_device_id()

REALM_PUBLIC_API bool Realm::Cuda::get_cuda_device_id	(	Processor	p,
		int *	device
	)

get_cuda_device_uuid()

REALM_PUBLIC_API bool Realm::Cuda::get_cuda_device_uuid	(	Processor	p,
		Uuid *	uuid
	)

get_task_cuda_stream()

REALM_PUBLIC_API CUstream_st * Realm::Cuda::get_task_cuda_stream

(

)

nvmlDeviceGetNvLinkRemoteDeviceType()

nvmlReturn_t Realm::Cuda::nvmlDeviceGetNvLinkRemoteDeviceType	(	nvmlDevice_t	device,
		unsigned int	link,
		nvmlIntNvLinkDeviceType_t *	pNvLinkDeviceType
	)

set_task_ctxsync_required()

REALM_PUBLIC_API void Realm::Cuda::set_task_ctxsync_required

(

bool

is_required

)

Variable Documentation

cuda_module_singleton

CudaModule* Realm::Cuda::cuda_module_singleton

extern

cudart_hijack_active

bool Realm::Cuda::cudart_hijack_active

extern

cudart_hijack_nongpu_sync

int Realm::Cuda::cudart_hijack_nongpu_sync

extern