Mila::Dnn::Compute::CudaDeviceMemoryResource Class Reference

CUDA device memory resource for GPU-accessible memory allocation. More...

Inheritance diagram for Mila::Dnn::Compute::CudaDeviceMemoryResource:

Public Member Functions
	CudaDeviceMemoryResource (int device_id)
	Constructs CUDA managed memory resource with device ID.
Public Member Functions inherited from Mila::Dnn::Compute::MemoryResource
virtual	~MemoryResource ()=default
	Virtual destructor for proper cleanup of derived classes.

Static Public Attributes
static constexpr DeviceType	device_type = DeviceType::Cuda
	Device type constant for CUDA memory resources.
static constexpr bool	is_device_accessible = DeviceAccessible::is_device_accessible
	Indicates CUDA device memory is accessible from device code.
static constexpr bool	is_host_accessible = false
	Indicates CUDA device memory is not accessible from host code.
Static Public Attributes inherited from Mila::Dnn::Compute::MemoryResource
static constexpr bool	is_device_accessible = false
	Checks if the memory is accessible from device code.
static constexpr bool	is_host_accessible = true
	Checks if the memory is accessible from host code.

Protected Member Functions
void *	do_allocate (std::size_t bytes, std::size_t) override
	Allocates memory on the CUDA device.
void	do_deallocate (void *ptr, std::size_t, std::size_t) override
	Deallocates CUDA device memory.
bool	do_is_equal (const std::pmr::memory_resource &other) const noexcept override
	Compares CUDA memory resources for equality.

Detailed Description

CUDA device memory resource for GPU-accessible memory allocation.

Provides optimized CUDA device memory allocation with proper device binding through device context integration. Focuses purely on memory allocation responsibilities without tensor-specific operations or type conversions.

Constructor & Destructor Documentation

CudaDeviceMemoryResource()

Mila::Dnn::Compute::CudaDeviceMemoryResource::CudaDeviceMemoryResource ( int device_id )

inlineexplicit

Constructs CUDA managed memory resource with device ID.

Parameters

device_id

CUDA device ID (0, 1, 2, etc.)

Exceptions

std::invalid_argument

If device_id is invalid

Member Function Documentation

do_allocate()

void * Mila::Dnn::Compute::CudaDeviceMemoryResource::do_allocate ( std::size_t bytes, std::size_t  )

inlineoverrideprotectedvirtual

Allocates memory on the CUDA device.

Ensures allocation occurs on the correct device by activating the device context before calling cudaMalloc. Provides detailed error information on allocation failure.

Parameters

bytes

Number of bytes to allocate The alignment argument is ignored: CUDA allocations are already suitably aligned.

Returns

Pointer to allocated device memory

Exceptions

CudaBadAlloc

If allocation fails

Implements Mila::Dnn::Compute::MemoryResource.

do_deallocate()

void Mila::Dnn::Compute::CudaDeviceMemoryResource::do_deallocate ( void * ptr, std::size_t , std::size_t  )

inlineoverrideprotectedvirtual

Deallocates CUDA device memory.

Ensures deallocation occurs on the correct device and provides detailed error information if deallocation fails.

Parameters

ptr	Pointer to device memory to deallocate

The size and alignment arguments are unused (kept for the memory-resource interface) and therefore intentionally unnamed.

Implements Mila::Dnn::Compute::MemoryResource.

do_is_equal()

bool Mila::Dnn::Compute::CudaDeviceMemoryResource::do_is_equal ( const std::pmr::memory_resource & other ) const

inlineoverrideprotectedvirtualnoexcept

Compares CUDA memory resources for equality.

CUDA memory resources are equal if they are both CudaDeviceMemoryResource instances. Device binding is handled at the tensor level through device context management.

Parameters

other

The other memory resource to compare with

Returns

true if both are CudaDeviceMemoryResource instances

Implements Mila::Dnn::Compute::MemoryResource.

---

The documentation for this class was generated from the following file:

• Mila/Src/Dnn/Compute/Devices/Cuda/CudaDeviceMemoryResource.ixx