Mila 0.13.48
Deep Neural Network Library
Loading...
Searching...
No Matches
Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource > Struct Referenceexport
module Compute.CudaDeviceMemoryResourceTraits

CUDA device memory resource traits providing detailed GPU backend characteristics. More...

Static Public Attributes

static constexpr bool device_accessible = true
static constexpr bool high_bandwidth = true
static constexpr bool host_accessible = false
static constexpr bool is_cpu_resource = false
static constexpr bool is_cuda_resource = true
static constexpr bool is_metal_resource = false
static constexpr bool is_opencl_resource = false
static constexpr bool is_vulkan_resource = false
static constexpr bool low_latency_access = false
static constexpr size_t preferred_alignment = 256
static constexpr bool requires_context_binding = true
static constexpr bool requires_host_synchronization = true
static constexpr bool supports_async_operations = true
static constexpr bool supports_atomic_operations = true
static constexpr bool supports_concurrent_kernels = true
static constexpr bool supports_constant_memory = true
static constexpr bool supports_peer_access = true
static constexpr bool supports_shared_memory = true
static constexpr bool supports_texture_memory = true
static constexpr bool supports_unified_memory = false
static constexpr bool supports_zero_copy = false
static constexpr size_t warp_size = 32

Detailed Description

CUDA device memory resource traits providing detailed GPU backend characteristics.

Extends base MemoryResourceTraits with CUDA-specific capabilities and optimizations. Used for CUDA backend dispatch and optimization decisions.

Member Data Documentation

◆ device_accessible

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::device_accessible = true
staticconstexpr

◆ high_bandwidth

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::high_bandwidth = true
staticconstexpr

◆ host_accessible

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::host_accessible = false
staticconstexpr

◆ is_cpu_resource

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::is_cpu_resource = false
staticconstexpr

◆ is_cuda_resource

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::is_cuda_resource = true
staticconstexpr

◆ is_metal_resource

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::is_metal_resource = false
staticconstexpr

◆ is_opencl_resource

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::is_opencl_resource = false
staticconstexpr

◆ is_vulkan_resource

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::is_vulkan_resource = false
staticconstexpr

◆ low_latency_access

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::low_latency_access = false
staticconstexpr

◆ preferred_alignment

size_t Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::preferred_alignment = 256
staticconstexpr

◆ requires_context_binding

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::requires_context_binding = true
staticconstexpr

◆ requires_host_synchronization

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::requires_host_synchronization = true
staticconstexpr

◆ supports_async_operations

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::supports_async_operations = true
staticconstexpr

◆ supports_atomic_operations

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::supports_atomic_operations = true
staticconstexpr

◆ supports_concurrent_kernels

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::supports_concurrent_kernels = true
staticconstexpr

◆ supports_constant_memory

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::supports_constant_memory = true
staticconstexpr

◆ supports_peer_access

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::supports_peer_access = true
staticconstexpr

◆ supports_shared_memory

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::supports_shared_memory = true
staticconstexpr

◆ supports_texture_memory

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::supports_texture_memory = true
staticconstexpr

◆ supports_unified_memory

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::supports_unified_memory = false
staticconstexpr

◆ supports_zero_copy

bool Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::supports_zero_copy = false
staticconstexpr

◆ warp_size

size_t Mila::Dnn::Compute::MemoryResourceTraits< CudaDeviceMemoryResource >::warp_size = 32
staticconstexpr
The documentation for this struct was generated from the following file: