Struct DeviceCapability 

pub struct DeviceCapability {
    pub compute_capability: (u32, u32),
    pub total_memory: usize,
    pub num_compute_units: u32,
    pub max_threads_per_block: u32,
    pub warp_size: u32,
    pub has_tensor_cores: bool,
    pub unified_memory: bool,
    pub max_shared_memory: usize,
    pub memory_bandwidth: f32,
    pub peak_tflops: f32,
}

Device capability information

Fields

compute_capability: (u32, u32)

Compute capability (e.g., 8.0 for A100)

total_memory: usize

Total memory in bytes

num_compute_units: u32

Number of SMs (for CUDA) or compute units

max_threads_per_block: u32

Maximum threads per block

warp_size: u32

Warp size (32 for CUDA, varies for others)

has_tensor_cores: bool

Whether tensor cores are available

unified_memory: bool

Whether unified memory is supported

max_shared_memory: usize

Maximum shared memory per block

memory_bandwidth: f32

Memory bandwidth (GB/s)

peak_tflops: f32

Peak FLOPS (TF32)

Implementations

impl DeviceCapability

pub fn cpu() -> Self

Create capability info for CPU

pub fn apple_silicon(total_memory: usize, gpu_cores: u32) -> Self

Create capability info for Apple Silicon (Metal)

pub fn can_fit_model(&self, model_size_bytes: usize) -> bool

Check if device can run a model of given size

Trait Implementations

impl Clone for DeviceCapability

fn clone(&self) -> DeviceCapability

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for DeviceCapability

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for DeviceCapability

fn default() -> Self

Returns the “default value” for a type.