Struct KernelFusionConfig 

pub struct KernelFusionConfig {
    pub compute_capability: (u32, u32),
    pub warp_size: usize,
    pub max_threads_per_block: usize,
    pub shared_memory_size: usize,
    pub mixed_precision: bool,
    pub use_tensor_cores: bool,
    pub coalescing_level: CoalescingLevel,
}

Configuration for GPU kernel fusion optimization.

Fields

compute_capability: (u32, u32)

Target GPU compute capability (e.g., 7.5 for V100, 8.0 for A100)

warp_size: usize

Warp size (typically 32 for NVIDIA GPUs)

max_threads_per_block: usize

Maximum threads per block

shared_memory_size: usize

Shared memory size per block in bytes

mixed_precision: bool

Enable mixed precision (FP16/FP32) kernels

use_tensor_cores: bool

Enable tensor core operations where possible

coalescing_level: CoalescingLevel

Memory coalescing optimization level

Implementations

impl KernelFusionConfig

pub fn a100() -> Self

Creates configuration for A100 GPUs.

pub fn h100() -> Self

Creates configuration for H100 GPUs.

pub fn rtx4090() -> Self

Creates configuration for RTX 4090.

pub fn optimal_block_size(&self, param_count: usize) -> usize

Gets optimal block size for given parameter count.

pub fn memory_alignment(&self) -> usize

Gets memory alignment requirement based on coalescing level.

Trait Implementations

impl Clone for KernelFusionConfig

fn clone(&self) -> KernelFusionConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for KernelFusionConfig

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

Formats the value using the given formatter.

impl Default for KernelFusionConfig

fn default() -> Self

Returns the “default value” for a type.