Enum GpuMuda 

pub enum GpuMuda {
    Transport {
        register_spills: u64,
        unnecessary_global_loads: u64,
        redundant_shared_stores: u64,
    },
    Waiting {
        barrier_stall_cycles: u64,
        memory_stall_cycles: u64,
        pipeline_bubbles: u64,
        warp_scheduler_idle_pct: f64,
    },
    Overprocessing {
        precision_waste_pct: f64,
        redundant_instructions: u64,
        unnecessary_bounds_checks: u64,
    },
    Inventory {
        unused_shared_memory_bytes: u64,
        unused_registers_per_thread: u32,
        occupancy_loss_pct: f64,
    },
    Motion {
        divergent_branches: u64,
        branch_efficiency_pct: f64,
        loop_overhead_cycles: u64,
    },
    Defects {
        nan_count: u64,
        inf_count: u64,
        precision_loss_bits: f64,
    },
    Overproduction {
        padding_waste_pct: f64,
        inactive_thread_pct: f64,
        unused_output_elements: u64,
    },
}

Seven Muda of GPU Compute, plus CPU-specific waste categories.

Variants

Transport

Muda of Transport: Data moved unnecessarily. Examples: register spills, redundant L2 traffic, unnecessary H2D copies.

Fields

register_spills: u64

unnecessary_global_loads: u64

redundant_shared_stores: u64

Waiting

Muda of Waiting: Hardware resources idle. Examples: barrier stalls, memory latency not hidden, pipeline bubbles.

Fields

barrier_stall_cycles: u64

memory_stall_cycles: u64

pipeline_bubbles: u64

warp_scheduler_idle_pct: f64

Overprocessing

Muda of Overprocessing: More work than necessary. Examples: FP32 when FP16 suffices, unneeded boundary checks, redundant instructions.

Fields

precision_waste_pct: f64

redundant_instructions: u64

unnecessary_bounds_checks: u64

Inventory

Muda of Inventory: Resources allocated but unused. Examples: shared memory allocated but not used, registers reserved but unused.

Fields

unused_shared_memory_bytes: u64

unused_registers_per_thread: u32

occupancy_loss_pct: f64

Motion

Muda of Motion: Excessive control flow. Examples: warp divergence, branch overhead, loop overhead.

Fields

divergent_branches: u64

branch_efficiency_pct: f64

loop_overhead_cycles: u64

Defects

Muda of Defects: Incorrect results requiring rework. Examples: NaN propagation, precision loss, numerical instability.

Fields

nan_count: u64

inf_count: u64

precision_loss_bits: f64

Overproduction

Muda of Overproduction: Computing results that aren’t needed. Examples: padding waste, inactive threads in partial tiles.

Fields

padding_waste_pct: f64

inactive_thread_pct: f64

unused_output_elements: u64

Trait Implementations

impl Clone for GpuMuda

fn clone(&self) -> GpuMuda

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GpuMuda

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for GpuMuda

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for GpuMuda

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.