Struct FullEncodeThroughput 

pub struct FullEncodeThroughput {
    pub n_rows: usize,
    pub encode_secs: f64,
    pub rows_per_sec: f64,
    pub device_encode_engaged: bool,
}

End-to-end throughput of the FULL exact per-row encode for one batch.

Distinct from ResidentSolveThroughput (which times only the inner solve): rows_per_sec here is n_rows / encode_secs for the entire production certified_encode_batch — routing, per-row Newton, certificate, fallback, and the per-row reconstruction selection included.

Fields

n_rows: usize

Rows encoded in the timed batch.

encode_secs: f64

Wall-clock seconds for the full encode of the batch.

rows_per_sec: f64

n_rows / encode_secs (0.0 for a degenerate / non-positive time).

device_encode_engaged: bool

true ONLY if a device-resident exact-encode kernel actually ran the encode. No such kernel exists yet, so this is false even on a GPU host — the flag is the false-routing guard that keeps the CPU encode rate from ever being reported as a device measurement.

Implementations

impl FullEncodeThroughput

pub fn from_elapsed( n_rows: usize, elapsed: Duration, device_encode_engaged: bool, ) -> Self

Build a throughput record from a measured elapsed time. engaged is the caller’s honest assertion that a device-resident encode kernel produced the result; pass false for the host encode path.

Trait Implementations

impl Clone for FullEncodeThroughput

fn clone(&self) -> FullEncodeThroughput

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for FullEncodeThroughput

impl Debug for FullEncodeThroughput

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

Formats the value using the given formatter.