Struct PerformanceProfile 

pub struct PerformanceProfile {

    pub iters: u32,
    pub catalogs: u32,
    pub ingest_us: u64,
    pub lowering_us: u64,
    pub contract_setup_us: u64,
    pub cuda_dispatch_run1_us: u64,
    pub cuda_dispatch_run2_us: u64,
    pub cuda_dispatch_extra_us: Vec<u64>,
    pub casefile_emit_us: u64,
    pub episodes_jsonl_emit_us: u64,
    pub audit_report_emit_us: u64,
    pub total_us: u64,
    pub events_emitted: u32,
    pub finite_cells: u32,
    pub fixture_byte_size: u64,
    pub catalogs_total_us: u64,
}

S-REAL.PERF per-dataset performance profile.

WHY: S-REAL.1 / .1.1 / .1.1.1 proved DSFB-GPU processes real datasets deterministically and emits human-readable artifacts. S-REAL.PERF answers the next honest question: how long does that take, and what dominates the wall? Every timing is host-Instant wall-clock microseconds; cudaEvent kernel-level timing remains S-PERF territory. The profile is runtime-dependent by design — the admission disclosure makes that explicit so an operator does not conflate timing-replay with byte-replay.

Fields

iters: u32

catalogs: u32

ingest_us: u64

lowering_us: u64

contract_setup_us: u64

cuda_dispatch_run1_us: u64

cuda_dispatch_run2_us: u64

cuda_dispatch_extra_us: Vec<u64>

casefile_emit_us: u64

episodes_jsonl_emit_us: u64

audit_report_emit_us: u64

total_us: u64

events_emitted: u32

finite_cells: u32

fixture_byte_size: u64

catalogs_total_us: u64

Sequential single-catalog total wall when –catalogs > 1. HONEST LABEL: NOT a batched dispatch; this is K sequential build_gpu calls on the same (events, contract). Reported so the operator can see launch-overhead amortization without being misled into thinking the dispatcher supports true K-batched mode.

Trait Implementations

impl Clone for PerformanceProfile

fn clone(&self) -> PerformanceProfile

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PerformanceProfile

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

Formats the value using the given formatter.

impl Default for PerformanceProfile

fn default() -> PerformanceProfile

Returns the “default value” for a type.