Skip to content
heliaPROFILER
Reference
HELIA

fixture_runtime

The compiled firmware a build produces and what it recorded: the build record, its memory plan, toolchain and PSRAM weight placement.

Every name on this page is imported from helia_profiler.

API tier: experimental

Generated from the src/helia_profiler tree f38e378c801d34438affb680257217e0b58d2cd1.

Machine-readable model

class

FixturePsramWeights

Python

Expected constant payloads; an uncompiled build has no image binding.

src/helia_profiler/fixtures/psram.py:43

FixturePsramWeights(
    payloads: tuple[FixturePsramPayload, ...],
    image_sha256: str | None = None,
    clock_hz: int = 96000000,
) -> None

dataclass

Expected constant payloads; an uncompiled build has no image binding.

API tier: experimental

class

FixtureBuild

Python

API tier: experimental

src/helia_profiler/fixtures/types.py:177

FixtureBuild(
    fixture_identity: str,
    app_dir: Path,
    binary: FixtureFile | None,
    generated_sources: tuple[FixtureFile, ...],
    expected: FixtureFile | None,
    built: bool,
    runtime_manifest: FixtureFile | None,
    flat_binary: FixtureFile | None,
    dependency_lock: FixtureFile | None,
    link_map: FixtureFile | None,
    timing_scope: FixtureTimingScope,
    engine: EngineType,
    build_identity: str | None,
    iterations: int,
    warmups: int,
    intent_identity: str,
    target: FixtureTarget,
    toolchain: ToolchainInfo | None,
    provider_provenance: str = 'manifest-declared; independently audit the pinned build/source record',
    planned_memory: MemoryPlan | None = None,
    planned_memory_reason: str | None = 'not_recorded',
    outputs: tuple[FixtureFile, ...] = (),
    capabilities: tuple[tuple[str, FixtureCapability], ...] = (),
    aot_arena_scan: tuple[tuple[int, int], ...] = (),
    aot_outputs: tuple[FixtureFile, ...] = (),
    engine_source: EngineSource | None = None,
    dependency_state: FixtureFile | None = None,
    state_policy: FixtureStatePolicy | None = None,
    source_initializer_present: bool = False,
    psram_weights: FixturePsramWeights | None = None,
    arena_location: Placement | None = None,
    weights_location: Placement | None = None,
    model: FixtureFile | None = None,
    inputs: tuple[FixtureIO, ...] = (),
    output_tensors: tuple[FixtureTensor, ...] = (),
    arena_capacity: int | None = None,
    clock_profile: str | None = None,
    cpu_hz: int | None = None,
    timer_hz: int | None = None,
    power_sync: FixturePowerSync | None = None,
    observe_ram: bool = False,
    kernel_source: FixtureKernelSource | None = None,
    runtime_source: FixtureRuntimeSource | None = None,
) -> None

dataclass

API tier: experimental

class

ToolchainInfo

Python

Build toolchain versions (populated by stage 4).

src/helia_profiler/results/models.py:254

ToolchainInfo(compiler: str = '', compiler_version: str = '', cmake_version: str = '') -> None

dataclass

Build toolchain versions (populated by stage 4).

API tier: experimental

class

MemoryPlan

Python

Engine-agnostic memory plan for a single profiling run.

src/helia_profiler/results/models.py:441

MemoryPlan(
    engine: EngineType,
    regions: tuple[MemoryRegionUsage, ...] = (),
    model_weight_bytes: int = 0,
    has_overflow: bool = False,
) -> None

dataclass

Engine-agnostic memory plan for a single profiling run.

Produced by the plan_memory stage by combining engine-specific knowledge (AOT arena regions, TFLM single-arena size, weight placement) with the SoC’s physical memory layout. Consumed by the report and by the firmware template generator for placement macros / linker hints.

API tier: experimental

Machine-readable: fixture_runtime.json · fixture_runtime.md, complete and lossless.