Crate gpuviewer_core 

gpuviewer-core — telemetry collection, data model, and event derivation.

Architecture (see CLAUDE.md and docs/research/04-synthesis.md):

• backend::GpuBackend: nvtop’s vendor-vtable split (static / dynamic / processes), per-field Option<T>, runtime-loaded vendor libs, failed init = skipped backend.
• events::EventEngine: derives the narrated “story” events from raw samples.
• mock::MockBackend: scripted simulation for CI/demos; the contract for real backends.

Re-exports

pub use backend::all_backends;

pub use backend::BackendError;

pub use backend::GpuBackend;

pub use events::Confidence;

pub use events::Event;

pub use events::EventEngine;

pub use events::EventKind;

pub use events::Severity;

pub use model::fmt_bytes;

pub use model::normalize_pci_id;

pub use model::now_ms;

pub use model::DeviceId;

pub use model::DynamicSample;

pub use model::ProcessKind;

pub use model::ProcessSample;

pub use model::StaticInfo;

pub use model::ThrottleReasons;

pub use model::Vendor;

pub use proc_meta::container_of;

pub use proc_meta::parse_cgroup;

pub use proc_meta::CpuTracker;

Modules

amd

AMD Linux backend — hand-rolled std::fs readers over sysfs/hwmon/fdinfo. No library linkage at all: librocm_smi64 is explicitly off the table (soname churn broke btop twice — btop #774) and libdrm ioctls are unnecessary for everything v1 needs.

backend

The vendor backend abstraction — nvtop’s struct gpu_vendor vtable, translated to Rust.

events

Event derivation — the “story” layer.

intel

Intel Linux backend — hand-rolled std::fs readers over sysfs/fdinfo, no library linkage (no Level Zero, no IGCL: per docs/research/02 their Linux coverage is gappy and root-gated; plain fdinfo+sysfs is the strong path).

mock

Mock backend: deterministic-ish simulation used for CI and demos (no GPU required).

model

Core data model. Every metric is Option<T>: absence (NOT_SUPPORTED, missing sysfs file, privilege wall) is a normal per-metric outcome, never an error.

nvidia

NVIDIA backend — direct NVML calls via nvml-wrapper (runtime-loads the driver’s own libnvidia-ml.so.1 / nvml.dll). No nvidia-smi anywhere: nvidia-smi is itself just a CLI wrapper over this same library.

proc_meta

Shared Linux per-process metadata helpers used across every Linux backend.

wddm

Windows cross-vendor WDDM backend (docs/design/cross-platform.md §3) — AMD and Intel on Windows (and NVIDIA when NVML is absent), entirely from OS surfaces: DXGI for enumeration and VRAM totals, PDH GPU counters for utilization and memory, D3DKMT for the LUID→PCI identity. pdh.dll/gdi32.dll/dxgi.dll are OS system libraries — linking them via the windows crate does NOT violate the no-vendor-SDK rule (that rule targets vendor SDKs with soname churn, not the OS itself).