Expand description
§Trueno-DB: GPU-First Embedded Analytics Database
Version: 0.1.0 (Phase 1 MVP)
Trueno-DB is a GPU-aware, compute-intensity-based embedded analytics database designed for high-performance aggregations with graceful degradation from GPU → SIMD → Scalar.
§Design Principles (Toyota Way Aligned)
- Muda elimination: Kernel fusion minimizes
PCIetransfers - Poka-Yoke safety: Out-of-core execution prevents VRAM OOM
- Genchi Genbutsu: Physics-based cost model (5x rule for GPU dispatch)
- Jidoka: Backend equivalence tests (GPU == SIMD == Scalar)
§Example Usage (Phase 1 MVP)
use trueno_db::storage::StorageEngine;
// Load Parquet file
let storage = StorageEngine::load_parquet("data/events.parquet")?;
// Iterate over 128MB morsels (out-of-core execution)
for morsel in storage.morsels() {
println!("Morsel: {} rows", morsel.num_rows());
}Re-exports§
Modules§
- backend
- Compute backend dispatcher
- error
- Error types for Trueno-DB
- experiment
- Experiment Tracking Schema (Phase 5: ENT-EPIC-001)
- gpu
- GPU compute backend using wgpu (WebGPU)
- kv
- Key-Value Store Module for PAIML Stack Integration (Phase 6)
- query
- Query parsing and execution
- storage
- Storage backend (Arrow/Parquet)
- topk
- Top-K selection algorithms
Structs§
- Database
- Database instance
- Database
Builder - Database builder
Enums§
- Backend
- Backend selection strategy