v0.3.0

What's Changed

• ci: bump actions/upload-pages-artifact from 3 to 4 by @dependabot[bot] in #107
• ci: bump actions/checkout from 4 to 6 by @dependabot[bot] in #108
• ci: bump actions/download-artifact from 4 to 7 by @dependabot[bot] in #109
• ci: bump actions/upload-artifact from 4 to 6 by @dependabot[bot] in #110
• deps: bump prost from 0.13.5 to 0.14.3 by @dependabot[bot] in #112
• deps: bump tract-onnx from 0.21.10 to 0.22.0 by @dependabot[bot] in #113
• Improve test coverage for core crates by @m96-chan in #114
• Add fuzz testing for WGSL parser by @m96-chan in #115
• Implement real vendor NPU backends (#104) by @m96-chan in #116
• feat: embed constant weights into ONNX output (#117) by @m96-chan in #118

New Contributors

• @dependabot[bot] made their first contribution in #107

Full Changelog: v0.2.0...v0.3.0

v0.2.0 — Test Coverage & CI Hardening

NxPU v0.2.0

テストカバレッジ拡充、CI強化、リリース品質改善のアップデートです。

Added

• Numerical correctness tests for tanh, conv2d, maxpool, and attention ops (#101)
• CLI unit tests for argument parsing, precision, and opt-level validation (#99)
• Workspace-level lint configuration with clippy pedantic (#98)
• cargo-deny license audit in CI (#97)
• Feature flags for backend selection (backend-onnx, backend-tflite, etc.) (#95)

Fixed

• Move miette fancy feature to CLI-only to reduce dependency footprint (#96)
• Fix MSRV CI and doc env for Windows (#102, #103)
• Fix security audit CI job permissions
• Resolve release quality issues #85-#92

Changed

• Annotate dev-only workspace dependencies (#93)

Full Changelog: v0.1.0...v0.2.0

v0.1.0 — Initial Release

NxPU v0.1.0

WGSL → NPU transpiler の初回リリースです。

Added

• WGSL-to-NPU transpiler with arena-based SSA IR
• nxpu-parser: WGSL parsing via naga with lowering to NxPU IR
• nxpu-ir: Intermediate representation with types, expressions, statements, and compute graphs
• nxpu-opt: Optimization pass framework with fixed-point iteration
  • Constant folding, FMA fusion, dead code elimination, IR validation
  • Quantization passes (F32 → F16, BF16, Int8, mixed-precision)
  • Shape inference and memory layout assignment (NHWC/NCHW)
• nxpu-analysis: Pattern classification (MatMul, Conv2D, ElementWise, Pool, Activation, Reduce, Transpose, Reshape, Normalization, Concat, Split, Attention) and fusion with tensor name connectivity
• nxpu-backend-core: Backend trait, plugin registry, IR dump backend
• Backend emitters: ONNX, TFLite, CoreML, StableHLO
• Vendor backends: Samsung, MediaTek, Intel, AMD, Qualcomm, Arm Ethos, CEVA, Rockchip
• nxpu-cli: CLI with target selection, optimization levels, precision control, --list-targets, RUST_LOG support
• End-to-end tests covering all backends, error paths, and numerical validation
• Documentation: architecture guide, backend implementer guide, CHANGELOG, CONTRIBUTING, SECURITY

Full Changelog: https://github.com/m96-chan/NxPU/commits/v0.1.0