dau build

Build tools for dau

Overview

dau-build turns declarative FPGA build specs into concrete artifacts: generated SystemVerilog, artlink.manifest/v0 artifact bundles, backend handoff manifests, Vivado Tcl, and ordered hardware command plans. Every operation is a typed ccflow.CallableModel composed from a Hydra config tree in dau_build/config, so the whole pipeline composes, inspects, and tests on a development machine, with the privileged steps (Vivado, JTAG, PCIe) gated behind an explicit execute=true.

Two synthesis engines are supported: Vivado (the FPGA bitstream flow) and yosys (open-source synthesis that runs in CI). The config plumbing is backend-agnostic, so more can be added.

Composition is the extension mechanism: any package can register its own config tree on the shared Hydra search path via a hydra.lernaplugins entry point, and its task/design/board groups compose through this CLI without dau-build importing that package. The core registry itself arrives this way — dau-core publishes every hardware building block as a CoreDefinition at /dau-core/<name>, and the synthesize-cores task resolves cores from that registry to run per-core out-of-context characterization (staged Tcl + command plan anywhere; model.execute=true on a synthesis host runs Vivado and flags envelope drift against the registered numbers):

dau-build task=tasks/build/synthesize-cores   'model.cores=[/dau-core/int32-predicate-filter]'   model.output_root=outputs/ooc model.part=xc7a200tfbg484-2

Quickstart

Build the checked-in identity example — no board or vendor tools required:

dau-build task=tasks/spec/inspect  model.spec_path=examples/identity/dau-build.yaml
dau-build task=tasks/spec/build     model.spec_path=examples/identity/dau-build.yaml model.output_root=outputs/identity
dau-build task=tasks/spec/validate  model.manifest_path=outputs/identity/dau-identity.manifest model.root=outputs/identity
dau-build task=tasks/sim/simulate   model.module=dau_identity_top model.spec_path=examples/identity/dau-build.yaml

Task and step names are path-style, mirroring the config tree (task=tasks/spec/inspect, step=steps/inspect).

Documentation

The documentation is organized into four sections:

• Tutorial — Build the identity example.
• How-to — Run a build end to end · Program a bitstream on dpv1 · Extend dau-build.
• Reference — Commands · Config groups · Task and step catalog.
• Explanation — Architecture: how Hydra composition, ccflow evaluation, and the search-path extension model work, and the current state of backend support.

[!NOTE] This library was generated using copier from the Base Python Project Template repository.