Spanda

Native deploy (LLVM)

Spanda can emit a linked native binary for field deployment when the CLI is built with the llvm feature (default). This path complements the interpreter (spanda run) and WASM deploy manifest (spanda deploy --target wasm).

Quick start

cargo build -p spanda --release --features llvm
spanda compile-native examples/showcase/killer_demo.sd
# → target/spanda-native/spanda-program

Or via deploy:

spanda deploy --target native examples/showcase/killer_demo.sd

Commands

Command	Output
`spanda llvm-ir <file.sd>`	LLVM IR (`.ll`) for inspection
`spanda compile-native <file.sd>`	LLVM IR + linked binary under `target/spanda-native/`
`spanda deploy --target native <file.sd>`	Same as `compile-native` with deploy-oriented defaults

Flags

Flag	Purpose
`--out <path>`	Output binary path
`--target-triple <triple>`	Cross-compile triple (e.g. `aarch64-unknown-linux-gnu`)
`--hal-profile <name>`	HAL profile baked into codegen metadata

Requirements

clang on PATH (links libspanda_rt)
Programs must lower to SIR successfully (spanda check first)

Embedded cross-build example (CI: llvm-embedded-golden-path):

spanda compile-native --target-triple aarch64-unknown-linux-gnu \
  --hal-profile jetson examples/showcase/killer_demo.sd

When to use native vs interpreter

Runtime	Best for
`spanda run` / `spanda sim`	Development, triggers, agents, full language surface
Native binary	Fixed behaviors, edge nodes with clang toolchain, HAL-tuned builds

Native codegen covers a subset of the language today. Use spanda check and known-limitations.md before relying on native output in production.

CI

Job	Script
`llvm-golden-path`	`scripts/llvm_golden_path.sh`
`llvm-embedded-golden-path`	`scripts/llvm_embedded_golden_path.sh`

This site is open source. Improve this page.