Syslab Deploy

Syslab Deploy is a template project for integrating Syslab/Julia programs into cross-platform GUI apps.

Features:

• Out-of-the-box support for Windows, macOS, Linux and Android (iOS not tested)
• One line build command to compile Julia code into a static library for each platform
• Support 32-bit ARM embedded devices (used by Flutter Android debug build)

Usage

• Build for Windows: flutter build windows
• Build for Android: flutter build apk --split-per-abi

Architecture

• julia/: julia/build.jl contains the entrypoint for building Julia code into a static library. Running julia/compile.sh will build the library into a C++ project located in native/syslabcc_lib/syslabcc_lib/.

• native/syslabcc_lib/: Integrating the generated C++ project using Rust's cargokit/rinf toolchain. If you export different symbols from Julia, you might modify the lib.rs file:

  #[link(name = "syslabcc_lib", kind = "static")]
  extern "C" {
  
      // you modify this according to the symbols you export from Julia
      #[allow(non_snake_case)]
      pub fn simulate(
          U: *mut f64,   // Nt x Ny x Nx array
          u: *const f64, // Ny x Nx matrix
          n_x: i64,      // number of rows
          n_y: i64,      // number of columns
          n_t: i64,      // number of time steps
          alpha: f64,    // thermal diffusivity
          dx: f64,       // spatial step size (x)
          dy: f64,       // spatial step size (y)
          dt: f64,       // time step size
      ) -> u8;
  }

• native/hub:

  This uses standard rinf structure, which implements the interops between Flutter and Julia exported functions.

• lib/: This is the actual Flutter project which calls functions exported from your Julia library.

Limitations

Julia programs bundled with Syslab Deploy do not have out-of-the-box support for BLAS-related functionalities on Android, iOS, or other embedded platforms.

Julia heavily uses shared libraries created by non-Julia programs, which is tricky to support when migrating the codebase to Android/iOS and embedded devices. This is the reason why we use --no-blas option for SyslabCC.

To use BLAS-related functionalties, you need to create platform-specific binaries for the following native dependencies:

• https://github.com/JuliaLinearAlgebra/libblastrampoline
• https://github.com/OpenMathLib/OpenBLAS