GSoC 2026 Final Report: A Reference Implementation for concore Library in Julia

[avinxshKD]

Project

My GSoC 2026 project added native Julia support to Concore.

The goal was to make Julia a first-class Concore language: no Python wrapper, the same wire format and synchronization model, and interoperability with existing Concore workflows.

This report only includes work completed during the GSoC coding period. Earlier prototype work and my pre-GSoC contributions are not included. All 16 PRs below were merged into dev.

Work completed

Core runtime and demos

• #552: Add Julia concore core runtime added the standalone concore.jl runtime with FileBackend, configuration and parameter loading, maxtime handling, synchronization, safe parsing, and wire-format support.
• #553: Add Julia mixed demo nodes added Julia controller and plant nodes, plus a mixed Python/Julia local study.

Tests and CI

• #554: Add Julia tests added parser, configuration, protocol, and synchronization tests.
• #555: Add interop and wire compatibility tests added bidirectional Julia/Python and Julia/C++ checks.
• #556: Add Julia test matrix to CI added Julia 1.10 and latest-stable testing on Ubuntu and Windows.
• #562: Fix SHM checks fixed the C++ SHM smoke compile and a timing assumption in the concurrency test found during CI validation.

Backends

• #557: Add MmapBackend added opt-in memory-mapped transport while keeping FileBackend as the default.
• #558: Add optional ZMQ backend added optional ZMQ.jl support, named ports, lifecycle handling, and a live REQ/REP test.

Docker support

• #561: Add Docker runtime wrapper added concoredocker.jl with Docker /in and /out path conventions.
• #563: Test the Docker runtime added focused tests for Docker paths, API aliases, read/write behavior, parameters, and maxtime.
• #564: Add Julia Docker generation added Dockerfile.jl and .jl handling to the existing study generator.
• #565: Add Julia mixed-language Docker study added a Python plant and Julia controller study. Both containers exchanged data and exited cleanly at maxtime.

Benchmarks and results

• #566: Add Julia backend comparisons added reproducible File/Mmap/ZMQ benchmarks and matched Julia/Python workloads.
• #567: Add Julia comparison results added the measured results, methodology, observations, and limitations.

On the tested Windows machine, Mmap reached 7.47 times the File rate and ZMQ reached 18.05 times the File rate. In the matched Julia/Python comparison, the observed Julia ratios were 3.20 for parsing, 1.15 for formatting, and 1.05 for File round trips. These are results from one machine, not general performance claims.

Documentation

• #569: Add Julia Documenter site added getting-started, API, backend selection, Docker, and wire-format documentation with a CI build check.

Optional interoperability

• #568: Add Julia MATLAB and Verilog interop completed the optional interoperability scope using real MATLAB and Verilog runtimes in both directions.

Final status

Julia now has an upstream native Concore runtime with File, Mmap, optional ZMQ, and Docker support. It has mixed-language demos, cross-language tests, CI coverage, benchmark evidence, and Documenter.jl documentation.

The final benchmark report is in benchmark/RESULTS.md.

I would like to sincerely thank my mentors @pradeeban, @mayureshkothare and @Rahuljagwani for all the reviews and guidance throughout the project.