Fortran-lang GSoC 2024 Project ideas & mentors

[gnikit]

We have applied for GSoC 2024 and now it would be a good time to start brainstorming about potential projects we want Fortran-lang to offer and willing mentors to supervise them. Ideas do not necessarily have to be for projects under the Fortran-lang GitHub organisation, e.g. fpm, stdlib, fortls, etc.

The GSoC projects are hosted in Fortran-lang’s wiki web page

GSoC 2024 Project ideas

[gnikit]

Project recommendation by @FedericoPerini

@FedericoPerini

Fortran Package Manager: Build Process Enhancements

Project Overview:
Fortran Package Manager (fpm) is pivotal for long-term Fortran success. This GSoC project aims to improve fpm’s build process by improving dependency detection, optimizing linking, implementing shared libraries, ensuring safe concurrent builds, and introducing external Makefile generation.

Project Tasks:

1. Dependency Detection:

• Enhance fpm’s dependency detection to minimize rebuilds by parsing or hashing module/submodule files or parsing procedure interfaces in module files. fpm should not rebuild dependencies to a module whose public interface has not changed.

2. Linking Optimization:

• Replace one-liner linking with static libraries to prevent line buffer overflow in Windows builds.

3. Shared Library Implementation:

• Introduce support for shared library targets for project flexibility.

4. Safe Concurrent Builds:

• Implement file locking for safe concurrent invocations, especially during OpenMP builds, to prevent data corruption.

5. External Makefile Generation:

• Enable generation of external Makefiles akin to cmake -G for advanced project configuration.

Expected Outcomes:

• Enhanced dependency tracking and reduced rebuild times.
• Improved reliability in linking, particularly in Windows.
• Increased project versatility with shared library support.
• Safer concurrent builds through file locking.
• Greater project configuration flexibility with external Makefile generation.

Difficulty:

Intermediate, 175 hours.

Mentors:

• Federico Perini @FedericoPerini, Jose Alvez @hkvzjal, [call for interest!]

Project Repository: GitHub - fortran-lang/fpm: Fortran Package Manager (fpm)

Prerequisites for Applicants:

• Proficiency in Fortran.
• Understanding of package management and build systems.
• Strong software development skills.

Conclusion:
This GSoC project invites contributions to improving the Fortran Package Manager strengthening and speeding up its build process.

---

Original post:

Will the Fortran Community participate at GSoC 2024?

I would like to propose the following for fpm and call for prospective mentors to fpm-related projects. If this is of interest, maybe it could replace the following '23 proposals: Source Code Distribution with Foreign Build Systems (fpm) → embedded Support for C++ (fpm) → was discarded last year Parallel builds (fpm) → embedded The focus would be on strengthening the fpm build process addressing the following reqiests from the community: Fortran Package Manager: Build Process Enhancements …

[gnikit]

Project recommendation by @hkvzjal

@hkvzjal

String to number conversion (stdlib)

This project will enhance stdlib’s string handling capabilities for fast number parsing in Fortran.

Recently, a new module was added to stdlib called stdlib_str2num which implements fast routines for converting strings to numerical types. The participant would get familiar with these implementations and subsequently:

• Create a full benchmark suite for the string to number conversion, across compiler vendors, operating systems, and CPU architectures.
• Explore ways to improve robustness and efficiency, e.g. error handling.
• Propose a shallow interface for the string_type facility in stdlib.
• Propose an enhancement to the loadtxt facility function to speed-up file reading.
• Depending on the advancement, the participant is also encouraged to include a roadmap for inclusion of the inverse conversion by following the initiative in this thread ryu-based to_string function

Relevant thread on Fortran Discrouse: Faster string to double

Expected outcomes: Enhancement of stdlib fast string to number conversion

Skills preferred: Fortran and C programming, understanding of floating-point arithmetic

Difficulty: intermediate, 175 hours

Mentors:
José Alves @hkvzjal
[call for mentors]
(@Carltoffel @everythingfunctional @ivanpribec you were listed in the previous proposal, do you want to stay in the list?)
@milancurcic (I added the shallow interface for string_type following your comment in the PR)

---

Original post:

Will the Fortran Community participate at GSoC 2024?

Here another idea in replacement of the Faster string to number conversion (stdlib) one String to number conversion (stdlib) This project will enhance stdlib’s string handling capabilities for fast number parsing in Fortran. Recently, a new module was added to stdlib called stdlib_str2num which implements fast routines for converting strings to numerical types. The participant would get familiar with these implementations and subsequently: Create a full benchmark suite for the string to num…

[certik]

Go ahead and put them all in.

[gnikit]

These are the projects I have on my to-do list.
My availability during the summer is questionable right now, but feel free to reach out either for mentoring or applying.

---

MPI support (fortls)

fortls has support for Fortran intrinsics, Standard modules and OpenMP.
It does not however support MPI. The goal of this project is to add full support
for completions, hover and signature help for MPI variables, subroutines and functions.

Due to the size of the MPI standard, the process of extracting the necessary
information from the standard such as names, interfaces and documentation will
be automated. The student will be responsible for creating a scraper/parser
to fetch the necessary information from the MPI standard and then create the
serialised data (JSON) to be used by fortls.

Discourse thread: MPI documentation and interfaces

Expected outcomes: fortls will have completion and hover support for MPI.

Skills preferred: Python, Fortran

Difficulty: intermediate, 175 hours

Mentors:
Giannis Nikiteas (@gnikit)

Semantic highlighting and collapsable scopes (fortls)

As part of this project the student will add support to fortls for the
Semantics Tokens request, which is used to provide improved syntax highlighting
and the Folding Range request, which is used to provide collapsable scopes.

Related Issues:

• fortls#56

Expected outcomes: fortls will support for semantic highlighting and collapsable scopes.

Skills preferred: Python, Fortran

Difficulty: intermediate, 175 hours

Mentors:
Giannis Nikiteas (@gnikit)

Replace explicit interface with pygls (fortls)

fortls uses explicit interfaces to the Language Server Protocol (LSP).
To decrease code duplication and increase maintainability, the work of
maintaining the explicit interfaces should be replaced with the use of
pygls’s types module.

Related Issues:

• fortls#96

Expected outcomes: fortls uses pygls’ to define LSP interfaces, types and requests.

Skills preferred: Python, Fortran

Difficulty: advanced, 350 hours

Mentors:
Giannis Nikiteas (@gnikit)

Python environment manager (vscode-fortran-support)

In the Modern Fortran for VS Code extension, the use of Python as a means to
install third party tools is essential. The goal of this project is to
create a robust Python environment manager for installing and running third
party tools such as fortls, fpm, findent, etc., taking into account
the user’s setup (venv, conda, system Python, etc.).

Expected outcomes: Modern Fortran for VS Code will have a robust Python environment manager for installing and running third party tools.

Skills preferred: Typescript, Python

Difficulty: advanced, 175 hours

Mentors:
Giannis Nikiteas (@gnikit)

vscode integration with fpm (vscode-fortran-support)

The goal of this project is to allow fpm integration with the Modern Fortran
extension for Visual Studio Code, similar to how CMake and Meson are integrated
in VS Code.

Using an Activity bar icon, the user will be able to build and run projects,
tests and examples. The student will be responsible for creating the GUI
integration and the necessary backend to communicate with fpm.

Expected outcomes: Modern Fortran for VS Code will have a GUI integration
with fpm to build and run projects, tests and examples.

Skills preferred: Typescript, Fortran

Difficulty: advanced, 350 hours

Mentors:
Giannis Nikiteas (@gnikit)

Export build order and compile_commands.json (fpm)

fpm has the ability to automatically determine the build order of a project’s
source files. This information is valuable to third party tools such as
language servers and code analysis tools. The goal of this project is to
export the build order of a project’s source files in the compile_commands.json.

An extension of this project is to implement the full syntax of
compile_commands.json as described in the Clang documentation. This would bring fpm a step closer
to being compatible with other build tools.

Expected outcomes: fpm will export a compile_commands.json file.

Skills preferred: Fortran

Difficulty: advanced, 175-350 hours

Mentors:
Giannis Nikiteas (@gnikit)

[rgoswami]

I have one relevant idea this year, beyond the ones in the list. I’m happy to be listed as a mentor for any of the projects which need a hand.

Binding PRIMA through to C++ and Python

• PRIMA is slated to be part of SciPy as soon as bindings are provided
• Given my personal interest I would prefer this project to use F2PY, but that will depend on the candidate, plain Python-C or even Pybind11 is acceptable for this as well

Skills: Fortran, mathematical optimization, Python, C
Outcome: Full set of bindings to PRIMA functions and associated unit tests in C, C++ and Python
Mentors: Rohit Goswami (@rgoswami)
Difficulty: Advanced, 350 hours

[jeremie.vandenplas]

Regarding fpm, could a full support of fypp be a GSoC topic for 2024?

[Beliavsky]

Anuj Mohite will be enhancing the f951 [gfortran] compiler’s DO CONCURRENT construct while mentored by Tobias Burnus and Thomas Schwinge.

This was a Fortran project that was funded.

[Jweber]

There has been an additional project funded that might affect gfortran:

Georgii Burunsuzian will be implementing OpenMP and OpenACC
offloading to a separate process on the same host, also mentored by
Thomas Schwinge and Tobias Burnus.