Combined Call 2025-12-03

Agenda

• Design docs and/or RFCs
  • [RFC] How to inform users why an executable stack is required
    • Currently trampolines are implemented on the stack requiring the stack to be executable and the linker emits a message about it, but this may be confusing to users
    • There are a few options to explore for how to improve this implementation (see RFC)
    • Reorganizing the table of contents for the documentation - seems a good idea and at least one person volunteered to review PRs related to this
    • Should trampolines be implemented on the stack or on the heap?
      • Kiran: Implementing on the heap is better, but do people have the bandwidth to re-implement on the heap?
    • This RFC is about improving the explanation to users, not necessarily about re-implementing trampolines in an improved fashion
  • `-ffp-contract=fast` Violates the Fortran Standard
    • Themos Tsikas: currently there is a disagreement about parentheses in rounding. We may need an interp request from the J3 committee.
    • Jean Perier: we need to account for behavior that apps rely on and provide a fallback mechanism if we make a change
  • [RFC] Command-line compatibility with gfortran
    • Related PR: https://github.com/llvm/llvm-project/pull/165579
    • Consensus: we should not try to aim for command-line compatibility with gfortran, and a list exists of specific options that are wanted by users for their specific cases. Support for options will be considered on a case-by-case basis.
    • Compatibility with Clang > with gfortran
• PRs of Note
  • [Flang][FIR] Introduce FIRToCoreMLIR pass. #168703
• Issues of Note
  • Invalid code for ‘where’
• FYI
  • CFP FOSDEM 2026 LLVM dev room
    • extended the submission deadline with 1 week until December 7th
    • Discourse post
  • Flang Liaison Report ot J3
  • Operational Maturity Round Table Notes
    • Possible switch to PR-only mechanism
• Other topics as time allows
  • Proposing an Interactive Fortran Workflow with Flang using Jupyter Notebooks
    • Anutosh Bhat - Fortran ecosystem lacks a modern interactive workflow
    • Something is available using LFortran, but it’s not perfect: https://github.com/lfortran/lfortran/blob/main/share/lfortran/nb/Demo1.ipynb
    • Flang seems a natural place to bring this into being
    • https://compiler-research.org/xeus-cpp-wasm/lab/index.html
    • There is clang-repl for this with C++
    • RFC coming
    • OpenMP notebook demo
    • CUDA notebook demo
    • prototype-debug.mp4
    • FOSDEM 2025 - O_o [ Flang + WASM ] o_O
  • Given the issues with the Call Notes document, this may be a good time to migrate to a Markdown document located in the repo

Details

• Consists of over 704,000 lines of code, documentation, build files, and test
• To date, over 11,430 commits have been made to Flang