• EDG open source C++ front end

    From Derek M Jones@derek@knosof.co.uk to comp.compilers on Fri Oct 2 23:05:30 2026
    From Newsgroup: comp.compilers

    All,

    EDG have open sourced their C++ front end
    https://edgcpp.org/

    Is there a commercial market for compilers anymore?

    The LLM generated compilers are creatures of their
    training data https://shape-of-code.com/2026/02/22/investigating-an-llm-generated-c-compiler/

    Is there a future for programming languages.
    Coding agents could cut out the middle man and go
    direct to assembler.

    --
    Derek M. Jones Evidence-based software engineering blog:https://shape-of-code.com
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From anton@anton@mips.complang.tuwien.ac.at to comp.compilers on Sat Oct 3 16:01:11 2026
    From Newsgroup: comp.compilers

    Derek M Jones <derek@knosof.co.uk> writes:
    All,

    EDG have open sourced their C++ front end
    https://edgcpp.org/

    Is there a commercial market for compilers anymore?

    There seems to be lots of money in gcc and LLVM development, and this
    money comes from commercial interests, so they are commercial
    compilers. It's just not the classic proprietary-software business
    model.

    Is there a future for programming languages.
    Coding agents could cut out the middle man and go
    direct to assembler.

    For now, they don't, for whatever reason. And the code they generate
    is stored in version control systems, not the prompts.

    - anton
    --
    M. Anton Ertl
    anton@mips.complang.tuwien.ac.at
    http://www.complang.tuwien.ac.at/anton/
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Cóilín Nioclásín Glostéir@thanks-to@Taf.com to comp.compilers on Mon Oct 5 16:55:40 2026
    From Newsgroup: comp.compilers

    Derek M Jones <derek@Knosof.co.UK> wrote: |---------------------------------------------|
    |"Is there a future for programming languages"| |---------------------------------------------|

    Of course. A person who uses an A.I. reports that to get non-trivial
    problems solved via an AI, that he needs to specify so many details
    that coding himself is easier.

    (S. HTTP://Gloucester.Insomnia247.NL/ fuer Kontaktdaten!)
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From anton@anton@mips.complang.tuwien.ac.at to comp.compilers on Mon Oct 5 17:33:29 2026
    From Newsgroup: comp.compilers

    Derek M Jones <derek@knosof.co.uk> writes:
    There seems to be lots of money in gcc and LLVM development, and this
    money comes from commercial interests, so they are commercial

    The gcc people and llvm people I spoke to about funding, some
    years ago, said that funding was hard to come by.

    Maybe. But it did come. They don't work on these compilers for free.

    Nowadays, there are lots of companies building new 'AI' chips,
    so demand has picked up. Or rather, job opportunities.

    If that has increased for them, that's an answer to:

    Is there a future for programming languages.
    Coding agents could cut out the middle man and go
    direct to assembler.

    For now, they don't, for whatever reason. And the code they generate
    is stored in version control systems, not the prompts.

    It does not matter where the code is stored.

    My sentence was phrased in a misleading way. What I meant is: Not the
    prompts (or not only), but primarily the generated code is stored in
    the version control systems. I.e., the companies that use AI in
    programming see a value in the code in a language that's higher-level
    than assembly language or machine code.

    To generate training data the model builders could use
    existing compilers to generate assembler, then feed this
    into model training.

    What would the advantage be? Can the LLM generate the assembly
    language more cheaply than what we get with the workflow that AI
    generates "high-level code and the compiler generates assembly/machine
    code"? What about human review that still happens AFAIK? What about
    programs that need to run on several architectures?

    - anton
    --
    M. Anton Ertl
    anton@mips.complang.tuwien.ac.at
    http://www.complang.tuwien.ac.at/anton/
    [I think that last point is important -- writing code that compiles
    cleanly to several modern architectures like amd64 and arm is easy,
    asking an LLM to do it, who knows? -John]
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Derek M Jones@derek@knosof.co.uk to comp.compilers on Tue Oct 6 22:00:27 2026
    From Newsgroup: comp.compilers

    Anton,

    My sentence was phrased in a misleading way. What I meant is: Not the prompts (or not only), but primarily the generated code is stored in
    the version control systems.

    Imagine if, when high level languages were new, that companies
    stored the compiler generated assembler in the version control system,
    rather than the source code.

    It is the specification used to generate the source that needs to
    be stored.

    The source is currently stored because coding agents are unreliable,
    and following existing practice is the lowest friction route.

    I.e., the companies that use AI in
    programming see a value in the code in a language that's higher-level
    than assembly language or machine code.

    If coding agents become reliable, the generated source becomes
    as interesting as the assembler generated by compilers.

    What would the advantage be? Can the LLM generate the assembly
    language more cheaply than what we get with the workflow that AI
    generates "high-level code and the compiler generates assembly/machine
    code"?

    Going to assembler will certainly require more tokens.
    Can coding agents generate smaller and/or more efficient assembler
    compared to going via source/compiler?

    We shall see.

    What about human review that still happens AFAIK? What about
    programs that need to run on several architectures?

    The review should happen at the specification stage.

    English is the new source code (probably some controlled).
    --
    Derek M. Jones Evidence-based software engineering blog:https://shape-of-code.com
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From anton@anton@mips.complang.tuwien.ac.at to comp.compilers on Wed Oct 7 07:10:28 2026
    From Newsgroup: comp.compilers

    Derek M Jones <derek@knosof.co.uk> writes:
    Anton,

    My sentence was phrased in a misleading way. What I meant is: Not the
    prompts (or not only), but primarily the generated code is stored in
    the version control systems.

    Imagine if, when high level languages were new, that companies
    stored the compiler generated assembler in the version control system,
    rather than the source code.

    Why would they? The compiler always produces the same binary from the
    source code.

    It is the specification used to generate the source that needs to
    be stored.

    The source is currently stored because coding agents are unreliable,

    The prompts are not a complete specification, and that's why
    unreliability is inherent in LLM code generation. The superpower of
    LLMs is that it takes a request that is missing a lot of things, and
    guesses the missing things based on the stuff in the training data.
    But the guesses can vary, so the unreliability is inherent in this
    development model.

    There is also the other unreliability where an LLM produces some stuff
    that is not aligned with what is explicitly asked for in the prompt,
    and maybe they will get that fixed (or, hopefully not; imagine what
    will happen once our oligarchs no longer need humans to build and
    control their machines, including swarms of killer drones, because the
    machines are reliably aligned with the oligarchs' prompts).

    Going to assembler will certainly require more tokens.

    That's a good reason to keep higher-level languages.

    Can coding agents generate smaller and/or more efficient assembler
    compared to going via source/compiler?

    Even if they can, for most programs, and for the largest part of the
    rest, efficiency is not important, and higher-level languages will be
    used for that code in order to reduce token costs.

    What about human review that still happens AFAIK? What about
    programs that need to run on several architectures?

    The review should happen at the specification stage.

    It's a benefit of LLMs that the need for a complete specification is
    avoided, but that means that reviewing at the prompt level is
    insufficient, even if the alignment problem is solved.

    - anton
    --
    M. Anton Ertl
    anton@mips.complang.tuwien.ac.at
    http://www.complang.tuwien.ac.at/anton/
    --- Synchronet 3.22a-Linux NewsLink 1.2