On Mon, Sep 14, 2026 at 10:22:17AM +0100, Lorenzo Stoakes (ARM) wrote:
> The threads are taken from make's jobserver, so a parallel build is not
> oversubscribed. It was found that benefits level off at 8 threads (16 was
> found to be around the same, and 32 slower).
I think it's worth noting that this comes from rustc itself (it has
built-in support for the make jobserver). See commit ecab4115c44c
("kbuild: mark `rustc` (and others) invocations as recursive").
What happens when there is no jobserver, e.g. "make" without -j, or
"make -j1"?
> [...]
> rustc has a parallel front end, enabled with -Zthreads, available since
> rust 1.84 [0] and therefore on every rustc the kernel supports.
>
> It is not yet stable and its output is not reproducible, so it cannot be
> [...]
> +The number of threads rustc's parallel front end may use, passed to it as
> +``-Zthreads``. Unset by default, as the parallel front end is not yet
> +stable and its output is not reproducible. The gain levels off at 8
> +threads.
Okay, unstable, not reproducible.
> Generated output was confirmed byte-for-byte identical.
But this says otherwise? I am confused. Reproducible or not?
--
Kees Cook
