On Wed, Jul 7, 2021 at 12:19 PM 'Marco Elver' via Clang Built Linux
<clang-built-linux@xxxxxxxxxxxxxxxx> wrote:
>
> What is the story with Rust and LTO? Intuitively, I would expect Rust
> code to only perform optimally if the kernel is built with LTO
> (currently only supported via Clang).
Cross-language LTO is possible and is used in other major projects.
For instance, Firefox seems to use it:
https://bugzilla.mozilla.org/show_bug.cgi?id=1486042
However, we have not taken a look at it yet for the kernel.
> Because if calls to every one of these helpers are real calls, I would
> expect performance to be pretty poor. There's probably a reason these
> are macros or inlinable functions.
Maybe -- we need to measure.
> I would almost go so far and suggest that CONFIG_RUST be modified as
> follows:
Even if the performance was unacceptable, there are other ways than
LTO to solve this, such as implementing the equivalent macro/function
on Rust's side.
Furthermore, we need to consider GCC too.
Thus I would not be premature in making it depend on LTO.
Cheers,
Miguel
