Re: [PATCH v15 5/6] rust: time: Add wrapper for fsleep() function

[Boqun Feng <boqun.feng@gmail.com>]

On Tue, Apr 29, 2025 at 06:11:02PM +0200, Arnd Bergmann wrote:
> On Tue, Apr 29, 2025, at 18:03, Boqun Feng wrote:
> > On Tue, Apr 29, 2025, at 8:51 AM, Arnd Bergmann wrote:
> >> On Tue, Apr 29, 2025, at 15:17, FUJITA Tomonori wrote:
> >>> On Mon, 28 Apr 2025 20:16:47 +0200 Andreas Hindborg <a.hindborg@kernel.org> wrote:
> >>>      /// Return the number of milliseconds in the [`Delta`].
> >>>      #[inline]
> >>> -    pub const fn as_millis(self) -> i64 {
> >>> -        self.as_nanos() / NSEC_PER_MSEC
> >>> +    pub fn as_millis(self) -> i64 {
> >>> +        math64::div64_s64(self.as_nanos(), NSEC_PER_MSEC)
> >>>      }
> >>>  }
> >>
> >> I think simply calling ktime_to_ms()/ktime_to_us() should result
> >> in reasonably efficient code, since the C version is able to
> >> convert the constant divisor into a multiply/shift operation.
> >>
> >
> > Well, before we jump into this, I would like to understand why
> > this is not optimized with multiply/shift operations on arm in
> > Rust code. Ideally all the dividing constants cases should not
> > need to call a C function.
> 
> I think it's just because nobody has rewritten the
> macros from include/asm-generic/div64.h into rust code.
> 
> The compiler could do the same transformation, but they
> generally just fall back to calling a libgcc function.
> 

FWIW, I found this:

	https://github.com/llvm/llvm-project/issues/63731

seems a WIP though.

Would it make sense if we rely on compiler optimization when it's
avaiable (for x86_64, arm64, riscv, etc), and only call ktime_to_ms() if
not? The downside of calling ktime_to_ms() are:

* it's a call function, and cannot be inlined with LTO or INLINE_HELPER:

	https://lore.kernel.org/all/20250319205141.3528424-1-gary@garyguo.net/

* it doesn't provide the overflow checking even if
  CONFIG_RUST_OVERFLOW_CHECKS=y

Thoughts?

Regards,
Boqun

>      Arnd