On Thu, Aug 1, 2024 at 12:28 PM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> On Wed, Jul 31, 2024 at 11:34:13PM +0200, Alice Ryhl wrote:
>
> > > Please work harder to not have to duplicate stuff like this.
> >
> > I really didn't want to duplicate it, but it's very hard to find a
> > performant alternative. Is there any way we could accept duplication
> > only in the cases where an 'i' parameter is used? I don't have the
> > choice of using a Rust helper for 'i' parameters.
> >
> > Perhaps one option could be to put the Rust code inside jump_label.h
> > and have the header file evaluate to either C or Rust depending on the
> > value of some #ifdefs?
> >
> > #ifndef RUST_ASM
> > /* existing C code goes here */
> > #endif
> > #ifdef RUST_ASM
> > // rust code goes here
> > #endif
> >
> > That way the duplication is all in a single file. It would also avoid
> > the need for duplicating the nop5 string, as the Rust case is still
> > going through the C preprocessor and can use the existing #define.
>
> I suppose that is slightly better, but ideally you generate the whole of
> the Rust thing from the C version. After all, Clang can already parse
> this.
>
> That said, with the below patch, I think you should be able to reuse the
> JUMP_TABLE_ENTRY macro like:
>
> JUMP_TABLE_ENTRY({0}, {1}, {2} + {3})
Yeah, I think this can work. I will submit a follow-up patch that
removes the duplication soon.
Alice
