xakep8 wrote:
> > > The results look right here, but it isn't clear to me how we are getting
> > > there. Can you elaborate on that?
> >
> > ```c++
> > int x;
> > unsigned long gx = (unsigned long)&x;
> > ```
> >
> > Here the base of x is int so tryEmitBase gives us `#cir.global_view<@x>`
> > then in tryEmit we retag that to `#cir.global_view<@x> : !u64i` similarly
> > for offset adjustment like for arrays we get `#cir.global_view<@arr, [2 :
> > i32]> : !u64i`.
> >
> > Then lowering CIR to LLVM emits
> >
> > ```llvm
> > llvm.mlir.global external @gx() : i64 {
> > %0 = llvm.mlir.addressof @x : !llvm.ptr
> > %1 = llvm.ptrtoint %0 : !llvm.ptr to i64
> > llvm.return %1 : i64
> > }
> > ```
> >
> > and the final IR is
> >
> > ```
> > @x = global i32 0, align 4
> > @gx = global i64 ptrtoint (ptr @x to i64), align 8
> > ```
>
> Thanks for the explanation. It's unfortunate that we're still lowering this
> to a region-based initializer, but I can live with that for now.
I think we could use the compact initializer for integer pointers but not for
array/offset-adjusted pointers. Should I raise a PR for that?
https://github.com/llvm/llvm-project/pull/220643
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
