Auto merge of #136410 - saethlin:clean-up-cgu-internal-copy, r=compiler-errors

Remove InstanceKind::generates_cgu_internal_copy

This PR should not contain any behavior changes. Before this PR, the logic for selecting instantiation mode is spread across all of
* `instantiation_mode`
* `cross_crate_inlinable`
* `generates_cgu_internal_copy`
* `requires_inline`

The last two of those functions are not well-designed. The function that actually decides if we generate a CGU-internal copy is `instantiation_mode`, _not_ `generates_cgu_internal_copy`. The function `requires_inline` documents that it is about the LLVM `inline` attribute and that it is a hint. The LLVM attribute is called `inlinehint`, this function is also used by other codegen backends, and since it is part of instantiation mode selection it is *not* a hint.

The goal of this PR is to start cleaning up the logic into a sequence of checks that have a more logical flow and are easier to customize in the future (to do things like improve incrementality or improve optimizations without causing obscure linker errors because you forgot to update another part of the compiler).

1 files changed, 0 insertions, 44 deletions

diff --git a/compiler/rustc_middle/src/ty/instance.rs b/compiler/rustc_middle/src/ty/instance.rs
index 980d20f9d3b..e3b3eccffb5 100644
--- a/compiler/rustc_middle/src/ty/instance.rs
+++ b/compiler/rustc_middle/src/ty/instance.rs
@@ -302,50 +302,6 @@ impl<'tcx> InstanceKind<'tcx> {
         )
     }
 
-    /// Returns `true` if the machine code for this instance is instantiated in
-    /// each codegen unit that references it.
-    /// Note that this is only a hint! The compiler can globally decide to *not*
-    /// do this in order to speed up compilation. CGU-internal copies are
-    /// only exist to enable inlining. If inlining is not performed (e.g. at
-    /// `-Copt-level=0`) then the time for generating them is wasted and it's
-    /// better to create a single copy with external linkage.
-    pub fn generates_cgu_internal_copy(&self, tcx: TyCtxt<'tcx>) -> bool {
-        if self.requires_inline(tcx) {
-            return true;
-        }
-        if let ty::InstanceKind::DropGlue(.., Some(ty))
-        | ty::InstanceKind::AsyncDropGlueCtorShim(.., Some(ty)) = *self
-        {
-            // Drop glue generally wants to be instantiated at every codegen
-            // unit, but without an #[inline] hint. We should make this
-            // available to normal end-users.
-            if tcx.sess.opts.incremental.is_none() {
-                return true;
-            }
-            // When compiling with incremental, we can generate a *lot* of
-            // codegen units. Including drop glue into all of them has a
-            // considerable compile time cost.
-            //
-            // We include enums without destructors to allow, say, optimizing
-            // drops of `Option::None` before LTO. We also respect the intent of
-            // `#[inline]` on `Drop::drop` implementations.
-            return ty.ty_adt_def().is_none_or(|adt_def| {
-                match *self {
-                    ty::InstanceKind::DropGlue(..) => adt_def.destructor(tcx).map(|dtor| dtor.did),
-                    ty::InstanceKind::AsyncDropGlueCtorShim(..) => {
-                        adt_def.async_destructor(tcx).map(|dtor| dtor.ctor)
-                    }
-                    _ => unreachable!(),
-                }
-                .map_or_else(|| adt_def.is_enum(), |did| tcx.cross_crate_inlinable(did))
-            });
-        }
-        if let ty::InstanceKind::ThreadLocalShim(..) = *self {
-            return false;
-        }
-        tcx.cross_crate_inlinable(self.def_id())
-    }
-
     pub fn requires_caller_location(&self, tcx: TyCtxt<'_>) -> bool {
         match *self {
             InstanceKind::Item(def_id) | InstanceKind::Virtual(def_id, _) => {