Auto merge of #142508 - Mark-Simulacrum:skip-noop-drop-glue, r=fee1-dead

Skip no-op drop glue

Since rust-lang/rust#122662 this no longer gets used in vtables, so we're safe to fully
drop generating functions from vtables. Those are eventually cleaned up
by LLVM, but it's wasteful to produce them in the first place.

This doesn't appear to be a significant win (and shows some slight regressions) but
seems like the right thing to do. At minimum it reduces noise in the LLVM IR we generate,
which seems like a good thing.

1 files changed, 10 insertions, 2 deletions

diff --git a/compiler/rustc_monomorphize/src/collector.rs b/compiler/rustc_monomorphize/src/collector.rs
index 173030e0326..e90e32ebebb 100644
--- a/compiler/rustc_monomorphize/src/collector.rs
+++ b/compiler/rustc_monomorphize/src/collector.rs
@@ -949,6 +949,9 @@ fn visit_instance_use<'tcx>(
         }
         ty::InstanceKind::DropGlue(_, None) => {
             // Don't need to emit noop drop glue if we are calling directly.
+            //
+            // Note that we also optimize away the call to visit_instance_use in vtable construction
+            // (see create_mono_items_for_vtable_methods).
             if !is_direct_call {
                 output.push(create_fn_mono_item(tcx, instance, source));
             }
@@ -1177,8 +1180,13 @@ fn create_mono_items_for_vtable_methods<'tcx>(
         output.extend(methods);
     }
 
-    // Also add the destructor.
-    visit_drop_use(tcx, impl_ty, false, source, output);
+    // Also add the destructor, if it's necessary.
+    //
+    // This matches the check in vtable_allocation_provider in middle/ty/vtable.rs,
+    // if we don't need drop we're not adding an actual pointer to the vtable.
+    if impl_ty.needs_drop(tcx, ty::TypingEnv::fully_monomorphized()) {
+        visit_drop_use(tcx, impl_ty, false, source, output);
+    }
 }
 
 /// Scans the CTFE alloc in order to find function pointers and statics that must be monomorphized.