Auto merge of #68414 - michaelwoerister:share-drop-glue, r=<try>

Also share drop-glue when compiling with -Zshare-generics (i.e. at opt-level=0)

This PR adds drop-glue to the set of monomorphizations that can be shared across crates via `-Zshare-generics`.

This version of the PR might have detrimental effects on performance as it makes lots of stuff dependent on a single query results (`upstream_monomorphizations_for(def_id_of_drop_in_place)`). That should be fixable but let's do a perf run first.

Potentially fixes issue #64140. (cc @alexcrichton)
The changes here are related to @matthewjasper's #67332 but should be mostly orthogonal.

r? @ghost

src/librustc/middle/exported_symbols.rs

@@ -3,7 +3,6 @@ use crate::ty::subst::SubstsRef;
 use crate::ty::{self, TyCtxt};
 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
 use rustc_hir::def_id::{DefId, LOCAL_CRATE};
-use std::cmp;
 use std::mem;
 
 /// The SymbolExportLevel of a symbols specifies from which kinds of crates
@@ -43,43 +42,6 @@ impl<'tcx> ExportedSymbol<'tcx> {
             ExportedSymbol::NoDefId(symbol_name) => symbol_name,
         }
     }
-
-    pub fn compare_stable(&self, tcx: TyCtxt<'tcx>, other: &ExportedSymbol<'tcx>) -> cmp::Ordering {
-        match *self {
-            ExportedSymbol::NonGeneric(self_def_id) => match *other {
-                ExportedSymbol::NonGeneric(other_def_id) => {
-                    tcx.def_path_hash(self_def_id).cmp(&tcx.def_path_hash(other_def_id))
-                }
-                ExportedSymbol::Generic(..) | ExportedSymbol::NoDefId(_) => cmp::Ordering::Less,
-            },
-            ExportedSymbol::Generic(self_def_id, self_substs) => match *other {
-                ExportedSymbol::NonGeneric(_) => cmp::Ordering::Greater,
-                ExportedSymbol::Generic(other_def_id, other_substs) => {
-                    // We compare the symbol names because they are cached as query
-                    // results which makes them relatively cheap to access repeatedly.
-                    //
-                    // It might be even faster to build a local cache of stable IDs
-                    // for sorting. Exported symbols are really only sorted once
-                    // in order to make the `exported_symbols` query result stable.
-                    let self_symbol_name =
-                        tcx.symbol_name(ty::Instance::new(self_def_id, self_substs));
-                    let other_symbol_name =
-                        tcx.symbol_name(ty::Instance::new(other_def_id, other_substs));
-
-                    self_symbol_name.cmp(&other_symbol_name)
-                }
-                ExportedSymbol::NoDefId(_) => cmp::Ordering::Less,
-            },
-            ExportedSymbol::NoDefId(self_symbol_name) => match *other {
-                ExportedSymbol::NonGeneric(_) | ExportedSymbol::Generic(..) => {
-                    cmp::Ordering::Greater
-                }
-                ExportedSymbol::NoDefId(ref other_symbol_name) => {
-                    self_symbol_name.cmp(other_symbol_name)
-                }
-            },
-        }
-    }
 }
 
 pub fn metadata_symbol_name(tcx: TyCtxt<'_>) -> String {

src/librustc/mir/mono.rs

@@ -79,7 +79,7 @@ impl<'tcx> MonoItem<'tcx> {
     }
 
     pub fn instantiation_mode(&self, tcx: TyCtxt<'tcx>) -> InstantiationMode {
-        let inline_in_all_cgus = tcx
+        let generate_cgu_internal_copies = tcx
             .sess
             .opts
             .debugging_opts
@@ -93,7 +93,7 @@ impl<'tcx> MonoItem<'tcx> {
                 // If this function isn't inlined or otherwise has explicit
                 // linkage, then we'll be creating a globally shared version.
                 if self.explicit_linkage(tcx).is_some()
-                    || !instance.def.requires_local(tcx)
+                    || !instance.def.generates_cgu_internal_copy(tcx)
                     || Some(instance.def_id()) == entry_def_id
                 {
                     return InstantiationMode::GloballyShared { may_conflict: false };
@@ -102,7 +102,7 @@ impl<'tcx> MonoItem<'tcx> {
                 // At this point we don't have explicit linkage and we're an
                 // inlined function. If we're inlining into all CGUs then we'll
                 // be creating a local copy per CGU
-                if inline_in_all_cgus {
+                if generate_cgu_internal_copies {
                     return InstantiationMode::LocalCopy;
                 }
 

src/librustc/query/mod.rs

@@ -783,6 +783,11 @@ rustc_queries! {
         }
         query upstream_monomorphizations_for(_: DefId)
             -> Option<&'tcx FxHashMap<SubstsRef<'tcx>, CrateNum>> {}
+
+        query upstream_drop_glue_for(substs: SubstsRef<'tcx>) -> Option<CrateNum> {
+            desc { "available upstream drop-glue for `{:?}`", substs }
+            no_force
+        }
     }
 
     Other {

src/librustc/ty/instance.rs

@@ -4,7 +4,7 @@ use crate::traits;
 use crate::ty::print::{FmtPrinter, Printer};
 use crate::ty::{self, SubstsRef, Ty, TyCtxt, TypeFoldable};
 use rustc_hir::def::Namespace;
-use rustc_hir::def_id::DefId;
+use rustc_hir::def_id::{CrateNum, DefId};
 use rustc_macros::HashStable;
 use rustc_target::spec::abi::Abi;
 
@@ -91,6 +91,38 @@ impl<'tcx> Instance<'tcx> {
         let ty = tcx.type_of(self.def.def_id());
         tcx.subst_and_normalize_erasing_regions(self.substs, param_env, &ty)
     }
+
+    pub fn upstream_monomorphization(&self, tcx: TyCtxt<'tcx>) -> Option<CrateNum> {
+        // If we are not in share generics mode, we don't link to upstream
+        // monomorphizations but always instantiate our own internal versions
+        // instead.
+        if !tcx.sess.opts.share_generics() {
+            return None;
+        }
+
+        // If this instance has non-erasable parameters, it cannot be a shared
+        // monomorphization. Non-generic instances are already handled above
+        // by `is_reachable_non_generic()`.
+        if self.substs.non_erasable_generics().next().is_none() {
+            return None;
+        }
+
+        if self.def_id().is_local() {
+            return None;
+        }
+
+        match self.def {
+            InstanceDef::Item(def_id) => {
+                tcx
+                    .upstream_monomorphizations_for(def_id)
+                    .and_then(|monos| monos.get(&self.substs).cloned())
+            }
+            InstanceDef::DropGlue(_, Some(_)) => {
+                tcx.upstream_drop_glue_for(self.substs)
+            }
+            _ => None,
+        }
+    }
 }
 
 impl<'tcx> InstanceDef<'tcx> {
@@ -114,7 +146,12 @@ impl<'tcx> InstanceDef<'tcx> {
         tcx.get_attrs(self.def_id())
     }
 
-    pub fn is_inline(&self, tcx: TyCtxt<'tcx>) -> bool {
+    /// Returns `true` if the LLVM version of this instance is unconditionally
+    /// marked with `inline`. This implies that a copy of this instance is
+    /// generated in every codegen unit.
+    /// Note that this is only a hint. See the documentation for
+    /// `generates_cgu_internal_copy` for more information.
+    pub fn requires_inline(&self, tcx: TyCtxt<'tcx>) -> bool {
         use crate::hir::map::DefPathData;
         let def_id = match *self {
             ty::InstanceDef::Item(def_id) => def_id,
@@ -127,8 +164,15 @@ impl<'tcx> InstanceDef<'tcx> {
         }
     }
 
-    pub fn requires_local(&self, tcx: TyCtxt<'tcx>) -> bool {
-        if self.is_inline(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::InstanceDef::DropGlue(..) = *self {

src/librustc/ty/query/keys.rs

@@ -116,6 +116,15 @@ impl Key for (DefId, SimplifiedType) {
     }
 }
 
+impl<'tcx> Key for SubstsRef<'tcx> {
+    fn query_crate(&self) -> CrateNum {
+        LOCAL_CRATE
+    }
+    fn default_span(&self, _: TyCtxt<'_>) -> Span {
+        DUMMY_SP
+    }
+}
+
 impl<'tcx> Key for (DefId, SubstsRef<'tcx>) {
     fn query_crate(&self) -> CrateNum {
         self.0.krate

src/librustc_codegen_llvm/attributes.rs

@@ -246,7 +246,7 @@ pub fn from_fn_attrs(
     }
 
     // FIXME(eddyb) consolidate these two `inline` calls (and avoid overwrites).
-    if instance.def.is_inline(cx.tcx) {
+    if instance.def.requires_inline(cx.tcx) {
         inline(cx, llfn, attributes::InlineAttr::Hint);
     }
 

src/librustc_codegen_llvm/callee.rs

@@ -130,12 +130,7 @@ pub fn get_fn(cx: &CodegenCx<'ll, 'tcx>, instance: Instance<'tcx>) -> &'ll Value
                     } else {
                         // This is a monomorphization of a generic function
                         // defined in an upstream crate.
-                        if cx
-                            .tcx
-                            .upstream_monomorphizations_for(instance_def_id)
-                            .map(|set| set.contains_key(instance.substs))
-                            .unwrap_or(false)
-                        {
+                        if instance.upstream_monomorphization(tcx).is_some() {
                             // This is instantiated in another crate. It cannot
                             // be `hidden`.
                         } else {

src/librustc_codegen_ssa/back/symbol_export.rs

@@ -5,7 +5,7 @@ use rustc::middle::codegen_fn_attrs::CodegenFnAttrFlags;
 use rustc::middle::exported_symbols::{metadata_symbol_name, ExportedSymbol, SymbolExportLevel};
 use rustc::session::config;
 use rustc::ty::query::Providers;
-use rustc::ty::subst::SubstsRef;
+use rustc::ty::subst::{GenericArgKind, SubstsRef};
 use rustc::ty::Instance;
 use rustc::ty::{SymbolName, TyCtxt};
 use rustc_codegen_utils::symbol_names;
@@ -17,8 +17,6 @@ use rustc_hir::Node;
 use rustc_index::vec::IndexVec;
 use syntax::expand::allocator::ALLOCATOR_METHODS;
 
-pub type ExportedSymbols = FxHashMap<CrateNum, Arc<Vec<(String, SymbolExportLevel)>>>;
-
 pub fn threshold(tcx: TyCtxt<'_>) -> SymbolExportLevel {
     crates_export_threshold(&tcx.sess.crate_types.borrow())
 }
@@ -96,7 +94,7 @@ fn reachable_non_generics_provider(
                     if !generics.requires_monomorphization(tcx) &&
                         // Functions marked with #[inline] are only ever codegened
                         // with "internal" linkage and are never exported.
-                        !Instance::mono(tcx, def_id).def.requires_local(tcx)
+                        !Instance::mono(tcx, def_id).def.generates_cgu_internal_copy(tcx)
                     {
                         Some(def_id)
                     } else {
@@ -240,19 +238,31 @@ fn exported_symbols_provider_local(
                 continue;
             }
 
-            if let &MonoItem::Fn(Instance { def: InstanceDef::Item(def_id), substs }) = mono_item {
-                if substs.non_erasable_generics().next().is_some() {
-                    symbols
-                        .push((ExportedSymbol::Generic(def_id, substs), SymbolExportLevel::Rust));
+            match *mono_item {
+                MonoItem::Fn(Instance { def: InstanceDef::Item(def_id), substs }) => {
+                    if substs.non_erasable_generics().next().is_some() {
+                        let symbol = ExportedSymbol::Generic(def_id, substs);
+                        symbols.push((symbol, SymbolExportLevel::Rust));
+                    }
+                }
+                MonoItem::Fn(Instance { def: InstanceDef::DropGlue(def_id, Some(ty)), substs }) => {
+                    // A little sanity-check
+                    debug_assert_eq!(
+                        substs.non_erasable_generics().next(),
+                        Some(GenericArgKind::Type(ty))
+                    );
+                    let symbol = ExportedSymbol::Generic(def_id, substs);
+                    symbols.push((symbol, SymbolExportLevel::Rust));
+                }
+                _ => {
+                    // Any other symbols don't qualify for sharing
                 }
             }
         }
     }
 
     // Sort so we get a stable incr. comp. hash.
-    symbols.sort_unstable_by(|&(ref symbol1, ..), &(ref symbol2, ..)| {
-        symbol1.compare_stable(tcx, symbol2)
-    });
+    symbols.sort_by_cached_key(|s| s.0.symbol_name_for_local_instance(tcx));
 
     Arc::new(symbols)
 }
@@ -311,6 +321,17 @@ fn upstream_monomorphizations_for_provider(
     tcx.upstream_monomorphizations(LOCAL_CRATE).get(&def_id)
 }
 
+fn upstream_drop_glue_for_provider<'tcx>(
+    tcx: TyCtxt<'tcx>,
+    substs: SubstsRef<'tcx>,
+) -> Option<CrateNum> {
+    if let Some(def_id) = tcx.lang_items().drop_in_place_fn() {
+        tcx.upstream_monomorphizations(LOCAL_CRATE).get(&def_id).and_then(|monos| monos.get(&substs).cloned())
+    } else {
+        None
+    }
+}
+
 fn is_unreachable_local_definition_provider(tcx: TyCtxt<'_>, def_id: DefId) -> bool {
     if let Some(hir_id) = tcx.hir().as_local_hir_id(def_id) {
         !tcx.reachable_set(LOCAL_CRATE).contains(&hir_id)
@@ -325,6 +346,7 @@ pub fn provide(providers: &mut Providers<'_>) {
     providers.exported_symbols = exported_symbols_provider_local;
     providers.upstream_monomorphizations = upstream_monomorphizations_provider;
     providers.is_unreachable_local_definition = is_unreachable_local_definition_provider;
+    providers.upstream_drop_glue_for = upstream_drop_glue_for_provider;
 }
 
 pub fn provide_extern(providers: &mut Providers<'_>) {

src/librustc_codegen_ssa/back/write.rs

@@ -2,7 +2,8 @@ use super::command::Command;
 use super::link::{self, get_linker, remove};
 use super::linker::LinkerInfo;
 use super::lto::{self, SerializedModule};
-use super::symbol_export::{symbol_name_for_instance_in_crate, ExportedSymbols};
+use super::symbol_export::symbol_name_for_instance_in_crate;
+
 use crate::{
     CachedModuleCodegen, CodegenResults, CompiledModule, CrateInfo, ModuleCodegen, ModuleKind,
     RLIB_BYTECODE_EXTENSION,
@@ -12,6 +13,7 @@ use crate::traits::*;
 use jobserver::{Acquired, Client};
 use rustc::dep_graph::{WorkProduct, WorkProductFileKind, WorkProductId};
 use rustc::middle::cstore::EncodedMetadata;
+use rustc::middle::exported_symbols::SymbolExportLevel;
 use rustc::session::config::{
     self, Lto, OutputFilenames, OutputType, Passes, Sanitizer, SwitchWithOptPath,
 };
@@ -205,6 +207,8 @@ impl<B: WriteBackendMethods> Clone for TargetMachineFactory<B> {
     }
 }
 
+pub type ExportedSymbols = FxHashMap<CrateNum, Arc<Vec<(String, SymbolExportLevel)>>>;
+
 /// Additional resources used by optimize_and_codegen (not module specific)
 #[derive(Clone)]
 pub struct CodegenContext<B: WriteBackendMethods> {

src/librustc_codegen_utils/symbol_names.rs

@@ -129,26 +129,8 @@ fn symbol_name_provider(tcx: TyCtxt<'tcx>, instance: Instance<'tcx>) -> ty::Symb
         //
         // For generics we might find re-usable upstream instances. For anything
         // else we rely on their being a local copy available.
-
         if is_generic(instance.substs) {
-            let def_id = instance.def_id();
-
-            if !def_id.is_local() && tcx.sess.opts.share_generics() {
-                // If we are re-using a monomorphization from another crate,
-                // we have to compute the symbol hash accordingly.
-                let upstream_monomorphizations = tcx.upstream_monomorphizations_for(def_id);
-
-                upstream_monomorphizations
-                    .and_then(|monos| monos.get(&instance.substs).cloned())
-                    // If there is no instance available upstream, there'll be
-                    // one in the current crate.
-                    .unwrap_or(LOCAL_CRATE)
-            } else {
-                // For generic functions defined in the current crate, there
-                // can be no upstream instances. Also, if we don't share
-                // generics, we'll instantiate a local copy too.
-                LOCAL_CRATE
-            }
+            instance.upstream_monomorphization(tcx).unwrap_or(LOCAL_CRATE)
         } else {
             // For non-generic things that need to avoid naming conflicts, we
             // always instantiate a copy in the local crate.