1 //! Partitioning Codegen Units for Incremental Compilation
2 //! ======================================================
4 //! The task of this module is to take the complete set of monomorphizations of
5 //! a crate and produce a set of codegen units from it, where a codegen unit
6 //! is a named set of (mono-item, linkage) pairs. That is, this module
7 //! decides which monomorphization appears in which codegen units with which
8 //! linkage. The following paragraphs describe some of the background on the
9 //! partitioning scheme.
11 //! The most important opportunity for saving on compilation time with
12 //! incremental compilation is to avoid re-codegenning and re-optimizing code.
13 //! Since the unit of codegen and optimization for LLVM is "modules" or, how
14 //! we call them "codegen units", the particulars of how much time can be saved
15 //! by incremental compilation are tightly linked to how the output program is
16 //! partitioned into these codegen units prior to passing it to LLVM --
17 //! especially because we have to treat codegen units as opaque entities once
18 //! they are created: There is no way for us to incrementally update an existing
19 //! LLVM module and so we have to build any such module from scratch if it was
20 //! affected by some change in the source code.
22 //! From that point of view it would make sense to maximize the number of
23 //! codegen units by, for example, putting each function into its own module.
24 //! That way only those modules would have to be re-compiled that were actually
25 //! affected by some change, minimizing the number of functions that could have
26 //! been re-used but just happened to be located in a module that is
29 //! However, since LLVM optimization does not work across module boundaries,
30 //! using such a highly granular partitioning would lead to very slow runtime
31 //! code since it would effectively prohibit inlining and other inter-procedure
32 //! optimizations. We want to avoid that as much as possible.
34 //! Thus we end up with a trade-off: The bigger the codegen units, the better
35 //! LLVM's optimizer can do its work, but also the smaller the compilation time
36 //! reduction we get from incremental compilation.
38 //! Ideally, we would create a partitioning such that there are few big codegen
39 //! units with few interdependencies between them. For now though, we use the
40 //! following heuristic to determine the partitioning:
42 //! - There are two codegen units for every source-level module:
43 //! - One for "stable", that is non-generic, code
44 //! - One for more "volatile" code, i.e., monomorphized instances of functions
45 //! defined in that module
47 //! In order to see why this heuristic makes sense, let's take a look at when a
48 //! codegen unit can get invalidated:
50 //! 1. The most straightforward case is when the BODY of a function or global
51 //! changes. Then any codegen unit containing the code for that item has to be
52 //! re-compiled. Note that this includes all codegen units where the function
55 //! 2. The next case is when the SIGNATURE of a function or global changes. In
56 //! this case, all codegen units containing a REFERENCE to that item have to be
57 //! re-compiled. This is a superset of case 1.
59 //! 3. The final and most subtle case is when a REFERENCE to a generic function
60 //! is added or removed somewhere. Even though the definition of the function
61 //! might be unchanged, a new REFERENCE might introduce a new monomorphized
62 //! instance of this function which has to be placed and compiled somewhere.
63 //! Conversely, when removing a REFERENCE, it might have been the last one with
64 //! that particular set of generic arguments and thus we have to remove it.
66 //! From the above we see that just using one codegen unit per source-level
67 //! module is not such a good idea, since just adding a REFERENCE to some
68 //! generic item somewhere else would invalidate everything within the module
69 //! containing the generic item. The heuristic above reduces this detrimental
70 //! side-effect of references a little by at least not touching the non-generic
71 //! code of the module.
73 //! A Note on Inlining
74 //! ------------------
75 //! As briefly mentioned above, in order for LLVM to be able to inline a
76 //! function call, the body of the function has to be available in the LLVM
77 //! module where the call is made. This has a few consequences for partitioning:
79 //! - The partitioning algorithm has to take care of placing functions into all
80 //! codegen units where they should be available for inlining. It also has to
81 //! decide on the correct linkage for these functions.
83 //! - The partitioning algorithm has to know which functions are likely to get
84 //! inlined, so it can distribute function instantiations accordingly. Since
85 //! there is no way of knowing for sure which functions LLVM will decide to
86 //! inline in the end, we apply a heuristic here: Only functions marked with
87 //! `#[inline]` are considered for inlining by the partitioner. The current
88 //! implementation will not try to determine if a function is likely to be
89 //! inlined by looking at the functions definition.
91 //! Note though that as a side-effect of creating a codegen units per
92 //! source-level module, functions from the same module will be available for
93 //! inlining, even when they are not marked `#[inline]`.
95 use std::collections::hash_map::Entry;
99 use syntax::symbol::Symbol;
100 use rustc::hir::CodegenFnAttrFlags;
101 use rustc::hir::def::DefKind;
102 use rustc::hir::def_id::{CrateNum, DefId, LOCAL_CRATE, CRATE_DEF_INDEX};
103 use rustc::mir::mono::{Linkage, Visibility, CodegenUnitNameBuilder, CodegenUnit};
104 use rustc::middle::exported_symbols::SymbolExportLevel;
105 use rustc::ty::{self, DefIdTree, TyCtxt, InstanceDef};
106 use rustc::ty::print::characteristic_def_id_of_type;
107 use rustc::ty::query::Providers;
108 use rustc::util::common::time;
109 use rustc::util::nodemap::{DefIdSet, FxHashMap, FxHashSet};
110 use rustc::mir::mono::{MonoItem, InstantiationMode};
112 use crate::monomorphize::collector::InliningMap;
113 use crate::monomorphize::collector::{self, MonoItemCollectionMode};
115 pub enum PartitioningStrategy {
116 /// Generates one codegen unit per source-level module.
119 /// Partition the whole crate into a fixed number of codegen units.
120 FixedUnitCount(usize)
123 // Anything we can't find a proper codegen unit for goes into this.
124 fn fallback_cgu_name(name_builder: &mut CodegenUnitNameBuilder<'_>) -> Symbol {
125 name_builder.build_cgu_name(LOCAL_CRATE, &["fallback"], Some("cgu"))
128 pub fn partition<'tcx, I>(
131 strategy: PartitioningStrategy,
132 inlining_map: &InliningMap<'tcx>,
133 ) -> Vec<CodegenUnit<'tcx>>
135 I: Iterator<Item = MonoItem<'tcx>>,
137 let _prof_timer = tcx.prof.generic_activity("cgu_partitioning");
139 // In the first step, we place all regular monomorphizations into their
140 // respective 'home' codegen unit. Regular monomorphizations are all
141 // functions and statics defined in the local crate.
142 let mut initial_partitioning = {
143 let _prof_timer = tcx.prof.generic_activity("cgu_partitioning_place_roots");
144 place_root_mono_items(tcx, mono_items)
147 initial_partitioning.codegen_units.iter_mut().for_each(|cgu| cgu.estimate_size(tcx));
149 debug_dump(tcx, "INITIAL PARTITIONING:", initial_partitioning.codegen_units.iter());
151 // If the partitioning should produce a fixed count of codegen units, merge
152 // until that count is reached.
153 if let PartitioningStrategy::FixedUnitCount(count) = strategy {
154 let _prof_timer = tcx.prof.generic_activity("cgu_partitioning_merge_cgus");
155 merge_codegen_units(tcx, &mut initial_partitioning, count);
156 debug_dump(tcx, "POST MERGING:", initial_partitioning.codegen_units.iter());
159 // In the next step, we use the inlining map to determine which additional
160 // monomorphizations have to go into each codegen unit. These additional
161 // monomorphizations can be drop-glue, functions from external crates, and
162 // local functions the definition of which is marked with `#[inline]`.
163 let mut post_inlining = {
165 tcx.prof.generic_activity("cgu_partitioning_place_inline_items");
166 place_inlined_mono_items(initial_partitioning, inlining_map)
169 post_inlining.codegen_units.iter_mut().for_each(|cgu| cgu.estimate_size(tcx));
171 debug_dump(tcx, "POST INLINING:", post_inlining.codegen_units.iter());
173 // Next we try to make as many symbols "internal" as possible, so LLVM has
174 // more freedom to optimize.
175 if !tcx.sess.opts.cg.link_dead_code {
177 tcx.prof.generic_activity("cgu_partitioning_internalize_symbols");
178 internalize_symbols(tcx, &mut post_inlining, inlining_map);
181 // Finally, sort by codegen unit name, so that we get deterministic results.
182 let PostInliningPartitioning {
183 codegen_units: mut result,
184 mono_item_placements: _,
185 internalization_candidates: _,
188 result.sort_by_cached_key(|cgu| cgu.name().as_str());
193 struct PreInliningPartitioning<'tcx> {
194 codegen_units: Vec<CodegenUnit<'tcx>>,
195 roots: FxHashSet<MonoItem<'tcx>>,
196 internalization_candidates: FxHashSet<MonoItem<'tcx>>,
199 /// For symbol internalization, we need to know whether a symbol/mono-item is
200 /// accessed from outside the codegen unit it is defined in. This type is used
201 /// to keep track of that.
202 #[derive(Clone, PartialEq, Eq, Debug)]
203 enum MonoItemPlacement {
204 SingleCgu { cgu_name: Symbol },
208 struct PostInliningPartitioning<'tcx> {
209 codegen_units: Vec<CodegenUnit<'tcx>>,
210 mono_item_placements: FxHashMap<MonoItem<'tcx>, MonoItemPlacement>,
211 internalization_candidates: FxHashSet<MonoItem<'tcx>>,
214 fn place_root_mono_items<'tcx, I>(tcx: TyCtxt<'tcx>, mono_items: I) -> PreInliningPartitioning<'tcx>
216 I: Iterator<Item = MonoItem<'tcx>>,
218 let mut roots = FxHashSet::default();
219 let mut codegen_units = FxHashMap::default();
220 let is_incremental_build = tcx.sess.opts.incremental.is_some();
221 let mut internalization_candidates = FxHashSet::default();
223 // Determine if monomorphizations instantiated in this crate will be made
224 // available to downstream crates. This depends on whether we are in
225 // share-generics mode and whether the current crate can even have
226 // downstream crates.
227 let export_generics = tcx.sess.opts.share_generics() &&
228 tcx.local_crate_exports_generics();
230 let cgu_name_builder = &mut CodegenUnitNameBuilder::new(tcx);
231 let cgu_name_cache = &mut FxHashMap::default();
233 for mono_item in mono_items {
234 match mono_item.instantiation_mode(tcx) {
235 InstantiationMode::GloballyShared { .. } => {}
236 InstantiationMode::LocalCopy => continue,
239 let characteristic_def_id = characteristic_def_id_of_mono_item(tcx, mono_item);
240 let is_volatile = is_incremental_build &&
241 mono_item.is_generic_fn();
243 let codegen_unit_name = match characteristic_def_id {
244 Some(def_id) => compute_codegen_unit_name(tcx,
249 None => fallback_cgu_name(cgu_name_builder),
252 let codegen_unit = codegen_units.entry(codegen_unit_name)
253 .or_insert_with(|| CodegenUnit::new(codegen_unit_name));
255 let mut can_be_internalized = true;
256 let (linkage, visibility) = mono_item_linkage_and_visibility(
259 &mut can_be_internalized,
262 if visibility == Visibility::Hidden && can_be_internalized {
263 internalization_candidates.insert(mono_item);
266 codegen_unit.items_mut().insert(mono_item, (linkage, visibility));
267 roots.insert(mono_item);
270 // Always ensure we have at least one CGU; otherwise, if we have a
271 // crate with just types (for example), we could wind up with no CGU.
272 if codegen_units.is_empty() {
273 let codegen_unit_name = fallback_cgu_name(cgu_name_builder);
274 codegen_units.insert(codegen_unit_name, CodegenUnit::new(codegen_unit_name));
277 PreInliningPartitioning {
278 codegen_units: codegen_units.into_iter()
279 .map(|(_, codegen_unit)| codegen_unit)
282 internalization_candidates,
286 fn mono_item_linkage_and_visibility(
288 mono_item: &MonoItem<'tcx>,
289 can_be_internalized: &mut bool,
290 export_generics: bool,
291 ) -> (Linkage, Visibility) {
292 if let Some(explicit_linkage) = mono_item.explicit_linkage(tcx) {
293 return (explicit_linkage, Visibility::Default)
295 let vis = mono_item_visibility(
301 (Linkage::External, vis)
304 fn mono_item_visibility(
306 mono_item: &MonoItem<'tcx>,
307 can_be_internalized: &mut bool,
308 export_generics: bool,
310 let instance = match mono_item {
311 // This is pretty complicated; see below.
312 MonoItem::Fn(instance) => instance,
314 // Misc handling for generics and such, but otherwise:
315 MonoItem::Static(def_id) => {
316 return if tcx.is_reachable_non_generic(*def_id) {
317 *can_be_internalized = false;
318 default_visibility(tcx, *def_id, false)
323 MonoItem::GlobalAsm(hir_id) => {
324 let def_id = tcx.hir().local_def_id(*hir_id);
325 return if tcx.is_reachable_non_generic(def_id) {
326 *can_be_internalized = false;
327 default_visibility(tcx, def_id, false)
334 let def_id = match instance.def {
335 InstanceDef::Item(def_id) => def_id,
337 // These are all compiler glue and such, never exported, always hidden.
338 InstanceDef::VtableShim(..) |
339 InstanceDef::ReifyShim(..) |
340 InstanceDef::FnPtrShim(..) |
341 InstanceDef::Virtual(..) |
342 InstanceDef::Intrinsic(..) |
343 InstanceDef::ClosureOnceShim { .. } |
344 InstanceDef::DropGlue(..) |
345 InstanceDef::CloneShim(..) => {
346 return Visibility::Hidden
350 // The `start_fn` lang item is actually a monomorphized instance of a
351 // function in the standard library, used for the `main` function. We don't
352 // want to export it so we tag it with `Hidden` visibility but this symbol
353 // is only referenced from the actual `main` symbol which we unfortunately
354 // don't know anything about during partitioning/collection. As a result we
355 // forcibly keep this symbol out of the `internalization_candidates` set.
357 // FIXME: eventually we don't want to always force this symbol to have
358 // hidden visibility, it should indeed be a candidate for
359 // internalization, but we have to understand that it's referenced
360 // from the `main` symbol we'll generate later.
362 // This may be fixable with a new `InstanceDef` perhaps? Unsure!
363 if tcx.lang_items().start_fn() == Some(def_id) {
364 *can_be_internalized = false;
365 return Visibility::Hidden
368 let is_generic = instance.substs.non_erasable_generics().next().is_some();
370 // Upstream `DefId` instances get different handling than local ones.
371 if !def_id.is_local() {
372 return if export_generics && is_generic {
373 // If it is a upstream monomorphization and we export generics, we must make
374 // it available to downstream crates.
375 *can_be_internalized = false;
376 default_visibility(tcx, def_id, true)
384 if tcx.is_unreachable_local_definition(def_id) {
385 // This instance cannot be used from another crate.
388 // This instance might be useful in a downstream crate.
389 *can_be_internalized = false;
390 default_visibility(tcx, def_id, true)
393 // We are not exporting generics or the definition is not reachable
394 // for downstream crates, we can internalize its instantiations.
399 // If this isn't a generic function then we mark this a `Default` if
400 // this is a reachable item, meaning that it's a symbol other crates may
401 // access when they link to us.
402 if tcx.is_reachable_non_generic(def_id) {
403 *can_be_internalized = false;
404 debug_assert!(!is_generic);
405 return default_visibility(tcx, def_id, false)
408 // If this isn't reachable then we're gonna tag this with `Hidden`
409 // visibility. In some situations though we'll want to prevent this
410 // symbol from being internalized.
412 // There's two categories of items here:
414 // * First is weak lang items. These are basically mechanisms for
415 // libcore to forward-reference symbols defined later in crates like
416 // the standard library or `#[panic_handler]` definitions. The
417 // definition of these weak lang items needs to be referenceable by
418 // libcore, so we're no longer a candidate for internalization.
419 // Removal of these functions can't be done by LLVM but rather must be
420 // done by the linker as it's a non-local decision.
422 // * Second is "std internal symbols". Currently this is primarily used
423 // for allocator symbols. Allocators are a little weird in their
424 // implementation, but the idea is that the compiler, at the last
425 // minute, defines an allocator with an injected object file. The
426 // `alloc` crate references these symbols (`__rust_alloc`) and the
427 // definition doesn't get hooked up until a linked crate artifact is
430 // The symbols synthesized by the compiler (`__rust_alloc`) are thin
431 // veneers around the actual implementation, some other symbol which
432 // implements the same ABI. These symbols (things like `__rg_alloc`,
433 // `__rdl_alloc`, `__rde_alloc`, etc), are all tagged with "std
434 // internal symbols".
436 // The std-internal symbols here **should not show up in a dll as an
437 // exported interface**, so they return `false` from
438 // `is_reachable_non_generic` above and we'll give them `Hidden`
439 // visibility below. Like the weak lang items, though, we can't let
440 // LLVM internalize them as this decision is left up to the linker to
441 // omit them, so prevent them from being internalized.
442 let attrs = tcx.codegen_fn_attrs(def_id);
443 if attrs.flags.contains(CodegenFnAttrFlags::RUSTC_STD_INTERNAL_SYMBOL) {
444 *can_be_internalized = false;
451 fn default_visibility(tcx: TyCtxt<'_>, id: DefId, is_generic: bool) -> Visibility {
452 if !tcx.sess.target.target.options.default_hidden_visibility {
453 return Visibility::Default
456 // Generic functions never have export-level C.
458 return Visibility::Hidden
461 // Things with export level C don't get instantiated in
462 // downstream crates.
464 return Visibility::Hidden
467 // C-export level items remain at `Default`, all other internal
468 // items become `Hidden`.
469 match tcx.reachable_non_generics(id.krate).get(&id) {
470 Some(SymbolExportLevel::C) => Visibility::Default,
471 _ => Visibility::Hidden,
475 fn merge_codegen_units<'tcx>(
477 initial_partitioning: &mut PreInliningPartitioning<'tcx>,
478 target_cgu_count: usize,
480 assert!(target_cgu_count >= 1);
481 let codegen_units = &mut initial_partitioning.codegen_units;
483 // Note that at this point in time the `codegen_units` here may not be in a
484 // deterministic order (but we know they're deterministically the same set).
485 // We want this merging to produce a deterministic ordering of codegen units
488 // Due to basically how we've implemented the merging below (merge the two
489 // smallest into each other) we're sure to start off with a deterministic
490 // order (sorted by name). This'll mean that if two cgus have the same size
491 // the stable sort below will keep everything nice and deterministic.
492 codegen_units.sort_by_cached_key(|cgu| cgu.name().as_str());
494 // Merge the two smallest codegen units until the target size is reached.
495 while codegen_units.len() > target_cgu_count {
496 // Sort small cgus to the back
497 codegen_units.sort_by_cached_key(|cgu| cmp::Reverse(cgu.size_estimate()));
498 let mut smallest = codegen_units.pop().unwrap();
499 let second_smallest = codegen_units.last_mut().unwrap();
501 second_smallest.modify_size_estimate(smallest.size_estimate());
502 for (k, v) in smallest.items_mut().drain() {
503 second_smallest.items_mut().insert(k, v);
505 debug!("CodegenUnit {} merged in to CodegenUnit {}",
507 second_smallest.name());
510 let cgu_name_builder = &mut CodegenUnitNameBuilder::new(tcx);
511 for (index, cgu) in codegen_units.iter_mut().enumerate() {
512 cgu.set_name(numbered_codegen_unit_name(cgu_name_builder, index));
516 fn place_inlined_mono_items<'tcx>(initial_partitioning: PreInliningPartitioning<'tcx>,
517 inlining_map: &InliningMap<'tcx>)
518 -> PostInliningPartitioning<'tcx> {
519 let mut new_partitioning = Vec::new();
520 let mut mono_item_placements = FxHashMap::default();
522 let PreInliningPartitioning {
523 codegen_units: initial_cgus,
525 internalization_candidates,
526 } = initial_partitioning;
528 let single_codegen_unit = initial_cgus.len() == 1;
530 for old_codegen_unit in initial_cgus {
531 // Collect all items that need to be available in this codegen unit.
532 let mut reachable = FxHashSet::default();
533 for root in old_codegen_unit.items().keys() {
534 follow_inlining(*root, inlining_map, &mut reachable);
537 let mut new_codegen_unit = CodegenUnit::new(old_codegen_unit.name());
539 // Add all monomorphizations that are not already there.
540 for mono_item in reachable {
541 if let Some(linkage) = old_codegen_unit.items().get(&mono_item) {
542 // This is a root, just copy it over.
543 new_codegen_unit.items_mut().insert(mono_item, *linkage);
545 if roots.contains(&mono_item) {
546 bug!("GloballyShared mono-item inlined into other CGU: \
550 // This is a CGU-private copy.
551 new_codegen_unit.items_mut().insert(
553 (Linkage::Internal, Visibility::Default),
557 if !single_codegen_unit {
558 // If there is more than one codegen unit, we need to keep track
559 // in which codegen units each monomorphization is placed.
560 match mono_item_placements.entry(mono_item) {
561 Entry::Occupied(e) => {
562 let placement = e.into_mut();
563 debug_assert!(match *placement {
564 MonoItemPlacement::SingleCgu { cgu_name } => {
565 cgu_name != new_codegen_unit.name()
567 MonoItemPlacement::MultipleCgus => true,
569 *placement = MonoItemPlacement::MultipleCgus;
571 Entry::Vacant(e) => {
572 e.insert(MonoItemPlacement::SingleCgu {
573 cgu_name: new_codegen_unit.name()
580 new_partitioning.push(new_codegen_unit);
583 return PostInliningPartitioning {
584 codegen_units: new_partitioning,
585 mono_item_placements,
586 internalization_candidates,
589 fn follow_inlining<'tcx>(mono_item: MonoItem<'tcx>,
590 inlining_map: &InliningMap<'tcx>,
591 visited: &mut FxHashSet<MonoItem<'tcx>>) {
592 if !visited.insert(mono_item) {
596 inlining_map.with_inlining_candidates(mono_item, |target| {
597 follow_inlining(target, inlining_map, visited);
602 fn internalize_symbols<'tcx>(
604 partitioning: &mut PostInliningPartitioning<'tcx>,
605 inlining_map: &InliningMap<'tcx>,
607 if partitioning.codegen_units.len() == 1 {
608 // Fast path for when there is only one codegen unit. In this case we
609 // can internalize all candidates, since there is nowhere else they
610 // could be accessed from.
611 for cgu in &mut partitioning.codegen_units {
612 for candidate in &partitioning.internalization_candidates {
613 cgu.items_mut().insert(*candidate,
614 (Linkage::Internal, Visibility::Default));
621 // Build a map from every monomorphization to all the monomorphizations that
623 let mut accessor_map: FxHashMap<MonoItem<'tcx>, Vec<MonoItem<'tcx>>> = Default::default();
624 inlining_map.iter_accesses(|accessor, accessees| {
625 for accessee in accessees {
626 accessor_map.entry(*accessee)
632 let mono_item_placements = &partitioning.mono_item_placements;
634 // For each internalization candidates in each codegen unit, check if it is
635 // accessed from outside its defining codegen unit.
636 for cgu in &mut partitioning.codegen_units {
637 let home_cgu = MonoItemPlacement::SingleCgu {
641 for (accessee, linkage_and_visibility) in cgu.items_mut() {
642 if !partitioning.internalization_candidates.contains(accessee) {
643 // This item is no candidate for internalizing, so skip it.
646 debug_assert_eq!(mono_item_placements[accessee], home_cgu);
648 if let Some(accessors) = accessor_map.get(accessee) {
650 .filter_map(|accessor| {
651 // Some accessors might not have been
652 // instantiated. We can safely ignore those.
653 mono_item_placements.get(accessor)
655 .any(|placement| *placement != home_cgu) {
656 // Found an accessor from another CGU, so skip to the next
657 // item without marking this one as internal.
662 // If we got here, we did not find any accesses from other CGUs,
663 // so it's fine to make this monomorphization internal.
664 *linkage_and_visibility = (Linkage::Internal, Visibility::Default);
669 fn characteristic_def_id_of_mono_item<'tcx>(
671 mono_item: MonoItem<'tcx>,
674 MonoItem::Fn(instance) => {
675 let def_id = match instance.def {
676 ty::InstanceDef::Item(def_id) => def_id,
677 ty::InstanceDef::VtableShim(..) |
678 ty::InstanceDef::ReifyShim(..) |
679 ty::InstanceDef::FnPtrShim(..) |
680 ty::InstanceDef::ClosureOnceShim { .. } |
681 ty::InstanceDef::Intrinsic(..) |
682 ty::InstanceDef::DropGlue(..) |
683 ty::InstanceDef::Virtual(..) |
684 ty::InstanceDef::CloneShim(..) => return None
687 // If this is a method, we want to put it into the same module as
688 // its self-type. If the self-type does not provide a characteristic
689 // DefId, we use the location of the impl after all.
691 if tcx.trait_of_item(def_id).is_some() {
692 let self_ty = instance.substs.type_at(0);
693 // This is an implementation of a trait method.
694 return characteristic_def_id_of_type(self_ty).or(Some(def_id));
697 if let Some(impl_def_id) = tcx.impl_of_method(def_id) {
698 // This is a method within an inherent impl, find out what the
700 let impl_self_ty = tcx.subst_and_normalize_erasing_regions(
702 ty::ParamEnv::reveal_all(),
703 &tcx.type_of(impl_def_id),
705 if let Some(def_id) = characteristic_def_id_of_type(impl_self_ty) {
712 MonoItem::Static(def_id) => Some(def_id),
713 MonoItem::GlobalAsm(hir_id) => Some(tcx.hir().local_def_id(hir_id)),
717 type CguNameCache = FxHashMap<(DefId, bool), Symbol>;
719 fn compute_codegen_unit_name(
721 name_builder: &mut CodegenUnitNameBuilder<'_>,
724 cache: &mut CguNameCache,
726 // Find the innermost module that is not nested within a function.
727 let mut current_def_id = def_id;
728 let mut cgu_def_id = None;
729 // Walk backwards from the item we want to find the module for.
731 if current_def_id.index == CRATE_DEF_INDEX {
732 if cgu_def_id.is_none() {
733 // If we have not found a module yet, take the crate root.
734 cgu_def_id = Some(DefId {
736 index: CRATE_DEF_INDEX,
740 } else if tcx.def_kind(current_def_id) == Some(DefKind::Mod) {
741 if cgu_def_id.is_none() {
742 cgu_def_id = Some(current_def_id);
745 // If we encounter something that is not a module, throw away
746 // any module that we've found so far because we now know that
747 // it is nested within something else.
751 current_def_id = tcx.parent(current_def_id).unwrap();
754 let cgu_def_id = cgu_def_id.unwrap();
756 cache.entry((cgu_def_id, volatile)).or_insert_with(|| {
757 let def_path = tcx.def_path(cgu_def_id);
759 let components = def_path
762 .map(|part| part.data.as_symbol());
764 let volatile_suffix = if volatile {
770 name_builder.build_cgu_name(def_path.krate, components, volatile_suffix)
774 fn numbered_codegen_unit_name(
775 name_builder: &mut CodegenUnitNameBuilder<'_>,
778 name_builder.build_cgu_name_no_mangle(LOCAL_CRATE, &["cgu"], Some(index))
781 fn debug_dump<'a, 'tcx, I>(tcx: TyCtxt<'tcx>, label: &str, cgus: I)
783 I: Iterator<Item = &'a CodegenUnit<'tcx>>,
786 if cfg!(debug_assertions) {
789 debug!("CodegenUnit {} estimated size {} :", cgu.name(), cgu.size_estimate());
791 for (mono_item, linkage) in cgu.items() {
792 let symbol_name = mono_item.symbol_name(tcx).name.as_str();
793 let symbol_hash_start = symbol_name.rfind('h');
794 let symbol_hash = symbol_hash_start.map(|i| &symbol_name[i ..])
795 .unwrap_or("<no hash>");
797 debug!(" - {} [{:?}] [{}] estimated size {}",
798 mono_item.to_string(tcx, true),
801 mono_item.size_estimate(tcx));
809 #[inline(never)] // give this a place in the profiler
810 fn assert_symbols_are_distinct<'a, 'tcx, I>(tcx: TyCtxt<'tcx>, mono_items: I)
812 I: Iterator<Item = &'a MonoItem<'tcx>>,
815 let mut symbols: Vec<_> = mono_items.map(|mono_item| {
816 (mono_item, mono_item.symbol_name(tcx))
819 symbols.sort_by_key(|sym| sym.1);
821 for pair in symbols.windows(2) {
822 let sym1 = &pair[0].1;
823 let sym2 = &pair[1].1;
826 let mono_item1 = pair[0].0;
827 let mono_item2 = pair[1].0;
829 let span1 = mono_item1.local_span(tcx);
830 let span2 = mono_item2.local_span(tcx);
832 // Deterministically select one of the spans for error reporting
833 let span = match (span1, span2) {
834 (Some(span1), Some(span2)) => {
835 Some(if span1.lo().0 > span2.lo().0 {
841 (span1, span2) => span1.or(span2),
844 let error_message = format!("symbol `{}` is already defined", sym1);
846 if let Some(span) = span {
847 tcx.sess.span_fatal(span, &error_message)
849 tcx.sess.fatal(&error_message)
855 fn collect_and_partition_mono_items(
858 ) -> (Arc<DefIdSet>, Arc<Vec<Arc<CodegenUnit<'_>>>>) {
859 assert_eq!(cnum, LOCAL_CRATE);
861 let collection_mode = match tcx.sess.opts.debugging_opts.print_mono_items {
863 let mode_string = s.to_lowercase();
864 let mode_string = mode_string.trim();
865 if mode_string == "eager" {
866 MonoItemCollectionMode::Eager
868 if mode_string != "lazy" {
869 let message = format!("Unknown codegen-item collection mode '{}'. \
870 Falling back to 'lazy' mode.",
872 tcx.sess.warn(&message);
875 MonoItemCollectionMode::Lazy
879 if tcx.sess.opts.cg.link_dead_code {
880 MonoItemCollectionMode::Eager
882 MonoItemCollectionMode::Lazy
887 let (items, inlining_map) =
888 time(tcx.sess, "monomorphization collection", || {
889 collector::collect_crate_mono_items(tcx, collection_mode)
892 tcx.sess.abort_if_errors();
894 assert_symbols_are_distinct(tcx, items.iter());
896 let strategy = if tcx.sess.opts.incremental.is_some() {
897 PartitioningStrategy::PerModule
899 PartitioningStrategy::FixedUnitCount(tcx.sess.codegen_units())
902 let codegen_units = time(tcx.sess, "codegen unit partitioning", || {
905 items.iter().cloned(),
914 let mono_items: DefIdSet = items.iter().filter_map(|mono_item| {
916 MonoItem::Fn(ref instance) => Some(instance.def_id()),
917 MonoItem::Static(def_id) => Some(def_id),
922 if tcx.sess.opts.debugging_opts.print_mono_items.is_some() {
923 let mut item_to_cgus: FxHashMap<_, Vec<_>> = Default::default();
925 for cgu in &codegen_units {
926 for (&mono_item, &linkage) in cgu.items() {
927 item_to_cgus.entry(mono_item)
929 .push((cgu.name(), linkage));
933 let mut item_keys: Vec<_> = items
936 let mut output = i.to_string(tcx, false);
937 output.push_str(" @@");
938 let mut empty = Vec::new();
939 let cgus = item_to_cgus.get_mut(i).unwrap_or(&mut empty);
940 cgus.sort_by_key(|(name, _)| *name);
942 for &(ref cgu_name, (linkage, _)) in cgus.iter() {
943 output.push_str(" ");
944 output.push_str(&cgu_name.as_str());
946 let linkage_abbrev = match linkage {
947 Linkage::External => "External",
948 Linkage::AvailableExternally => "Available",
949 Linkage::LinkOnceAny => "OnceAny",
950 Linkage::LinkOnceODR => "OnceODR",
951 Linkage::WeakAny => "WeakAny",
952 Linkage::WeakODR => "WeakODR",
953 Linkage::Appending => "Appending",
954 Linkage::Internal => "Internal",
955 Linkage::Private => "Private",
956 Linkage::ExternalWeak => "ExternalWeak",
957 Linkage::Common => "Common",
960 output.push_str("[");
961 output.push_str(linkage_abbrev);
962 output.push_str("]");
970 for item in item_keys {
971 println!("MONO_ITEM {}", item);
975 (Arc::new(mono_items), Arc::new(codegen_units))
978 pub fn provide(providers: &mut Providers<'_>) {
979 providers.collect_and_partition_mono_items =
980 collect_and_partition_mono_items;
982 providers.is_codegened_item = |tcx, def_id| {
983 let (all_mono_items, _) =
984 tcx.collect_and_partition_mono_items(LOCAL_CRATE);
985 all_mono_items.contains(&def_id)
988 providers.codegen_unit = |tcx, name| {
989 let (_, all) = tcx.collect_and_partition_mono_items(LOCAL_CRATE);
991 .find(|cgu| cgu.name() == name)
993 .unwrap_or_else(|| panic!("failed to find cgu with name {:?}", name))