1 //! Mono Item Collection
2 //! ====================
4 //! This module is responsible for discovering all items that will contribute
5 //! to code generation of the crate. The important part here is that it not only
6 //! needs to find syntax-level items (functions, structs, etc) but also all
7 //! their monomorphized instantiations. Every non-generic, non-const function
8 //! maps to one LLVM artifact. Every generic function can produce
9 //! from zero to N artifacts, depending on the sets of type arguments it
10 //! is instantiated with.
11 //! This also applies to generic items from other crates: A generic definition
12 //! in crate X might produce monomorphizations that are compiled into crate Y.
13 //! We also have to collect these here.
15 //! The following kinds of "mono items" are handled here:
23 //! The following things also result in LLVM artifacts, but are not collected
24 //! here, since we instantiate them locally on demand when needed in a given
34 //! Let's define some terms first:
36 //! - A "mono item" is something that results in a function or global in
37 //! the LLVM IR of a codegen unit. Mono items do not stand on their
38 //! own, they can reference other mono items. For example, if function
39 //! `foo()` calls function `bar()` then the mono item for `foo()`
40 //! references the mono item for function `bar()`. In general, the
41 //! definition for mono item A referencing a mono item B is that
42 //! the LLVM artifact produced for A references the LLVM artifact produced
45 //! - Mono items and the references between them form a directed graph,
46 //! where the mono items are the nodes and references form the edges.
47 //! Let's call this graph the "mono item graph".
49 //! - The mono item graph for a program contains all mono items
50 //! that are needed in order to produce the complete LLVM IR of the program.
52 //! The purpose of the algorithm implemented in this module is to build the
53 //! mono item graph for the current crate. It runs in two phases:
55 //! 1. Discover the roots of the graph by traversing the HIR of the crate.
56 //! 2. Starting from the roots, find neighboring nodes by inspecting the MIR
57 //! representation of the item corresponding to a given node, until no more
58 //! new nodes are found.
60 //! ### Discovering roots
62 //! The roots of the mono item graph correspond to the public non-generic
63 //! syntactic items in the source code. We find them by walking the HIR of the
64 //! crate, and whenever we hit upon a public function, method, or static item,
65 //! we create a mono item consisting of the items DefId and, since we only
66 //! consider non-generic items, an empty type-substitution set. (In eager
67 //! collection mode, during incremental compilation, all non-generic functions
68 //! are considered as roots, as well as when the `-Clink-dead-code` option is
69 //! specified. Functions marked `#[no_mangle]` and functions called by inlinable
70 //! functions also always act as roots.)
72 //! ### Finding neighbor nodes
73 //! Given a mono item node, we can discover neighbors by inspecting its
74 //! MIR. We walk the MIR and any time we hit upon something that signifies a
75 //! reference to another mono item, we have found a neighbor. Since the
76 //! mono item we are currently at is always monomorphic, we also know the
77 //! concrete type arguments of its neighbors, and so all neighbors again will be
78 //! monomorphic. The specific forms a reference to a neighboring node can take
79 //! in MIR are quite diverse. Here is an overview:
81 //! #### Calling Functions/Methods
82 //! The most obvious form of one mono item referencing another is a
83 //! function or method call (represented by a CALL terminator in MIR). But
84 //! calls are not the only thing that might introduce a reference between two
85 //! function mono items, and as we will see below, they are just a
86 //! specialization of the form described next, and consequently will not get any
87 //! special treatment in the algorithm.
89 //! #### Taking a reference to a function or method
90 //! A function does not need to actually be called in order to be a neighbor of
91 //! another function. It suffices to just take a reference in order to introduce
92 //! an edge. Consider the following example:
95 //! fn print_val<T: Display>(x: T) {
96 //! println!("{}", x);
99 //! fn call_fn(f: &Fn(i32), x: i32) {
104 //! let print_i32 = print_val::<i32>;
105 //! call_fn(&print_i32, 0);
108 //! The MIR of none of these functions will contain an explicit call to
109 //! `print_val::<i32>`. Nonetheless, in order to mono this program, we need
110 //! an instance of this function. Thus, whenever we encounter a function or
111 //! method in operand position, we treat it as a neighbor of the current
112 //! mono item. Calls are just a special case of that.
115 //! In a way, closures are a simple case. Since every closure object needs to be
116 //! constructed somewhere, we can reliably discover them by observing
117 //! `RValue::Aggregate` expressions with `AggregateKind::Closure`. This is also
118 //! true for closures inlined from other crates.
121 //! Drop glue mono items are introduced by MIR drop-statements. The
122 //! generated mono item will again have drop-glue item neighbors if the
123 //! type to be dropped contains nested values that also need to be dropped. It
124 //! might also have a function item neighbor for the explicit `Drop::drop`
125 //! implementation of its type.
127 //! #### Unsizing Casts
128 //! A subtle way of introducing neighbor edges is by casting to a trait object.
129 //! Since the resulting fat-pointer contains a reference to a vtable, we need to
130 //! instantiate all object-save methods of the trait, as we need to store
131 //! pointers to these functions even if they never get called anywhere. This can
132 //! be seen as a special case of taking a function reference.
135 //! Since `Box` expression have special compiler support, no explicit calls to
136 //! `exchange_malloc()` and `box_free()` may show up in MIR, even if the
137 //! compiler will generate them. We have to observe `Rvalue::Box` expressions
138 //! and Box-typed drop-statements for that purpose.
141 //! Interaction with Cross-Crate Inlining
142 //! -------------------------------------
143 //! The binary of a crate will not only contain machine code for the items
144 //! defined in the source code of that crate. It will also contain monomorphic
145 //! instantiations of any extern generic functions and of functions marked with
147 //! The collection algorithm handles this more or less mono. If it is
148 //! about to create a mono item for something with an external `DefId`,
149 //! it will take a look if the MIR for that item is available, and if so just
150 //! proceed normally. If the MIR is not available, it assumes that the item is
151 //! just linked to and no node is created; which is exactly what we want, since
152 //! no machine code should be generated in the current crate for such an item.
154 //! Eager and Lazy Collection Mode
155 //! ------------------------------
156 //! Mono item collection can be performed in one of two modes:
158 //! - Lazy mode means that items will only be instantiated when actually
159 //! referenced. The goal is to produce the least amount of machine code
162 //! - Eager mode is meant to be used in conjunction with incremental compilation
163 //! where a stable set of mono items is more important than a minimal
164 //! one. Thus, eager mode will instantiate drop-glue for every drop-able type
165 //! in the crate, even if no drop call for that type exists (yet). It will
166 //! also instantiate default implementations of trait methods, something that
167 //! otherwise is only done on demand.
172 //! Some things are not yet fully implemented in the current version of this
176 //! Ideally, no mono item should be generated for const fns unless there
177 //! is a call to them that cannot be evaluated at compile time. At the moment
178 //! this is not implemented however: a mono item will be produced
179 //! regardless of whether it is actually needed or not.
181 use rustc_data_structures::fx::{FxHashMap, FxHashSet};
182 use rustc_data_structures::sync::{par_iter, MTLock, MTRef, ParallelIterator};
183 use rustc_hir as hir;
184 use rustc_hir::def::DefKind;
185 use rustc_hir::def_id::{DefId, DefIdMap, LocalDefId, LOCAL_CRATE};
186 use rustc_hir::lang_items::LangItem;
187 use rustc_index::bit_set::GrowableBitSet;
188 use rustc_middle::mir::interpret::{AllocId, ConstValue};
189 use rustc_middle::mir::interpret::{ErrorHandled, GlobalAlloc, Scalar};
190 use rustc_middle::mir::mono::{InstantiationMode, MonoItem};
191 use rustc_middle::mir::visit::Visitor as MirVisitor;
192 use rustc_middle::mir::{self, Local, Location};
193 use rustc_middle::ty::adjustment::{CustomCoerceUnsized, PointerCast};
194 use rustc_middle::ty::print::with_no_trimmed_paths;
195 use rustc_middle::ty::subst::{GenericArgKind, InternalSubsts};
196 use rustc_middle::ty::{self, GenericParamDefKind, Instance, Ty, TyCtxt, TypeFoldable, VtblEntry};
197 use rustc_middle::{middle::codegen_fn_attrs::CodegenFnAttrFlags, mir::visit::TyContext};
198 use rustc_session::config::EntryFnType;
199 use rustc_session::lint::builtin::LARGE_ASSIGNMENTS;
200 use rustc_session::Limit;
201 use rustc_span::source_map::{dummy_spanned, respan, Span, Spanned, DUMMY_SP};
202 use rustc_target::abi::Size;
203 use smallvec::SmallVec;
206 use std::path::PathBuf;
209 pub enum MonoItemCollectionMode {
214 /// Maps every mono item to all mono items it references in its
216 pub struct InliningMap<'tcx> {
217 // Maps a source mono item to the range of mono items
219 // The range selects elements within the `targets` vecs.
220 index: FxHashMap<MonoItem<'tcx>, Range<usize>>,
221 targets: Vec<MonoItem<'tcx>>,
223 // Contains one bit per mono item in the `targets` field. That bit
224 // is true if that mono item needs to be inlined into every CGU.
225 inlines: GrowableBitSet<usize>,
228 impl<'tcx> InliningMap<'tcx> {
229 fn new() -> InliningMap<'tcx> {
231 index: FxHashMap::default(),
233 inlines: GrowableBitSet::with_capacity(1024),
237 fn record_accesses(&mut self, source: MonoItem<'tcx>, new_targets: &[(MonoItem<'tcx>, bool)]) {
238 let start_index = self.targets.len();
239 let new_items_count = new_targets.len();
240 let new_items_count_total = new_items_count + self.targets.len();
242 self.targets.reserve(new_items_count);
243 self.inlines.ensure(new_items_count_total);
245 for (i, (target, inline)) in new_targets.iter().enumerate() {
246 self.targets.push(*target);
248 self.inlines.insert(i + start_index);
252 let end_index = self.targets.len();
253 assert!(self.index.insert(source, start_index..end_index).is_none());
256 // Internally iterate over all items referenced by `source` which will be
257 // made available for inlining.
258 pub fn with_inlining_candidates<F>(&self, source: MonoItem<'tcx>, mut f: F)
260 F: FnMut(MonoItem<'tcx>),
262 if let Some(range) = self.index.get(&source) {
263 for (i, candidate) in self.targets[range.clone()].iter().enumerate() {
264 if self.inlines.contains(range.start + i) {
271 // Internally iterate over all items and the things each accesses.
272 pub fn iter_accesses<F>(&self, mut f: F)
274 F: FnMut(MonoItem<'tcx>, &[MonoItem<'tcx>]),
276 for (&accessor, range) in &self.index {
277 f(accessor, &self.targets[range.clone()])
282 pub fn collect_crate_mono_items(
284 mode: MonoItemCollectionMode,
285 ) -> (FxHashSet<MonoItem<'_>>, InliningMap<'_>) {
286 let _prof_timer = tcx.prof.generic_activity("monomorphization_collector");
289 tcx.sess.time("monomorphization_collector_root_collections", || collect_roots(tcx, mode));
291 debug!("building mono item graph, beginning at roots");
293 let mut visited = MTLock::new(FxHashSet::default());
294 let mut inlining_map = MTLock::new(InliningMap::new());
295 let recursion_limit = tcx.recursion_limit();
298 let visited: MTRef<'_, _> = &mut visited;
299 let inlining_map: MTRef<'_, _> = &mut inlining_map;
301 tcx.sess.time("monomorphization_collector_graph_walk", || {
302 par_iter(roots).for_each(|root| {
303 let mut recursion_depths = DefIdMap::default();
308 &mut recursion_depths,
316 (visited.into_inner(), inlining_map.into_inner())
319 // Find all non-generic items by walking the HIR. These items serve as roots to
320 // start monomorphizing from.
321 fn collect_roots(tcx: TyCtxt<'_>, mode: MonoItemCollectionMode) -> Vec<MonoItem<'_>> {
322 debug!("collecting roots");
323 let mut roots = Vec::new();
326 let entry_fn = tcx.entry_fn(());
328 debug!("collect_roots: entry_fn = {:?}", entry_fn);
330 let mut collector = RootCollector { tcx, mode, entry_fn, output: &mut roots };
332 let crate_items = tcx.hir_crate_items(());
334 for id in crate_items.items() {
335 collector.process_item(id);
338 for id in crate_items.impl_items() {
339 collector.process_impl_item(id);
342 collector.push_extra_entry_roots();
345 // We can only codegen items that are instantiable - items all of
346 // whose predicates hold. Luckily, items that aren't instantiable
347 // can't actually be used, so we can just skip codegenning them.
350 .filter_map(|root| root.node.is_instantiable(tcx).then_some(root.node))
354 /// Collect all monomorphized items reachable from `starting_point`, and emit a note diagnostic if a
355 /// post-monorphization error is encountered during a collection step.
356 fn collect_items_rec<'tcx>(
358 starting_point: Spanned<MonoItem<'tcx>>,
359 visited: MTRef<'_, MTLock<FxHashSet<MonoItem<'tcx>>>>,
360 recursion_depths: &mut DefIdMap<usize>,
361 recursion_limit: Limit,
362 inlining_map: MTRef<'_, MTLock<InliningMap<'tcx>>>,
364 if !visited.lock_mut().insert(starting_point.node) {
365 // We've been here already, no need to search again.
368 debug!("BEGIN collect_items_rec({})", starting_point.node);
370 let mut neighbors = Vec::new();
371 let recursion_depth_reset;
374 // Post-monomorphization errors MVP
376 // We can encounter errors while monomorphizing an item, but we don't have a good way of
377 // showing a complete stack of spans ultimately leading to collecting the erroneous one yet.
378 // (It's also currently unclear exactly which diagnostics and information would be interesting
379 // to report in such cases)
381 // This leads to suboptimal error reporting: a post-monomorphization error (PME) will be
382 // shown with just a spanned piece of code causing the error, without information on where
383 // it was called from. This is especially obscure if the erroneous mono item is in a
384 // dependency. See for example issue #85155, where, before minimization, a PME happened two
385 // crates downstream from libcore's stdarch, without a way to know which dependency was the
388 // If such an error occurs in the current crate, its span will be enough to locate the
389 // source. If the cause is in another crate, the goal here is to quickly locate which mono
390 // item in the current crate is ultimately responsible for causing the error.
392 // To give at least _some_ context to the user: while collecting mono items, we check the
393 // error count. If it has changed, a PME occurred, and we trigger some diagnostics about the
394 // current step of mono items collection.
396 let error_count = tcx.sess.diagnostic().err_count();
398 match starting_point.node {
399 MonoItem::Static(def_id) => {
400 let instance = Instance::mono(tcx, def_id);
402 // Sanity check whether this ended up being collected accidentally
403 debug_assert!(should_codegen_locally(tcx, &instance));
405 let ty = instance.ty(tcx, ty::ParamEnv::reveal_all());
406 visit_drop_use(tcx, ty, true, starting_point.span, &mut neighbors);
408 recursion_depth_reset = None;
410 if let Ok(alloc) = tcx.eval_static_initializer(def_id) {
411 for &id in alloc.inner().relocations().values() {
412 collect_miri(tcx, id, &mut neighbors);
416 MonoItem::Fn(instance) => {
417 // Sanity check whether this ended up being collected accidentally
418 debug_assert!(should_codegen_locally(tcx, &instance));
420 // Keep track of the monomorphization recursion depth
421 recursion_depth_reset = Some(check_recursion_limit(
428 check_type_length_limit(tcx, instance);
430 rustc_data_structures::stack::ensure_sufficient_stack(|| {
431 collect_neighbours(tcx, instance, &mut neighbors);
434 MonoItem::GlobalAsm(item_id) => {
435 recursion_depth_reset = None;
437 let item = tcx.hir().item(item_id);
438 if let hir::ItemKind::GlobalAsm(asm) = item.kind {
439 for (op, op_sp) in asm.operands {
441 hir::InlineAsmOperand::Const { .. } => {
442 // Only constants which resolve to a plain integer
443 // are supported. Therefore the value should not
444 // depend on any other items.
446 hir::InlineAsmOperand::SymFn { anon_const } => {
447 let def_id = tcx.hir().body_owner_def_id(anon_const.body).to_def_id();
448 if let Ok(val) = tcx.const_eval_poly(def_id) {
449 rustc_data_structures::stack::ensure_sufficient_stack(|| {
450 collect_const_value(tcx, val, &mut neighbors);
454 hir::InlineAsmOperand::SymStatic { path: _, def_id } => {
455 let instance = Instance::mono(tcx, *def_id);
456 if should_codegen_locally(tcx, &instance) {
457 trace!("collecting static {:?}", def_id);
458 neighbors.push(dummy_spanned(MonoItem::Static(*def_id)));
461 hir::InlineAsmOperand::In { .. }
462 | hir::InlineAsmOperand::Out { .. }
463 | hir::InlineAsmOperand::InOut { .. }
464 | hir::InlineAsmOperand::SplitInOut { .. } => {
465 span_bug!(*op_sp, "invalid operand type for global_asm!")
470 span_bug!(item.span, "Mismatch between hir::Item type and MonoItem type")
475 // Check for PMEs and emit a diagnostic if one happened. To try to show relevant edges of the
476 // mono item graph where the PME diagnostics are currently the most problematic (e.g. ones
477 // involving a dependency, and the lack of context is confusing) in this MVP, we focus on
478 // diagnostics on edges crossing a crate boundary: the collected mono items which are not
479 // defined in the local crate.
480 if tcx.sess.diagnostic().err_count() > error_count
481 && starting_point.node.krate() != LOCAL_CRATE
482 && starting_point.node.is_user_defined()
484 let formatted_item = with_no_trimmed_paths!(starting_point.node.to_string());
485 tcx.sess.span_note_without_error(
487 &format!("the above error was encountered while instantiating `{}`", formatted_item),
491 record_accesses(tcx, starting_point.node, neighbors.iter().map(|i| &i.node), inlining_map);
493 for neighbour in neighbors {
494 collect_items_rec(tcx, neighbour, visited, recursion_depths, recursion_limit, inlining_map);
497 if let Some((def_id, depth)) = recursion_depth_reset {
498 recursion_depths.insert(def_id, depth);
501 debug!("END collect_items_rec({})", starting_point.node);
504 fn record_accesses<'a, 'tcx: 'a>(
506 caller: MonoItem<'tcx>,
507 callees: impl Iterator<Item = &'a MonoItem<'tcx>>,
508 inlining_map: MTRef<'_, MTLock<InliningMap<'tcx>>>,
510 let is_inlining_candidate = |mono_item: &MonoItem<'tcx>| {
511 mono_item.instantiation_mode(tcx) == InstantiationMode::LocalCopy
514 // We collect this into a `SmallVec` to avoid calling `is_inlining_candidate` in the lock.
515 // FIXME: Call `is_inlining_candidate` when pushing to `neighbors` in `collect_items_rec`
516 // instead to avoid creating this `SmallVec`.
517 let accesses: SmallVec<[_; 128]> =
518 callees.map(|mono_item| (*mono_item, is_inlining_candidate(mono_item))).collect();
520 inlining_map.lock_mut().record_accesses(caller, &accesses);
523 /// Format instance name that is already known to be too long for rustc.
524 /// Show only the first and last 32 characters to avoid blasting
525 /// the user's terminal with thousands of lines of type-name.
527 /// If the type name is longer than before+after, it will be written to a file.
528 fn shrunk_instance_name<'tcx>(
530 instance: &Instance<'tcx>,
533 ) -> (String, Option<PathBuf>) {
534 let s = instance.to_string();
536 // Only use the shrunk version if it's really shorter.
537 // This also avoids the case where before and after slices overlap.
538 if s.chars().nth(before + after + 1).is_some() {
539 // An iterator of all byte positions including the end of the string.
540 let positions = || s.char_indices().map(|(i, _)| i).chain(iter::once(s.len()));
542 let shrunk = format!(
543 "{before}...{after}",
544 before = &s[..positions().nth(before).unwrap_or(s.len())],
545 after = &s[positions().rev().nth(after).unwrap_or(0)..],
548 let path = tcx.output_filenames(()).temp_path_ext("long-type.txt", None);
549 let written_to_path = std::fs::write(&path, s).ok().map(|_| path);
551 (shrunk, written_to_path)
557 fn check_recursion_limit<'tcx>(
559 instance: Instance<'tcx>,
561 recursion_depths: &mut DefIdMap<usize>,
562 recursion_limit: Limit,
563 ) -> (DefId, usize) {
564 let def_id = instance.def_id();
565 let recursion_depth = recursion_depths.get(&def_id).cloned().unwrap_or(0);
566 debug!(" => recursion depth={}", recursion_depth);
568 let adjusted_recursion_depth = if Some(def_id) == tcx.lang_items().drop_in_place_fn() {
569 // HACK: drop_in_place creates tight monomorphization loops. Give
576 // Code that needs to instantiate the same function recursively
577 // more than the recursion limit is assumed to be causing an
578 // infinite expansion.
579 if !recursion_limit.value_within_limit(adjusted_recursion_depth) {
580 let (shrunk, written_to_path) = shrunk_instance_name(tcx, &instance, 32, 32);
581 let error = format!("reached the recursion limit while instantiating `{}`", shrunk);
582 let mut err = tcx.sess.struct_span_fatal(span, &error);
584 tcx.def_span(def_id),
585 &format!("`{}` defined here", tcx.def_path_str(def_id)),
587 if let Some(path) = written_to_path {
588 err.note(&format!("the full type name has been written to '{}'", path.display()));
593 recursion_depths.insert(def_id, recursion_depth + 1);
595 (def_id, recursion_depth)
598 fn check_type_length_limit<'tcx>(tcx: TyCtxt<'tcx>, instance: Instance<'tcx>) {
599 let type_length = instance
602 .flat_map(|arg| arg.walk())
603 .filter(|arg| match arg.unpack() {
604 GenericArgKind::Type(_) | GenericArgKind::Const(_) => true,
605 GenericArgKind::Lifetime(_) => false,
608 debug!(" => type length={}", type_length);
610 // Rust code can easily create exponentially-long types using only a
611 // polynomial recursion depth. Even with the default recursion
612 // depth, you can easily get cases that take >2^60 steps to run,
613 // which means that rustc basically hangs.
615 // Bail out in these cases to avoid that bad user experience.
616 if !tcx.type_length_limit().value_within_limit(type_length) {
617 let (shrunk, written_to_path) = shrunk_instance_name(tcx, &instance, 32, 32);
618 let msg = format!("reached the type-length limit while instantiating `{}`", shrunk);
619 let mut diag = tcx.sess.struct_span_fatal(tcx.def_span(instance.def_id()), &msg);
620 if let Some(path) = written_to_path {
621 diag.note(&format!("the full type name has been written to '{}'", path.display()));
624 "consider adding a `#![type_length_limit=\"{}\"]` attribute to your crate",
631 struct MirNeighborCollector<'a, 'tcx> {
633 body: &'a mir::Body<'tcx>,
634 output: &'a mut Vec<Spanned<MonoItem<'tcx>>>,
635 instance: Instance<'tcx>,
638 impl<'a, 'tcx> MirNeighborCollector<'a, 'tcx> {
639 pub fn monomorphize<T>(&self, value: T) -> T
641 T: TypeFoldable<'tcx>,
643 debug!("monomorphize: self.instance={:?}", self.instance);
644 self.instance.subst_mir_and_normalize_erasing_regions(
646 ty::ParamEnv::reveal_all(),
652 impl<'a, 'tcx> MirVisitor<'tcx> for MirNeighborCollector<'a, 'tcx> {
653 fn visit_rvalue(&mut self, rvalue: &mir::Rvalue<'tcx>, location: Location) {
654 debug!("visiting rvalue {:?}", *rvalue);
656 let span = self.body.source_info(location).span;
659 // When doing an cast from a regular pointer to a fat pointer, we
660 // have to instantiate all methods of the trait being cast to, so we
661 // can build the appropriate vtable.
663 mir::CastKind::Pointer(PointerCast::Unsize),
667 let target_ty = self.monomorphize(target_ty);
668 let source_ty = operand.ty(self.body, self.tcx);
669 let source_ty = self.monomorphize(source_ty);
670 let (source_ty, target_ty) =
671 find_vtable_types_for_unsizing(self.tcx, source_ty, target_ty);
672 // This could also be a different Unsize instruction, like
673 // from a fixed sized array to a slice. But we are only
674 // interested in things that produce a vtable.
675 if target_ty.is_trait() && !source_ty.is_trait() {
676 create_mono_items_for_vtable_methods(
686 mir::CastKind::Pointer(PointerCast::ReifyFnPointer),
690 let fn_ty = operand.ty(self.body, self.tcx);
691 let fn_ty = self.monomorphize(fn_ty);
692 visit_fn_use(self.tcx, fn_ty, false, span, &mut self.output);
695 mir::CastKind::Pointer(PointerCast::ClosureFnPointer(_)),
699 let source_ty = operand.ty(self.body, self.tcx);
700 let source_ty = self.monomorphize(source_ty);
701 match *source_ty.kind() {
702 ty::Closure(def_id, substs) => {
703 let instance = Instance::resolve_closure(
707 ty::ClosureKind::FnOnce,
709 if should_codegen_locally(self.tcx, &instance) {
710 self.output.push(create_fn_mono_item(self.tcx, instance, span));
716 mir::Rvalue::ThreadLocalRef(def_id) => {
717 assert!(self.tcx.is_thread_local_static(def_id));
718 let instance = Instance::mono(self.tcx, def_id);
719 if should_codegen_locally(self.tcx, &instance) {
720 trace!("collecting thread-local static {:?}", def_id);
721 self.output.push(respan(span, MonoItem::Static(def_id)));
724 _ => { /* not interesting */ }
727 self.super_rvalue(rvalue, location);
730 /// This does not walk the constant, as it has been handled entirely here and trying
731 /// to walk it would attempt to evaluate the `ty::Const` inside, which doesn't necessarily
732 /// work, as some constants cannot be represented in the type system.
733 fn visit_constant(&mut self, constant: &mir::Constant<'tcx>, location: Location) {
734 let literal = self.monomorphize(constant.literal);
735 let val = match literal {
736 mir::ConstantKind::Val(val, _) => val,
737 mir::ConstantKind::Ty(ct) => match ct.val() {
738 ty::ConstKind::Value(val) => val,
739 ty::ConstKind::Unevaluated(ct) => {
740 let param_env = ty::ParamEnv::reveal_all();
741 match self.tcx.const_eval_resolve(param_env, ct, None) {
742 // The `monomorphize` call should have evaluated that constant already.
744 Err(ErrorHandled::Reported(_) | ErrorHandled::Linted) => return,
745 Err(ErrorHandled::TooGeneric) => span_bug!(
746 self.body.source_info(location).span,
747 "collection encountered polymorphic constant: {:?}",
755 collect_const_value(self.tcx, val, self.output);
756 self.visit_ty(literal.ty(), TyContext::Location(location));
759 fn visit_const(&mut self, constant: ty::Const<'tcx>, location: Location) {
760 debug!("visiting const {:?} @ {:?}", constant, location);
762 let substituted_constant = self.monomorphize(constant);
763 let param_env = ty::ParamEnv::reveal_all();
765 match substituted_constant.val() {
766 ty::ConstKind::Value(val) => collect_const_value(self.tcx, val, self.output),
767 ty::ConstKind::Unevaluated(unevaluated) => {
768 match self.tcx.const_eval_resolve(param_env, unevaluated, None) {
769 // The `monomorphize` call should have evaluated that constant already.
770 Ok(val) => span_bug!(
771 self.body.source_info(location).span,
772 "collection encountered the unevaluated constant {} which evaluated to {:?}",
773 substituted_constant,
776 Err(ErrorHandled::Reported(_) | ErrorHandled::Linted) => {}
777 Err(ErrorHandled::TooGeneric) => span_bug!(
778 self.body.source_info(location).span,
779 "collection encountered polymorphic constant: {}",
787 self.super_const(constant);
790 fn visit_terminator(&mut self, terminator: &mir::Terminator<'tcx>, location: Location) {
791 debug!("visiting terminator {:?} @ {:?}", terminator, location);
792 let source = self.body.source_info(location).span;
795 match terminator.kind {
796 mir::TerminatorKind::Call { ref func, .. } => {
797 let callee_ty = func.ty(self.body, tcx);
798 let callee_ty = self.monomorphize(callee_ty);
799 visit_fn_use(self.tcx, callee_ty, true, source, &mut self.output);
801 mir::TerminatorKind::Drop { ref place, .. }
802 | mir::TerminatorKind::DropAndReplace { ref place, .. } => {
803 let ty = place.ty(self.body, self.tcx).ty;
804 let ty = self.monomorphize(ty);
805 visit_drop_use(self.tcx, ty, true, source, self.output);
807 mir::TerminatorKind::InlineAsm { ref operands, .. } => {
810 mir::InlineAsmOperand::SymFn { ref value } => {
811 let fn_ty = self.monomorphize(value.literal.ty());
812 visit_fn_use(self.tcx, fn_ty, false, source, &mut self.output);
814 mir::InlineAsmOperand::SymStatic { def_id } => {
815 let instance = Instance::mono(self.tcx, def_id);
816 if should_codegen_locally(self.tcx, &instance) {
817 trace!("collecting asm sym static {:?}", def_id);
818 self.output.push(respan(source, MonoItem::Static(def_id)));
825 mir::TerminatorKind::Assert { ref msg, .. } => {
826 let lang_item = match msg {
827 mir::AssertKind::BoundsCheck { .. } => LangItem::PanicBoundsCheck,
828 _ => LangItem::Panic,
830 let instance = Instance::mono(tcx, tcx.require_lang_item(lang_item, Some(source)));
831 if should_codegen_locally(tcx, &instance) {
832 self.output.push(create_fn_mono_item(tcx, instance, source));
835 mir::TerminatorKind::Abort { .. } => {
836 let instance = Instance::mono(
838 tcx.require_lang_item(LangItem::PanicNoUnwind, Some(source)),
840 if should_codegen_locally(tcx, &instance) {
841 self.output.push(create_fn_mono_item(tcx, instance, source));
844 mir::TerminatorKind::Goto { .. }
845 | mir::TerminatorKind::SwitchInt { .. }
846 | mir::TerminatorKind::Resume
847 | mir::TerminatorKind::Return
848 | mir::TerminatorKind::Unreachable => {}
849 mir::TerminatorKind::GeneratorDrop
850 | mir::TerminatorKind::Yield { .. }
851 | mir::TerminatorKind::FalseEdge { .. }
852 | mir::TerminatorKind::FalseUnwind { .. } => bug!(),
855 self.super_terminator(terminator, location);
858 fn visit_operand(&mut self, operand: &mir::Operand<'tcx>, location: Location) {
859 self.super_operand(operand, location);
860 let limit = self.tcx.move_size_limit().0;
864 let limit = Size::from_bytes(limit);
865 let ty = operand.ty(self.body, self.tcx);
866 let ty = self.monomorphize(ty);
867 let layout = self.tcx.layout_of(ty::ParamEnv::reveal_all().and(ty));
868 if let Ok(layout) = layout {
869 if layout.size > limit {
871 let source_info = self.body.source_info(location);
872 debug!(?source_info);
873 let lint_root = source_info.scope.lint_root(&self.body.source_scopes);
875 let Some(lint_root) = lint_root else {
876 // This happens when the issue is in a function from a foreign crate that
877 // we monomorphized in the current crate. We can't get a `HirId` for things
879 // FIXME: Find out where to report the lint on. Maybe simply crate-level lint root
880 // but correct span? This would make the lint at least accept crate-level lint attributes.
883 self.tcx.struct_span_lint_hir(
888 let mut err = lint.build(&format!("moving {} bytes", layout.size.bytes()));
889 err.span_label(source_info.span, "value moved from here");
890 err.note(&format!(r#"The current maximum size is {}, but it can be customized with the move_size_limit attribute: `#![move_size_limit = "..."]`"#, limit.bytes()));
900 _place_local: &Local,
901 _context: mir::visit::PlaceContext,
907 fn visit_drop_use<'tcx>(
910 is_direct_call: bool,
912 output: &mut Vec<Spanned<MonoItem<'tcx>>>,
914 let instance = Instance::resolve_drop_in_place(tcx, ty);
915 visit_instance_use(tcx, instance, is_direct_call, source, output);
918 fn visit_fn_use<'tcx>(
921 is_direct_call: bool,
923 output: &mut Vec<Spanned<MonoItem<'tcx>>>,
925 if let ty::FnDef(def_id, substs) = *ty.kind() {
926 let instance = if is_direct_call {
927 ty::Instance::resolve(tcx, ty::ParamEnv::reveal_all(), def_id, substs).unwrap().unwrap()
929 ty::Instance::resolve_for_fn_ptr(tcx, ty::ParamEnv::reveal_all(), def_id, substs)
932 visit_instance_use(tcx, instance, is_direct_call, source, output);
936 fn visit_instance_use<'tcx>(
938 instance: ty::Instance<'tcx>,
939 is_direct_call: bool,
941 output: &mut Vec<Spanned<MonoItem<'tcx>>>,
943 debug!("visit_item_use({:?}, is_direct_call={:?})", instance, is_direct_call);
944 if !should_codegen_locally(tcx, &instance) {
949 ty::InstanceDef::Virtual(..) | ty::InstanceDef::Intrinsic(_) => {
951 bug!("{:?} being reified", instance);
954 ty::InstanceDef::DropGlue(_, None) => {
955 // Don't need to emit noop drop glue if we are calling directly.
957 output.push(create_fn_mono_item(tcx, instance, source));
960 ty::InstanceDef::DropGlue(_, Some(_))
961 | ty::InstanceDef::VtableShim(..)
962 | ty::InstanceDef::ReifyShim(..)
963 | ty::InstanceDef::ClosureOnceShim { .. }
964 | ty::InstanceDef::Item(..)
965 | ty::InstanceDef::FnPtrShim(..)
966 | ty::InstanceDef::CloneShim(..) => {
967 output.push(create_fn_mono_item(tcx, instance, source));
972 /// Returns `true` if we should codegen an instance in the local crate, or returns `false` if we
973 /// can just link to the upstream crate and therefore don't need a mono item.
974 fn should_codegen_locally<'tcx>(tcx: TyCtxt<'tcx>, instance: &Instance<'tcx>) -> bool {
975 let Some(def_id) = instance.def.def_id_if_not_guaranteed_local_codegen() else {
979 if tcx.is_foreign_item(def_id) {
980 // Foreign items are always linked against, there's no way of instantiating them.
984 if def_id.is_local() {
985 // Local items cannot be referred to locally without monomorphizing them locally.
989 if tcx.is_reachable_non_generic(def_id)
990 || instance.polymorphize(tcx).upstream_monomorphization(tcx).is_some()
992 // We can link to the item in question, no instance needed in this crate.
996 if !tcx.is_mir_available(def_id) {
997 bug!("no MIR available for {:?}", def_id);
1003 /// For a given pair of source and target type that occur in an unsizing coercion,
1004 /// this function finds the pair of types that determines the vtable linking
1007 /// For example, the source type might be `&SomeStruct` and the target type
1008 /// might be `&SomeTrait` in a cast like:
1010 /// let src: &SomeStruct = ...;
1011 /// let target = src as &SomeTrait;
1013 /// Then the output of this function would be (SomeStruct, SomeTrait) since for
1014 /// constructing the `target` fat-pointer we need the vtable for that pair.
1016 /// Things can get more complicated though because there's also the case where
1017 /// the unsized type occurs as a field:
1020 /// struct ComplexStruct<T: ?Sized> {
1027 /// In this case, if `T` is sized, `&ComplexStruct<T>` is a thin pointer. If `T`
1028 /// is unsized, `&SomeStruct` is a fat pointer, and the vtable it points to is
1029 /// for the pair of `T` (which is a trait) and the concrete type that `T` was
1030 /// originally coerced from:
1032 /// let src: &ComplexStruct<SomeStruct> = ...;
1033 /// let target = src as &ComplexStruct<SomeTrait>;
1035 /// Again, we want this `find_vtable_types_for_unsizing()` to provide the pair
1036 /// `(SomeStruct, SomeTrait)`.
1038 /// Finally, there is also the case of custom unsizing coercions, e.g., for
1039 /// smart pointers such as `Rc` and `Arc`.
1040 fn find_vtable_types_for_unsizing<'tcx>(
1042 source_ty: Ty<'tcx>,
1043 target_ty: Ty<'tcx>,
1044 ) -> (Ty<'tcx>, Ty<'tcx>) {
1045 let ptr_vtable = |inner_source: Ty<'tcx>, inner_target: Ty<'tcx>| {
1046 let param_env = ty::ParamEnv::reveal_all();
1047 let type_has_metadata = |ty: Ty<'tcx>| -> bool {
1048 if ty.is_sized(tcx.at(DUMMY_SP), param_env) {
1051 let tail = tcx.struct_tail_erasing_lifetimes(ty, param_env);
1053 ty::Foreign(..) => false,
1054 ty::Str | ty::Slice(..) | ty::Dynamic(..) => true,
1055 _ => bug!("unexpected unsized tail: {:?}", tail),
1058 if type_has_metadata(inner_source) {
1059 (inner_source, inner_target)
1061 tcx.struct_lockstep_tails_erasing_lifetimes(inner_source, inner_target, param_env)
1065 match (&source_ty.kind(), &target_ty.kind()) {
1066 (&ty::Ref(_, a, _), &ty::Ref(_, b, _) | &ty::RawPtr(ty::TypeAndMut { ty: b, .. }))
1067 | (&ty::RawPtr(ty::TypeAndMut { ty: a, .. }), &ty::RawPtr(ty::TypeAndMut { ty: b, .. })) => {
1070 (&ty::Adt(def_a, _), &ty::Adt(def_b, _)) if def_a.is_box() && def_b.is_box() => {
1071 ptr_vtable(source_ty.boxed_ty(), target_ty.boxed_ty())
1074 (&ty::Adt(source_adt_def, source_substs), &ty::Adt(target_adt_def, target_substs)) => {
1075 assert_eq!(source_adt_def, target_adt_def);
1077 let CustomCoerceUnsized::Struct(coerce_index) =
1078 crate::custom_coerce_unsize_info(tcx, source_ty, target_ty);
1080 let source_fields = &source_adt_def.non_enum_variant().fields;
1081 let target_fields = &target_adt_def.non_enum_variant().fields;
1084 coerce_index < source_fields.len() && source_fields.len() == target_fields.len()
1087 find_vtable_types_for_unsizing(
1089 source_fields[coerce_index].ty(tcx, source_substs),
1090 target_fields[coerce_index].ty(tcx, target_substs),
1094 "find_vtable_types_for_unsizing: invalid coercion {:?} -> {:?}",
1101 fn create_fn_mono_item<'tcx>(
1103 instance: Instance<'tcx>,
1105 ) -> Spanned<MonoItem<'tcx>> {
1106 debug!("create_fn_mono_item(instance={})", instance);
1108 let def_id = instance.def_id();
1109 if tcx.sess.opts.debugging_opts.profile_closures && def_id.is_local() && tcx.is_closure(def_id)
1111 crate::util::dump_closure_profile(tcx, instance);
1114 respan(source, MonoItem::Fn(instance.polymorphize(tcx)))
1117 /// Creates a `MonoItem` for each method that is referenced by the vtable for
1118 /// the given trait/impl pair.
1119 fn create_mono_items_for_vtable_methods<'tcx>(
1124 output: &mut Vec<Spanned<MonoItem<'tcx>>>,
1126 assert!(!trait_ty.has_escaping_bound_vars() && !impl_ty.has_escaping_bound_vars());
1128 if let ty::Dynamic(ref trait_ty, ..) = trait_ty.kind() {
1129 if let Some(principal) = trait_ty.principal() {
1130 let poly_trait_ref = principal.with_self_ty(tcx, impl_ty);
1131 assert!(!poly_trait_ref.has_escaping_bound_vars());
1133 // Walk all methods of the trait, including those of its supertraits
1134 let entries = tcx.vtable_entries(poly_trait_ref);
1135 let methods = entries
1137 .filter_map(|entry| match entry {
1138 VtblEntry::MetadataDropInPlace
1139 | VtblEntry::MetadataSize
1140 | VtblEntry::MetadataAlign
1141 | VtblEntry::Vacant => None,
1142 VtblEntry::TraitVPtr(_) => {
1143 // all super trait items already covered, so skip them.
1146 VtblEntry::Method(instance) => {
1147 Some(*instance).filter(|instance| should_codegen_locally(tcx, instance))
1150 .map(|item| create_fn_mono_item(tcx, item, source));
1151 output.extend(methods);
1154 // Also add the destructor.
1155 visit_drop_use(tcx, impl_ty, false, source, output);
1159 //=-----------------------------------------------------------------------------
1161 //=-----------------------------------------------------------------------------
1163 struct RootCollector<'a, 'tcx> {
1165 mode: MonoItemCollectionMode,
1166 output: &'a mut Vec<Spanned<MonoItem<'tcx>>>,
1167 entry_fn: Option<(DefId, EntryFnType)>,
1170 impl<'v> RootCollector<'_, 'v> {
1171 fn process_item(&mut self, id: hir::ItemId) {
1172 match self.tcx.hir().def_kind(id.def_id) {
1173 DefKind::Enum | DefKind::Struct | DefKind::Union => {
1174 let item = self.tcx.hir().item(id);
1176 hir::ItemKind::Enum(_, ref generics)
1177 | hir::ItemKind::Struct(_, ref generics)
1178 | hir::ItemKind::Union(_, ref generics) => {
1179 if generics.params.is_empty() {
1180 if self.mode == MonoItemCollectionMode::Eager {
1182 "RootCollector: ADT drop-glue for {}",
1183 self.tcx.def_path_str(item.def_id.to_def_id())
1187 Instance::new(item.def_id.to_def_id(), InternalSubsts::empty())
1188 .ty(self.tcx, ty::ParamEnv::reveal_all());
1189 visit_drop_use(self.tcx, ty, true, DUMMY_SP, self.output);
1196 DefKind::GlobalAsm => {
1198 "RootCollector: ItemKind::GlobalAsm({})",
1199 self.tcx.def_path_str(id.def_id.to_def_id())
1201 self.output.push(dummy_spanned(MonoItem::GlobalAsm(id)));
1203 DefKind::Static(..) => {
1205 "RootCollector: ItemKind::Static({})",
1206 self.tcx.def_path_str(id.def_id.to_def_id())
1208 self.output.push(dummy_spanned(MonoItem::Static(id.def_id.to_def_id())));
1211 // const items only generate mono items if they are
1212 // actually used somewhere. Just declaring them is insufficient.
1214 // but even just declaring them must collect the items they refer to
1215 if let Ok(val) = self.tcx.const_eval_poly(id.def_id.to_def_id()) {
1216 collect_const_value(self.tcx, val, &mut self.output);
1220 if self.mode == MonoItemCollectionMode::Eager {
1221 let item = self.tcx.hir().item(id);
1222 create_mono_items_for_default_impls(self.tcx, item, self.output);
1226 self.push_if_root(id.def_id);
1232 fn process_impl_item(&mut self, id: hir::ImplItemId) {
1233 if matches!(self.tcx.hir().def_kind(id.def_id), DefKind::AssocFn) {
1234 self.push_if_root(id.def_id);
1238 fn is_root(&self, def_id: LocalDefId) -> bool {
1239 !item_requires_monomorphization(self.tcx, def_id)
1240 && match self.mode {
1241 MonoItemCollectionMode::Eager => true,
1242 MonoItemCollectionMode::Lazy => {
1243 self.entry_fn.and_then(|(id, _)| id.as_local()) == Some(def_id)
1244 || self.tcx.is_reachable_non_generic(def_id)
1247 .codegen_fn_attrs(def_id)
1249 .contains(CodegenFnAttrFlags::RUSTC_STD_INTERNAL_SYMBOL)
1254 /// If `def_id` represents a root, pushes it onto the list of
1255 /// outputs. (Note that all roots must be monomorphic.)
1256 fn push_if_root(&mut self, def_id: LocalDefId) {
1257 if self.is_root(def_id) {
1258 debug!("RootCollector::push_if_root: found root def_id={:?}", def_id);
1260 let instance = Instance::mono(self.tcx, def_id.to_def_id());
1261 self.output.push(create_fn_mono_item(self.tcx, instance, DUMMY_SP));
1265 /// As a special case, when/if we encounter the
1266 /// `main()` function, we also have to generate a
1267 /// monomorphized copy of the start lang item based on
1268 /// the return type of `main`. This is not needed when
1269 /// the user writes their own `start` manually.
1270 fn push_extra_entry_roots(&mut self) {
1271 let Some((main_def_id, EntryFnType::Main)) = self.entry_fn else {
1275 let start_def_id = match self.tcx.lang_items().require(LangItem::Start) {
1277 Err(err) => self.tcx.sess.fatal(&err),
1279 let main_ret_ty = self.tcx.fn_sig(main_def_id).output();
1281 // Given that `main()` has no arguments,
1282 // then its return type cannot have
1283 // late-bound regions, since late-bound
1284 // regions must appear in the argument
1286 let main_ret_ty = self.tcx.normalize_erasing_regions(
1287 ty::ParamEnv::reveal_all(),
1288 main_ret_ty.no_bound_vars().unwrap(),
1291 let start_instance = Instance::resolve(
1293 ty::ParamEnv::reveal_all(),
1295 self.tcx.intern_substs(&[main_ret_ty.into()]),
1300 self.output.push(create_fn_mono_item(self.tcx, start_instance, DUMMY_SP));
1304 fn item_requires_monomorphization(tcx: TyCtxt<'_>, def_id: LocalDefId) -> bool {
1305 let generics = tcx.generics_of(def_id);
1306 generics.requires_monomorphization(tcx)
1309 fn create_mono_items_for_default_impls<'tcx>(
1311 item: &'tcx hir::Item<'tcx>,
1312 output: &mut Vec<Spanned<MonoItem<'tcx>>>,
1315 hir::ItemKind::Impl(ref impl_) => {
1316 for param in impl_.generics.params {
1318 hir::GenericParamKind::Lifetime { .. } => {}
1319 hir::GenericParamKind::Type { .. } | hir::GenericParamKind::Const { .. } => {
1326 "create_mono_items_for_default_impls(item={})",
1327 tcx.def_path_str(item.def_id.to_def_id())
1330 if let Some(trait_ref) = tcx.impl_trait_ref(item.def_id) {
1331 let param_env = ty::ParamEnv::reveal_all();
1332 let trait_ref = tcx.normalize_erasing_regions(param_env, trait_ref);
1333 let overridden_methods = tcx.impl_item_implementor_ids(item.def_id);
1334 for method in tcx.provided_trait_methods(trait_ref.def_id) {
1335 if overridden_methods.contains_key(&method.def_id) {
1339 if tcx.generics_of(method.def_id).own_requires_monomorphization() {
1344 InternalSubsts::for_item(tcx, method.def_id, |param, _| match param.kind {
1345 GenericParamDefKind::Lifetime => tcx.lifetimes.re_erased.into(),
1346 GenericParamDefKind::Type { .. }
1347 | GenericParamDefKind::Const { .. } => {
1348 trait_ref.substs[param.index as usize]
1351 let instance = ty::Instance::resolve(tcx, param_env, method.def_id, substs)
1355 let mono_item = create_fn_mono_item(tcx, instance, DUMMY_SP);
1356 if mono_item.node.is_instantiable(tcx) && should_codegen_locally(tcx, &instance)
1358 output.push(mono_item);
1367 /// Scans the miri alloc in order to find function calls, closures, and drop-glue.
1368 fn collect_miri<'tcx>(
1371 output: &mut Vec<Spanned<MonoItem<'tcx>>>,
1373 match tcx.global_alloc(alloc_id) {
1374 GlobalAlloc::Static(def_id) => {
1375 assert!(!tcx.is_thread_local_static(def_id));
1376 let instance = Instance::mono(tcx, def_id);
1377 if should_codegen_locally(tcx, &instance) {
1378 trace!("collecting static {:?}", def_id);
1379 output.push(dummy_spanned(MonoItem::Static(def_id)));
1382 GlobalAlloc::Memory(alloc) => {
1383 trace!("collecting {:?} with {:#?}", alloc_id, alloc);
1384 for &inner in alloc.inner().relocations().values() {
1385 rustc_data_structures::stack::ensure_sufficient_stack(|| {
1386 collect_miri(tcx, inner, output);
1390 GlobalAlloc::Function(fn_instance) => {
1391 if should_codegen_locally(tcx, &fn_instance) {
1392 trace!("collecting {:?} with {:#?}", alloc_id, fn_instance);
1393 output.push(create_fn_mono_item(tcx, fn_instance, DUMMY_SP));
1399 /// Scans the MIR in order to find function calls, closures, and drop-glue.
1400 fn collect_neighbours<'tcx>(
1402 instance: Instance<'tcx>,
1403 output: &mut Vec<Spanned<MonoItem<'tcx>>>,
1405 debug!("collect_neighbours: {:?}", instance.def_id());
1406 let body = tcx.instance_mir(instance.def);
1408 MirNeighborCollector { tcx, body: &body, output, instance }.visit_body(&body);
1411 fn collect_const_value<'tcx>(
1413 value: ConstValue<'tcx>,
1414 output: &mut Vec<Spanned<MonoItem<'tcx>>>,
1417 ConstValue::Scalar(Scalar::Ptr(ptr, _size)) => collect_miri(tcx, ptr.provenance, output),
1418 ConstValue::Slice { data: alloc, start: _, end: _ } | ConstValue::ByRef { alloc, .. } => {
1419 for &id in alloc.inner().relocations().values() {
1420 collect_miri(tcx, id, output);