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]`.
98 use rustc_data_structures::fx::{FxHashMap, FxHashSet};
99 use rustc_data_structures::sync;
100 use rustc_hir::def_id::{CrateNum, DefIdSet, LOCAL_CRATE};
101 use rustc_middle::mir::mono::MonoItem;
102 use rustc_middle::mir::mono::{CodegenUnit, Linkage};
103 use rustc_middle::ty::query::Providers;
104 use rustc_middle::ty::TyCtxt;
105 use rustc_span::symbol::Symbol;
107 use crate::monomorphize::collector::InliningMap;
108 use crate::monomorphize::collector::{self, MonoItemCollectionMode};
110 trait Partitioner<'tcx> {
111 fn place_root_mono_items(
114 mono_items: &mut dyn Iterator<Item = MonoItem<'tcx>>,
115 ) -> PreInliningPartitioning<'tcx>;
117 fn merge_codegen_units(
120 initial_partitioning: &mut PreInliningPartitioning<'tcx>,
121 target_cgu_count: usize,
124 fn place_inlined_mono_items(
126 initial_partitioning: PreInliningPartitioning<'tcx>,
127 inlining_map: &InliningMap<'tcx>,
128 ) -> PostInliningPartitioning<'tcx>;
130 fn internalize_symbols(
133 partitioning: &mut PostInliningPartitioning<'tcx>,
134 inlining_map: &InliningMap<'tcx>,
138 fn get_partitioner<'tcx>(tcx: TyCtxt<'tcx>) -> Box<dyn Partitioner<'tcx>> {
139 let strategy = match &tcx.sess.opts.debugging_opts.cgu_partitioning_strategy {
145 "default" => Box::new(default::DefaultPartitioning),
146 _ => tcx.sess.fatal("unknown partitioning strategy"),
150 pub fn partition<'tcx>(
152 mono_items: &mut dyn Iterator<Item = MonoItem<'tcx>>,
153 max_cgu_count: usize,
154 inlining_map: &InliningMap<'tcx>,
155 ) -> Vec<CodegenUnit<'tcx>> {
156 let _prof_timer = tcx.prof.generic_activity("cgu_partitioning");
158 let mut partitioner = get_partitioner(tcx);
159 // In the first step, we place all regular monomorphizations into their
160 // respective 'home' codegen unit. Regular monomorphizations are all
161 // functions and statics defined in the local crate.
162 let mut initial_partitioning = {
163 let _prof_timer = tcx.prof.generic_activity("cgu_partitioning_place_roots");
164 partitioner.place_root_mono_items(tcx, mono_items)
167 initial_partitioning.codegen_units.iter_mut().for_each(|cgu| cgu.estimate_size(tcx));
169 debug_dump(tcx, "INITIAL PARTITIONING:", initial_partitioning.codegen_units.iter());
171 // Merge until we have at most `max_cgu_count` codegen units.
173 let _prof_timer = tcx.prof.generic_activity("cgu_partitioning_merge_cgus");
174 partitioner.merge_codegen_units(tcx, &mut initial_partitioning, max_cgu_count);
175 debug_dump(tcx, "POST MERGING:", initial_partitioning.codegen_units.iter());
178 // In the next step, we use the inlining map to determine which additional
179 // monomorphizations have to go into each codegen unit. These additional
180 // monomorphizations can be drop-glue, functions from external crates, and
181 // local functions the definition of which is marked with `#[inline]`.
182 let mut post_inlining = {
183 let _prof_timer = tcx.prof.generic_activity("cgu_partitioning_place_inline_items");
184 partitioner.place_inlined_mono_items(initial_partitioning, inlining_map)
187 post_inlining.codegen_units.iter_mut().for_each(|cgu| cgu.estimate_size(tcx));
189 debug_dump(tcx, "POST INLINING:", post_inlining.codegen_units.iter());
191 // Next we try to make as many symbols "internal" as possible, so LLVM has
192 // more freedom to optimize.
193 if tcx.sess.opts.cg.link_dead_code != Some(true) {
194 let _prof_timer = tcx.prof.generic_activity("cgu_partitioning_internalize_symbols");
195 partitioner.internalize_symbols(tcx, &mut post_inlining, inlining_map);
198 // Finally, sort by codegen unit name, so that we get deterministic results.
199 let PostInliningPartitioning {
200 codegen_units: mut result,
201 mono_item_placements: _,
202 internalization_candidates: _,
205 result.sort_by_cached_key(|cgu| cgu.name().as_str());
210 pub struct PreInliningPartitioning<'tcx> {
211 codegen_units: Vec<CodegenUnit<'tcx>>,
212 roots: FxHashSet<MonoItem<'tcx>>,
213 internalization_candidates: FxHashSet<MonoItem<'tcx>>,
216 /// For symbol internalization, we need to know whether a symbol/mono-item is
217 /// accessed from outside the codegen unit it is defined in. This type is used
218 /// to keep track of that.
219 #[derive(Clone, PartialEq, Eq, Debug)]
220 enum MonoItemPlacement {
221 SingleCgu { cgu_name: Symbol },
225 struct PostInliningPartitioning<'tcx> {
226 codegen_units: Vec<CodegenUnit<'tcx>>,
227 mono_item_placements: FxHashMap<MonoItem<'tcx>, MonoItemPlacement>,
228 internalization_candidates: FxHashSet<MonoItem<'tcx>>,
231 fn debug_dump<'a, 'tcx, I>(tcx: TyCtxt<'tcx>, label: &str, cgus: I)
233 I: Iterator<Item = &'a CodegenUnit<'tcx>>,
236 if cfg!(debug_assertions) {
239 debug!("CodegenUnit {} estimated size {} :", cgu.name(), cgu.size_estimate());
241 for (mono_item, linkage) in cgu.items() {
242 let symbol_name = mono_item.symbol_name(tcx).name;
243 let symbol_hash_start = symbol_name.rfind('h');
245 symbol_hash_start.map(|i| &symbol_name[i..]).unwrap_or("<no hash>");
248 " - {} [{:?}] [{}] estimated size {}",
249 mono_item.to_string(tcx, true),
252 mono_item.size_estimate(tcx)
261 #[inline(never)] // give this a place in the profiler
262 fn assert_symbols_are_distinct<'a, 'tcx, I>(tcx: TyCtxt<'tcx>, mono_items: I)
264 I: Iterator<Item = &'a MonoItem<'tcx>>,
267 let _prof_timer = tcx.prof.generic_activity("assert_symbols_are_distinct");
269 let mut symbols: Vec<_> =
270 mono_items.map(|mono_item| (mono_item, mono_item.symbol_name(tcx))).collect();
272 symbols.sort_by_key(|sym| sym.1);
274 for pair in symbols.windows(2) {
275 let sym1 = &pair[0].1;
276 let sym2 = &pair[1].1;
279 let mono_item1 = pair[0].0;
280 let mono_item2 = pair[1].0;
282 let span1 = mono_item1.local_span(tcx);
283 let span2 = mono_item2.local_span(tcx);
285 // Deterministically select one of the spans for error reporting
286 let span = match (span1, span2) {
287 (Some(span1), Some(span2)) => {
288 Some(if span1.lo().0 > span2.lo().0 { span1 } else { span2 })
290 (span1, span2) => span1.or(span2),
293 let error_message = format!("symbol `{}` is already defined", sym1);
295 if let Some(span) = span {
296 tcx.sess.span_fatal(span, &error_message)
298 tcx.sess.fatal(&error_message)
304 fn collect_and_partition_mono_items<'tcx>(
307 ) -> (&'tcx DefIdSet, &'tcx [CodegenUnit<'tcx>]) {
308 assert_eq!(cnum, LOCAL_CRATE);
310 let collection_mode = match tcx.sess.opts.debugging_opts.print_mono_items {
312 let mode_string = s.to_lowercase();
313 let mode_string = mode_string.trim();
314 if mode_string == "eager" {
315 MonoItemCollectionMode::Eager
317 if mode_string != "lazy" {
318 let message = format!(
319 "Unknown codegen-item collection mode '{}'. \
320 Falling back to 'lazy' mode.",
323 tcx.sess.warn(&message);
326 MonoItemCollectionMode::Lazy
330 if tcx.sess.opts.cg.link_dead_code == Some(true) {
331 MonoItemCollectionMode::Eager
333 MonoItemCollectionMode::Lazy
338 let (items, inlining_map) = collector::collect_crate_mono_items(tcx, collection_mode);
340 tcx.sess.abort_if_errors();
342 let (codegen_units, _) = tcx.sess.time("partition_and_assert_distinct_symbols", || {
345 &*tcx.arena.alloc_from_iter(partition(
347 &mut items.iter().cloned(),
348 tcx.sess.codegen_units(),
352 || assert_symbols_are_distinct(tcx, items.iter()),
356 let mono_items: DefIdSet = items
358 .filter_map(|mono_item| match *mono_item {
359 MonoItem::Fn(ref instance) => Some(instance.def_id()),
360 MonoItem::Static(def_id) => Some(def_id),
365 if tcx.sess.opts.debugging_opts.print_mono_items.is_some() {
366 let mut item_to_cgus: FxHashMap<_, Vec<_>> = Default::default();
368 for cgu in codegen_units {
369 for (&mono_item, &linkage) in cgu.items() {
370 item_to_cgus.entry(mono_item).or_default().push((cgu.name(), linkage));
374 let mut item_keys: Vec<_> = items
377 let mut output = i.to_string(tcx, false);
378 output.push_str(" @@");
379 let mut empty = Vec::new();
380 let cgus = item_to_cgus.get_mut(i).unwrap_or(&mut empty);
381 cgus.sort_by_key(|(name, _)| *name);
383 for &(ref cgu_name, (linkage, _)) in cgus.iter() {
384 output.push_str(" ");
385 output.push_str(&cgu_name.as_str());
387 let linkage_abbrev = match linkage {
388 Linkage::External => "External",
389 Linkage::AvailableExternally => "Available",
390 Linkage::LinkOnceAny => "OnceAny",
391 Linkage::LinkOnceODR => "OnceODR",
392 Linkage::WeakAny => "WeakAny",
393 Linkage::WeakODR => "WeakODR",
394 Linkage::Appending => "Appending",
395 Linkage::Internal => "Internal",
396 Linkage::Private => "Private",
397 Linkage::ExternalWeak => "ExternalWeak",
398 Linkage::Common => "Common",
401 output.push_str("[");
402 output.push_str(linkage_abbrev);
403 output.push_str("]");
411 for item in item_keys {
412 println!("MONO_ITEM {}", item);
416 (tcx.arena.alloc(mono_items), codegen_units)
419 pub fn provide(providers: &mut Providers) {
420 providers.collect_and_partition_mono_items = collect_and_partition_mono_items;
422 providers.is_codegened_item = |tcx, def_id| {
423 let (all_mono_items, _) = tcx.collect_and_partition_mono_items(LOCAL_CRATE);
424 all_mono_items.contains(&def_id)
427 providers.codegen_unit = |tcx, name| {
428 let (_, all) = tcx.collect_and_partition_mono_items(LOCAL_CRATE);
430 .find(|cgu| cgu.name() == name)
431 .unwrap_or_else(|| panic!("failed to find cgu with name {:?}", name))