rust/src/librustc_trans/partitioning.rs
2017-11-08 11:44:55 +01:00

654 lines
27 KiB
Rust

// Copyright 2016 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
//! Partitioning Codegen Units for Incremental Compilation
//! ======================================================
//!
//! The task of this module is to take the complete set of translation items of
//! a crate and produce a set of codegen units from it, where a codegen unit
//! is a named set of (translation-item, linkage) pairs. That is, this module
//! decides which translation item appears in which codegen units with which
//! linkage. The following paragraphs describe some of the background on the
//! partitioning scheme.
//!
//! The most important opportunity for saving on compilation time with
//! incremental compilation is to avoid re-translating and re-optimizing code.
//! Since the unit of translation and optimization for LLVM is "modules" or, how
//! we call them "codegen units", the particulars of how much time can be saved
//! by incremental compilation are tightly linked to how the output program is
//! partitioned into these codegen units prior to passing it to LLVM --
//! especially because we have to treat codegen units as opaque entities once
//! they are created: There is no way for us to incrementally update an existing
//! LLVM module and so we have to build any such module from scratch if it was
//! affected by some change in the source code.
//!
//! From that point of view it would make sense to maximize the number of
//! codegen units by, for example, putting each function into its own module.
//! That way only those modules would have to be re-compiled that were actually
//! affected by some change, minimizing the number of functions that could have
//! been re-used but just happened to be located in a module that is
//! re-compiled.
//!
//! However, since LLVM optimization does not work across module boundaries,
//! using such a highly granular partitioning would lead to very slow runtime
//! code since it would effectively prohibit inlining and other inter-procedure
//! optimizations. We want to avoid that as much as possible.
//!
//! Thus we end up with a trade-off: The bigger the codegen units, the better
//! LLVM's optimizer can do its work, but also the smaller the compilation time
//! reduction we get from incremental compilation.
//!
//! Ideally, we would create a partitioning such that there are few big codegen
//! units with few interdependencies between them. For now though, we use the
//! following heuristic to determine the partitioning:
//!
//! - There are two codegen units for every source-level module:
//! - One for "stable", that is non-generic, code
//! - One for more "volatile" code, i.e. monomorphized instances of functions
//! defined in that module
//!
//! In order to see why this heuristic makes sense, let's take a look at when a
//! codegen unit can get invalidated:
//!
//! 1. The most straightforward case is when the BODY of a function or global
//! changes. Then any codegen unit containing the code for that item has to be
//! re-compiled. Note that this includes all codegen units where the function
//! has been inlined.
//!
//! 2. The next case is when the SIGNATURE of a function or global changes. In
//! this case, all codegen units containing a REFERENCE to that item have to be
//! re-compiled. This is a superset of case 1.
//!
//! 3. The final and most subtle case is when a REFERENCE to a generic function
//! is added or removed somewhere. Even though the definition of the function
//! might be unchanged, a new REFERENCE might introduce a new monomorphized
//! instance of this function which has to be placed and compiled somewhere.
//! Conversely, when removing a REFERENCE, it might have been the last one with
//! that particular set of generic arguments and thus we have to remove it.
//!
//! From the above we see that just using one codegen unit per source-level
//! module is not such a good idea, since just adding a REFERENCE to some
//! generic item somewhere else would invalidate everything within the module
//! containing the generic item. The heuristic above reduces this detrimental
//! side-effect of references a little by at least not touching the non-generic
//! code of the module.
//!
//! A Note on Inlining
//! ------------------
//! As briefly mentioned above, in order for LLVM to be able to inline a
//! function call, the body of the function has to be available in the LLVM
//! module where the call is made. This has a few consequences for partitioning:
//!
//! - The partitioning algorithm has to take care of placing functions into all
//! codegen units where they should be available for inlining. It also has to
//! decide on the correct linkage for these functions.
//!
//! - The partitioning algorithm has to know which functions are likely to get
//! inlined, so it can distribute function instantiations accordingly. Since
//! there is no way of knowing for sure which functions LLVM will decide to
//! inline in the end, we apply a heuristic here: Only functions marked with
//! #[inline] are considered for inlining by the partitioner. The current
//! implementation will not try to determine if a function is likely to be
//! inlined by looking at the functions definition.
//!
//! Note though that as a side-effect of creating a codegen units per
//! source-level module, functions from the same module will be available for
//! inlining, even when they are not marked #[inline].
use collector::InliningMap;
use common;
use rustc::dep_graph::WorkProductId;
use rustc::hir::def_id::DefId;
use rustc::hir::map::DefPathData;
use rustc::middle::trans::{Linkage, Visibility};
use rustc::ty::{self, TyCtxt, InstanceDef};
use rustc::ty::item_path::characteristic_def_id_of_type;
use rustc::util::nodemap::{FxHashMap, FxHashSet};
use std::collections::hash_map::Entry;
use syntax::ast::NodeId;
use syntax::symbol::{Symbol, InternedString};
use trans_item::{TransItem, BaseTransItemExt, TransItemExt, InstantiationMode};
pub use rustc::middle::trans::CodegenUnit;
pub enum PartitioningStrategy {
/// Generate one codegen unit per source-level module.
PerModule,
/// Partition the whole crate into a fixed number of codegen units.
FixedUnitCount(usize)
}
pub trait CodegenUnitExt<'tcx> {
fn as_codegen_unit(&self) -> &CodegenUnit<'tcx>;
fn contains_item(&self, item: &TransItem<'tcx>) -> bool {
self.items().contains_key(item)
}
fn name<'a>(&'a self) -> &'a InternedString
where 'tcx: 'a,
{
&self.as_codegen_unit().name()
}
fn items(&self) -> &FxHashMap<TransItem<'tcx>, (Linkage, Visibility)> {
&self.as_codegen_unit().items()
}
fn work_product_id(&self) -> WorkProductId {
WorkProductId::from_cgu_name(self.name())
}
fn items_in_deterministic_order<'a>(&self,
tcx: TyCtxt<'a, 'tcx, 'tcx>)
-> Vec<(TransItem<'tcx>,
(Linkage, Visibility))> {
// The codegen tests rely on items being process in the same order as
// they appear in the file, so for local items, we sort by node_id first
#[derive(PartialEq, Eq, PartialOrd, Ord)]
pub struct ItemSortKey(Option<NodeId>, ty::SymbolName);
fn item_sort_key<'a, 'tcx>(tcx: TyCtxt<'a, 'tcx, 'tcx>,
item: TransItem<'tcx>) -> ItemSortKey {
ItemSortKey(match item {
TransItem::Fn(ref instance) => {
match instance.def {
// We only want to take NodeIds of user-defined
// instances into account. The others don't matter for
// the codegen tests and can even make item order
// unstable.
InstanceDef::Item(def_id) => {
tcx.hir.as_local_node_id(def_id)
}
InstanceDef::Intrinsic(..) |
InstanceDef::FnPtrShim(..) |
InstanceDef::Virtual(..) |
InstanceDef::ClosureOnceShim { .. } |
InstanceDef::DropGlue(..) |
InstanceDef::CloneShim(..) => {
None
}
}
}
TransItem::Static(node_id) |
TransItem::GlobalAsm(node_id) => {
Some(node_id)
}
}, item.symbol_name(tcx))
}
let items: Vec<_> = self.items().iter().map(|(&i, &l)| (i, l)).collect();
let mut items : Vec<_> = items.iter()
.map(|il| (il, item_sort_key(tcx, il.0))).collect();
items.sort_by(|&(_, ref key1), &(_, ref key2)| key1.cmp(key2));
items.into_iter().map(|(&item_linkage, _)| item_linkage).collect()
}
}
impl<'tcx> CodegenUnitExt<'tcx> for CodegenUnit<'tcx> {
fn as_codegen_unit(&self) -> &CodegenUnit<'tcx> {
self
}
}
// Anything we can't find a proper codegen unit for goes into this.
const FALLBACK_CODEGEN_UNIT: &'static str = "__rustc_fallback_codegen_unit";
pub fn partition<'a, 'tcx, I>(tcx: TyCtxt<'a, 'tcx, 'tcx>,
trans_items: I,
strategy: PartitioningStrategy,
inlining_map: &InliningMap<'tcx>)
-> Vec<CodegenUnit<'tcx>>
where I: Iterator<Item = TransItem<'tcx>>
{
// In the first step, we place all regular translation items into their
// respective 'home' codegen unit. Regular translation items are all
// functions and statics defined in the local crate.
let mut initial_partitioning = place_root_translation_items(tcx,
trans_items);
debug_dump(tcx, "INITIAL PARTITIONING:", initial_partitioning.codegen_units.iter());
// If the partitioning should produce a fixed count of codegen units, merge
// until that count is reached.
if let PartitioningStrategy::FixedUnitCount(count) = strategy {
merge_codegen_units(&mut initial_partitioning, count, &tcx.crate_name.as_str());
debug_dump(tcx, "POST MERGING:", initial_partitioning.codegen_units.iter());
}
// In the next step, we use the inlining map to determine which additional
// translation items have to go into each codegen unit. These additional
// translation items can be drop-glue, functions from external crates, and
// local functions the definition of which is marked with #[inline].
let mut post_inlining = place_inlined_translation_items(initial_partitioning,
inlining_map);
debug_dump(tcx, "POST INLINING:", post_inlining.codegen_units.iter());
// Next we try to make as many symbols "internal" as possible, so LLVM has
// more freedom to optimize.
internalize_symbols(tcx, &mut post_inlining, inlining_map);
// Finally, sort by codegen unit name, so that we get deterministic results
let PostInliningPartitioning {
codegen_units: mut result,
trans_item_placements: _,
internalization_candidates: _,
} = post_inlining;
result.sort_by(|cgu1, cgu2| {
cgu1.name().cmp(cgu2.name())
});
result
}
struct PreInliningPartitioning<'tcx> {
codegen_units: Vec<CodegenUnit<'tcx>>,
roots: FxHashSet<TransItem<'tcx>>,
internalization_candidates: FxHashSet<TransItem<'tcx>>,
}
/// For symbol internalization, we need to know whether a symbol/trans-item is
/// accessed from outside the codegen unit it is defined in. This type is used
/// to keep track of that.
#[derive(Clone, PartialEq, Eq, Debug)]
enum TransItemPlacement {
SingleCgu { cgu_name: InternedString },
MultipleCgus,
}
struct PostInliningPartitioning<'tcx> {
codegen_units: Vec<CodegenUnit<'tcx>>,
trans_item_placements: FxHashMap<TransItem<'tcx>, TransItemPlacement>,
internalization_candidates: FxHashSet<TransItem<'tcx>>,
}
fn place_root_translation_items<'a, 'tcx, I>(tcx: TyCtxt<'a, 'tcx, 'tcx>,
trans_items: I)
-> PreInliningPartitioning<'tcx>
where I: Iterator<Item = TransItem<'tcx>>
{
let mut roots = FxHashSet();
let mut codegen_units = FxHashMap();
let is_incremental_build = tcx.sess.opts.incremental.is_some();
let mut internalization_candidates = FxHashSet();
for trans_item in trans_items {
match trans_item.instantiation_mode(tcx) {
InstantiationMode::GloballyShared { .. } => {}
InstantiationMode::LocalCopy => continue,
}
let characteristic_def_id = characteristic_def_id_of_trans_item(tcx, trans_item);
let is_volatile = is_incremental_build &&
trans_item.is_generic_fn();
let codegen_unit_name = match characteristic_def_id {
Some(def_id) => compute_codegen_unit_name(tcx, def_id, is_volatile),
None => Symbol::intern(FALLBACK_CODEGEN_UNIT).as_str(),
};
let make_codegen_unit = || {
CodegenUnit::new(codegen_unit_name.clone())
};
let codegen_unit = codegen_units.entry(codegen_unit_name.clone())
.or_insert_with(make_codegen_unit);
let (linkage, visibility) = match trans_item.explicit_linkage(tcx) {
Some(explicit_linkage) => (explicit_linkage, Visibility::Default),
None => {
match trans_item {
TransItem::Fn(ref instance) => {
let visibility = match instance.def {
InstanceDef::Item(def_id) => {
if def_id.is_local() {
if tcx.is_exported_symbol(def_id) {
Visibility::Default
} else {
Visibility::Hidden
}
} else {
Visibility::Hidden
}
}
InstanceDef::FnPtrShim(..) |
InstanceDef::Virtual(..) |
InstanceDef::Intrinsic(..) |
InstanceDef::ClosureOnceShim { .. } |
InstanceDef::DropGlue(..) |
InstanceDef::CloneShim(..) => {
Visibility::Hidden
}
};
(Linkage::External, visibility)
}
TransItem::Static(node_id) |
TransItem::GlobalAsm(node_id) => {
let def_id = tcx.hir.local_def_id(node_id);
let visibility = if tcx.is_exported_symbol(def_id) {
Visibility::Default
} else {
Visibility::Hidden
};
(Linkage::External, visibility)
}
}
}
};
if visibility == Visibility::Hidden {
internalization_candidates.insert(trans_item);
}
codegen_unit.items_mut().insert(trans_item, (linkage, visibility));
roots.insert(trans_item);
}
// always ensure we have at least one CGU; otherwise, if we have a
// crate with just types (for example), we could wind up with no CGU
if codegen_units.is_empty() {
let codegen_unit_name = Symbol::intern(FALLBACK_CODEGEN_UNIT).as_str();
codegen_units.insert(codegen_unit_name.clone(),
CodegenUnit::new(codegen_unit_name.clone()));
}
PreInliningPartitioning {
codegen_units: codegen_units.into_iter()
.map(|(_, codegen_unit)| codegen_unit)
.collect(),
roots,
internalization_candidates,
}
}
fn merge_codegen_units<'tcx>(initial_partitioning: &mut PreInliningPartitioning<'tcx>,
target_cgu_count: usize,
crate_name: &str) {
assert!(target_cgu_count >= 1);
let codegen_units = &mut initial_partitioning.codegen_units;
// Merge the two smallest codegen units until the target size is reached.
// Note that "size" is estimated here rather inaccurately as the number of
// translation items in a given unit. This could be improved on.
while codegen_units.len() > target_cgu_count {
// Sort small cgus to the back
codegen_units.sort_by_key(|cgu| -(cgu.items().len() as i64));
let mut smallest = codegen_units.pop().unwrap();
let second_smallest = codegen_units.last_mut().unwrap();
for (k, v) in smallest.items_mut().drain() {
second_smallest.items_mut().insert(k, v);
}
}
for (index, cgu) in codegen_units.iter_mut().enumerate() {
cgu.set_name(numbered_codegen_unit_name(crate_name, index));
}
}
fn place_inlined_translation_items<'tcx>(initial_partitioning: PreInliningPartitioning<'tcx>,
inlining_map: &InliningMap<'tcx>)
-> PostInliningPartitioning<'tcx> {
let mut new_partitioning = Vec::new();
let mut trans_item_placements = FxHashMap();
let PreInliningPartitioning {
codegen_units: initial_cgus,
roots,
internalization_candidates,
} = initial_partitioning;
let single_codegen_unit = initial_cgus.len() == 1;
for old_codegen_unit in initial_cgus {
// Collect all items that need to be available in this codegen unit
let mut reachable = FxHashSet();
for root in old_codegen_unit.items().keys() {
follow_inlining(*root, inlining_map, &mut reachable);
}
let mut new_codegen_unit = CodegenUnit::new(old_codegen_unit.name().clone());
// Add all translation items that are not already there
for trans_item in reachable {
if let Some(linkage) = old_codegen_unit.items().get(&trans_item) {
// This is a root, just copy it over
new_codegen_unit.items_mut().insert(trans_item, *linkage);
} else {
if roots.contains(&trans_item) {
bug!("GloballyShared trans-item inlined into other CGU: \
{:?}", trans_item);
}
// This is a cgu-private copy
new_codegen_unit.items_mut().insert(
trans_item,
(Linkage::Internal, Visibility::Default),
);
}
if !single_codegen_unit {
// If there is more than one codegen unit, we need to keep track
// in which codegen units each translation item is placed:
match trans_item_placements.entry(trans_item) {
Entry::Occupied(e) => {
let placement = e.into_mut();
debug_assert!(match *placement {
TransItemPlacement::SingleCgu { ref cgu_name } => {
*cgu_name != *new_codegen_unit.name()
}
TransItemPlacement::MultipleCgus => true,
});
*placement = TransItemPlacement::MultipleCgus;
}
Entry::Vacant(e) => {
e.insert(TransItemPlacement::SingleCgu {
cgu_name: new_codegen_unit.name().clone()
});
}
}
}
}
new_partitioning.push(new_codegen_unit);
}
return PostInliningPartitioning {
codegen_units: new_partitioning,
trans_item_placements,
internalization_candidates,
};
fn follow_inlining<'tcx>(trans_item: TransItem<'tcx>,
inlining_map: &InliningMap<'tcx>,
visited: &mut FxHashSet<TransItem<'tcx>>) {
if !visited.insert(trans_item) {
return;
}
inlining_map.with_inlining_candidates(trans_item, |target| {
follow_inlining(target, inlining_map, visited);
});
}
}
fn internalize_symbols<'a, 'tcx>(_tcx: TyCtxt<'a, 'tcx, 'tcx>,
partitioning: &mut PostInliningPartitioning<'tcx>,
inlining_map: &InliningMap<'tcx>) {
if partitioning.codegen_units.len() == 1 {
// Fast path for when there is only one codegen unit. In this case we
// can internalize all candidates, since there is nowhere else they
// could be accessed from.
for cgu in &mut partitioning.codegen_units {
for candidate in &partitioning.internalization_candidates {
cgu.items_mut().insert(*candidate,
(Linkage::Internal, Visibility::Default));
}
}
return;
}
// Build a map from every translation item to all the translation items that
// reference it.
let mut accessor_map: FxHashMap<TransItem<'tcx>, Vec<TransItem<'tcx>>> = FxHashMap();
inlining_map.iter_accesses(|accessor, accessees| {
for accessee in accessees {
accessor_map.entry(*accessee)
.or_insert(Vec::new())
.push(accessor);
}
});
let trans_item_placements = &partitioning.trans_item_placements;
// For each internalization candidates in each codegen unit, check if it is
// accessed from outside its defining codegen unit.
for cgu in &mut partitioning.codegen_units {
let home_cgu = TransItemPlacement::SingleCgu {
cgu_name: cgu.name().clone()
};
for (accessee, linkage_and_visibility) in cgu.items_mut() {
if !partitioning.internalization_candidates.contains(accessee) {
// This item is no candidate for internalizing, so skip it.
continue
}
debug_assert_eq!(trans_item_placements[accessee], home_cgu);
if let Some(accessors) = accessor_map.get(accessee) {
if accessors.iter()
.filter_map(|accessor| {
// Some accessors might not have been
// instantiated. We can safely ignore those.
trans_item_placements.get(accessor)
})
.any(|placement| *placement != home_cgu) {
// Found an accessor from another CGU, so skip to the next
// item without marking this one as internal.
continue
}
}
// If we got here, we did not find any accesses from other CGUs,
// so it's fine to make this translation item internal.
*linkage_and_visibility = (Linkage::Internal, Visibility::Default);
}
}
}
fn characteristic_def_id_of_trans_item<'a, 'tcx>(tcx: TyCtxt<'a, 'tcx, 'tcx>,
trans_item: TransItem<'tcx>)
-> Option<DefId> {
match trans_item {
TransItem::Fn(instance) => {
let def_id = match instance.def {
ty::InstanceDef::Item(def_id) => def_id,
ty::InstanceDef::FnPtrShim(..) |
ty::InstanceDef::ClosureOnceShim { .. } |
ty::InstanceDef::Intrinsic(..) |
ty::InstanceDef::DropGlue(..) |
ty::InstanceDef::Virtual(..) |
ty::InstanceDef::CloneShim(..) => return None
};
// If this is a method, we want to put it into the same module as
// its self-type. If the self-type does not provide a characteristic
// DefId, we use the location of the impl after all.
if tcx.trait_of_item(def_id).is_some() {
let self_ty = instance.substs.type_at(0);
// This is an implementation of a trait method.
return characteristic_def_id_of_type(self_ty).or(Some(def_id));
}
if let Some(impl_def_id) = tcx.impl_of_method(def_id) {
// This is a method within an inherent impl, find out what the
// self-type is:
let impl_self_ty = common::def_ty(tcx, impl_def_id, instance.substs);
if let Some(def_id) = characteristic_def_id_of_type(impl_self_ty) {
return Some(def_id);
}
}
Some(def_id)
}
TransItem::Static(node_id) |
TransItem::GlobalAsm(node_id) => Some(tcx.hir.local_def_id(node_id)),
}
}
fn compute_codegen_unit_name<'a, 'tcx>(tcx: TyCtxt<'a, 'tcx, 'tcx>,
def_id: DefId,
volatile: bool)
-> InternedString {
// Unfortunately we cannot just use the `ty::item_path` infrastructure here
// because we need paths to modules and the DefIds of those are not
// available anymore for external items.
let mut mod_path = String::with_capacity(64);
let def_path = tcx.def_path(def_id);
mod_path.push_str(&tcx.crate_name(def_path.krate).as_str());
for part in tcx.def_path(def_id)
.data
.iter()
.take_while(|part| {
match part.data {
DefPathData::Module(..) => true,
_ => false,
}
}) {
mod_path.push_str("-");
mod_path.push_str(&part.data.as_interned_str());
}
if volatile {
mod_path.push_str(".volatile");
}
return Symbol::intern(&mod_path[..]).as_str();
}
fn numbered_codegen_unit_name(crate_name: &str, index: usize) -> InternedString {
Symbol::intern(&format!("{}{}", crate_name, index)).as_str()
}
fn debug_dump<'a, 'b, 'tcx, I>(tcx: TyCtxt<'a, 'tcx, 'tcx>,
label: &str,
cgus: I)
where I: Iterator<Item=&'b CodegenUnit<'tcx>>,
'tcx: 'a + 'b
{
if cfg!(debug_assertions) {
debug!("{}", label);
for cgu in cgus {
debug!("CodegenUnit {}:", cgu.name());
for (trans_item, linkage) in cgu.items() {
let symbol_name = trans_item.symbol_name(tcx);
let symbol_hash_start = symbol_name.rfind('h');
let symbol_hash = symbol_hash_start.map(|i| &symbol_name[i ..])
.unwrap_or("<no hash>");
debug!(" - {} [{:?}] [{}]",
trans_item.to_string(tcx),
linkage,
symbol_hash);
}
debug!("");
}
}
}