Introduce `ChunkedBitSet` and use it for some dataflow analyses.

This reduces peak memory usage significantly for some programs with very
large functions, such as:
- `keccak`, `unicode_normalization`, and `match-stress-enum`, from
  the `rustc-perf` benchmark suite;
- `http-0.2.6` from crates.io.

The new type is used in the analyses where the bitsets can get huge
(e.g. 10s of thousands of bits): `MaybeInitializedPlaces`,
`MaybeUninitializedPlaces`, and `EverInitializedPlaces`.

Some refactoring was required in `rustc_mir_dataflow`. All existing
analysis domains are either `BitSet` or a trivial wrapper around
`BitSet`, and access in a few places is done via `Borrow<BitSet>` or
`BorrowMut<BitSet>`. Now that some of these domains are `ClusterBitSet`,
that no longer works. So this commit replaces the `Borrow`/`BorrowMut`
usage with a new trait `BitSetExt` containing the needed bitset
operations. The impls just forward these to the underlying bitset type.
This required fiddling with trait bounds in a few places.

The commit also:
- Moves `static_assert_size` from `rustc_data_structures` to
  `rustc_index` so it can be used in the latter; the former now
  re-exports it so existing users are unaffected.
- Factors out some common "clear excess bits in the final word"
  functionality in `bit_set.rs`.
- Uses `fill` in a few places instead of loops.

7 files changed, 121 insertions, 33 deletions

diff --git a/compiler/rustc_mir_dataflow/src/framework/cursor.rs b/compiler/rustc_mir_dataflow/src/framework/cursor.rs
index ba6b566a304..2de5a9d6991 100644
--- a/compiler/rustc_mir_dataflow/src/framework/cursor.rs
+++ b/compiler/rustc_mir_dataflow/src/framework/cursor.rs
@@ -1,10 +1,12 @@
 //! Random access inspection of the results of a dataflow analysis.
 
+use crate::framework::BitSetExt;
+
 use std::borrow::Borrow;
 use std::cmp::Ordering;
 
+#[cfg(debug_assertions)]
 use rustc_index::bit_set::BitSet;
-use rustc_index::vec::Idx;
 use rustc_middle::mir::{self, BasicBlock, Location};
 
 use super::{Analysis, Direction, Effect, EffectIndex, Results};
@@ -209,13 +211,13 @@ where
     }
 }
 
-impl<'mir, 'tcx, A, R, T> ResultsCursor<'mir, 'tcx, A, R>
+impl<'mir, 'tcx, A, R> ResultsCursor<'mir, 'tcx, A, R>
 where
-    A: Analysis<'tcx, Domain = BitSet<T>>,
-    T: Idx,
+    A: crate::GenKillAnalysis<'tcx>,
+    A::Domain: BitSetExt<A::Idx>,
     R: Borrow<Results<'tcx, A>>,
 {
-    pub fn contains(&self, elem: T) -> bool {
+    pub fn contains(&self, elem: A::Idx) -> bool {
         self.get().contains(elem)
     }
 }
diff --git a/compiler/rustc_mir_dataflow/src/framework/engine.rs b/compiler/rustc_mir_dataflow/src/framework/engine.rs
index e20358ba989..88ed0128a1f 100644
--- a/compiler/rustc_mir_dataflow/src/framework/engine.rs
+++ b/compiler/rustc_mir_dataflow/src/framework/engine.rs
@@ -1,6 +1,7 @@
 //! A solver for dataflow problems.
 
-use std::borrow::BorrowMut;
+use crate::framework::BitSetExt;
+
 use std::ffi::OsString;
 use std::path::PathBuf;
 
@@ -91,7 +92,7 @@ where
 impl<'a, 'tcx, A, D, T> Engine<'a, 'tcx, A>
 where
     A: GenKillAnalysis<'tcx, Idx = T, Domain = D>,
-    D: Clone + JoinSemiLattice + GenKill<T> + BorrowMut<BitSet<T>>,
+    D: Clone + JoinSemiLattice + GenKill<T> + BitSetExt<T>,
     T: Idx,
 {
     /// Creates a new `Engine` to solve a gen-kill dataflow problem.
@@ -106,7 +107,7 @@ where
 
         // Otherwise, compute and store the cumulative transfer function for each block.
 
-        let identity = GenKillSet::identity(analysis.bottom_value(body).borrow().domain_size());
+        let identity = GenKillSet::identity(analysis.bottom_value(body).domain_size());
         let mut trans_for_block = IndexVec::from_elem(identity, body.basic_blocks());
 
         for (block, block_data) in body.basic_blocks().iter_enumerated() {
@@ -115,7 +116,7 @@ where
         }
 
         let apply_trans = Box::new(move |bb: BasicBlock, state: &mut A::Domain| {
-            trans_for_block[bb].apply(state.borrow_mut());
+            trans_for_block[bb].apply(state);
         });
 
         Self::new(tcx, body, analysis, Some(apply_trans as Box<_>))
diff --git a/compiler/rustc_mir_dataflow/src/framework/fmt.rs b/compiler/rustc_mir_dataflow/src/framework/fmt.rs
index 1d1553bbbd9..99735673f4d 100644
--- a/compiler/rustc_mir_dataflow/src/framework/fmt.rs
+++ b/compiler/rustc_mir_dataflow/src/framework/fmt.rs
@@ -1,7 +1,7 @@
 //! Custom formatting traits used when outputting Graphviz diagrams with the results of a dataflow
 //! analysis.
 
-use rustc_index::bit_set::{BitSet, HybridBitSet};
+use rustc_index::bit_set::{BitSet, ChunkedBitSet, HybridBitSet};
 use rustc_index::vec::Idx;
 use std::fmt;
 
@@ -133,6 +133,19 @@ where
     }
 }
 
+impl<T, C> DebugWithContext<C> for ChunkedBitSet<T>
+where
+    T: Idx + DebugWithContext<C>,
+{
+    fn fmt_with(&self, _ctxt: &C, _f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        unimplemented!("implement when/if needed");
+    }
+
+    fn fmt_diff_with(&self, _old: &Self, _ctxt: &C, _f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        unimplemented!("implement when/if needed");
+    }
+}
+
 impl<T, C> DebugWithContext<C> for &'_ T
 where
     T: DebugWithContext<C>,
diff --git a/compiler/rustc_mir_dataflow/src/framework/lattice.rs b/compiler/rustc_mir_dataflow/src/framework/lattice.rs
index f937b31f4cf..9a462f6e1a4 100644
--- a/compiler/rustc_mir_dataflow/src/framework/lattice.rs
+++ b/compiler/rustc_mir_dataflow/src/framework/lattice.rs
@@ -38,7 +38,8 @@
 //! [Hasse diagram]: https://en.wikipedia.org/wiki/Hasse_diagram
 //! [poset]: https://en.wikipedia.org/wiki/Partially_ordered_set
 
-use rustc_index::bit_set::BitSet;
+use crate::framework::BitSetExt;
+use rustc_index::bit_set::{BitSet, ChunkedBitSet, HybridBitSet};
 use rustc_index::vec::{Idx, IndexVec};
 use std::iter;
 
@@ -145,6 +146,18 @@ impl<T: Idx> MeetSemiLattice for BitSet<T> {
     }
 }
 
+impl<T: Idx> JoinSemiLattice for ChunkedBitSet<T> {
+    fn join(&mut self, other: &Self) -> bool {
+        self.union(other)
+    }
+}
+
+impl<T: Idx> MeetSemiLattice for ChunkedBitSet<T> {
+    fn meet(&mut self, other: &Self) -> bool {
+        self.intersect(other)
+    }
+}
+
 /// The counterpart of a given semilattice `T` using the [inverse order].
 ///
 /// The dual of a join-semilattice is a meet-semilattice and vice versa. For example, the dual of a
@@ -155,15 +168,21 @@ impl<T: Idx> MeetSemiLattice for BitSet<T> {
 #[derive(Clone, Copy, Debug, PartialEq, Eq)]
 pub struct Dual<T>(pub T);
 
-impl<T> std::borrow::Borrow<T> for Dual<T> {
-    fn borrow(&self) -> &T {
-        &self.0
+impl<T: Idx> BitSetExt<T> for Dual<BitSet<T>> {
+    fn domain_size(&self) -> usize {
+        self.0.domain_size()
+    }
+
+    fn contains(&self, elem: T) -> bool {
+        self.0.contains(elem)
+    }
+
+    fn union(&mut self, other: &HybridBitSet<T>) {
+        self.0.union(other);
     }
-}
 
-impl<T> std::borrow::BorrowMut<T> for Dual<T> {
-    fn borrow_mut(&mut self) -> &mut T {
-        &mut self.0
+    fn subtract(&mut self, other: &HybridBitSet<T>) {
+        self.0.subtract(other);
     }
 }
 
diff --git a/compiler/rustc_mir_dataflow/src/framework/mod.rs b/compiler/rustc_mir_dataflow/src/framework/mod.rs
index dac6720a6e5..c51dd06de25 100644
--- a/compiler/rustc_mir_dataflow/src/framework/mod.rs
+++ b/compiler/rustc_mir_dataflow/src/framework/mod.rs
@@ -30,10 +30,9 @@
 //!
 //! [gen-kill]: https://en.wikipedia.org/wiki/Data-flow_analysis#Bit_vector_problems
 
-use std::borrow::BorrowMut;
 use std::cmp::Ordering;
 
-use rustc_index::bit_set::{BitSet, HybridBitSet};
+use rustc_index::bit_set::{BitSet, ChunkedBitSet, HybridBitSet};
 use rustc_index::vec::Idx;
 use rustc_middle::mir::{self, BasicBlock, Location};
 use rustc_middle::ty::TyCtxt;
@@ -52,6 +51,51 @@ pub use self::engine::{Engine, Results};
 pub use self::lattice::{JoinSemiLattice, MeetSemiLattice};
 pub use self::visitor::{visit_results, ResultsVisitable, ResultsVisitor};
 
+/// Analysis domains are all bitsets of various kinds. This trait holds
+/// operations needed by all of them.
+pub trait BitSetExt<T> {
+    fn domain_size(&self) -> usize;
+    fn contains(&self, elem: T) -> bool;
+    fn union(&mut self, other: &HybridBitSet<T>);
+    fn subtract(&mut self, other: &HybridBitSet<T>);
+}
+
+impl<T: Idx> BitSetExt<T> for BitSet<T> {
+    fn domain_size(&self) -> usize {
+        self.domain_size()
+    }
+
+    fn contains(&self, elem: T) -> bool {
+        self.contains(elem)
+    }
+
+    fn union(&mut self, other: &HybridBitSet<T>) {
+        self.union(other);
+    }
+
+    fn subtract(&mut self, other: &HybridBitSet<T>) {
+        self.subtract(other);
+    }
+}
+
+impl<T: Idx> BitSetExt<T> for ChunkedBitSet<T> {
+    fn domain_size(&self) -> usize {
+        self.domain_size()
+    }
+
+    fn contains(&self, elem: T) -> bool {
+        self.contains(elem)
+    }
+
+    fn union(&mut self, other: &HybridBitSet<T>) {
+        self.union(other);
+    }
+
+    fn subtract(&mut self, other: &HybridBitSet<T>) {
+        self.subtract(other);
+    }
+}
+
 /// Define the domain of a dataflow problem.
 ///
 /// This trait specifies the lattice on which this analysis operates (the domain) as well as its
@@ -303,7 +347,7 @@ pub trait GenKillAnalysis<'tcx>: Analysis<'tcx> {
 impl<'tcx, A> Analysis<'tcx> for A
 where
     A: GenKillAnalysis<'tcx>,
-    A::Domain: GenKill<A::Idx> + BorrowMut<BitSet<A::Idx>>,
+    A::Domain: GenKill<A::Idx> + BitSetExt<A::Idx>,
 {
     fn apply_statement_effect(
         &self,
@@ -435,7 +479,7 @@ impl<T: Idx> GenKillSet<T> {
         }
     }
 
-    pub fn apply(&self, state: &mut BitSet<T>) {
+    pub fn apply(&self, state: &mut impl BitSetExt<T>) {
         state.union(&self.gen);
         state.subtract(&self.kill);
     }
@@ -463,6 +507,16 @@ impl<T: Idx> GenKill<T> for BitSet<T> {
     }
 }
 
+impl<T: Idx> GenKill<T> for ChunkedBitSet<T> {
+    fn gen(&mut self, elem: T) {
+        self.insert(elem);
+    }
+
+    fn kill(&mut self, elem: T) {
+        self.remove(elem);
+    }
+}
+
 impl<T: Idx> GenKill<T> for lattice::Dual<BitSet<T>> {
     fn gen(&mut self, elem: T) {
         self.0.insert(elem);
diff --git a/compiler/rustc_mir_dataflow/src/impls/mod.rs b/compiler/rustc_mir_dataflow/src/impls/mod.rs
index 63d935db8ca..9502c5d57d6 100644
--- a/compiler/rustc_mir_dataflow/src/impls/mod.rs
+++ b/compiler/rustc_mir_dataflow/src/impls/mod.rs
@@ -2,7 +2,7 @@
 //! bitvectors attached to each basic block, represented via a
 //! zero-sized structure.
 
-use rustc_index::bit_set::BitSet;
+use rustc_index::bit_set::{BitSet, ChunkedBitSet};
 use rustc_index::vec::Idx;
 use rustc_middle::mir::visit::{MirVisitable, Visitor};
 use rustc_middle::mir::{self, Body, Location};
@@ -286,12 +286,12 @@ impl<'a, 'tcx> DefinitelyInitializedPlaces<'a, 'tcx> {
 }
 
 impl<'tcx> AnalysisDomain<'tcx> for MaybeInitializedPlaces<'_, 'tcx> {
-    type Domain = BitSet<MovePathIndex>;
+    type Domain = ChunkedBitSet<MovePathIndex>;
     const NAME: &'static str = "maybe_init";
 
     fn bottom_value(&self, _: &mir::Body<'tcx>) -> Self::Domain {
         // bottom = uninitialized
-        BitSet::new_empty(self.move_data().move_paths.len())
+        ChunkedBitSet::new_empty(self.move_data().move_paths.len())
     }
 
     fn initialize_start_block(&self, _: &mir::Body<'tcx>, state: &mut Self::Domain) {
@@ -417,13 +417,13 @@ impl<'tcx> GenKillAnalysis<'tcx> for MaybeInitializedPlaces<'_, 'tcx> {
 }
 
 impl<'tcx> AnalysisDomain<'tcx> for MaybeUninitializedPlaces<'_, 'tcx> {
-    type Domain = BitSet<MovePathIndex>;
+    type Domain = ChunkedBitSet<MovePathIndex>;
 
     const NAME: &'static str = "maybe_uninit";
 
     fn bottom_value(&self, _: &mir::Body<'tcx>) -> Self::Domain {
         // bottom = initialized (start_block_effect counters this at outset)
-        BitSet::new_empty(self.move_data().move_paths.len())
+        ChunkedBitSet::new_empty(self.move_data().move_paths.len())
     }
 
     // sets on_entry bits for Arg places
@@ -606,13 +606,13 @@ impl<'tcx> GenKillAnalysis<'tcx> for DefinitelyInitializedPlaces<'_, 'tcx> {
 }
 
 impl<'tcx> AnalysisDomain<'tcx> for EverInitializedPlaces<'_, 'tcx> {
-    type Domain = BitSet<InitIndex>;
+    type Domain = ChunkedBitSet<InitIndex>;
 
     const NAME: &'static str = "ever_init";
 
     fn bottom_value(&self, _: &mir::Body<'tcx>) -> Self::Domain {
         // bottom = no initialized variables by default
-        BitSet::new_empty(self.move_data().inits.len())
+        ChunkedBitSet::new_empty(self.move_data().inits.len())
     }
 
     fn initialize_start_block(&self, body: &mir::Body<'tcx>, state: &mut Self::Domain) {
diff --git a/compiler/rustc_mir_dataflow/src/rustc_peek.rs b/compiler/rustc_mir_dataflow/src/rustc_peek.rs
index 1746d5ee38b..51ab5b43bff 100644
--- a/compiler/rustc_mir_dataflow/src/rustc_peek.rs
+++ b/compiler/rustc_mir_dataflow/src/rustc_peek.rs
@@ -1,5 +1,3 @@
-use std::borrow::Borrow;
-
 use rustc_ast::ast;
 use rustc_span::symbol::sym;
 use rustc_span::Span;
@@ -10,6 +8,7 @@ use rustc_middle::mir::MirPass;
 use rustc_middle::mir::{self, Body, Local, Location};
 use rustc_middle::ty::{self, Ty, TyCtxt};
 
+use crate::framework::BitSetExt;
 use crate::impls::{
     DefinitelyInitializedPlaces, MaybeInitializedPlaces, MaybeLiveLocals, MaybeUninitializedPlaces,
 };
@@ -248,7 +247,7 @@ pub trait RustcPeekAt<'tcx>: Analysis<'tcx> {
 impl<'tcx, A, D> RustcPeekAt<'tcx> for A
 where
     A: Analysis<'tcx, Domain = D> + HasMoveData<'tcx>,
-    D: JoinSemiLattice + Clone + Borrow<BitSet<MovePathIndex>>,
+    D: JoinSemiLattice + Clone + BitSetExt<MovePathIndex>,
 {
     fn peek_at(
         &self,
@@ -259,7 +258,7 @@ where
     ) {
         match self.move_data().rev_lookup.find(place.as_ref()) {
             LookupResult::Exact(peek_mpi) => {
-                let bit_state = flow_state.borrow().contains(peek_mpi);
+                let bit_state = flow_state.contains(peek_mpi);
                 debug!("rustc_peek({:?} = &{:?}) bit_state: {}", call.arg, place, bit_state);
                 if !bit_state {
                     tcx.sess.span_err(call.span, "rustc_peek: bit not set");