1
Fork 0
rust/src/librustc_mir/dataflow/mod.rs

822 lines
31 KiB
Rust
Raw Normal View History

use syntax::ast::{self, MetaItem};
use rustc_data_structures::bit_set::{BitSet, BitSetOperator, HybridBitSet};
use rustc_data_structures::indexed_vec::Idx;
2018-07-03 00:18:54 +05:30
use rustc_data_structures::work_queue::WorkQueue;
use rustc::hir::HirId;
use rustc::ty::{self, TyCtxt};
use rustc::mir::{self, Mir, BasicBlock, BasicBlockData, Location, Statement, Terminator};
2018-07-05 09:06:04 +08:00
use rustc::mir::traversal;
use rustc::session::Session;
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
2018-03-05 02:44:10 -05:00
use std::borrow::Borrow;
use std::fmt;
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
use std::io;
use std::path::PathBuf;
use std::usize;
pub use self::impls::{MaybeStorageLive};
pub use self::impls::{MaybeInitializedPlaces, MaybeUninitializedPlaces};
2018-08-30 18:54:32 -03:00
pub use self::impls::DefinitelyInitializedPlaces;
pub use self::impls::EverInitializedPlaces;
pub use self::impls::borrows::Borrows;
pub use self::impls::HaveBeenBorrowedLocals;
pub use self::at_location::{FlowAtLocation, FlowsAtLocation};
pub(crate) use self::drop_flag_effects::*;
use self::move_paths::MoveData;
mod at_location;
2018-08-30 18:54:32 -03:00
pub mod drop_flag_effects;
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
mod graphviz;
mod impls;
pub mod move_paths;
pub(crate) use self::move_paths::indexes;
pub(crate) struct DataflowBuilder<'a, 'tcx: 'a, BD>
where
BD: BitDenotation<'tcx>
{
hir_id: HirId,
flow_state: DataflowAnalysis<'a, 'tcx, BD>,
print_preflow_to: Option<String>,
print_postflow_to: Option<String>,
}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// `DebugFormatted` encapsulates the "{:?}" rendering of some
/// arbitrary value. This way: you pay cost of allocating an extra
/// string (as well as that of rendering up-front); in exchange, you
/// don't have to hand over ownership of your value or deal with
/// borrowing it.
pub(crate) struct DebugFormatted(String);
impl DebugFormatted {
2018-02-23 10:15:26 -08:00
pub fn new(input: &dyn fmt::Debug) -> DebugFormatted {
DebugFormatted(format!("{:?}", input))
}
}
impl fmt::Debug for DebugFormatted {
2019-02-08 06:28:15 +09:00
fn fmt(&self, w: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(w, "{}", self.0)
}
}
pub(crate) trait Dataflow<'tcx, BD: BitDenotation<'tcx>> {
/// Sets up and runs the dataflow problem, using `p` to render results if
/// implementation so chooses.
fn dataflow<P>(&mut self, p: P) where P: Fn(&BD, BD::Idx) -> DebugFormatted {
let _ = p; // default implementation does not instrument process.
self.build_sets();
self.propagate();
}
/// Sets up the entry, gen, and kill sets for this instance of a dataflow problem.
fn build_sets(&mut self);
/// Finds a fixed-point solution to this instance of a dataflow problem.
fn propagate(&mut self);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
impl<'a, 'tcx: 'a, BD> Dataflow<'tcx, BD> for DataflowBuilder<'a, 'tcx, BD>
where
BD: BitDenotation<'tcx>
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
fn dataflow<P>(&mut self, p: P) where P: Fn(&BD, BD::Idx) -> DebugFormatted {
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
self.flow_state.build_sets();
self.pre_dataflow_instrumentation(|c,i| p(c,i)).unwrap();
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
self.flow_state.propagate();
self.post_dataflow_instrumentation(|c,i| p(c,i)).unwrap();
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
fn build_sets(&mut self) { self.flow_state.build_sets(); }
fn propagate(&mut self) { self.flow_state.propagate(); }
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
pub(crate) fn has_rustc_mir_with(attrs: &[ast::Attribute], name: &str) -> Option<MetaItem> {
for attr in attrs {
if attr.check_name("rustc_mir") {
let items = attr.meta_item_list();
for item in items.iter().flat_map(|l| l.iter()) {
match item.meta_item() {
Some(mi) if mi.check_name(name) => return Some(mi.clone()),
_ => continue
}
}
}
}
return None;
}
pub struct MoveDataParamEnv<'gcx, 'tcx> {
pub(crate) move_data: MoveData<'tcx>,
pub(crate) param_env: ty::ParamEnv<'gcx>,
}
pub(crate) fn do_dataflow<'a, 'gcx, 'tcx, BD, P>(tcx: TyCtxt<'a, 'gcx, 'tcx>,
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
mir: &'a Mir<'tcx>,
hir_id: HirId,
attributes: &[ast::Attribute],
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
dead_unwinds: &BitSet<BasicBlock>,
bd: BD,
p: P)
-> DataflowResults<'tcx, BD>
where BD: BitDenotation<'tcx> + InitialFlow,
P: Fn(&BD, BD::Idx) -> DebugFormatted
{
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
let flow_state = DataflowAnalysis::new(mir, dead_unwinds, bd);
flow_state.run(tcx, hir_id, attributes, p)
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
}
impl<'a, 'gcx: 'tcx, 'tcx: 'a, BD> DataflowAnalysis<'a, 'tcx, BD> where BD: BitDenotation<'tcx>
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
{
pub(crate) fn run<P>(self,
tcx: TyCtxt<'a, 'gcx, 'tcx>,
hir_id: HirId,
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
attributes: &[ast::Attribute],
p: P) -> DataflowResults<'tcx, BD>
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
where P: Fn(&BD, BD::Idx) -> DebugFormatted
{
let name_found = |sess: &Session, attrs: &[ast::Attribute], name| -> Option<String> {
if let Some(item) = has_rustc_mir_with(attrs, name) {
if let Some(s) = item.value_str() {
return Some(s.to_string())
} else {
sess.span_err(
item.span,
&format!("{} attribute requires a path", item.ident));
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
return None;
}
}
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
return None;
};
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
let print_preflow_to =
name_found(tcx.sess, attributes, "borrowck_graphviz_preflow");
let print_postflow_to =
name_found(tcx.sess, attributes, "borrowck_graphviz_postflow");
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
let mut mbcx = DataflowBuilder {
hir_id,
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
print_preflow_to, print_postflow_to, flow_state: self,
};
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
mbcx.dataflow(p);
mbcx.flow_state.results()
}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
struct PropagationContext<'b, 'a: 'b, 'tcx: 'a, O> where O: 'b + BitDenotation<'tcx>
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
builder: &'b mut DataflowAnalysis<'a, 'tcx, O>,
}
impl<'a, 'tcx: 'a, BD> DataflowAnalysis<'a, 'tcx, BD> where BD: BitDenotation<'tcx>
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
fn propagate(&mut self) {
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
let mut temp = BitSet::new_empty(self.flow_state.sets.bits_per_block);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
let mut propcx = PropagationContext {
builder: self,
};
2018-07-03 00:18:54 +05:30
propcx.walk_cfg(&mut temp);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
fn build_sets(&mut self) {
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
// First we need to build the entry-, gen- and kill-sets.
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
2016-09-19 23:50:00 +03:00
let sets = &mut self.flow_state.sets.for_block(mir::START_BLOCK.index());
self.flow_state.operator.start_block_effect(&mut sets.on_entry);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
2016-06-07 21:20:50 +03:00
for (bb, data) in self.mir.basic_blocks().iter_enumerated() {
2016-09-19 23:50:00 +03:00
let &mir::BasicBlockData { ref statements, ref terminator, is_cleanup: _ } = data;
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
let mut interim_state;
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
let sets = &mut self.flow_state.sets.for_block(bb.index());
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
let track_intrablock = BD::accumulates_intrablock_state();
if track_intrablock {
debug!("swapping in mutable on_entry, initially {:?}", sets.on_entry);
interim_state = sets.on_entry.to_owned();
sets.on_entry = &mut interim_state;
}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
for j_stmt in 0..statements.len() {
let location = Location { block: bb, statement_index: j_stmt };
2017-12-24 00:45:53 +02:00
self.flow_state.operator.before_statement_effect(sets, location);
self.flow_state.operator.statement_effect(sets, location);
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
if track_intrablock {
sets.apply_local_effect();
}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
if terminator.is_some() {
let location = Location { block: bb, statement_index: statements.len() };
2017-12-24 00:45:53 +02:00
self.flow_state.operator.before_terminator_effect(sets, location);
self.flow_state.operator.terminator_effect(sets, location);
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
if track_intrablock {
sets.apply_local_effect();
}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
}
}
}
impl<'b, 'a: 'b, 'tcx: 'a, BD> PropagationContext<'b, 'a, 'tcx, BD> where BD: BitDenotation<'tcx>
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
fn walk_cfg(&mut self, in_out: &mut BitSet<BD::Idx>) {
2018-07-03 00:18:54 +05:30
let mut dirty_queue: WorkQueue<mir::BasicBlock> =
WorkQueue::with_all(self.builder.mir.basic_blocks().len());
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
let mir = self.builder.mir;
2018-07-03 00:18:54 +05:30
while let Some(bb) = dirty_queue.pop() {
let bb_data = &mir[bb];
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
2018-07-03 00:18:54 +05:30
let sets = self.builder.flow_state.sets.for_block(bb.index());
debug_assert!(in_out.words().len() == sets.on_entry.words().len());
in_out.overwrite(sets.on_entry);
in_out.union(sets.gen_set);
in_out.subtract(sets.kill_set);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
2018-07-03 00:18:54 +05:30
self.builder.propagate_bits_into_graph_successors_of(
in_out, (bb, bb_data), &mut dirty_queue);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
}
}
fn dataflow_path(context: &str, path: &str) -> PathBuf {
let mut path = PathBuf::from(path);
let new_file_name = {
let orig_file_name = path.file_name().unwrap().to_str().unwrap();
format!("{}_{}", context, orig_file_name)
};
path.set_file_name(new_file_name);
path
}
impl<'a, 'tcx: 'a, BD> DataflowBuilder<'a, 'tcx, BD> where BD: BitDenotation<'tcx>
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
fn pre_dataflow_instrumentation<P>(&self, p: P) -> io::Result<()>
where P: Fn(&BD, BD::Idx) -> DebugFormatted
{
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
if let Some(ref path_str) = self.print_preflow_to {
let path = dataflow_path(BD::name(), path_str);
graphviz::print_borrowck_graph_to(self, &path, p)
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
} else {
Ok(())
}
}
fn post_dataflow_instrumentation<P>(&self, p: P) -> io::Result<()>
where P: Fn(&BD, BD::Idx) -> DebugFormatted
{
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
if let Some(ref path_str) = self.print_postflow_to {
let path = dataflow_path(BD::name(), path_str);
graphviz::print_borrowck_graph_to(self, &path, p)
} else {
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
Ok(())
}
}
}
/// DataflowResultsConsumer abstracts over walking the MIR with some
/// already constructed dataflow results.
///
/// It abstracts over the FlowState and also completely hides the
/// underlying flow analysis results, because it needs to handle cases
/// where we are combining the results of *multiple* flow analyses
/// (e.g., borrows + inits + uninits).
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
pub(crate) trait DataflowResultsConsumer<'a, 'tcx: 'a> {
type FlowState: FlowsAtLocation;
// Observation Hooks: override (at least one of) these to get analysis feedback.
fn visit_block_entry(&mut self,
_bb: BasicBlock,
_flow_state: &Self::FlowState) {}
fn visit_statement_entry(&mut self,
_loc: Location,
_stmt: &Statement<'tcx>,
_flow_state: &Self::FlowState) {}
fn visit_terminator_entry(&mut self,
_loc: Location,
_term: &Terminator<'tcx>,
_flow_state: &Self::FlowState) {}
// Main entry point: this drives the processing of results.
fn analyze_results(&mut self, flow_uninit: &mut Self::FlowState) {
let flow = flow_uninit;
2018-07-05 09:06:04 +08:00
for (bb, _) in traversal::reverse_postorder(self.mir()) {
flow.reset_to_entry_of(bb);
self.process_basic_block(bb, flow);
}
}
fn process_basic_block(&mut self, bb: BasicBlock, flow_state: &mut Self::FlowState) {
let BasicBlockData { ref statements, ref terminator, is_cleanup: _ } =
self.mir()[bb];
let mut location = Location { block: bb, statement_index: 0 };
for stmt in statements.iter() {
flow_state.reconstruct_statement_effect(location);
self.visit_statement_entry(location, stmt, flow_state);
flow_state.apply_local_effect(location);
location.statement_index += 1;
}
if let Some(ref term) = *terminator {
flow_state.reconstruct_terminator_effect(location);
self.visit_terminator_entry(location, term, flow_state);
// We don't need to apply the effect of the terminator,
// since we are only visiting dataflow state on control
// flow entry to the various nodes. (But we still need to
// reconstruct the effect, because the visit method might
// inspect it.)
}
}
// Delegated Hooks: Provide access to the MIR and process the flow state.
fn mir(&self) -> &'a Mir<'tcx>;
}
pub fn state_for_location<'tcx, T: BitDenotation<'tcx>>(loc: Location,
analysis: &T,
result: &DataflowResults<'tcx, T>,
mir: &Mir<'tcx>)
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
-> BitSet<T::Idx> {
let mut on_entry = result.sets().on_entry_set_for(loc.block.index()).to_owned();
let mut kill_set = on_entry.to_hybrid();
let mut gen_set = kill_set.clone();
{
let mut sets = BlockSets {
on_entry: &mut on_entry,
kill_set: &mut kill_set,
gen_set: &mut gen_set,
};
for stmt in 0..loc.statement_index {
let mut stmt_loc = loc;
stmt_loc.statement_index = stmt;
2017-12-24 00:45:53 +02:00
analysis.before_statement_effect(&mut sets, stmt_loc);
analysis.statement_effect(&mut sets, stmt_loc);
}
2017-12-24 00:45:53 +02:00
// Apply the pre-statement effect of the statement we're evaluating.
if loc.statement_index == mir[loc.block].statements.len() {
analysis.before_terminator_effect(&mut sets, loc);
} else {
analysis.before_statement_effect(&mut sets, loc);
}
}
gen_set.to_dense()
}
pub struct DataflowAnalysis<'a, 'tcx: 'a, O> where O: BitDenotation<'tcx>
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
flow_state: DataflowState<'tcx, O>,
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
dead_unwinds: &'a BitSet<mir::BasicBlock>,
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
mir: &'a Mir<'tcx>,
}
impl<'a, 'tcx: 'a, O> DataflowAnalysis<'a, 'tcx, O> where O: BitDenotation<'tcx>
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
pub fn results(self) -> DataflowResults<'tcx, O> {
DataflowResults(self.flow_state)
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
pub fn mir(&self) -> &'a Mir<'tcx> { self.mir }
}
pub struct DataflowResults<'tcx, O>(pub(crate) DataflowState<'tcx, O>) where O: BitDenotation<'tcx>;
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
impl<'tcx, O: BitDenotation<'tcx>> DataflowResults<'tcx, O> {
2016-05-27 15:07:08 +03:00
pub fn sets(&self) -> &AllSets<O::Idx> {
&self.0.sets
}
pub fn operator(&self) -> &O {
&self.0.operator
}
}
/// State of a dataflow analysis; couples a collection of bit sets
/// with operator used to initialize and merge bits during analysis.
pub struct DataflowState<'tcx, O: BitDenotation<'tcx>>
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
/// All the sets for the analysis. (Factored into its
/// own structure so that we can borrow it mutably
/// on its own separate from other fields.)
pub sets: AllSets<O::Idx>,
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// operator used to initialize, combine, and interpret bits.
pub(crate) operator: O,
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
impl<'tcx, O: BitDenotation<'tcx>> DataflowState<'tcx, O> {
pub(crate) fn interpret_set<'c, P>(&self,
o: &'c O,
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
set: &BitSet<O::Idx>,
render_idx: &P)
-> Vec<DebugFormatted>
where P: Fn(&O, O::Idx) -> DebugFormatted
{
set.iter().map(|i| render_idx(o, i)).collect()
}
pub(crate) fn interpret_hybrid_set<'c, P>(&self,
o: &'c O,
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
set: &HybridBitSet<O::Idx>,
render_idx: &P)
-> Vec<DebugFormatted>
where P: Fn(&O, O::Idx) -> DebugFormatted
{
set.iter().map(|i| render_idx(o, i)).collect()
}
}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
#[derive(Debug)]
pub struct AllSets<E: Idx> {
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// Analysis bitwidth for each block.
bits_per_block: usize,
/// For each block, bits valid on entry to the block.
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
on_entry_sets: Vec<BitSet<E>>,
/// For each block, bits generated by executing the statements +
/// terminator in the block -- with one caveat. In particular, for
/// *call terminators*, the effect of storing the destination is
/// not included, since that only takes effect on the **success**
/// edge (and not the unwind edge).
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
gen_sets: Vec<HybridBitSet<E>>,
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// For each block, bits killed by executing the statements +
/// terminator in the block -- with one caveat. In particular, for
/// *call terminators*, the effect of storing the destination is
/// not included, since that only takes effect on the **success**
/// edge (and not the unwind edge).
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
kill_sets: Vec<HybridBitSet<E>>,
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
/// Triple of sets associated with a given block.
///
/// Generally, one sets up `on_entry`, `gen_set`, and `kill_set` for
/// each block individually, and then runs the dataflow analysis which
/// iteratively modifies the various `on_entry` sets (but leaves the
/// other two sets unchanged, since they represent the effect of the
/// block, which should be invariant over the course of the analysis).
///
/// It is best to ensure that the intersection of `gen_set` and
/// `kill_set` is empty; otherwise the results of the dataflow will
/// have a hidden dependency on what order the bits are generated and
/// killed during the iteration. (This is such a good idea that the
/// `fn gen` and `fn kill` methods that set their state enforce this
/// for you.)
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
#[derive(Debug)]
pub struct BlockSets<'a, E: Idx> {
/// Dataflow state immediately before control flow enters the given block.
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
pub(crate) on_entry: &'a mut BitSet<E>,
/// Bits that are set to 1 by the time we exit the given block. Hybrid
/// because it usually contains only 0 or 1 elements.
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
pub(crate) gen_set: &'a mut HybridBitSet<E>,
/// Bits that are set to 0 by the time we exit the given block. Hybrid
/// because it usually contains only 0 or 1 elements.
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
pub(crate) kill_set: &'a mut HybridBitSet<E>,
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
impl<'a, E:Idx> BlockSets<'a, E> {
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
fn gen(&mut self, e: E) {
self.gen_set.insert(e);
self.kill_set.remove(e);
}
2017-12-04 01:00:46 +02:00
fn gen_all<I>(&mut self, i: I)
where I: IntoIterator,
I::Item: Borrow<E>
{
for j in i {
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
self.gen(*j.borrow());
2017-12-04 01:00:46 +02:00
}
}
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
fn kill(&mut self, e: E) {
self.gen_set.remove(e);
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
self.kill_set.insert(e);
}
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
2017-12-04 01:00:46 +02:00
fn kill_all<I>(&mut self, i: I)
where I: IntoIterator,
I::Item: Borrow<E>
{
for j in i {
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
self.kill(*j.borrow());
2017-12-04 01:00:46 +02:00
}
}
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
fn apply_local_effect(&mut self) {
self.on_entry.union(self.gen_set);
self.on_entry.subtract(self.kill_set);
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
}
}
impl<E:Idx> AllSets<E> {
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
pub fn bits_per_block(&self) -> usize { self.bits_per_block }
2019-02-08 06:28:15 +09:00
pub fn for_block(&mut self, block_idx: usize) -> BlockSets<'_, E> {
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
BlockSets {
on_entry: &mut self.on_entry_sets[block_idx],
gen_set: &mut self.gen_sets[block_idx],
kill_set: &mut self.kill_sets[block_idx],
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
}
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
pub fn on_entry_set_for(&self, block_idx: usize) -> &BitSet<E> {
&self.on_entry_sets[block_idx]
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
pub fn gen_set_for(&self, block_idx: usize) -> &HybridBitSet<E> {
&self.gen_sets[block_idx]
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
pub fn kill_set_for(&self, block_idx: usize) -> &HybridBitSet<E> {
&self.kill_sets[block_idx]
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
}
/// Parameterization for the precise form of data flow that is used.
/// `InitialFlow` handles initializing the bitvectors before any
/// code is inspected by the analysis. Analyses that need more nuanced
/// initialization (e.g., they need to consult the results of some other
/// dataflow analysis to set up the initial bitvectors) should not
/// implement this.
pub trait InitialFlow {
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// Specifies the initial value for each bit in the `on_entry` set
fn bottom_value() -> bool;
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
pub trait BitDenotation<'tcx>: BitSetOperator {
/// Specifies what index type is used to access the bitvector.
type Idx: Idx;
/// Some analyses want to accumulate knowledge within a block when
/// analyzing its statements for building the gen/kill sets. Override
/// this method to return true in such cases.
///
/// When this returns true, the statement-effect (re)construction
/// will clone the `on_entry` state and pass along a reference via
/// `sets.on_entry` to that local clone into `statement_effect` and
/// `terminator_effect`).
///
2018-08-19 15:30:23 +02:00
/// When it's false, no local clone is constructed; instead a
/// reference directly into `on_entry` is passed along via
/// `sets.on_entry` instead, which represents the flow state at
/// the block's start, not necessarily the state immediately prior
/// to the statement/terminator under analysis.
///
2019-02-08 14:53:55 +01:00
/// In either case, the passed reference is mutable, but this is a
/// wart from using the `BlockSets` type in the API; the intention
/// is that the `statement_effect` and `terminator_effect` methods
/// mutate only the gen/kill sets.
2019-02-08 14:53:55 +01:00
//
// FIXME: we should consider enforcing the intention described in
// the previous paragraph by passing the three sets in separate
// parameters to encode their distinct mutabilities.
fn accumulates_intrablock_state() -> bool { false }
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// A name describing the dataflow analysis that this
2019-02-08 14:53:55 +01:00
/// `BitDenotation` is supporting. The name should be something
/// suitable for plugging in as part of a filename (i.e., avoid
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// space-characters or other things that tend to look bad on a
2019-02-08 14:53:55 +01:00
/// file system, like slashes or periods). It is also better for
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// the name to be reasonably short, again because it will be
/// plugged into a filename.
fn name() -> &'static str;
/// Size of each bitvector allocated for each block in the analysis.
fn bits_per_block(&self) -> usize;
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// Mutates the entry set according to the effects that
/// have been established *prior* to entering the start
/// block. This can't access the gen/kill sets, because
/// these won't be accounted for correctly.
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
///
/// (For example, establishing the call arguments.)
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
fn start_block_effect(&self, entry_set: &mut BitSet<Self::Idx>);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
2017-12-24 00:45:53 +02:00
/// Similar to `statement_effect`, except it applies
/// *just before* the statement rather than *just after* it.
///
/// This matters for "dataflow at location" APIs, because the
/// before-statement effect is visible while visiting the
/// statement, while the after-statement effect only becomes
/// visible at the next statement.
///
/// Both the before-statement and after-statement effects are
/// applied, in that order, before moving for the next
/// statement.
fn before_statement_effect(&self,
2019-02-08 06:28:15 +09:00
_sets: &mut BlockSets<'_, Self::Idx>,
2017-12-24 00:45:53 +02:00
_location: Location) {}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// Mutates the block-sets (the flow sets for the given
/// basic block) according to the effects of evaluating statement.
///
/// This is used, in particular, for building up the
2017-08-11 20:34:14 +02:00
/// "transfer-function" representing the overall-effect of the
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// block, represented via GEN and KILL sets.
///
2016-05-25 14:52:40 +02:00
/// The statement is identified as `bb_data[idx_stmt]`, where
2017-08-11 20:34:14 +02:00
/// `bb_data` is the sequence of statements identified by `bb` in
2016-05-25 14:52:40 +02:00
/// the MIR.
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
fn statement_effect(&self,
2019-02-08 06:28:15 +09:00
sets: &mut BlockSets<'_, Self::Idx>,
location: Location);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
2017-12-24 00:45:53 +02:00
/// Similar to `terminator_effect`, except it applies
/// *just before* the terminator rather than *just after* it.
///
/// This matters for "dataflow at location" APIs, because the
/// before-terminator effect is visible while visiting the
/// terminator, while the after-terminator effect only becomes
/// visible at the terminator's successors.
///
/// Both the before-terminator and after-terminator effects are
/// applied, in that order, before moving for the next
/// terminator.
fn before_terminator_effect(&self,
2019-02-08 06:28:15 +09:00
_sets: &mut BlockSets<'_, Self::Idx>,
2017-12-24 00:45:53 +02:00
_location: Location) {}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// Mutates the block-sets (the flow sets for the given
/// basic block) according to the effects of evaluating
/// the terminator.
///
/// This is used, in particular, for building up the
2017-08-11 20:34:14 +02:00
/// "transfer-function" representing the overall-effect of the
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// block, represented via GEN and KILL sets.
///
/// The effects applied here cannot depend on which branch the
/// terminator took.
fn terminator_effect(&self,
2019-02-08 06:28:15 +09:00
sets: &mut BlockSets<'_, Self::Idx>,
location: Location);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// Mutates the block-sets according to the (flow-dependent)
/// effect of a successful return from a Call terminator.
///
/// If basic-block BB_x ends with a call-instruction that, upon
/// successful return, flows to BB_y, then this method will be
/// called on the exit flow-state of BB_x in order to set up the
/// entry flow-state of BB_y.
///
/// This is used, in particular, as a special case during the
/// "propagate" loop where all of the basic blocks are repeatedly
/// visited. Since the effects of a Call terminator are
/// flow-dependent, the current MIR cannot encode them via just
/// GEN and KILL sets attached to the block, and so instead we add
/// this extra machinery to represent the flow-dependent effect.
2019-02-08 14:53:55 +01:00
//
// FIXME: right now this is a bit of a wart in the API. It might
// be better to represent this as an additional gen- and
// kill-sets associated with each edge coming out of the basic
// block.
fn propagate_call_return(
&self,
in_out: &mut BitSet<Self::Idx>,
call_bb: mir::BasicBlock,
dest_bb: mir::BasicBlock,
dest_place: &mir::Place<'tcx>,
);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
impl<'a, 'tcx, D> DataflowAnalysis<'a, 'tcx, D> where D: BitDenotation<'tcx>
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
pub fn new(mir: &'a Mir<'tcx>,
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
dead_unwinds: &'a BitSet<mir::BasicBlock>,
denotation: D) -> Self where D: InitialFlow {
let bits_per_block = denotation.bits_per_block();
let num_blocks = mir.basic_blocks().len();
2018-03-05 02:44:10 -05:00
let on_entry_sets = if D::bottom_value() {
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
vec![BitSet::new_filled(bits_per_block); num_blocks]
} else {
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
vec![BitSet::new_empty(bits_per_block); num_blocks]
};
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
let gen_sets = vec![HybridBitSet::new_empty(bits_per_block); num_blocks];
let kill_sets = gen_sets.clone();
DataflowAnalysis {
mir,
dead_unwinds,
flow_state: DataflowState {
sets: AllSets {
bits_per_block,
on_entry_sets,
gen_sets,
kill_sets,
},
operator: denotation,
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
New `ActiveBorrows` dataflow for two-phase `&mut`; not yet borrowed-checked. High-level picture: The old `Borrows` analysis is now called `Reservations` (implemented as a newtype wrapper around `Borrows`); this continues to compute whether a `Rvalue::Ref` can reach a statement without an intervening `EndRegion`. In addition, we also track what `Place` each such `Rvalue::Ref` was immediately assigned to in a given borrow (yay for MIR-structural properties!). The new `ActiveBorrows` analysis then tracks the initial use of any of those assigned `Places` for a given borrow. I.e. a borrow becomes "active" immediately after it starts being "used" in some way. (This is conservative in the sense that we will treat a copy `x = y;` as a use of `y`; in principle one might further delay activation in such cases.) The new `ActiveBorrows` analysis needs to take the `Reservations` results as an initial input, because the reservation state influences the gen/kill sets for `ActiveBorrows`. In particular, a use of `a` activates a borrow `a = &b` if and only if there exists a path (in the control flow graph) from the borrow to that use. So we need to know if the borrow reaches a given use to know if it really gets a gen-bit or not. * Incorporating the output from one dataflow analysis into the input of another required more changes to the infrastructure than I had expected, and even after those changes, the resulting code is still a bit subtle. * In particular, Since we need to know the intrablock reservation state, we need to dynamically update a bitvector for the reservations as we are also trying to compute the gen/kills bitvector for the active borrows. * The way I ended up deciding to do this (after also toying with at least two other designs) is to put both the reservation state and the active borrow state into a single bitvector. That is why we now have separate (but related) `BorrowIndex` and `ReserveOrActivateIndex`: each borrow index maps to a pair of neighboring reservation and activation indexes. As noted above, these changes are solely adding the active borrows dataflow analysis (and updating the existing code to cope with the switch from `Borrows` to `Reservations`). The code to process the bitvector in the borrow checker currently just skips over all of the active borrow bits. But atop this commit, one *can* observe the analysis results by looking at the graphviz output, e.g. via ```rust #[rustc_mir(borrowck_graphviz_preflow="pre_two_phase.dot", borrowck_graphviz_postflow="post_two_phase.dot")] ``` Includes doc for `FindPlaceUses`, as well as `Reservations` and `ActiveBorrows` structs, which are wrappers are the `Borrows` struct that dictate which flow analysis should be performed.
2017-12-01 12:32:51 +01:00
}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
impl<'a, 'tcx: 'a, D> DataflowAnalysis<'a, 'tcx, D> where D: BitDenotation<'tcx> {
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// Propagates the bits of `in_out` into all the successors of `bb`,
/// using bitwise operator denoted by `self.operator`.
///
/// For most blocks, this is entirely uniform. However, for blocks
/// that end with a call terminator, the effect of the call on the
/// dataflow state may depend on whether the call returned
/// successfully or unwound.
///
/// To reflect this, the `propagate_call_return` method of the
/// `BitDenotation` mutates `in_out` when propagating `in_out` via
/// a call terminator; such mutation is performed *last*, to
/// ensure its side-effects do not leak elsewhere (e.g., into
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
/// unwind target).
fn propagate_bits_into_graph_successors_of(
&mut self,
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
in_out: &mut BitSet<D::Idx>,
(bb, bb_data): (mir::BasicBlock, &mir::BasicBlockData<'tcx>),
2018-07-03 00:18:54 +05:30
dirty_list: &mut WorkQueue<mir::BasicBlock>)
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
{
match bb_data.terminator().kind {
2016-09-19 23:50:00 +03:00
mir::TerminatorKind::Return |
mir::TerminatorKind::Resume |
mir::TerminatorKind::Abort |
2016-12-26 14:34:03 +01:00
mir::TerminatorKind::GeneratorDrop |
2016-09-19 23:50:00 +03:00
mir::TerminatorKind::Unreachable => {}
2018-07-03 00:18:54 +05:30
mir::TerminatorKind::Goto { target } |
mir::TerminatorKind::Assert { target, cleanup: None, .. } |
mir::TerminatorKind::Yield { resume: target, drop: None, .. } |
mir::TerminatorKind::Drop { target, location: _, unwind: None } |
2016-09-19 23:50:00 +03:00
mir::TerminatorKind::DropAndReplace {
2018-07-03 00:18:54 +05:30
target, value: _, location: _, unwind: None
} => {
2018-07-03 00:18:54 +05:30
self.propagate_bits_into_entry_set_for(in_out, target, dirty_list);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
2018-07-03 00:18:54 +05:30
mir::TerminatorKind::Yield { resume: target, drop: Some(drop), .. } => {
self.propagate_bits_into_entry_set_for(in_out, target, dirty_list);
self.propagate_bits_into_entry_set_for(in_out, drop, dirty_list);
2016-12-26 14:34:03 +01:00
}
2018-07-03 00:18:54 +05:30
mir::TerminatorKind::Assert { target, cleanup: Some(unwind), .. } |
mir::TerminatorKind::Drop { target, location: _, unwind: Some(unwind) } |
2016-09-19 23:50:00 +03:00
mir::TerminatorKind::DropAndReplace {
2018-07-03 00:18:54 +05:30
target, value: _, location: _, unwind: Some(unwind)
} => {
2018-07-03 00:18:54 +05:30
self.propagate_bits_into_entry_set_for(in_out, target, dirty_list);
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
if !self.dead_unwinds.contains(bb) {
2018-07-03 00:18:54 +05:30
self.propagate_bits_into_entry_set_for(in_out, unwind, dirty_list);
}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
2016-09-19 23:50:00 +03:00
mir::TerminatorKind::SwitchInt { ref targets, .. } => {
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
for target in targets {
2018-07-03 00:18:54 +05:30
self.propagate_bits_into_entry_set_for(in_out, *target, dirty_list);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
}
mir::TerminatorKind::Call { cleanup, ref destination, .. } => {
2018-07-03 00:18:54 +05:30
if let Some(unwind) = cleanup {
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
if !self.dead_unwinds.contains(bb) {
2018-07-03 00:18:54 +05:30
self.propagate_bits_into_entry_set_for(in_out, unwind, dirty_list);
}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
2018-07-03 00:18:54 +05:30
if let Some((ref dest_place, dest_bb)) = *destination {
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
// N.B.: This must be done *last*, after all other
// propagation, as documented in comment above.
self.flow_state.operator.propagate_call_return(
2018-07-03 00:18:54 +05:30
in_out, bb, dest_bb, dest_place);
self.propagate_bits_into_entry_set_for(in_out, dest_bb, dirty_list);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
}
2018-07-03 00:18:54 +05:30
mir::TerminatorKind::FalseEdges { real_target, ref imaginary_targets } => {
self.propagate_bits_into_entry_set_for(in_out, real_target, dirty_list);
for target in imaginary_targets {
2018-07-03 00:18:54 +05:30
self.propagate_bits_into_entry_set_for(in_out, *target, dirty_list);
}
}
2018-07-03 00:18:54 +05:30
mir::TerminatorKind::FalseUnwind { real_target, unwind } => {
self.propagate_bits_into_entry_set_for(in_out, real_target, dirty_list);
if let Some(unwind) = unwind {
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
if !self.dead_unwinds.contains(bb) {
2018-07-03 00:18:54 +05:30
self.propagate_bits_into_entry_set_for(in_out, unwind, dirty_list);
}
}
}
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
}
fn propagate_bits_into_entry_set_for(&mut self,
Merge indexed_set.rs into bitvec.rs, and rename it bit_set.rs. Currently we have two files implementing bitsets (and 2D bit matrices). This commit combines them into one, taking the best features from each. This involves renaming a lot of things. The high level changes are as follows. - bitvec.rs --> bit_set.rs - indexed_set.rs --> (removed) - BitArray + IdxSet --> BitSet (merged, see below) - BitVector --> GrowableBitSet - {,Sparse,Hybrid}IdxSet --> {,Sparse,Hybrid}BitSet - BitMatrix --> BitMatrix - SparseBitMatrix --> SparseBitMatrix The changes within the bitset types themselves are as follows. ``` OLD OLD NEW BitArray<C> IdxSet<T> BitSet<T> -------- ------ ------ grow - grow new - (remove) new_empty new_empty new_empty new_filled new_filled new_filled - to_hybrid to_hybrid clear clear clear set_up_to set_up_to set_up_to clear_above - clear_above count - count contains(T) contains(&T) contains(T) contains_all - superset is_empty - is_empty insert(T) add(&T) insert(T) insert_all - insert_all() remove(T) remove(&T) remove(T) words words words words_mut words_mut words_mut - overwrite overwrite merge union union - subtract subtract - intersect intersect iter iter iter ``` In general, when choosing names I went with: - names that are more obvious (e.g. `BitSet` over `IdxSet`). - names that are more like the Rust libraries (e.g. `T` over `C`, `insert` over `add`); - names that are more set-like (e.g. `union` over `merge`, `superset` over `contains_all`, `domain_size` over `num_bits`). Also, using `T` for index arguments seems more sensible than `&T` -- even though the latter is standard in Rust collection types -- because indices are always copyable. It also results in fewer `&` and `*` sigils in practice.
2018-09-14 15:07:25 +10:00
in_out: &BitSet<D::Idx>,
2018-07-03 00:18:54 +05:30
bb: mir::BasicBlock,
dirty_queue: &mut WorkQueue<mir::BasicBlock>) {
let entry_set = &mut self.flow_state.sets.for_block(bb.index()).on_entry;
let set_changed = self.flow_state.operator.join(entry_set, &in_out);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
if set_changed {
2018-07-03 00:18:54 +05:30
dirty_queue.insert(bb);
Revised mir-dataflow. Incorporates many fixes contributed by arielb1. ---- revise borrowck::mir::dataflow code to allow varying domain for bitvectors. This particular code implements the `BitDenotation` trait for three analyses: * `MovingOutStatements`, which, like `borrowck::move_data`, maps each bit-index to a move instruction, and a 1 means "the effect of this move reaches this point" (and the assigned l-value, if a scoped declaration, is still in scope). * `MaybeInitializedLvals`, which maps each bit-index to an l-value. A 1 means "there exists a control flow path to this point that initializes the associated l-value." * `MaybeUninitializedLvals`, which maps each bit-index to an l-value A 1 means "there exists a control flow path to this point that de-initializes the associated l-value." ---- Revised `graphviz` dataflow-rendering support in `borrowck::mir`. One big difference is that this code is now parameterized over the `BitDenotation`, so that it can be used to render dataflow results independent of how the dataflow bitvectors are interpreted; see where reference to `MoveOut` is replaced by the type parameter `D`. ---- Factor out routine to query subattributes in `#[rustc_mir(..)]`. (Later commits build upon this for some unit testing and instrumentation.) ---- thread through a tcx so that I can query types of lvalues as part of analysis. ---- Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`. The main motivation is to ease threading through a `TyCtxt`. (In hindsight it might have been better to instead attach the `TyCtxt` to each of the different dataflow implementations, but that would require e.g. switching away from having a `Default` impl, so I am leaving that experiment for another time.)
2016-05-02 15:50:27 +02:00
}
}
}