Auto merge of #122568 - RalfJung:mentioned-items, r=oli-obk
recursively evaluate the constants in everything that is 'mentioned' This is another attempt at fixing https://github.com/rust-lang/rust/issues/107503. The previous attempt at https://github.com/rust-lang/rust/pull/112879 seems stuck in figuring out where the [perf regression](https://perf.rust-lang.org/compare.html?start=c55d1ee8d4e3162187214692229a63c2cc5e0f31&end=ec8de1ebe0d698b109beeaaac83e60f4ef8bb7d1&stat=instructions:u) comes from. In https://github.com/rust-lang/rust/pull/122258 I learned some things, which informed the approach this PR is taking. Quoting from the new collector docs, which explain the high-level idea: ```rust //! One important role of collection is to evaluate all constants that are used by all the items //! which are being collected. Codegen can then rely on only encountering constants that evaluate //! successfully, and if a constant fails to evaluate, the collector has much better context to be //! able to show where this constant comes up. //! //! However, the exact set of "used" items (collected as described above), and therefore the exact //! set of used constants, can depend on optimizations. Optimizing away dead code may optimize away //! a function call that uses a failing constant, so an unoptimized build may fail where an //! optimized build succeeds. This is undesirable. //! //! To fix this, the collector has the concept of "mentioned" items. Some time during the MIR //! pipeline, before any optimization-level-dependent optimizations, we compute a list of all items //! that syntactically appear in the code. These are considered "mentioned", and even if they are in //! dead code and get optimized away (which makes them no longer "used"), they are still //! "mentioned". For every used item, the collector ensures that all mentioned items, recursively, //! do not use a failing constant. This is reflected via the [`CollectionMode`], which determines //! whether we are visiting a used item or merely a mentioned item. //! //! The collector and "mentioned items" gathering (which lives in `rustc_mir_transform::mentioned_items`) //! need to stay in sync in the following sense: //! //! - For every item that the collector gather that could eventually lead to build failure (most //! likely due to containing a constant that fails to evaluate), a corresponding mentioned item //! must be added. This should use the exact same strategy as the ecollector to make sure they are //! in sync. However, while the collector works on monomorphized types, mentioned items are //! collected on generic MIR -- so any time the collector checks for a particular type (such as //! `ty::FnDef`), we have to just onconditionally add this as a mentioned item. //! - In `visit_mentioned_item`, we then do with that mentioned item exactly what the collector //! would have done during regular MIR visiting. Basically you can think of the collector having //! two stages, a pre-monomorphization stage and a post-monomorphization stage (usually quite //! literally separated by a call to `self.monomorphize`); the pre-monomorphizationn stage is //! duplicated in mentioned items gathering and the post-monomorphization stage is duplicated in //! `visit_mentioned_item`. //! - Finally, as a performance optimization, the collector should fill `used_mentioned_item` during //! its MIR traversal with exactly what mentioned item gathering would have added in the same //! situation. This detects mentioned items that have *not* been optimized away and hence don't //! need a dedicated traversal. enum CollectionMode { /// Collect items that are used, i.e., actually needed for codegen. /// /// Which items are used can depend on optimization levels, as MIR optimizations can remove /// uses. UsedItems, /// Collect items that are mentioned. The goal of this mode is that it is independent of /// optimizations: the set of "mentioned" items is computed before optimizations are run. /// /// The exact contents of this set are *not* a stable guarantee. (For instance, it is currently /// computed after drop-elaboration. If we ever do some optimizations even in debug builds, we /// might decide to run them before computing mentioned items.) The key property of this set is /// that it is optimization-independent. MentionedItems, } ``` And the `mentioned_items` MIR body field docs: ```rust /// Further items that were mentioned in this function and hence *may* become monomorphized, /// depending on optimizations. We use this to avoid optimization-dependent compile errors: the /// collector recursively traverses all "mentioned" items and evaluates all their /// `required_consts`. /// /// This is *not* soundness-critical and the contents of this list are *not* a stable guarantee. /// All that's relevant is that this set is optimization-level-independent, and that it includes /// everything that the collector would consider "used". (For example, we currently compute this /// set after drop elaboration, so some drop calls that can never be reached are not considered /// "mentioned".) See the documentation of `CollectionMode` in /// `compiler/rustc_monomorphize/src/collector.rs` for more context. pub mentioned_items: Vec<Spanned<MentionedItem<'tcx>>>, ``` Fixes #107503
This commit is contained in:
commit
df8ac8f1d7
52 changed files with 1376 additions and 386 deletions
|
@ -565,7 +565,8 @@ impl<'tcx> Inliner<'tcx> {
|
|||
mut callee_body: Body<'tcx>,
|
||||
) {
|
||||
let terminator = caller_body[callsite.block].terminator.take().unwrap();
|
||||
let TerminatorKind::Call { args, destination, unwind, target, .. } = terminator.kind else {
|
||||
let TerminatorKind::Call { func, args, destination, unwind, target, .. } = terminator.kind
|
||||
else {
|
||||
bug!("unexpected terminator kind {:?}", terminator.kind);
|
||||
};
|
||||
|
||||
|
@ -717,6 +718,24 @@ impl<'tcx> Inliner<'tcx> {
|
|||
Const::Val(..) | Const::Unevaluated(..) => true,
|
||||
},
|
||||
));
|
||||
// Now that we incorporated the callee's `required_consts`, we can remove the callee from
|
||||
// `mentioned_items` -- but we have to take their `mentioned_items` in return. This does
|
||||
// some extra work here to save the monomorphization collector work later. It helps a lot,
|
||||
// since monomorphization can avoid a lot of work when the "mentioned items" are similar to
|
||||
// the actually used items. By doing this we can entirely avoid visiting the callee!
|
||||
// We need to reconstruct the `required_item` for the callee so that we can find and
|
||||
// remove it.
|
||||
let callee_item = MentionedItem::Fn(func.ty(caller_body, self.tcx));
|
||||
if let Some(idx) =
|
||||
caller_body.mentioned_items.iter().position(|item| item.node == callee_item)
|
||||
{
|
||||
// We found the callee, so remove it and add its items instead.
|
||||
caller_body.mentioned_items.remove(idx);
|
||||
caller_body.mentioned_items.extend(callee_body.mentioned_items);
|
||||
} else {
|
||||
// If we can't find the callee, there's no point in adding its items.
|
||||
// Probably it already got removed by being inlined elsewhere in the same function.
|
||||
}
|
||||
}
|
||||
|
||||
fn make_call_args(
|
||||
|
|
|
@ -88,6 +88,7 @@ mod lint;
|
|||
mod lower_intrinsics;
|
||||
mod lower_slice_len;
|
||||
mod match_branches;
|
||||
mod mentioned_items;
|
||||
mod multiple_return_terminators;
|
||||
mod normalize_array_len;
|
||||
mod nrvo;
|
||||
|
@ -565,6 +566,10 @@ fn run_optimization_passes<'tcx>(tcx: TyCtxt<'tcx>, body: &mut Body<'tcx>) {
|
|||
tcx,
|
||||
body,
|
||||
&[
|
||||
// Before doing anything, remember which items are being mentioned so that the set of items
|
||||
// visited does not depend on the optimization level.
|
||||
&mentioned_items::MentionedItems,
|
||||
// Add some UB checks before any UB gets optimized away.
|
||||
&check_alignment::CheckAlignment,
|
||||
// Before inlining: trim down MIR with passes to reduce inlining work.
|
||||
|
||||
|
|
117
compiler/rustc_mir_transform/src/mentioned_items.rs
Normal file
117
compiler/rustc_mir_transform/src/mentioned_items.rs
Normal file
|
@ -0,0 +1,117 @@
|
|||
use rustc_middle::mir::visit::Visitor;
|
||||
use rustc_middle::mir::{self, Location, MentionedItem, MirPass};
|
||||
use rustc_middle::ty::{self, adjustment::PointerCoercion, TyCtxt};
|
||||
use rustc_session::Session;
|
||||
use rustc_span::source_map::Spanned;
|
||||
|
||||
pub struct MentionedItems;
|
||||
|
||||
struct MentionedItemsVisitor<'a, 'tcx> {
|
||||
tcx: TyCtxt<'tcx>,
|
||||
body: &'a mir::Body<'tcx>,
|
||||
mentioned_items: &'a mut Vec<Spanned<MentionedItem<'tcx>>>,
|
||||
}
|
||||
|
||||
impl<'tcx> MirPass<'tcx> for MentionedItems {
|
||||
fn is_enabled(&self, _sess: &Session) -> bool {
|
||||
// If this pass is skipped the collector assume that nothing got mentioned! We could
|
||||
// potentially skip it in opt-level 0 if we are sure that opt-level will never *remove* uses
|
||||
// of anything, but that still seems fragile. Furthermore, even debug builds use level 1, so
|
||||
// special-casing level 0 is just not worth it.
|
||||
true
|
||||
}
|
||||
|
||||
fn run_pass(&self, tcx: TyCtxt<'tcx>, body: &mut mir::Body<'tcx>) {
|
||||
debug_assert!(body.mentioned_items.is_empty());
|
||||
let mut mentioned_items = Vec::new();
|
||||
MentionedItemsVisitor { tcx, body, mentioned_items: &mut mentioned_items }.visit_body(body);
|
||||
body.mentioned_items = mentioned_items;
|
||||
}
|
||||
}
|
||||
|
||||
// This visitor is carefully in sync with the one in `rustc_monomorphize::collector`. We are
|
||||
// visiting the exact same places but then instead of monomorphizing and creating `MonoItems`, we
|
||||
// have to remain generic and just recording the relevant information in `mentioned_items`, where it
|
||||
// will then be monomorphized later during "mentioned items" collection.
|
||||
impl<'tcx> Visitor<'tcx> for MentionedItemsVisitor<'_, 'tcx> {
|
||||
fn visit_terminator(&mut self, terminator: &mir::Terminator<'tcx>, location: Location) {
|
||||
self.super_terminator(terminator, location);
|
||||
let span = || self.body.source_info(location).span;
|
||||
match &terminator.kind {
|
||||
mir::TerminatorKind::Call { func, .. } => {
|
||||
let callee_ty = func.ty(self.body, self.tcx);
|
||||
self.mentioned_items
|
||||
.push(Spanned { node: MentionedItem::Fn(callee_ty), span: span() });
|
||||
}
|
||||
mir::TerminatorKind::Drop { place, .. } => {
|
||||
let ty = place.ty(self.body, self.tcx).ty;
|
||||
self.mentioned_items.push(Spanned { node: MentionedItem::Drop(ty), span: span() });
|
||||
}
|
||||
mir::TerminatorKind::InlineAsm { ref operands, .. } => {
|
||||
for op in operands {
|
||||
match *op {
|
||||
mir::InlineAsmOperand::SymFn { ref value } => {
|
||||
self.mentioned_items.push(Spanned {
|
||||
node: MentionedItem::Fn(value.const_.ty()),
|
||||
span: span(),
|
||||
});
|
||||
}
|
||||
_ => {}
|
||||
}
|
||||
}
|
||||
}
|
||||
_ => {}
|
||||
}
|
||||
}
|
||||
|
||||
fn visit_rvalue(&mut self, rvalue: &mir::Rvalue<'tcx>, location: Location) {
|
||||
self.super_rvalue(rvalue, location);
|
||||
let span = || self.body.source_info(location).span;
|
||||
match *rvalue {
|
||||
// We need to detect unsizing casts that required vtables.
|
||||
mir::Rvalue::Cast(
|
||||
mir::CastKind::PointerCoercion(PointerCoercion::Unsize),
|
||||
ref operand,
|
||||
target_ty,
|
||||
)
|
||||
| mir::Rvalue::Cast(mir::CastKind::DynStar, ref operand, target_ty) => {
|
||||
// This isn't monomorphized yet so we can't tell what the actual types are -- just
|
||||
// add everything that may involve a vtable.
|
||||
let source_ty = operand.ty(self.body, self.tcx);
|
||||
let may_involve_vtable = match (
|
||||
source_ty.builtin_deref(true).map(|t| t.ty.kind()),
|
||||
target_ty.builtin_deref(true).map(|t| t.ty.kind()),
|
||||
) {
|
||||
(Some(ty::Array(..)), Some(ty::Str | ty::Slice(..))) => false, // &str/&[T] unsizing
|
||||
_ => true,
|
||||
};
|
||||
if may_involve_vtable {
|
||||
self.mentioned_items.push(Spanned {
|
||||
node: MentionedItem::UnsizeCast { source_ty, target_ty },
|
||||
span: span(),
|
||||
});
|
||||
}
|
||||
}
|
||||
// Similarly, record closures that are turned into function pointers.
|
||||
mir::Rvalue::Cast(
|
||||
mir::CastKind::PointerCoercion(PointerCoercion::ClosureFnPointer(_)),
|
||||
ref operand,
|
||||
_,
|
||||
) => {
|
||||
let source_ty = operand.ty(self.body, self.tcx);
|
||||
self.mentioned_items
|
||||
.push(Spanned { node: MentionedItem::Closure(source_ty), span: span() });
|
||||
}
|
||||
// And finally, function pointer reification casts.
|
||||
mir::Rvalue::Cast(
|
||||
mir::CastKind::PointerCoercion(PointerCoercion::ReifyFnPointer),
|
||||
ref operand,
|
||||
_,
|
||||
) => {
|
||||
let fn_ty = operand.ty(self.body, self.tcx);
|
||||
self.mentioned_items.push(Spanned { node: MentionedItem::Fn(fn_ty), span: span() });
|
||||
}
|
||||
_ => {}
|
||||
}
|
||||
}
|
||||
}
|
|
@ -17,7 +17,7 @@ use std::iter;
|
|||
|
||||
use crate::{
|
||||
abort_unwinding_calls, add_call_guards, add_moves_for_packed_drops, deref_separator,
|
||||
pass_manager as pm, remove_noop_landing_pads, simplify,
|
||||
mentioned_items, pass_manager as pm, remove_noop_landing_pads, simplify,
|
||||
};
|
||||
use rustc_middle::mir::patch::MirPatch;
|
||||
use rustc_mir_dataflow::elaborate_drops::{self, DropElaborator, DropFlagMode, DropStyle};
|
||||
|
@ -112,6 +112,7 @@ fn make_shim<'tcx>(tcx: TyCtxt<'tcx>, instance: ty::InstanceDef<'tcx>) -> Body<'
|
|||
tcx,
|
||||
&mut body,
|
||||
&[
|
||||
&mentioned_items::MentionedItems,
|
||||
&abort_unwinding_calls::AbortUnwindingCalls,
|
||||
&add_call_guards::CriticalCallEdges,
|
||||
],
|
||||
|
@ -143,6 +144,7 @@ fn make_shim<'tcx>(tcx: TyCtxt<'tcx>, instance: ty::InstanceDef<'tcx>) -> Body<'
|
|||
tcx,
|
||||
&mut result,
|
||||
&[
|
||||
&mentioned_items::MentionedItems,
|
||||
&add_moves_for_packed_drops::AddMovesForPackedDrops,
|
||||
&deref_separator::Derefer,
|
||||
&remove_noop_landing_pads::RemoveNoopLandingPads,
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue