Render pattern types nicely in mir dumps
avoid falling through to the fallback rendering that just does a hex dump
r? ``@scottmcm``
best reviewed commit by commit
Add `#[optimize(none)]`
cc #54882
This extends the `optimize` attribute to add `none`, which corresponds to the LLVM `OptimizeNone` attribute.
Not sure if an MCP is required for this, happy to file one if so.
Don't drop types with no drop glue when building drops for tailcalls
this is required as otherwise drops of `&mut` refs count as a usage of a
'two-phase temporary' causing an ICE.
fixes#128097
The underlying issue is that the current code generates drops for `&mut` which are later counted as a second use of a two-phase temporary:
`bat t.rs -p`
```rust
#![expect(incomplete_features)]
#![feature(explicit_tail_calls)]
fn f(x: &mut ()) {
let _y = String::new();
become f(x);
}
fn main() {}
```
`rustc t.rs -Zdump_mir=f`
```text
error: internal compiler error: compiler/rustc_borrowck/src/borrow_set.rs:298:17: found two uses for 2-phase borrow temporary _4: bb2[1] and bb3[0]
--> t.rs:6:5
|
6 | become f(x);
| ^^^^^^^^^^^
thread 'rustc' panicked at compiler/rustc_borrowck/src/borrow_set.rs:298:17:
Box<dyn Any>
stack backtrace:
[REDACTED]
error: aborting due to 1 previous error
```
`bat ./mir_dump/t.f.-------.renumber.0.mir -p -lrust`
```rust
// MIR for `f` 0 renumber
fn f(_1: &mut ()) -> () {
debug x => _1;
let mut _0: ();
let mut _2: !;
let _3: std::string::String;
let mut _4: &mut ();
scope 1 {
debug _y => _3;
}
bb0: {
StorageLive(_3);
_3 = String::new() -> [return: bb1, unwind: bb4];
}
bb1: {
FakeRead(ForLet(None), _3);
StorageLive(_4);
_4 = &mut (*_1);
drop(_3) -> [return: bb2, unwind: bb3];
}
bb2: {
StorageDead(_3);
tailcall f(Spanned { node: move _4, span: t.rs:6:14: 6:15 (#0) });
}
bb3 (cleanup): {
drop(_4) -> [return: bb4, unwind terminate(cleanup)];
}
bb4 (cleanup): {
resume;
}
}
```
Note how `_4 is moved into the tail call in `bb2` and dropped in `bb3`.
This PR adds a check that the locals we drop need dropping.
r? `@oli-obk` (feel free to reassign, I'm not sure who would be a good reviewer, but thought you might have an idea)
cc `@beepster4096,` since you wrote the original drop implementation.
Reexport likely/unlikely in std::hint
Since `likely`/`unlikely` should be working now, we could reexport them in `std::hint`. I'm not sure if this is already approved or if it requires approval
Tracking issue: #26179
coverage: Completely overhaul counter assignment, using node-flow graphs
The existing code for choosing where to put physical counter-increments gets the job done, but is very ad-hoc and hard to modify without introducing tricky regressions.
This PR replaces all of that with a more principled approach, based on the algorithm described in "Optimal measurement points for program frequency counts" (Knuth & Stevenson, 1973).
---
We start by ensuring that our graph has “balanced flow”, i.e. each node's flow (execution count) is equal to the sum of all its in-edge flows, and equal to the sum of all its out-edge flows. That isn't naturally true of control-flow graphs, so we introduce a wrapper type `BalancedFlowGraph` to fix that by introducing synthetic nodes and edges as needed.
Once our graph has balanced flow, the next step is to create another view of that graph in which each node's successors have all been merged into one “supernode”. Consequently, each node's out-edges can be coalesced into a single out-edge to one of those supernodes. Because of the balanced-flow property, the flow of that coalesced edge is equal to the flow of the original node.
Having expressed all of our node flows as edge flows, we can then analyze node flows using techniques for analyzing edge flows. We incrementally build a spanning tree over the merged supernodes, such that each new edge in the spanning tree represents a node whose flow can be computed from that of other nodes.
When this is done, we end up with a list of “counter terms” for each node, describing which nodes need physical counters, and how the remaining nodes can have their flow calculated by adding and subtracting those physical counters.
---
The re-blessed coverage tests show that this results in modest or major improvements for our test programs. Some tests need fewer physical counters, some tests need fewer expression nodes for the same number of physical counters, and some tests show striking reductions in both.
Fix cycle error only occurring with -Zdump-mir
fixes#134205
During mir dumping, we evaluate static items to render their allocations. If a static item refers to itself, its own MIR will have a reference to itself, so during mir dumping we end up evaluating the static again, causing us to try to build MIR again (mir dumping happens during MIR building).
Thus I disabled evaluation of statics during MIR dumps in case the MIR body isn't far enough along yet to be able to be guaranteed cycle free.
Make MIR cleanup for functions with impossible predicates into a real MIR pass
It's a bit jarring to see the body of a function with an impossible-to-satisfy where clause suddenly go to a single `unreachable` terminator when looking at the MIR dump output in order, and I discovered it's because we manually replace the body outside of a MIR pass.
Let's make it into a fully flegded MIR pass so it's more clear what it's doing and when it's being applied.
Add an InstSimplify for repetitive array expressions
I noticed in https://github.com/rust-lang/rust/pull/135068#issuecomment-2569955426 that GVN's implementation of this same transform was quite profitable on the deep-vector benchmark. But of course GVN doesn't run in unoptimized builds, so this is my attempt to write a version of this transform that benefits the deep-vector case and is fast enough to run in InstSimplify.
The benchmark suite indicates that this is effective.
Adds `#[rustc_force_inline]` which is similar to always inlining but
reports an error if the inlining was not possible, and which always
attempts to inline annotated items, regardless of optimisation levels.
It can only be applied to free functions to guarantee that the MIR
inliner will be able to resolve calls.
We already did `Transmute`-then-`PtrToPtr`; this adds the nearly-identical `PtrToPtr`-then-`Transmute`.
It also adds `transmute(Foo(x))` → `transmute(x)`, when `Foo` is a single-field transparent type. That's useful for things like `NonNull { pointer: p }.as_ptr()`.
Found these as I was looking at MCP807-related changes.
Use `PtrMetadata` instead of `Len` in slice drop shims
I tried to do a bigger change in #134297 which didn't work, so here's the part I really wanted: Removing another use of `Len`, in favour of `PtrMetadata`.
Split into two commits where the first just adds a test, so you can look at the second commit to see how the drop shim for an array changes with this PR.
Reusing the same reviewer from the last one:
r? BoxyUwU
Make sure we handle `backwards_incompatible_lint` drops appropriately in drop elaboration
In #131326, a new kind of scheduled drop (`drop_kind: DropKind::Value` + `backwards_incompatible_lint: true`) was added so that we could insert a new kind of no-op MIR statement (`backward incompatible drop`) for linting purposes.
These drops were intended to have *no side-effects*, but drop elaboration code forgot to handle these drops specially and they were handled otherwise as normal drops in most of the code. This ends up being **unsound** since we insert more than one drop call for some values, which means that `Drop::drop` could be called more than once.
This PR fixes this by splitting out the `DropKind::ForLint` and adjusting the code. I'm not totally certain if all of the places I've adjusted are either reachable or correct, but I'm pretty certain that it's *more* correct than it was previously.
cc `@dingxiangfei2009`
r? nikomatsakis
Fixes#134482
coverage: Store coverage source regions as `Span` until codegen (take 2)
This is an attempt to re-land #133418:
> Historically, coverage spans were converted into line/column coordinates during the MIR instrumentation pass.
> This PR moves that conversion step into codegen, so that coverage spans spend most of their time stored as Span instead.
> In addition to being conceptually nicer, this also reduces the size of coverage mappings in MIR, because Span is smaller than 4x u32.
That PR was reverted by #133608, because in some circumstances not covered by our test suite we were emitting coverage metadata that was causing `llvm-cov` to exit with an error (#133606).
---
The implementation here is *mostly* the same, but adapted for subsequent changes in the relevant code (e.g. #134163).
I believe that the changes in #134163 should be sufficient to prevent the problem that required the original PR to be reverted. But I haven't been able to reproduce the original breakage in a regression test, and the `llvm-cov` error message is extremely unhelpful, so I can't completely rule out the possibility of this breaking again.
r? jieyouxu (reviewer of the original PR)
Variants::Single: do not use invalid VariantIdx for uninhabited enums
~~Stacked on top of https://github.com/rust-lang/rust/pull/133681, only the last commit is new.~~
Currently, `Variants::Single` for an empty enum contains a `VariantIdx` of 0; looking that up in the enum variant list will ICE. That's quite confusing. So let's fix that by adding a new `Variants::Empty` case for types that have 0 variants.
try-job: i686-msvc
