See through aggregates in GVN
This PR is extracted from https://github.com/rust-lang/rust/pull/111344
The first 2 commit are cleanups to avoid repeated work. I propose to stop removing useless assignments as part of this pass, and let a later `SimplifyLocals` do it. This makes tests easier to read (among others).
The next 3 commits add a constant folding mechanism to the GVN pass, presented in https://github.com/rust-lang/rust/pull/116012. ~This pass is designed to only use global allocations, to avoid any risk of accidental modification of the stored state.~
The following commits implement opportunistic simplifications, in particular:
- projections of aggregates: `MyStruct { x: a }.x` gets replaced by `a`, works with enums too;
- projections of arrays: `[a, b][0]` becomes `a`;
- projections of repeat expressions: `[a; N][x]` becomes `a`;
- transform arrays of equal operands into a repeat rvalue.
Fixes https://github.com/rust-lang/miri/issues/3090
r? `@oli-obk`
Allow partially moved values in match
This PR attempts to unify the behaviour between `let _ = PLACE`, `let _: TY = PLACE;` and `match PLACE { _ => {} }`.
The logical conclusion is that the `match` version should not check for uninitialised places nor check that borrows are still live.
The `match PLACE {}` case is handled by keeping a `FakeRead` in the unreachable fallback case to verify that `PLACE` has a legal value.
Schematically, `match PLACE { arms }` in surface rust becomes in MIR:
```rust
PlaceMention(PLACE)
match PLACE {
// Decision tree for the explicit arms
arms,
// An extra fallback arm
_ => {
FakeRead(ForMatchedPlace, PLACE);
unreachable
}
}
```
`match *borrow { _ => {} }` continues to check that `*borrow` is live, but does not read the value.
`match *borrow {}` both checks that `*borrow` is live, and fake-reads the value.
Continuation of ~https://github.com/rust-lang/rust/pull/102256~ ~https://github.com/rust-lang/rust/pull/104844~
Fixes https://github.com/rust-lang/rust/issues/99180https://github.com/rust-lang/rust/issues/53114
Never consider raw pointer casts to be trival
HIR typeck tries to figure out which casts are trivial by doing them as
coercions and seeing whether this works. Since HIR typeck is oblivious
of lifetimes, this doesn't work for pointer casts that only change the
lifetime of the pointee, which are, as borrowck will tell you, not
trivial.
This change makes it so that raw pointer casts are never considered
trivial.
This also incidentally fixes the "trivial cast" lint false positive on
the same code. Unfortunately, "trivial cast" lints are now never emitted
on raw pointer casts, even if they truly are trivial. This could be
fixed by also doing the lint in borrowck for raw pointers specifically.
fixes#113257
HIR typeck tries to figure out which casts are trivial by doing them as
coercions and seeing whether this works. Since HIR typeck is oblivious
of lifetimes, this doesn't work for pointer casts that only change the
lifetime of the pointee, which are, as borrowck will tell you, not
trivial.
This change makes it so that raw pointer casts are never considered
trivial.
This also incidentally fixes the "trivial cast" lint false positive on
the same code. Unfortunately, "trivial cast" lints are now never emitted
on raw pointer casts, even if they truly are trivial. This could be
fixed by also doing the lint in borrowck for raw pointers specifically.
Require target features to match exactly during inlining
In general it is not correct to inline a callee with a target features
that are subset of the callee. Require target features to match exactly
during inlining.
The exact match could be potentially relaxed, but this would require
identifying specific feature that are allowed to differ, those that need
to match, and those that can be present in caller but not in callee.
This resolves MIR part of #116573. For other concerns with respect to
the previous implementation also see areInlineCompatible in LLVM.
In general it is not correct to inline a callee with a target features
that are subset of the callee. Require target features to match exactly
during inlining.
The exact match could be potentially relaxed, but this would require
identifying specific feature that are allowed to differ, those that need
to match, and those that can be present in caller but not in callee.
This resolves MIR part of #116573. For other concerns with respect to
the previous implementation also see areInlineCompatible in LLVM.
Separate move path tracking between borrowck and drop elaboration.
The primary goal of this PR is to skip creating a `MovePathIndex` for path that do not need dropping in drop elaboration.
The 2 first commits are cleanups.
The next 2 commits displace `move` errors from move-path builder to borrowck. Move-path builder keeps the same logic, but does not carry error information any more.
The remaining commits allow to filter `MovePathIndex` creation according to types. This is used in drop elaboration, to avoid computing dataflow for paths that do not need dropping.
Even though expression details are now stored in the info structure, we still
need to inject `ExpressionUsed` statements into MIR, because if one is missing
during codegen then we know that it was optimized out and we can remap all of
its associated code regions to zero.
Previously, mappings were attached to individual coverage statements in MIR.
That necessitated special handling in MIR optimizations to avoid deleting those
statements, since otherwise codegen would be unable to reassemble the original
list of mappings.
With this change, a function's list of mappings is now attached to its MIR
body, and survives intact even if individual statements are deleted by
optimizations.
coverage: Allow each coverage statement to have multiple code regions
The original implementation of coverage instrumentation was built around the assumption that a coverage counter/expression would be associated with *up to one* code region. When it was discovered that *multiple* regions would sometimes need to share a counter, a workaround was found: for the remaining regions, the instrumentor would create a fresh expression that adds zero to the existing counter/expression.
That got the job done, but resulted in some awkward code, and produces unnecessarily complicated coverage maps in the final binary.
---
This PR removes that tension by changing `StatementKind::Coverage`'s code region field from `Option<CodeRegion>` to `Vec<CodeRegion>`.
The changes on the codegen side are fairly straightforward. As long as each `CoverageKind::Counter` only injects one `llvm.instrprof.increment`, the rest of coverage codegen is happy to handle multiple regions mapped to the same counter/expression, with only minor option-to-vec adjustments.
On the instrumentor/mir-transform side, we can get rid of the code that creates extra (x + 0) expressions. Instead we gather all of the code regions associated with a single BCB, and inject them all into one coverage statement.
---
There are several patches here but they can be divided in to three phases:
- Preparatory work
- Actually switching over to multiple regions per coverage statement
- Cleaning up
So viewing the patches individually may be easier.
If a BCB has more than one code region, those extra regions can now all be
stored in the same coverage statement, instead of being stored in additional
statements.
