| Age | Commit message (Collapse) | Author | Lines |
|---|
|
Allow more MIR SROA
This removes some guards on SROA that are no longer needed:
- With https://github.com/rust-lang/compiler-team/issues/838 it no longer needs to check for SIMD
- With https://github.com/rust-lang/compiler-team/issues/807 it no longer needs to check for niches
- This means that `Wrapper(char)` and `Pin<&mut T>` can get SRoA'd now, where previously they weren't because the check was banning SRaA for anything with a niche -- not just things with `#[rustc_layout_scalar_valid_range_*]`.
- Technically rust-lang/rust#133652 isn't complete yet, but `NonZero` and `NonNull` have already moved over, so this is fine. At worst this will mean that LLVM gets less `!range` metadata on something that wasn't already fixed by rust-lang/rust#133651 or rust-lang/rust#135236, but that's still sound, and unblocking general SRoA is worth that tradeoff.
|
|
|
|
Rollup of 7 pull requests
Successful merges:
- rust-lang/rust#144072 (update `Atomic*::from_ptr` and `Atomic*::as_ptr` docs)
- rust-lang/rust#144151 (`tests/ui/issues/`: The Issues Strike Back [1/N])
- rust-lang/rust#144300 (Clippy fixes for miropt-test-tools)
- rust-lang/rust#144399 (Add a ratchet for moving all standard library tests to separate packages)
- rust-lang/rust#144472 (str: Mark unstable `round_char_boundary` feature functions as const)
- rust-lang/rust#144503 (Various refactors to the codegen coordinator code (part 3))
- rust-lang/rust#144530 (coverage: Infer `instances_used` from `pgo_func_name_var_map`)
r? `@ghost`
`@rustbot` modify labels: rollup
|
|
coverage: Infer `instances_used` from `pgo_func_name_var_map`
In obscure circumstances involving macro-expanded spans, we would sometimes emit a covfun record for a function with no physical coverage counters, and therefore no corresponding entry in the “PGO names” section of the binary. The absence of that name entry causes `llvm-cov` to fail with the cryptic error message:
```text
malformed instrumentation profile data: function name is empty
```
We can eliminate this mismatch by removing `instances_used` entirely, and instead inferring its contents from the keys of `pgo_func_name_var_map`.
This makes it impossible for a "used" function to lack a PGO name entry.
---
This is an attempt to eliminate the cause of rust-lang/rust#141577 when re-landing changes like rust-lang/rust#144298 in the future.
I haven't been able to reproduce the underlying issue in an in-tree test, because the only known repro involves a non-trivial derive proc-macro that relies on `syn` and `proc-macro2`. But I have manually verified in a separate branch that this change would have prevented the reoccurrence of https://github.com/rust-lang/rust/issues/141577#issuecomment-3120667286.
|
|
Various refactors to the codegen coordinator code (part 3)
Continuing from https://github.com/rust-lang/rust/pull/144062 this removes an option without any known users, uses the object crate in favor of LLVM for getting the LTO bitcode and improves the coordinator channel handling.
|
|
|
|
Some `let chains` clean-up
Not sure if this kind of clean-up is welcoming because of size, but I decided to try out one
r? compiler
|
|
Rollup of 6 pull requests
Successful merges:
- rust-lang/rust#143607 (Port the proc macro attributes to the new attribute parsing infrastructure)
- rust-lang/rust#144471 (Remove `compiler-builtins-{no-asm,mangled-names}`)
- rust-lang/rust#144495 (bump cargo_metadata)
- rust-lang/rust#144523 (rustdoc: save target modifiers)
- rust-lang/rust#144534 (check_static_item: explain should_check_for_sync choices)
- rust-lang/rust#144535 (miri: for ABI mismatch errors, say which argument is the problem)
Failed merges:
- rust-lang/rust#144536 (miri subtree update)
r? `@ghost`
`@rustbot` modify labels: rollup
|
|
|
|
|
|
|
|
|
|
miri: for ABI mismatch errors, say which argument is the problem
|
|
check_static_item: explain should_check_for_sync choices
Follow-up to https://github.com/rust-lang/rust/pull/144226.
r? ``@oli-obk``
|
|
r=jdonszelmann,traviscross
Port the proc macro attributes to the new attribute parsing infrastructure
Ports `#[proc_macro]`, `#[proc_macro_attribute]`, `#[proc_macro_derive]` and `#[rustc_builtin_macro]` to the new attribute parsing infrastructure for https://github.com/rust-lang/rust/issues/131229#issuecomment-2971351163
I've split this PR into commits for reviewability, and left some comments to clarify things
I did 4 related attributes in one PR because they share a lot of their code and logic, and doing them separately is kind of annoying as I need to leave both the old and new parsing in place then.
r? ``@oli-obk``
cc ``@jdonszelmann``
|
|
Don't special-case llvm.* as nounwind
Certain LLVM intrinsics, such as `llvm.wasm.throw`, can unwind. Marking them as nounwind causes us to skip cleanup of locals and optimize out `catch_unwind` under inlining or when `llvm.wasm.throw` is used directly by user code.
The motivation for forcibly marking llvm.* as nounwind is no longer present: most intrinsics are linked as `extern "C"` or other non-unwinding ABIs, so we won't codegen `invoke` for them anyway.
Closes rust-lang/rust#132416.
`@rustbot` label +T-compiler +A-panic
|
|
In rustc it doesn't really matter what the order of the witnesses is, but I'm planning to use the witnesses for implementing the "add missing match arms" assist in rust-analyzer, and there `true` before `false` is the natural order (like `Some` before `None`), and also what the current assist does.
The current order doesn't seem to be intentional; the code was created when bool ctors became their own thing, not just int ctors, but for integer, 0 before 1 is indeed the natural order.
|
|
|
|
|
|
|
|
|
|
Preintern some `TyKind::Bound` values
The new trait solver produces a lot of these.
r? `@compiler-errors`
|
|
default alignments.
|
|
Avoid unnecessary `new_adt`/`new_fn_def` calls.
They can be skipped if there are no arguments, avoiding the "relate" operation work and also the subsequent interning.
r? `@ghost`
|
|
|
|
Resolve: refactor `define` into `define_local` and `define_extern`
Follow up on rust-lang/rust#143550 and part of [#gsoc > Project: Parallel Macro Expansion](https://rust-lang.zulipchat.com/#narrow/channel/421156-gsoc/topic/Project.3A.20Parallel.20Macro.20Expansion/with/527348747).
Split up `define` into `define_local` and `define_extern`. Refactor usages of `define` into either one where it's "correct" (idk if everything is correct atm). Big part of this is that `resolution` can now take a `&Resolver` instead of a mutable one.
r? `@petrochenkov`
|
|
|
|
In obscure circumstances, we would sometimes emit a covfun record for a
function with no physical coverage counters, causing `llvm-cov` to fail with
the cryptic error message:
malformed instrumentation profile data: function name is empty
We can eliminate this mismatch by removing `instances_used` entirely, and
instead inferring its contents from the keys of `pgo_func_name_var_map`.
This makes it impossible for a "used" function to lack a PGO name entry.
|
|
recursive non-enum types
|
|
|
|
Rollup of 4 pull requests
Successful merges:
- rust-lang/rust#144226 (Do not assert layout in KnownPanicsLint.)
- rust-lang/rust#144385 (Optimize performance by inline in macro hygiene system)
- rust-lang/rust#144454 (move uefi test to run-make)
- rust-lang/rust#144455 (Unify LLVM ctlz/cttz intrinsic generation)
r? `@ghost`
`@rustbot` modify labels: rollup
|
|
Unify LLVM ctlz/cttz intrinsic generation
The type signature for `llvm.ctlz` is the same as the one for `llvm.cttz`, which enables a small simplification.
|
|
Optimize performance by inline in macro hygiene system
I inline some small method in `rustc_span/src/hygiene.rs` and so on.
|
|
Do not assert layout in KnownPanicsLint.
Fixes rust-lang/rust#121176
Fixes rust-lang/rust#129109
Fixes rust-lang/rust#130970
Fixes rust-lang/rust#131347
Fixes rust-lang/rust#139872
Fixes rust-lang/rust#140332
|
|
No longer need `alloca`s for consuming `Result<!, i32>` and similar
In optimized builds GVN gets rid of these already, but in `opt-level=0` we actually make `alloca`s for this, which particularly impacts `?`-style things that use actually-only-one-variant types like this.
While doing so, rewrite `LocalAnalyzer::process_place` to be non-recursive, solving a 6+ year old FIXME.
r? codegen
|
|
Revert "coverage: Enlarge empty spans during MIR instrumentation, not codegen"
Surprisingly, rust-lang/rust#144298 alone (extracted from rust-lang/rust#140847) was enough to re-trigger the failures observed in https://github.com/rust-lang/rust/issues/141577#issuecomment-3120667286.
---
This reverts commit f877aa7d14916f71a2f88c6d4c009e7ded7684c4.
---
r? ghost
|
|
Enable outline-atomics for aarch64-unknown-linux-musl
They were disabled in bd287fa5084f2e153c1044632f9f3d190f090d69 and haven't been causing problems for a while anymore, see https://github.com/rust-lang/rust/issues/89626 for details.
The entire testsuite still passes on `aarch64-unknown-linux-musl` with this feature enabled.
|
|
Stop compilation early if macro expansion failed
Fixes rust-lang/rust#116180.
So there isn't really a type that is central for macro expansion and some errors are actually emitted (because the resolution happens after the expansion I suppose) after the expansion pass (like "not found macro"). Sometimes, errors are only emitted on the second "try" (to improve error output). So I couldn't reach a similar solution than what was done in https://github.com/rust-lang/rust/pull/133937 and suggested by ````@estebank```` in https://github.com/rust-lang/rust/issues/116180#issuecomment-3109468922. But maybe I missed something?
So in the end, I realized that there is method called every time (except one, described below) a macro error is actually emitted: `ExtCtxt::trace_macros_diag`. Considering I updated what it did, I renamed it into `macro_error_and_trace_macros_diag` to better reflect it.
There is only one call of `trace_macros_diag` which isn't reporting an error but just used for `macro_trace` feature, so I kept it as is.
r? ````@oli-obk````
|
|
disable cfg.has_reliable_f128 on amdgcn
I was experimenting with compiling a few kernels for amd while working on std::offload. It seems like the logic in https://github.com/rust-lang/compiler-builtins/pull/737 got removed, so I re-introduce it here. Probably should have a test to avoid another regression and make sure that f128 doesn't show up as target feature for amdgcn. It looks like currently we neither check that for nvptx, nor amdgpu. Maybe I could add two revisions to https://github.com/rust-lang/rust/blob/master/tests/ui/float/target-has-reliable-nightly-float.rs?
r? ````@Flakebi````
fixes: https://github.com/rust-lang/rust/issues/144381
|
|
Let `codegen_transmute_operand` just handle everything
When combined with rust-lang/rust#143720, this means `rvalue_creates_operand` can just return `true` for *every* `Rvalue`. (A future PR could consider removing it, though just letting it optimize out is fine for now.)
It's nicer anyway, IMHO, because it avoids needing the layout checks to be consistent in the two places, and thus is an overall reduction in code. Plus it's a more helpful building block when used in other places this way.
(TBH, it probably would have been better to have it this way the whole time, but I clearly didn't understand `rvalue_creates_operand` when I originally wrote rust-lang/rust#109843.)
|
|
|
|
|
|
Signed-off-by: Jonathan Brouwer <jonathantbrouwer@gmail.com>
|
|
Signed-off-by: Jonathan Brouwer <jonathantbrouwer@gmail.com>
|
|
Signed-off-by: Jonathan Brouwer <jonathantbrouwer@gmail.com>
|
|
|
|
|
|
Skip walking into param-env component if it has no placeholder/re-var
Although it only provides a minor perf improvement, it seems like it could matter in more pathological cases.
|
|
Nobody seems to actually use this, while still adding some extra
complexity to the already rather complex codegen coordinator code.
It is also not supported by any backend other than the LLVM backend.
|
|
resolve: Do not create `NameResolutions` on access unless necessary
`fn resolution` now just performs the access, and `fn resolution_or_default` will insert a default entry if the entry is missing.
|
