| Age | Commit message (Collapse) | Author | Lines |
|---|
|
compiler: Apply target features to the entry function
Fixes rust-lang/rust#146143
|
|
|
|
|
|
Rework how the codegen coordinator code handles the allocator shim
Continuing from https://github.com/rust-lang/rust/pull/144503 this centralizes most handling of the allocator shim to a single 4 line block in the codegen coordinator. The allocator shim is small enough that making it go through the main codegen loop and spawning a worker thread for it is wasted effort.
|
|
A lot of places had special handling just in case they would get an
allocator module even though most of these places could never get one or
would have a trivial implementation for the allocator module. Moving all
handling of the allocator module to a single place simplifies things a
fair bit.
|
|
Implement Integer funnel shifts
Tracking issue: rust-lang/rust#145686
ACP: https://github.com/rust-lang/libs-team/issues/642
This implements funnel shifts on primitive integer types. Implements this for cg_llvm, with a fallback impl for everything else
Thanks `@folkertdev` for the fixes and tests
cc `@rust-lang/libs-api`
|
|
llvm: nvptx: Layout update to match LLVM
LLVM upstream switched layouts to support 256-bit vector load/store.
``````@rustbot`````` label llvm-main
r? durin42
|
|
Allow `inline(always)` with a target feature behind a unstable feature `target_feature_inline_always`.
Rather than adding the inline always attribute to the function definition, we add it to the callsite. We can then check that the target features match and that the call would be safe to inline. If the function isn't inlined due to a mismatch, we emit a warning informing the user that the function can't be inlined due to the target feature mismatch.
See tracking issue rust-lang/rust#145574
|
|
- Add a fallback implementation for the intrinsics
- Add LLVM backend support for funnel shifts
Co-Authored-By: folkertdev <folkert@folkertdev.nl>
|
|
LLVM upstream switched layouts to support 256-bit vector load/store.
|
|
explicitly end the lifetime of `va_list`
tracking issue: https://github.com/rust-lang/rust/issues/44930
split out from: https://github.com/rust-lang/rust/pull/144549
The `va_list` is created in the compiler itself when the variable argument list `...` is desugared, and hence the lifetime end is not inserted automatically. The value can't outlive the function in which it was created, so it is correct to end the lifetime here. Ending the lifetime explicitly also appears to give slightly better codegen in https://github.com/rust-lang/rust/pull/144549.
I also included a little drive-by improvement to not cast pointers to integers and back again.
r? codegen
|
|
Update to ar_archive_writer 0.5
This updates `ar_archive_writer` to 0.5, which in turn was updated to match LLVM 20.1.8: <https://github.com/rust-lang/ar_archive_writer/pull/24>
As part of this, I refactored part of `SymbolWrapper.cpp` to pull common code that I was about to duplicate again into a new function.
NOTE: `ar_archive_writer` does include a breaking change where it no longer supports mangling C++ mangled names for Arm64EC. Since we don't need the mangled name (it's not the "exported name" which we're trying to load from the external dll), I'm setting the `import_name` when building for Arm64EC to prevent error when failing to mangle.
r? `@bjorn3`
|
|
|
|
We have a few ui tests to ensure we emit an error if we encounter too
big arrays. Before this fix, compiling the tests with `-Cdebuginfo=2`
would not include the spans of the instantiation sites, because the
error is then emitted from a different code path that does not include
the span.
Propagate the span to the error also in the debuginfo case, so the tests
passes regardless of debuginfo level.
|
|
compiler: Include span of too huge enum with `-Cdebuginfo=2`
We have the ui test `tests/ui/limits/huge-enum.rs` to ensure we emit an error if we encounter too big enums. Before this fix, compiling the test with `-Cdebuginfo=2` would not include the span of the instantiation site, because the error is then emitted from a different code path that does not include the span.
Propagate the span to the error also in the debuginfo case, so the test passes regardless of debuginfo level. I'm sure we can propagate spans in more places, but let's start small.
## Test failure without the fix
Here is what the failure looks like if you run the test without the fix:
```
[ui] tests/ui/limits/huge-enum.rs#full-debuginfo ... F
.
failures:
---- [ui] tests/ui/limits/huge-enum.rs#full-debuginfo stdout ----
Saved the actual stderr to `/home/martin/src/rust/build/x86_64-unknown-linux-gnu/test/ui/limits/huge-enum.full-debuginfo/huge-enum.full-debuginfo.stderr`
diff of stderr:
1 error: values of the type `Option<TYPE>` are too big for the target architecture
- --> $DIR/huge-enum.rs:17:9
- |
- LL | let big: BIG = None;
- | ^^^
6
7 error: aborting due to 1 previous error
8
The actual stderr differed from the expected stderr
To update references, rerun the tests and pass the `--bless` flag
To only update this specific test, also pass `--test-args limits/huge-enum.rs`
```
as can be seen, the `span` used to be missing with `debuginfo=2`.
## See also
This is one small step towards resolving rust-lang/rust#61117.
cc https://github.com/rust-lang/rust/pull/144499 which began running UI tests with `rust.debuginfo-level-tests=1`. This PR is part of preparing for increasing that to debuglevel 2.
|
|
|
|
All other sanitizer symbols are handled in prepare_lto already.
|
|
Don't export them from cdylibs. There is no need to do so and it
complicates exported_non_generic_symbols. In addition the GCC backend
likely uses different symbols and may potentially not even need us to
explicitly tell it to export the symbols it needs.
|
|
We have a ui test to ensure we emit an error if we encounter too big
enums. Before this fix, compiling the test with `-Cdebuginfo=2` would
not include the span of the instantiation site, because the error is
then emitted from a different code path that does not include the span.
Propagate the span to the error also in the debuginfo case, so the test
passes regardless of debuginfo level.
|
|
Use captures(address) instead of captures(none) for indirect args
While provenance cannot be captured through these arguments, the address / object identity can.
Fixes https://github.com/rust-lang/rust/issues/137668.
r? `@ghost`
|
|
Co-authored-by: Jamie Cunliffe <Jamie.Cunliffe@arm.com>
|
|
cg_llvm: Assert that LLVM range-attribute values don't exceed 128 bits
The underlying implementation of `LLVMCreateConstantRangeAttribute` assumes that each of `LowerWords` and `UpperWords` points to enough u64 values to define an integer of the specified bit-length, and will encounter UB if that is not the case.
Our safe wrapper function always passes pointers to `[u64; 2]` arrays, regardless of the bit-length specified. That's fine in practice, because scalar primitives never exceed 128 bits, but it is technically a soundness hole in a safe function.
We can close the soundness hole by explicitly asserting `size_bits <= 128`. This is effectively just a stricter version of the existing check that the value must be small enough to fit in `c_uint`.
---
This is a narrower version of the fix in rust-lang/rust#145846.
|
|
r=davidtwco
Add new Tier-3 target: riscv64a23-unknown-linux-gnu
MCP: [Tier 3 target proposal: riscv64a23-unknown-linux-gnu](https://github.com/rust-lang/compiler-team/issues/894)
Changes:
- add new target: riscv64a23-unknown-linux-gnu
- add target page
|
|
While provenance cannot be captured through these arguments, the
address / object identity can.
|
|
Handle unwinding fatal errors in codegen workers
Also directly unwind on fatal errors at the point they are emitted inside the codegen backends.
Fixes the coordinator ICE of https://github.com/rust-lang/rust/issues/132240, https://github.com/rust-lang/rust/issues/135075 and https://github.com/rust-lang/rust/issues/145800.
|
|
The underlying implementation of `LLVMCreateConstantRangeAttribute` assumes
that each of `LowerWords` and `UpperWords` points to enough u64 values to
define an integer of the specified bit-length, and will encounter UB if that is
not the case.
Our safe wrapper function always passes pointers to `[u64; 2]` arrays,
regardless of the bit-length specified. That's fine in practice, because scalar
primitives never exceed 128 bits, but it is technically a soundness hole in a
safe function.
We can close the soundness hole by explicitly asserting `size_bits <= 128`.
This is effectively just a stricter version of the existing check that the
value must be small enough to fit in `c_uint`.
|
|
This avoids the need for `#![allow(non_upper_case_globals)]`.
|
|
|
|
As opposed to passing it around through Result.
|
|
remove an `as` cast in prefetch codegen
r? `@RalfJung`
|
|
|
|
fix(debuginfo): handle false positives in overflow check
Fixes rust-lang/rust#144636.
Duplicate wrappers and normal recursive types can lead to false positives.
```rust
struct Recursive {
a: Box<Box<Recursive>>,
}
```
The ADT stack can be:
- `Box<Recursive>`
- `Recursive`
- `Box<Box<Recursive>>` (`Box` now detected as expanding)
We can filter them out by tracing the generic arg back through the stack, as true expanding recursive types must have their expanding arg used as generic arg throughout.
r? ````@wesleywiser````
|
|
Tell LLVM about read-only captures
`&Freeze` parameters are not only `readonly` within the function, but any captures of the pointer can also only be used for reads. This can now be encoded using the `captures(address, read_provenance)` attribute.
|
|
Pass `alloc-variant-zeroed` to LLVM
Makes use of https://github.com/llvm/llvm-project/pull/138299 (once we pull in a version of LLVM with this attribute). ~~Unfortunately also requires https://github.com/llvm/llvm-project/pull/149336 to work.~~
Closes rust-lang/rust#104847
|
|
`&Freeze` parameters are not only `readonly` within the function,
but any captures of the pointer can also only be used for reads.
This can now be encoded using the `captures(address, read_provenance)`
attribute.
|
|
|
|
add a fallback implementation for the `prefetch_*` intrinsics
related ACP: https://github.com/rust-lang/libs-team/issues/638
The fallback is to just ignore the arguments. That is a valid implementation because this intrinsic is just a hint.
I also added the `miri::intrinsic_fallback_is_spec` annotation, so that miri now supports these operations. A prefetch intrinsic call is valid on any pointer. (specifically LLVM guarantees this https://llvm.org/docs/LangRef.html#llvm-prefetch-intrinsic)
Next, I made the `LOCALITY` argument a const generic. That argument must be const (otherwise LLVM crashes), but that was not reflected in the type.
Finally, with these changes, the intrinsic can be safe and `const` (a prefetch at const evaluation time is just a no-op).
cc `@Amanieu`
r? `@RalfJung`
|
|
|
|
Rollup of 15 pull requests
Successful merges:
- rust-lang/rust#145338 (actually provide the correct args to coroutine witnesses)
- rust-lang/rust#145429 (Couple of codegen_fn_attrs improvements)
- rust-lang/rust#145452 (Do not strip binaries in bootstrap everytime if they are unchanged)
- rust-lang/rust#145464 (Stabilize `const_pathbuf_osstring_new` feature)
- rust-lang/rust#145474 (Properly recover from parenthesized use-bounds (precise capturing lists) plus small cleanups)
- rust-lang/rust#145486 (Fix `unicode_data.rs` mention message)
- rust-lang/rust#145490 (Trace some basic I/O operations in bootstrap)
- rust-lang/rust#145493 (remove `should_render` in `PrintAttribute` derive)
- rust-lang/rust#145500 (Port must_use to the new target checking)
- rust-lang/rust#145505 (Simplify span caches)
- rust-lang/rust#145510 (Visit and print async_fut local for async drop.)
- rust-lang/rust#145511 (Rust build fails on OpenBSD after using file_lock feature)
- rust-lang/rust#145532 (resolve: debug for block module)
- rust-lang/rust#145533 (Reorder `lto` options from most to least optimizing)
- rust-lang/rust#145537 (Do not consider a `T: !Sized` candidate to satisfy a `T: !MetaSized` obligation.)
r? `@ghost`
`@rustbot` modify labels: rollup
|
|
Couple of codegen_fn_attrs improvements
As noted in https://github.com/rust-lang/rust/pull/144678#discussion_r2245060329 here is no need to keep link_name and export_name separate, which the third commit fixes by merging them. The second commit removes some dead code and the first commit merges two ifs with equivalent conditions. The last commit is an unrelated change which removes an unused `feature(autodiff)`.
|
|
rustc_target: Add the `32s` target feature for LoongArch
LLVM: https://github.com/llvm/llvm-project/pull/139695
|
|
Add `-Zindirect-branch-cs-prefix`
Cc: ``@azhogin`` ``@Darksonn``
This goes on top of https://github.com/rust-lang/rust/pull/135927, i.e. please skip the first commit here. Please feel free to inherit it there.
In fact, I am not sure if there is any use case for the flag without `-Zretpoline*`. GCC and Clang allow it, though.
There is a `FIXME` for two `ignore`s in the test that I took from another test I did in the past -- they may be needed or not here since I didn't run the full CI. Either way, it is not critical.
Tracking issue: https://github.com/rust-lang/rust/issues/116852.
MCP: https://github.com/rust-lang/compiler-team/issues/868.
|
|
Remove `LlvmArchiveBuilder` and supporting code/bindings
Switching over to the newer Rust-based `ArArchiveBuilder` happened in rust-lang/rust#128936, a year ago.
Per the comment in `new_archive_builder`, that seems like enough time to justify removing the older, unused `LlvmArchiveBuilder` implementation and its associated bindings.
Fixes rust-lang/rust#128955.
|
|
cg_llvm: Small cleanups to `owned_target_machine`
This PR contains a few tiny cleanups to the `owned_target_machine` code.
Each individual commit should be fairly straightforward.
|
|
|
|
cg_llvm: Use LLVM-C bindings for `LLVMSetTailCallKind`, `LLVMGetTypeKind`
This PR replaces two existing `LLVMRust` bindings with equivalent calls to the LLVM-C API.
For `LLVMGetTypeKind`, we avoid the UB hazard by declaring the foreign function to return `RawEnum<TypeKind>` (which is a wrapper around `u32`), and then perform checked conversion from `u32` to `TypeKind`.
|
|
Fix `-Zregparm` for LLVM builtins
This fixes the issue where `-Zregparm=N` was not working correctly when calling LLVM intrinsics
By default on `x86-32`, arguments are passed on the stack. The `-Zregparm=N` flag allows the first `N` arguments to be passed in registers instead.
When calling intrinsics like `memset`, LLVM still passes parameters on the stack, which prevents optimizations like tail calls.
As proposed by ````@tgross35,```` I fixed this by setting the `NumRegisterParameters` LLVM module flag to `N` when the `-Zregparm=N` is set.
```rust
// compiler/rust_codegen_llvm/src/context.rs#375-382
if let Some(regparm_count) = sess.opts.unstable_opts.regparm {
llvm::add_module_flag_u32(
llmod,
llvm::ModuleFlagMergeBehavior::Error,
"NumRegisterParameters",
regparm_count,
);
}
```
[Here](https://rust.godbolt.org/z/YMezreo48) is a before/after compiler explorer.
Here is the final result for the code snippet in the original issue:
```asm
entrypoint:
push esi
mov esi, eax
mov eax, ecx
mov ecx, esi
pop esi
jmp memset ; Tail call parameters in registers
```
Fixes: https://github.com/rust-lang/rust/issues/145271
|
|
|
|
|
|
|
