Clang changed to representing tuples of scalable vectors as
structs rather than as wide vectors (that is, scalable vector types
where the `N` part of the `<vscale x N x ty>` type was multiplied by
the number of vectors). rustc mirrored this in the initial implementation
of scalable vectors.
Earlier versions of our patches used the wide vector representation and
our intrinsic patches used the legacy
`llvm.aarch64.sve.tuple.{create,get,set}{2,3,4}` intrinsics for creating
these tuples/getting/setting the vectors, which were only supported
due to LLVM's `AutoUpgrade` pass converting these intrinsics into
`llvm.vector.insert`. `AutoUpgrade` only supports these legacy intrinsics
with the wide vector representation.
With the current struct representation, Clang has special handling in
codegen for generating `insertvalue`/`extractvalue` instructions for
these operations, which must be replicated by rustc's codegen for our
intrinsics to use. This patch implements new intrinsics in
`core::intrinsics::scalable` (mirroring the structure of
`core::intrinsics::simd`) which rustc lowers to the appropriate
`insertvalue`/`extractvalue` instructions.
Unify wording of resolve error
Remove "failed to resolve" from the main error message and use the same format we use in other resolution errors "cannot find `name`":
```
error[E0433]: cannot find `nonexistent` in `existent`
--> $DIR/custom_attr_multisegment_error.rs:5:13
|
LL | #[existent::nonexistent]
| ^^^^^^^^^^^ could not find `nonexistent` in `existent`
```
The intent behind this is to end up with all resolve errors eventually be on the form of
```
error[ECODE]: cannot find `{NAME}` in {SCOPE}
--> $DIR/file.rs:5:13
|
LL | #[existent::nonexistent]
| ^^^^^^^^^^^ {SPECIFIC LABEL}
```
A category of errors that is interest are those that involve keywords. For example:
```
error[E0433]: cannot find `Self` in this scope
--> $DIR/issue-97194.rs:2:35
|
LL | fn bget(&self, index: [usize; Self::DIM]) -> bool {
| ^^^^ `Self` is only available in impls, traits, and type definitions
```
and
```
error[E0433]: cannot find `super` in this scope
--> $DIR/keyword-super.rs:2:9
|
LL | let super: isize;
| ^^^^^ there are too many leading `super` keywords
```
For these the label provides the actual help, while the message is less informative beyond telling you "couldn't find `name`".
This is an off-shoot of https://github.com/rust-lang/rust/pull/126810 and https://github.com/rust-lang/rust/pull/128086, a subset of the intended changes there with review comments applied.
r? @petrochenkov
Remove "failed to resolve" and use the same format we use in other resolution errors "cannot find `name`".
```
error[E0433]: cannot find `nonexistent` in `existent`
--> $DIR/custom_attr_multisegment_error.rs:5:13
|
LL | #[existent::nonexistent]
| ^^^^^^^^^^^ could not find `nonexistent` in `existent`
```
add `simd_splat` intrinsic
Add `simd_splat` which lowers to the LLVM canonical splat sequence.
```llvm
insertelement <N x elem> poison, elem %x, i32 0
shufflevector <N x elem> v0, <N x elem> poison, <N x i32> zeroinitializer
```
Right now we try to fake it using one of
```rust
fn splat(x: u32) -> u32x8 {
u32x8::from_array([x; 8])
}
```
or (in `stdarch`)
```rust
fn splat(value: $elem_type) -> $name {
#[derive(Copy, Clone)]
#[repr(simd)]
struct JustOne([$elem_type; 1]);
let one = JustOne([value]);
// SAFETY: 0 is always in-bounds because we're shuffling
// a simd type with exactly one element.
unsafe { simd_shuffle!(one, one, [0; $len]) }
}
```
Both of these can confuse the LLVM optimizer, producing sub-par code. Some examples:
- https://github.com/rust-lang/rust/issues/60637
- https://github.com/rust-lang/rust/issues/137407
- https://github.com/rust-lang/rust/issues/122623
- https://github.com/rust-lang/rust/issues/97804
---
As far as I can tell there is no way to provide a fallback implementation for this intrinsic, because there is no `const` way of evaluating the number of elements (there might be issues beyond that, too). So, I added implementations for all 4 backends.
Both GCC and const-eval appear to have some issues with simd vectors containing pointers. I have a workaround for GCC, but haven't yet been able to make const-eval work. See the comments below.
Currently this just adds the intrinsic, it does not actually use it anywhere yet.
Fix ICE when using zero-length SIMD type in extern static
before my fix using a zero-length SIMD type in an extern static would cause an internal compiler error. now it properly shows a diagnostic error instead of panicking. it was because `LayoutError::InvalidSimd` wasn't handled in `check_static_inhabited` and fell through to a generic `delayed_bug`.
i added handling for `InvalidSimd` in `check_static_inhabited` (similar to `SizeOverflow`): when a SIMD type has an invalid layout, we call `emit_err` with `Spanned` to emit a normal error instead of an ICE. compiler now emits a clear error `"the SIMD type Simd<u8, 0> has zero elements"` with the correct span on the type, matching expected compiler behavior.
fixesrust-lang/rust#151451
Previously, using a zero-length SIMD type in an extern static would
cause an internal compiler error. Now it properly emits a diagnostic
error instead of panicking.
This reverts commit 1eeb8e8b15, reversing
changes made to 324bf2b9fd.
Unfortunately the assert desugaring change is not backwards compatible,
see RUST-145770.
Code such as
```rust
#[derive(Debug)]
struct F {
data: bool
}
impl std::ops::Not for F {
type Output = bool;
fn not(self) -> Self::Output { !self.data }
}
fn main() {
let f = F { data: true };
assert!(f);
}
```
would be broken by the assert desugaring change. We may need to land
the change over an edition boundary, or limit the editions that the
desugaring change impacts.
In the desugaring of `assert!`, we now expand to a `match` expression
instead of `if !cond {..}`.
The span of incorrect conditions will point only at the expression, and not
the whole `assert!` invocation.
```
error[E0308]: mismatched types
--> $DIR/issue-14091.rs:2:13
|
LL | assert!(1,1);
| ^ expected `bool`, found integer
```
We no longer mention the expression needing to implement the `Not` trait.
```
error[E0308]: mismatched types
--> $DIR/issue-14091-2.rs:15:13
|
LL | assert!(x, x);
| ^ expected `bool`, found `BytePos`
```
`assert!(val)` now desugars to:
```rust
match val {
true => {},
_ => $crate::panic::panic_2021!(),
}
```
Fix#122159.
We make some minor changes to some diagnostics to avoid span overlap on
type mismatch or inverted "expected"/"found" on type errors.
We remove some unnecessary parens from core, alloc and miri.
address review comments
`TyCtxt::short_string` ensures that user visible type paths aren't overwhelming on the terminal output, and properly saves the long name to disk as a side-channel. We already use these throughout the compiler and have been using them as needed when users find cases where the output is verbose. This is a proactive search of some cases to use `short_string`.
We add support for shortening the path of "trait path only".
Every manual use of `short_string` is a bright marker that that error should be using structured diagnostics instead (as they have proper handling of long types without the maintainer having to think abou tthem).
When we don't actually print out a shortened type we don't need the "use `--verbose`" note.
On E0599 show type identity to avoid expanding the receiver's generic parameters.
Unify wording on `long_ty_path` everywhere.
Add SIMD funnel shift and round-to-even intrinsics
This PR adds 3 new SIMD intrinsics
- `simd_funnel_shl` - funnel shift left
- `simd_funnel_shr` - funnel shift right
- `simd_round_ties_even` (vector version of `round_ties_even_fN`)
TODO (future PR): implement `simd_fsh{l,r}` in miri, cg_gcc and cg_clif (it is surprisingly hard to implement without branches, the common tricks that rotate uses doesn't work because we have 2 elements now. e.g, the `-n&31` trick used by cg_gcc to implement rotate doesn't work with this because then `fshl(a, b, 0)` will be `a | b`)
[#t-compiler > More SIMD intrinsics](https://rust-lang.zulipchat.com/#narrow/channel/131828-t-compiler/topic/More.20SIMD.20intrinsics/with/522130286)
`@rustbot` label T-compiler T-libs A-intrinsics F-core_intrinsics
r? `@workingjubilee`
`concat_idents` has been around unstably for a long time, but there is
now a better (but still unstable) way to join identifiers using
`${concat(...)}` syntax with `macro_metavar_expr_concat`. This resolves
a lot of the problems with `concat_idents` and is on a better track
toward stabilization, so there is no need to keep both versions around.
`concat_idents!` still has a lot of use in the ecosystem so deprecate it
before removing, as discussed in [1].
Link: https://github.com/rust-lang/rust/issues/124225
[1]: https://rust-lang.zulipchat.com/#narrow/channel/219381-t-libs/topic/Removing.20.60concat_idents.60
simd intrinsics with mask: accept unsigned integer masks, and fix some of the errors
It's not clear at all why the mask would have to be signed, it is anyway interpreted bitwise. The backend should just make sure that works no matter the surface-level type; our LLVM backend already does this correctly. The note of "the mask may be widened, which only has the correct behavior for signed integers" explains... nothing? Why can't the code do the widening correctly? If necessary, just cast to the signed type first...
Also while we are at it, fix the errors. For simd_masked_load/store, the errors talked about the "third argument" but they meant the first argument (the mask is the first argument there). They also used the wrong type for `expected_element`.
I have extremely low confidence in the GCC part of this PR.
See [discussion on Zulip](https://rust-lang.zulipchat.com/#narrow/channel/257879-project-portable-simd/topic/On.20the.20sign.20of.20masks)
add `core::intrinsics::simd::{simd_extract_dyn, simd_insert_dyn}`
fixes https://github.com/rust-lang/rust/issues/137372
adds `core::intrinsics::simd::{simd_extract_dyn, simd_insert_dyn}`, which contrary to their non-dyn counterparts allow a non-const index. Many platforms (but notably not x86_64 or aarch64) have dedicated instructions for this operation, which stdarch can emit with this change.
Future work is to also make the `Index` operation on the `Simd` type emit this operation, but the intrinsic can't be used directly. We'll need some MIR shenanigans for that.
r? `@ghost`
Ralf Jung