Adds concrete overloads of the following SIMD operations:
- Comparisons: .==, .!=, .<, .<=, .>, .>=
- Logical operations on masks: .!, .&, .^, .|
- Integer arithmetic: &+, &-, &, &+=, &-=, &=
This makes some simple benchmarks 10-100x faster, which is basically a no-brainer, while staying away from the most heavily used operators, so hopefully doesn't impact compilation performance too badly.
* Casting from AnyHashable to AnyHashable should never create another wrapper
This adds a conformance for _HasCustomAnyHashableRepresentation to
AnyHashable that simply returns self. This ensures that anytime
you try to create a new AnyHashable wrapper for an existing
AnyHashable, you just get back the original.
Resolves rdar://75180619
* Move the `Struct AnyHashable` change to `without-asserts` list
As suggested by @lorentey
* [stdlib][SR-13883] Avoid advancing past representable bounds when striding.
* [stdlib] Expand a test and add a comment to ensure correct floating-point stride bounds checking.
* [stdlib][NFC] Clarify a comment in a test.
* [stdlib][NFC] Adjust copyright notices, clarify comments, delete '-swift-version=3' for tests.
* [stdlib] Add implementations for fixed-width integer strides for performance.
* [stdlib] Document `Strideable._step` and modify overflow checking behavior of `Stride*Iterator`.
* [stdlib] Address reviewer comments, postpone documentation changes
* [stdlib][NFC] Update documentation for '_step(after:from:by:)'
* [stdlib][NFC] Use 'nil' instead of an arbitrary value for integer striding '_step' index
For this special case we copied the objects out of the cocoa array without retaining them.
This lead to a double-free crash.
Unfortunately I could not come up with an isolated test case.
rdar://74624065
* [stdlib] Deprecate MutableCollection._withUnsafeMutableBufferPointerIfSupported
In Swift 5.0, [SE-0237] introduced the public `MutableCollection.withContiguousMutableStorageIfAvailable` method. It’s time we migrated off the old, underscored variant and deprecated it.
The default `MutableCollection.sort` and `.partition(by:)` implementations are currently calling this hidden method rather than the documented interface, preventing custom Collection implementations from achieving good performance, even if they have contiguous storage.
[SE-0237]: https://github.com/apple/swift-evolution/blob/master/proposals/0237-contiguous-collection.md
* [test] Update tests for stdlib behavior changes
* Update stdlib/private/StdlibCollectionUnittest/CheckMutableCollectionType.swift
Co-authored-by: Nate Cook <natecook@apple.com>
* Update stdlib/private/StdlibCollectionUnittest/CheckMutableCollectionType.swift
Co-authored-by: Nate Cook <natecook@apple.com>
* Apply suggestions from code review
Co-authored-by: Nate Cook <natecook@apple.com>
* [test] LoggingMutableCollection: Fix logging targets
* [stdlib] Fix warning by restoring original workaround
Co-authored-by: Nate Cook <natecook@apple.com>
Array’s _copyContents implementation traps if given too small a buffer. This prevents Sequence._copyContents from being used to speed up copying data into discontiguous memory. (We have to require a Collection conformance and run _copyContents on slices instead).
Document the underlying requirement.
Make both Error and CodingKey conform to ConcurrentValue, so that
thrown errors always conform to ConcurrentValue. Downgrade (to
warnings) and ConcurrentValue-related diagnostics that are triggered
by this change in existing Error and CodingKey-conforming types to
reduce the impact on source compatibility.
There's nothing useful to say about the type of .zero -- it's
constrained by the protocol to Self, so you don't have any choice what
it is when conforming to AdditiveArithmetic.
Fixes <rdar://problem/72592923> (SR-13983).
Because the attribute is part of the declaration, putting a doc comment
between the attribute and "public" ends up with the comment in the
middle of the declaration. This results in SourceKit skipping over the
comment, and the docs not being shown.
Fixes <rdar://problem/58716408>.
The default `partition(by:)` implementation cannot currently be specialized, which makes it ~64-100x slower than it could be for common cases. (E.g., `Array<Int>.partition(by: { $0 >= foo})` is slower than `.sort()` by a factor of 2-32.)
Make the implementation inlinable.
Even though Float16 is unavailable on macOS, the type metadata can
still be referenced, for example from another unavailable declaration.
Make sure it has the correct OS version annotation so that it can be
weak linked.
Part of <rdar://problem/72151067>.
Call through to _swift_modifyAtReferenceWritableKeyPath_impl in that case. This fixes an assertion failure (or worse) when upcasting a ReferenceWritableKeyPath and then using subscript(keyPath:) to modify a value with it.
rdar://74191390
Introduce checking of ConcurrentValue conformances:
- For structs, check that each stored property conforms to ConcurrentValue
- For enums, check that each associated value conforms to ConcurrentValue
- For classes, check that each stored property is immutable and conforms
to ConcurrentValue
Because all of the stored properties / associated values need to be
visible for this check to work, limit ConcurrentValue conformances to
be in the same source file as the type definition.
This checking can be disabled by conforming to a new marker protocol,
UnsafeConcurrentValue, that refines ConcurrentValue.
UnsafeConcurrentValue otherwise his no specific meaning. This allows
both "I know what I'm doing" for types that manage concurrent access
themselves as well as enabling retroactive conformance, both of which
are fundamentally unsafe but also quite necessary.
The bulk of this change ended up being to the standard library, because
all conformances of standard library types to the ConcurrentValue
protocol needed to be sunk down into the standard library so they
would benefit from the checking above. There were numerous little
mistakes in the initial pass through the stsandard library types that
have now been corrected.
When referring to an actor-isolated declaration from outside of the
actor, ensure that the types involved conform to the `ConcurrentValue`
protocol. Otherwise, produce a diagnostic stating that it is unsafe to
pass such types across actors.
Apply the same rule to local captures within concurrent code.
By making ConcurrentValue a marker protocol, it no longer has any ABI
impact whatsoever. Sink it down into the Standard Library so it's
easier to use, even from code that isn't building with concurrency
enabled.
Add a missing fix_lifetime. This miscompiles with OSSA because
`object` is destroyed before `bridgeObject` is retained.
Reinterpreting a reference to a trivial type always requires a
fix_lifetime.
When the stdlib says not to check, it’s a good idea to actually not have any checking, so that we leave existing code paths unchanged. (And because we do trust that the stdlib is responsible about such things.)
