mirror of
https://github.com/apple/swift.git
synced 2025-12-14 20:36:38 +01:00
Inspired by #84826, I've dusted off and completely reworked a native implementation of integer-to-string conversion. Besides existing tests in this repository, the core of the implementation has been comprehensively tested in a separate package for all bases between 2–36 to demonstrate identical output for all 8-bit and 16-bit values, and for randomly generated 32-bit, 64-bit, and 128-bit values. Resolves #51902. <!-- If this pull request is targeting a release branch, please fill out the following form: https://github.com/swiftlang/.github/blob/main/PULL_REQUEST_TEMPLATE/release.md?plain=1 Otherwise, replace this comment with a description of your changes and rationale. Provide links to external references/discussions if appropriate. If this pull request resolves any GitHub issues, link them like so: Resolves <link to issue>, resolves <link to another issue>. For more information about linking a pull request to an issue, see: https://docs.github.com/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue --> <!-- Before merging this pull request, you must run the Swift continuous integration tests. For information about triggering CI builds via @swift-ci, see: https://github.com/apple/swift/blob/main/docs/ContinuousIntegration.md#swift-ci Thank you for your contribution to Swift! -->
365 lines
12 KiB
Swift
365 lines
12 KiB
Swift
//===----------------------------------------------------------*- swift -*-===//
|
|
//
|
|
// This source file is part of the Swift.org open source project
|
|
//
|
|
// Copyright (c) 2014 - 2019 Apple Inc. and the Swift project authors
|
|
// Licensed under Apache License v2.0 with Runtime Library Exception
|
|
//
|
|
// See https://swift.org/LICENSE.txt for license information
|
|
// See https://swift.org/CONTRIBUTORS.txt for the list of Swift project authors
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
///
|
|
/// This file contains Swift wrappers for functions defined in the C++ runtime.
|
|
///
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
import SwiftShims
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
// Atomics
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
@_transparent
|
|
public // @testable
|
|
func _stdlib_atomicCompareExchangeStrongPtr(
|
|
object target: UnsafeMutablePointer<UnsafeRawPointer?>,
|
|
expected: UnsafeMutablePointer<UnsafeRawPointer?>,
|
|
desired: UnsafeRawPointer?
|
|
) -> Bool {
|
|
// We use Builtin.Word here because Builtin.RawPointer can't be nil.
|
|
let (oldValue, won) = unsafe Builtin.cmpxchg_seqcst_seqcst_Word(
|
|
target._rawValue,
|
|
UInt(bitPattern: expected.pointee)._builtinWordValue,
|
|
UInt(bitPattern: desired)._builtinWordValue)
|
|
unsafe expected.pointee = UnsafeRawPointer(bitPattern: Int(oldValue))
|
|
return Bool(won)
|
|
}
|
|
|
|
/// Atomic compare and exchange of `UnsafeMutablePointer<T>` with sequentially
|
|
/// consistent memory ordering. Precise semantics are defined in C++11 or C11.
|
|
///
|
|
/// - Warning: This operation is extremely tricky to use correctly because of
|
|
/// writeback semantics.
|
|
///
|
|
/// It is best to use it directly on an
|
|
/// `UnsafeMutablePointer<UnsafeMutablePointer<T>>` that is known to point
|
|
/// directly to the memory where the value is stored.
|
|
///
|
|
/// In a call like this:
|
|
///
|
|
/// _stdlib_atomicCompareExchangeStrongPtr(&foo.property1.property2, ...)
|
|
///
|
|
/// you need to manually make sure that:
|
|
///
|
|
/// - all properties in the chain are physical (to make sure that no writeback
|
|
/// happens; the compare-and-exchange instruction should operate on the
|
|
/// shared memory); and
|
|
///
|
|
/// - the shared memory that you are accessing is located inside a heap
|
|
/// allocation (a class instance property, a `_BridgingBuffer`, a pointer to
|
|
/// an `Array` element etc.)
|
|
///
|
|
/// If the conditions above are not met, the code will still compile, but the
|
|
/// compare-and-exchange instruction will operate on the writeback buffer, and
|
|
/// you will get a *race* while doing writeback into shared memory.
|
|
@_transparent
|
|
public // @testable
|
|
func _stdlib_atomicCompareExchangeStrongPtr<T>(
|
|
object target: UnsafeMutablePointer<UnsafeMutablePointer<T>>,
|
|
expected: UnsafeMutablePointer<UnsafeMutablePointer<T>>,
|
|
desired: UnsafeMutablePointer<T>
|
|
) -> Bool {
|
|
let rawTarget = unsafe UnsafeMutableRawPointer(target).assumingMemoryBound(
|
|
to: Optional<UnsafeRawPointer>.self)
|
|
let rawExpected = unsafe UnsafeMutableRawPointer(expected).assumingMemoryBound(
|
|
to: Optional<UnsafeRawPointer>.self)
|
|
return unsafe _stdlib_atomicCompareExchangeStrongPtr(
|
|
object: rawTarget,
|
|
expected: rawExpected,
|
|
desired: UnsafeRawPointer(desired))
|
|
}
|
|
|
|
/// Atomic compare and exchange of `UnsafeMutablePointer<T>` with sequentially
|
|
/// consistent memory ordering. Precise semantics are defined in C++11 or C11.
|
|
///
|
|
/// - Warning: This operation is extremely tricky to use correctly because of
|
|
/// writeback semantics.
|
|
///
|
|
/// It is best to use it directly on an
|
|
/// `UnsafeMutablePointer<UnsafeMutablePointer<T>>` that is known to point
|
|
/// directly to the memory where the value is stored.
|
|
///
|
|
/// In a call like this:
|
|
///
|
|
/// _stdlib_atomicCompareExchangeStrongPtr(&foo.property1.property2, ...)
|
|
///
|
|
/// you need to manually make sure that:
|
|
///
|
|
/// - all properties in the chain are physical (to make sure that no writeback
|
|
/// happens; the compare-and-exchange instruction should operate on the
|
|
/// shared memory); and
|
|
///
|
|
/// - the shared memory that you are accessing is located inside a heap
|
|
/// allocation (a class instance property, a `_BridgingBuffer`, a pointer to
|
|
/// an `Array` element etc.)
|
|
///
|
|
/// If the conditions above are not met, the code will still compile, but the
|
|
/// compare-and-exchange instruction will operate on the writeback buffer, and
|
|
/// you will get a *race* while doing writeback into shared memory.
|
|
@_transparent
|
|
public // @testable
|
|
func _stdlib_atomicCompareExchangeStrongPtr<T>(
|
|
object target: UnsafeMutablePointer<UnsafeMutablePointer<T>?>,
|
|
expected: UnsafeMutablePointer<UnsafeMutablePointer<T>?>,
|
|
desired: UnsafeMutablePointer<T>?
|
|
) -> Bool {
|
|
let rawTarget = unsafe UnsafeMutableRawPointer(target).assumingMemoryBound(
|
|
to: Optional<UnsafeRawPointer>.self)
|
|
let rawExpected = unsafe UnsafeMutableRawPointer(expected).assumingMemoryBound(
|
|
to: Optional<UnsafeRawPointer>.self)
|
|
return unsafe _stdlib_atomicCompareExchangeStrongPtr(
|
|
object: rawTarget,
|
|
expected: rawExpected,
|
|
desired: UnsafeRawPointer(desired))
|
|
}
|
|
|
|
@_transparent
|
|
@discardableResult
|
|
@_unavailableInEmbedded
|
|
public // @testable
|
|
func _stdlib_atomicInitializeARCRef(
|
|
object target: UnsafeMutablePointer<AnyObject?>,
|
|
desired: AnyObject
|
|
) -> Bool {
|
|
// Note: this assumes that AnyObject? is layout-compatible with a RawPointer
|
|
// that simply points to the same memory.
|
|
var expected: UnsafeRawPointer? = nil
|
|
let unmanaged = unsafe Unmanaged.passRetained(desired)
|
|
let desiredPtr = unsafe unmanaged.toOpaque()
|
|
let rawTarget = unsafe UnsafeMutableRawPointer(target).assumingMemoryBound(
|
|
to: Optional<UnsafeRawPointer>.self)
|
|
let wonRace = unsafe withUnsafeMutablePointer(to: &expected) {
|
|
unsafe _stdlib_atomicCompareExchangeStrongPtr(
|
|
object: rawTarget, expected: $0, desired: desiredPtr
|
|
)
|
|
}
|
|
if !wonRace {
|
|
// Some other thread initialized the value. Balance the retain that we
|
|
// performed on 'desired'.
|
|
unsafe unmanaged.release()
|
|
}
|
|
return wonRace
|
|
}
|
|
|
|
@_transparent
|
|
@_unavailableInEmbedded
|
|
public // @testable
|
|
func _stdlib_atomicLoadARCRef(
|
|
object target: UnsafeMutablePointer<AnyObject?>
|
|
) -> AnyObject? {
|
|
let value = Builtin.atomicload_seqcst_Word(target._rawValue)
|
|
if let unwrapped = unsafe UnsafeRawPointer(bitPattern: Int(value)) {
|
|
return unsafe Unmanaged<AnyObject>.fromOpaque(unwrapped).takeUnretainedValue()
|
|
}
|
|
return nil
|
|
}
|
|
|
|
@_transparent
|
|
@_alwaysEmitIntoClient
|
|
@discardableResult
|
|
public func _stdlib_atomicAcquiringInitializeARCRef<T: AnyObject>(
|
|
object target: UnsafeMutablePointer<T?>,
|
|
desired: __owned T
|
|
) -> Unmanaged<T> {
|
|
// Note: this assumes that AnyObject? is layout-compatible with a RawPointer
|
|
// that simply points to the same memory, and that `nil` is represented by an
|
|
// all-zero bit pattern.
|
|
let unmanaged = unsafe Unmanaged.passRetained(desired)
|
|
let desiredPtr = unsafe unmanaged.toOpaque()
|
|
|
|
let (value, won) = Builtin.cmpxchg_acqrel_acquire_Word(
|
|
target._rawValue,
|
|
0._builtinWordValue,
|
|
Builtin.ptrtoint_Word(desiredPtr._rawValue))
|
|
|
|
if Bool(won) { return unsafe unmanaged }
|
|
|
|
// Some other thread initialized the value before us. Balance the retain that
|
|
// we performed on 'desired', and return what we loaded.
|
|
unsafe unmanaged.release()
|
|
let ptr = UnsafeRawPointer(Builtin.inttoptr_Word(value))
|
|
return unsafe Unmanaged<T>.fromOpaque(ptr)
|
|
}
|
|
|
|
@_alwaysEmitIntoClient
|
|
@_transparent
|
|
public func _stdlib_atomicAcquiringLoadARCRef<T: AnyObject>(
|
|
object target: UnsafeMutablePointer<T?>
|
|
) -> Unmanaged<T>? {
|
|
let value = Builtin.atomicload_acquire_Word(target._rawValue)
|
|
if Int(value) == 0 { return nil }
|
|
let opaque = UnsafeRawPointer(Builtin.inttoptr_Word(value))
|
|
return unsafe Unmanaged<T>.fromOpaque(opaque)
|
|
}
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
// Conversion of primitive types to `String`
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
@inlinable
|
|
internal func _rawPointerToString(_ value: Builtin.RawPointer) -> String {
|
|
var result = _uint64ToString(
|
|
UInt64(UInt(bitPattern: UnsafeRawPointer(value))),
|
|
radix: 16,
|
|
uppercase: false
|
|
)
|
|
for _ in unsafe 0..<(2 * MemoryLayout<UnsafeRawPointer>.size - result.utf16.count) {
|
|
result = "0" + result
|
|
}
|
|
return "0x" + result
|
|
}
|
|
|
|
#if _runtime(_ObjC)
|
|
// At runtime, these classes are derived from `__SwiftNativeNSXXXBase`,
|
|
// which are derived from `NSXXX`.
|
|
//
|
|
// The @swift_native_objc_runtime_base attribute
|
|
// allows us to subclass an Objective-C class and still use the fast Swift
|
|
// memory allocator.
|
|
//
|
|
// NOTE: older runtimes called these _SwiftNativeNSXXX. The two must
|
|
// coexist, so they were renamed. The old names must not be used in the
|
|
// new runtime.
|
|
|
|
@_fixed_layout
|
|
@usableFromInline
|
|
@objc @_swift_native_objc_runtime_base(__SwiftNativeNSArrayBase)
|
|
internal class __SwiftNativeNSArray {
|
|
@inlinable
|
|
@nonobjc
|
|
internal init() {}
|
|
// @objc public init(coder: AnyObject) {}
|
|
@inlinable
|
|
deinit {}
|
|
}
|
|
|
|
@available(*, unavailable)
|
|
extension __SwiftNativeNSArray: Sendable {}
|
|
|
|
@_fixed_layout
|
|
@usableFromInline
|
|
@objc @_swift_native_objc_runtime_base(__SwiftNativeNSMutableArrayBase)
|
|
internal class _SwiftNativeNSMutableArray {
|
|
@inlinable
|
|
@nonobjc
|
|
internal init() {}
|
|
// @objc public init(coder: AnyObject) {}
|
|
@inlinable
|
|
deinit {}
|
|
}
|
|
|
|
@available(*, unavailable)
|
|
extension _SwiftNativeNSMutableArray: Sendable {}
|
|
|
|
@_fixed_layout
|
|
@usableFromInline
|
|
@objc @_swift_native_objc_runtime_base(__SwiftNativeNSDictionaryBase)
|
|
internal class __SwiftNativeNSDictionary {
|
|
@nonobjc
|
|
internal init() {}
|
|
@objc public init(coder: AnyObject) {}
|
|
deinit {}
|
|
}
|
|
|
|
@available(*, unavailable)
|
|
extension __SwiftNativeNSDictionary: Sendable {}
|
|
|
|
@_fixed_layout
|
|
@usableFromInline
|
|
@objc @_swift_native_objc_runtime_base(__SwiftNativeNSSetBase)
|
|
internal class __SwiftNativeNSSet {
|
|
@nonobjc
|
|
internal init() {}
|
|
@objc public init(coder: AnyObject) {}
|
|
deinit {}
|
|
}
|
|
|
|
@available(*, unavailable)
|
|
extension __SwiftNativeNSSet: Sendable {}
|
|
|
|
@objc
|
|
@_swift_native_objc_runtime_base(__SwiftNativeNSEnumeratorBase)
|
|
internal class __SwiftNativeNSEnumerator {
|
|
@nonobjc
|
|
internal init() {}
|
|
@objc public init(coder: AnyObject) {}
|
|
deinit {}
|
|
}
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
// Support for reliable testing of the return-autoreleased optimization
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
@objc
|
|
internal class __stdlib_ReturnAutoreleasedDummy {
|
|
@objc
|
|
internal init() {}
|
|
|
|
// Use 'dynamic' to force Objective-C dispatch, which uses the
|
|
// return-autoreleased call sequence.
|
|
@objc
|
|
internal dynamic func returnsAutoreleased(_ x: AnyObject) -> AnyObject {
|
|
return x
|
|
}
|
|
}
|
|
|
|
/// This function ensures that the return-autoreleased optimization works.
|
|
///
|
|
/// On some platforms (for example, x86_64), the first call to
|
|
/// `objc_autoreleaseReturnValue` will always autorelease because it would fail
|
|
/// to verify the instruction sequence in the caller. On x86_64 certain PLT
|
|
/// entries would be still pointing to the resolver function, and sniffing
|
|
/// the call sequence would fail.
|
|
///
|
|
/// This code should live in the core stdlib dylib because PLT tables are
|
|
/// separate for each dylib.
|
|
///
|
|
/// Call this function in a fresh autorelease pool.
|
|
public func _stdlib_initializeReturnAutoreleased() {
|
|
#if arch(x86_64)
|
|
// On x86_64 it is sufficient to perform one cycle of return-autoreleased
|
|
// call sequence in order to initialize all required PLT entries.
|
|
let dummy = __stdlib_ReturnAutoreleasedDummy()
|
|
_ = dummy.returnsAutoreleased(dummy)
|
|
#endif
|
|
}
|
|
#else
|
|
|
|
@_fixed_layout
|
|
@usableFromInline
|
|
internal class __SwiftNativeNSArray {
|
|
@inlinable
|
|
internal init() {}
|
|
@inlinable
|
|
deinit {}
|
|
}
|
|
@_fixed_layout
|
|
@usableFromInline
|
|
internal class __SwiftNativeNSDictionary {
|
|
@inlinable
|
|
internal init() {}
|
|
@inlinable
|
|
deinit {}
|
|
}
|
|
@_fixed_layout
|
|
@usableFromInline
|
|
internal class __SwiftNativeNSSet {
|
|
@inlinable
|
|
internal init() {}
|
|
@inlinable
|
|
deinit {}
|
|
}
|
|
|
|
#endif
|