Merge pull request #42401 from glessard/se-unaligned-5.7

[5.7][SE-0349] Implementation for unaligned loads from raw memory

Author: glessard

CHANGELOG.md

@@ -5,6 +5,30 @@ _**Note:** This is in reverse chronological order, so newer entries are added to
 
 ## Swift 5.7
 
+* [SE-0349][]:
+
+  Loading data from raw memory represented by `UnsafeRawPointer`,
+  `UnsafeRawBufferPointer` and their mutable counterparts now supports unaligned
+  accesses. This previously required a workaround involving an intermediate
+  copy:
+
+  ```swift
+  let result = unalignedData.withUnsafeBytes { buffer -> UInt32 in
+    var storage = UInt32.zero
+    withUnsafeMutableBytes(of: &storage) {
+      $0.copyBytes(from: buffer.prefix(MemoryLayout<UInt32>.size))
+    }
+    return storage
+  }
+  ```
+  Now:
+  ```swift
+  let result = unalignedData.withUnsafeBytes { $0.loadUnaligned(as: UInt32.self) }
+  ```
+  Additionally, the counterpart `storeBytes(of:toByteOffset:as:)` had its
+  alignment restriction lifted, so that storing to arbitrary offsets of raw
+  memory can now succeed.
+
 * References to `optional` methods on a protocol metatype, as well as references to dynamically looked up methods on the `AnyObject` metatype are now supported. These references always have the type of a function that accepts a single argument and returns an optional value of function type:
 
   ```swift
@@ -9203,6 +9227,7 @@ Swift 1.0
 [SE-0345]: <https://github.com/apple/swift-evolution/blob/main/proposals/0345-if-let-shorthand.md>
 [SE-0326]: <https://github.com/apple/swift-evolution/blob/main/proposals/0326-extending-multi-statement-closure-inference.md>
 [SE-0347]: <https://github.com/apple/swift-evolution/blob/main/proposals/0347-type-inference-from-default-exprs.md>
+[SE-0349]: <https://github.com/apple/swift-evolution/blob/main/proposals/0349-unaligned-loads-and-stores.md>
 [SE-0352]: <https://github.com/apple/swift-evolution/blob/main/proposals/0352-implicit-open-existentials.md>
 
 [SR-75]: <https://bugs.swift.org/browse/SR-75>

stdlib/public/core/UnsafeRawBufferPointer.swift.gyb

@@ -381,6 +381,49 @@ extension Unsafe${Mutable}RawBufferPointer {
     return baseAddress!.load(fromByteOffset: offset, as: T.self)
   }
 
+  /// Returns a new instance of the given type, constructed from the raw memory
+  /// at the specified offset.
+  ///
+  /// This function only supports loading trivial types.
+  /// A trivial type does not contain any reference-counted property
+  /// within its in-memory stored representation.
+  /// The memory at `offset` bytes into the buffer must be laid out
+  /// identically to the in-memory representation of `T`.
+  ///
+  /// You can use this method to create new values from the buffer pointer's
+  /// underlying bytes. The following example creates two new `Int32`
+  /// instances from the memory referenced by the buffer pointer `someBytes`.
+  /// The bytes for `a` are copied from the first four bytes of `someBytes`,
+  /// and the bytes for `b` are copied from the next four bytes.
+  ///
+  ///     let a = someBytes.load(as: Int32.self)
+  ///     let b = someBytes.load(fromByteOffset: 4, as: Int32.self)
+  ///
+  /// The memory to read for the new instance must not extend beyond the buffer
+  /// pointer's memory region---that is, `offset + MemoryLayout<T>.size` must
+  /// be less than or equal to the buffer pointer's `count`.
+  ///
+  /// - Parameters:
+  ///   - offset: The offset, in bytes, into the buffer pointer's memory at
+  ///     which to begin reading data for the new instance. The buffer pointer
+  ///     plus `offset` must be properly aligned for accessing an instance of
+  ///     type `T`. The default is zero.
+  ///   - type: The type to use for the newly constructed instance. The memory
+  ///     must be initialized to a value of a type that is layout compatible
+  ///     with `type`.
+  /// - Returns: A new instance of type `T`, copied from the buffer pointer's
+  ///   memory.
+  @_alwaysEmitIntoClient
+  public func loadUnaligned<T>(
+    fromByteOffset offset: Int = 0,
+    as type: T.Type
+  ) -> T {
+    _debugPrecondition(offset >= 0, "${Self}.load with negative offset")
+    _debugPrecondition(offset + MemoryLayout<T>.size <= self.count,
+      "${Self}.load out of bounds")
+    return baseAddress!.loadUnaligned(fromByteOffset: offset, as: T.self)
+  }
+
 %  if mutable:
   /// Stores a value's bytes into the buffer pointer's raw memory at the
   /// specified byte offset.
@@ -399,6 +442,16 @@ extension Unsafe${Mutable}RawBufferPointer {
   /// type `U`. Calling `storeBytes(of:toByteOffset:as:)` does not change the
   /// bound type of the memory.
   ///
+  /// - Note: A trivial type can be copied with just a bit-for-bit copy without
+  ///   any indirection or reference-counting operations. Generally, native
+  ///   Swift types that do not contain strong or weak references or other
+  ///   forms of indirection are trivial, as are imported C structs and enums.
+  ///
+  /// If you need to store into memory a copy of a value of a type that isn't
+  /// trivial, you cannot use the `storeBytes(of:toByteOffset:as:)` method.
+  /// Instead, you must know either initialize the memory or,
+  /// if you know the memory was already bound to `type`, assign to the memory.
+  ///
   /// - Parameters:
   ///   - value: The value to store as raw bytes.
   ///   - offset: The offset in bytes into the buffer pointer's memory to begin
@@ -409,14 +462,34 @@ extension Unsafe${Mutable}RawBufferPointer {
   ///     must be initialized to a value of a type that is layout compatible
   ///     with `type`.
   @inlinable
+  @_alwaysEmitIntoClient
+  @_silgen_name("_swift_se0349_UnsafeMutableRawBufferPointer_storeBytes")
   public func storeBytes<T>(
-    of value: T, toByteOffset offset: Int = 0, as: T.Type
+    of value: T, toByteOffset offset: Int = 0, as type: T.Type
+  ) {
+    _debugPrecondition(offset >= 0, "${Self}.storeBytes with negative offset")
+    _debugPrecondition(offset + MemoryLayout<T>.size <= self.count,
+      "${Self}.storeBytes out of bounds")
+
+    let pointer = baseAddress._unsafelyUnwrappedUnchecked
+    pointer.storeBytes(of: value, toByteOffset: offset, as: T.self)
+  }
+
+  // This unavailable implementation uses the expected mangled name
+  // of `storeBytes<T>(of:toByteOffset:as:)`, and provides an entry point for
+  // any binary compiled against the stlib binary for Swift 5.6 and older.
+  @available(*, unavailable)
+  @_silgen_name("$sSw10storeBytes2of12toByteOffset2asyx_SixmtlF")
+  @usableFromInline func _legacy_se0349_storeBytes<T>(
+    of value: T, toByteOffset offset: Int = 0, as type: T.Type
   ) {
     _debugPrecondition(offset >= 0, "${Self}.storeBytes with negative offset")
     _debugPrecondition(offset + MemoryLayout<T>.size <= self.count,
       "${Self}.storeBytes out of bounds")
 
-    baseAddress!.storeBytes(of: value, toByteOffset: offset, as: T.self)
+    baseAddress!._legacy_se0349_storeBytes_internal(
+      of: value, toByteOffset: offset, as: T.self
+    )
   }
 
   /// Copies the bytes from the given buffer to this buffer's memory.

stdlib/public/core/UnsafeRawPointer.swift

@@ -424,7 +424,6 @@ public struct UnsafeRawPointer: _Pointer {
     let rawPointer = (self + offset)._rawValue
 
 #if compiler(>=5.5) && $BuiltinAssumeAlignment
-    // TODO: to support misaligned raw loads, simply remove this assumption.
     let alignedPointer =
       Builtin.assumeAlignment(rawPointer,
                               MemoryLayout<T>.alignment._builtinWordValue)
@@ -434,6 +433,36 @@ public struct UnsafeRawPointer: _Pointer {
 #endif
   }
 
+  /// Returns a new instance of the given type, constructed from the raw memory
+  /// at the specified offset.
+  ///
+  /// This function only supports loading trivial types,
+  /// and will trap if this precondition is not met.
+  /// A trivial type does not contain any reference-counted property
+  /// within its in-memory representation.
+  /// The memory at this pointer plus `offset` must be laid out
+  /// identically to the in-memory representation of `T`.
+  ///
+  /// - Note: A trivial type can be copied with just a bit-for-bit copy without
+  ///   any indirection or reference-counting operations. Generally, native
+  ///   Swift types that do not contain strong or weak references or other
+  ///   forms of indirection are trivial, as are imported C structs and enums.
+  ///
+  /// - Parameters:
+  ///   - offset: The offset from this pointer, in bytes. `offset` must be
+  ///     nonnegative. The default is zero.
+  ///   - type: The type of the instance to create.
+  /// - Returns: A new instance of type `T`, read from the raw bytes at
+  ///   `offset`. The returned instance isn't associated
+  ///   with the value in the range of memory referenced by this pointer.
+  @_alwaysEmitIntoClient
+  public func loadUnaligned<T>(
+    fromByteOffset offset: Int = 0,
+    as type: T.Type
+  ) -> T {
+    _debugPrecondition(_isPOD(T.self))
+    return Builtin.loadRaw((self + offset)._rawValue)
+  }
 }
 
 extension UnsafeRawPointer: Strideable {
@@ -1120,7 +1149,6 @@ public struct UnsafeMutableRawPointer: _Pointer {
     let rawPointer = (self + offset)._rawValue
 
 #if compiler(>=5.5) && $BuiltinAssumeAlignment
-    // TODO: to support misaligned raw loads, simply remove this assumption.
     let alignedPointer =
       Builtin.assumeAlignment(rawPointer,
                               MemoryLayout<T>.alignment._builtinWordValue)
@@ -1130,11 +1158,41 @@ public struct UnsafeMutableRawPointer: _Pointer {
 #endif
   }
 
+  /// Returns a new instance of the given type, constructed from the raw memory
+  /// at the specified offset.
+  ///
+  /// This function only supports loading trivial types,
+  /// and will trap if this precondition is not met.
+  /// A trivial type does not contain any reference-counted property
+  /// within its in-memory representation.
+  /// The memory at this pointer plus `offset` must be laid out
+  /// identically to the in-memory representation of `T`.
+  ///
+  /// - Note: A trivial type can be copied with just a bit-for-bit copy without
+  ///   any indirection or reference-counting operations. Generally, native
+  ///   Swift types that do not contain strong or weak references or other
+  ///   forms of indirection are trivial, as are imported C structs and enums.
+  ///
+  /// - Parameters:
+  ///   - offset: The offset from this pointer, in bytes. `offset` must be
+  ///     nonnegative. The default is zero.
+  ///   - type: The type of the instance to create.
+  /// - Returns: A new instance of type `T`, read from the raw bytes at
+  ///   `offset`. The returned instance isn't associated
+  ///   with the value in the range of memory referenced by this pointer.
+  @_alwaysEmitIntoClient
+  public func loadUnaligned<T>(
+    fromByteOffset offset: Int = 0,
+    as type: T.Type
+  ) -> T {
+    _debugPrecondition(_isPOD(T.self))
+    return Builtin.loadRaw((self + offset)._rawValue)
+  }
+
   /// Stores the given value's bytes into raw memory at the specified offset.
   ///
-  /// The type `T` to be stored must be a trivial type. The memory at this
-  /// pointer plus `offset` must be properly aligned for accessing `T`. The
-  /// memory must also be uninitialized, initialized to `T`, or initialized to
+  /// The type `T` to be stored must be a trivial type. The memory
+  /// must also be uninitialized, initialized to `T`, or initialized to
   /// another trivial type that is layout compatible with `T`.
   ///
   /// After calling `storeBytes(of:toByteOffset:as:)`, the memory is
@@ -1148,14 +1206,14 @@ public struct UnsafeMutableRawPointer: _Pointer {
   ///   Swift types that do not contain strong or weak references or other
   ///   forms of indirection are trivial, as are imported C structs and enums.
   ///
-  /// If you need to store a copy of a nontrivial value into memory, or to
-  /// store a value into memory that contains a nontrivial value, you cannot
-  /// use the `storeBytes(of:toByteOffset:as:)` method. Instead, you must know
-  /// the type of value previously in memory and initialize or assign the
-  /// memory. For example, to replace a value stored in a raw pointer `p`,
+  /// If you need to store into memory a copy of a value of a type that isn't
+  /// trivial, you cannot use the `storeBytes(of:toByteOffset:as:)` method.
+  /// Instead, you must know either initialize the memory or,
+  /// if you know the memory was already bound to `type`, assign to the memory.
+  /// For example, to replace a value stored in a raw pointer `p`,
   /// where `U` is the current type and `T` is the new type, use a typed
   /// pointer to access and deinitialize the current value before initializing
-  /// the memory with a new value.
+  /// the memory with a new value:
   ///
   ///     let typedPointer = p.bindMemory(to: U.self, capacity: 1)
   ///     typedPointer.deinitialize(count: 1)
@@ -1167,8 +1225,41 @@ public struct UnsafeMutableRawPointer: _Pointer {
   ///     nonnegative. The default is zero.
   ///   - type: The type of `value`.
   @inlinable
+  @_alwaysEmitIntoClient
+  @_silgen_name("_swift_se0349_UnsafeMutableRawPointer_storeBytes")
   public func storeBytes<T>(
     of value: T, toByteOffset offset: Int = 0, as type: T.Type
+  ) {
+    _debugPrecondition(_isPOD(T.self))
+
+    withUnsafePointer(to: value) { source in
+      // FIXME: to be replaced by _memcpy when conversions are implemented.
+      Builtin.int_memcpy_RawPointer_RawPointer_Int64(
+        (self + offset)._rawValue,
+        source._rawValue,
+        UInt64(MemoryLayout<T>.size)._value,
+        /*volatile:*/ false._value
+      )
+    }
+  }
+
+  // This unavailable implementation uses the expected mangled name
+  // of `storeBytes<T>(of:toByteOffset:as:)`, and provides an entry point for
+  // any binary compiled against the stlib binary for Swift 5.6 and older.
+  @available(*, unavailable)
+  @_silgen_name("sSv10storeBytes2of12toByteOffset2asyx_SixmtlF")
+  @usableFromInline func _legacy_se0349_storeBytes<T>(
+    of value: T, toByteOffset offset: Int = 0, as type: T.Type
+  ) {
+    _legacy_se0349_storeBytes_internal(
+      of: value, toByteOffset: offset, as: T.self
+    )
+  }
+
+  // This is the implementation of `storeBytes` from SwiftStdlib 5.6
+  @_alwaysEmitIntoClient
+  internal func _legacy_se0349_storeBytes_internal<T>(
+    of value: T, toByteOffset offset: Int = 0, as type: T.Type
   ) {
     _debugPrecondition(0 == (UInt(bitPattern: self + offset)
         & (UInt(MemoryLayout<T>.alignment) - 1)),
@@ -1183,7 +1274,7 @@ public struct UnsafeMutableRawPointer: _Pointer {
         /*volatile:*/ false._value)
     }
   }
-  
+
   /// Copies the specified number of bytes from the given raw pointer's memory
   /// into this pointer's memory.
   ///

test/api-digester/stability-stdlib-abi-without-asserts.test

@@ -89,5 +89,7 @@ Func UnsafeBufferPointer.withMemoryRebound(to:_:) has been removed
 Func UnsafeMutableBufferPointer.withMemoryRebound(to:_:) has been removed
 Func UnsafeMutablePointer.withMemoryRebound(to:capacity:_:) has been removed
 Func UnsafePointer.withMemoryRebound(to:capacity:_:) has been removed
+Func UnsafeMutableRawBufferPointer.storeBytes(of:toByteOffset:as:) has been removed
+Func UnsafeMutableRawPointer.storeBytes(of:toByteOffset:as:) has been removed
 
 // *** DO NOT DISABLE OR XFAIL THIS TEST. *** (See comment above.)