Previously, metadata prespecialization for classes only occurred when
all of a specialized generic class's ancestors were themselves generic.
Here, that requirement is lifted so that generic classes with concrete
ancestors are also eligible for prespecialization.
Use the new builtins for COW representation in Array, ContiguousArray and ArraySlice.
The basic idea is to strictly separate code which mutates an array buffer from code which reads from an array.
The concept is explained in more detail in docs/SIL.rst, section "Copy-on-Write Representation".
The main change is to use beginCOWMutation() instead of isUniquelyReferenced() and insert endCOWMutation() at the end of all mutating functions. Also, reading from the array buffer must be done differently, depending on if the buffer is in a mutable or immutable state.
All the required invariants are enforced by runtime checks - but only in an assert-build of the library: a bit in the buffer object side-table indicates if the buffer is mutable or not.
Along with the library changes, also two optimizations needed to be updated: COWArrayOpt and ObjectOutliner.
In an assert built of the library, store an extra boolean flag (isImmutable) in the object side-buffer table.
This flag can be set and get by the Array implementation to sanity check the immutability status of the buffer object.
For COW support in SIL it's required to "finalize" array literals.
_finalizeUninitializedArray is a compiler known stdlib function which is called after all elements of an array literal are stored.
This runtime function marks the array literal as finished.
%uninitialized_result_tuple = apply %_allocateUninitializedArray(%count)
%mutable_array = tuple_extract %uninitialized_result_tuple, 0
%elem_base_address = tuple_extract %uninitialized_result_tuple, 1
...
store %elem_0 to %elem_addr_0
store %elem_1 to %elem_addr_1
...
%final_array = apply %_finalizeUninitializedArray(%mutable_array)
In this commit _finalizeUninitializedArray is still a no-op because the COW support is not used in the Array implementation yet.
`swiftDemangling` was built three times:
1. swiftc
2. swiftRuntime
3. swiftReflection
Fold the last two instances into a single build, sharing the objects
across both the target libraries. This ensures that `swiftDemangling`
is built with the same compiler as the target libraries and that the
target library build remains self-contained.
This has two advantages:
1. It does not force the Array in memory (to pass it as inout self to the non-inlinable _createNewBuffer).
2. The new _consumeAndCreateNew is annotated to consume self. This helps to reduce unnecessary retains/releases.
The change applies for Array and ContiguousArray.
In the corner case of a 0 sized replacement in an empty array, we did write the 0 count back to the array singleton.
This is not a big problem right now, because it would just overwrite a 0 with a 0, but it shouldn't be done.
But with COW representation in Array, it would break the sanity checks.
In order to fully optimize OptionSet literals, it's important that this function is inlined and fully optimized.
So far this was done by chance, but with COW representation it needs a hint to the optimizer.
As seen in SR-12881, `DefaultIndices` was not properly dispatching certain `Collection` requirements through conditional conformances to `BidirectionalCollection` and `RandomAccessCollection`. This should fix that.
The previous documentation of this operator in `Range` context was lost sometime during 2017. This new version is a simplified version of the original copy, with substantial inspiration taken from the present-day documentation of the same operator on `Optional`.
Fixes [SR-12842](https://bugs.swift.org/browse/SR-12842).
Rather than build multiple copies of LLVMSupport (4x!) build it one and
merge it into the various targets. This would ideally not be needed to
be named explicitly everywhere, but that requires using `add_library`
rather than `add_swift_target_library`.
