Many, many, many types in the Swift compiler are intended to only be allocated in the ASTContext. We have previously implemented this by writing several `operator new` and `operator delete` implementations into these types. Factor those out into a new base class instead.
After converting a function type to its corresponding SIL type by our usual rules, extract
the substituted generic signature and common interface type for all concrete function types
that the function could be called as. We don't currently have any use cases for distinguishing
types at finer granularity than generic class-constrained/generic unconstrained/not-generic, so
to reduce the amount of conversion noise in SIL, generate the most general generic signature
possible, by extracting every position with a dependent generic parameter or associated type thereof
into a separate generic parameter in the substituted signature, discarding all constraints except
for `AnyObject` layout constraints. This way, if `foo<T>(x: (T, T) -> ())` calls
`bar<T, U>(x: (T, U) -> ())` and forwards `x` along, the two closures that only vary in genericity
don't need a conversion even with substituted function types, because they'll both have the same
abstract lowering `<A, B> (A, B) -> ()` with different substitutions.
By convention, most structs and classes in the Swift compiler include a `dump()` method which prints debugging information. This method is meant to be called only from the debugger, but this means they’re often unused and may be eliminated from optimized binaries. On the other hand, some parts of the compiler call `dump()` methods directly despite them being intended as a pure debugging aid. clang supports attributes which can be used to avoid these problems, but they’re used very inconsistently across the compiler.
This commit adds `SWIFT_DEBUG_DUMP` and `SWIFT_DEBUG_DUMPER(<name>(<params>))` macros to declare `dump()` methods with the appropriate set of attributes and adopts this macro throughout the frontend. It does not pervasively adopt this macro in SILGen, SILOptimizer, or IRGen; these components use `dump()` methods in a different way where they’re frequently called from debugging code. Nor does it adopt it in runtime components like swiftRuntime and swiftReflection, because I’m a bit worried about size.
Despite the large number of files and lines affected, this change is NFC.
Introduce a request to form an abstract generic signature given a
base signature, additional generic parameters, and additional
requirements. It is meant to provide a caching layer in front of the
generic signature builder.
Switch one direct client of the generic signature builder over to this
mechanism, the formation of a generic signature for an existential
type.
Introduce a new form of request that produces the ith Requirement of the
where clause of a given entity, which could be a protocol, associated type,
or anything with generic parameters. The requirement can be evaluated
at any type resolution stage; the “interface” type stage will be cached
in the existing RequirementReprs.
Fixes SR-8119 / rdar://problem/41498944.
We compile with a pedantic warning that complains about these things,
and the massive flood of warnings is actually causing problems for the
build infrastructure.
This PR addresses TODOs from #8241.
- It supports merging for layout constraints, e.g., if both a _Trivial constraint and a _Trivial(64) constraint appear on a type parameter, we keep only _Trivial(64) as a more specific layout constraint. We do a similar thing for ref-counted/native-ref-counted. The overall idea is to keep the more specific of two compatible layout constraints.
- The presence of a superclass constraint implies a layout constraint, e.g., a superclass constraint implies _Class or _NativeClass
- Add support for _Class and _NativeClass layout constraints, which are supposed to represent T: Superclass and P: class constraints in the future.
- Use the re-factoring to also reduce the number of dynamic allocations when creating layout constraints. Simple non-parametrized layout constraints are now represented as statically allocated singletons (static members of LayoutConstraintInfo).
This biggest change is:
- LayoutConstraintInfo is now a FoldingSetNode, which allows for proper canonicalization of LayoutConstraints. This is important for the correctness of type comparisons if types contain layout constraints.
No functionality changes from the client's point of view.
The new `@_specialize` attribute has a syntax like this:
```swift
@_specialize(exported: true, kind: full, where K == Int, V == Int)
@_specialize(exported: false, kind: partial, where K: _Trivial64)
func dictFunction<K, V>(dict: Dictionary<K, V>) {
}
```
If `exported` is set, the corresponding specialization would have a public visibility and can be used by clients.
If `exported` is omitted, it's value is assumed to be `false`.
If `kind` is `full` it means a full specialization.
If `kind` is `partial` it means a partial specialization.
If `kind` is omitted, its value is assumed to be `full`.
