Previously even if a type's metadata was optimized away, we would still
emit a field descriptor, which in turn could reference nominal type
descriptors for other types via symbolic references, etc.
Instead of a wholly separate lazyness mechanism for foreign metadata where
the first call to getAddrOfForeignTypeMetadataCandidate() would emit the
metadata, emit it using the lazy metadata mechanism.
This eliminates some code duplication. It also ensures that foreign
metadata is only emitted once per SIL module, and not once per LLVM
module, avoiding duplicate copies that must be ODR'd away in multi-threaded
mode.
This fixes the test case from <rdar://problem/49710077>.
Use the `IRLInkage::InternalLinkOnceODR` linkage rather than computing
that in a couple of sites. This gives us semantic meaning to the
linkage being applied. NFC.
When we emit "false" symbolic references to accesors for conformances or
type metadata, ensure that we end up with two-byte-aligned symbolic
references. This addresses a problem with ARM+Thumb compilation where the
low bit wasn't getting set for Thumb code.
Fixes rdar://problem/46067353.
Switch key path metadata over to mangled names for each of the places it
refers to either a type metadata accessor or a witness table accessor. For
now, the mangled name is a symbolic reference to the existing accessors.
Part of rdar://problem/38038799.
The current representation of an associated conformance in a witness
tables (e.g., Iterator: IteratorProtocol within a witness table for
Sequence) is a function that the client calls.
Replace this with something more like what we do for associated types:
an associated conformance is either a pointer to the witness table (once
it is known) or a pointer to a mangled name that describes that
conformance. On first access, demangle the mangled name and replace the
entry with the resulting witness table. This will give us a more compact
representation of associated conformances, as well as always caching
them.
For now, the mangled name is a sham: it’s a mangled relative reference to
the existing witness table accessors, not a true mangled name. In time,
we’ll extend the support here to handle proper mangled names.
Part of rdar://problem/38038799.
Bitwise takability is now part of the layout of a type, because
non-bitwise takable types are never stored inline in an
existential or resilient global's buffer, even if they would
fit.
The basic rule is that weak references, unknown-refcounted
unowned references, and aggregates that contain them, are not
bitwise takable, whereas everything else is bitwise takable.
Also, since the bitwise takable for an unowned reference
depends on the reference counting style, we have to record the
superclass of a protocol, if any, to correctly determine the
reference counting style of the protocol existential.
In order to handle LinkOnceODR semantics correctly across various object
formats, introduce a new helper ApplyIRLinkage. This abstracts the need
to create a COMDAT group and set it on the GlobalValue. Adjust all
sites where we set the IR linkage attributes to use this mechanism
instead to avoid having to track down symbols not being added to a
COMDAT group.
Encode default associated type witnesses using a sentinel prefix byte
(0xFF) in the mangled name rather than as a second low bit on the
reference. Align all of the mangled names used for type references to
2 bytes (so we get that low bit regardless) and separate the symbol
names for default associated type witnesses vs. other kinds of
metadata or reflection metadata.
Rather than storing associated type metadata access functions in
witness tables, initially store a pointer to a mangled type name.
On first access, demangle that type name and replace the witness
table entry with the resulting type metadata.
This reduces the code size of protocol conformances, because we no
longer need to create associated type metadata access functions for
every associated type, and the mangled names are much smaller (and
sharable). The same code size improvements apply to defaulted
associated types for resilient protocols, although those are more
rare. Witness tables themselves are slightly smaller, because we
don’t need separate private entries in them to act as caches.
On the caller side, associated type metadata is always produced via
a call to swift_getAssociatedTypeWitness(), which handles the demangling
and caching behavior.
In all, this reduces the size of the standard library by ~70k. There
are additional code-size wins that are possible with follow-on work:
* We can stop emitting type metadata access functions for non-resilient
types that have constant metadata (like `Int`), because they’re only
currently used as associated type metadata access functions.
* We can stop emitting separate associated type reflection metadata,
because the reflection infrastructure can use these mangled names
directly.
This makes it easier to grep for and eventually remove the
remaining usages.
It also allows you to write FunctionType::get({}, ...) to call the
ArrayRef overload empty parameter list, instead of picking the Type
overload and calling it with an empty Type() value.
While I"m at it, in a few places instead of renaming just clean up
usages where it was completely mechanical to do so.
This saves us some expensive cross-referencing and caching in the runtime, and lets us reclaim the `isReflectable` bit from the context descriptor flags (since a null field descriptor is a suitable and more accurate indicator of whether a type is reflectable).
We’re not using this parameter, and don’t expect to do so in the future,
so remove it. Also fold away TypeBase::usesNativeReferenceCounting()
and irgen::getReferenceCountingForType(), both of which are trivial.
The TypeDecoder doesn't support the new box mangling yet and instead
just decodes it as Builtin.NativeObject, but that's OK because the
Remote Mirrors lowered the old box mangling as Builtin.NativeObject
anyway.
This makes resolving mangled names to nominal types in the same module more efficient, and for eventual secrecy improvements, also allows types in the same module to be referenced from mangled typerefs without encoding any source-level name information about them.
Update IRGen to trigger generation of type metadata for foreign
struct types found in fields. And fix TypeRefBuilder to handle
the case where struct has fields but at the same time has opaque
metadata.
The mangled name of protocol descriptors was the “protocol composition”
type consisting of a single protocol, which is a little odd. Instead,
use a bare protocol reference (e.g., “6Module5ProtoP”) with the “$S”
prefer to be more in line with nominal type descriptor names while still
making it clear that this is a Swift (not an Objective-C) protocol.
Restructure the COFF metadata handling to use the linker section
grouping to emit section start/stop markers in the appropriate location.
This allows us to lookup the sections statically without having to the
walk the entire image structure.
Introduce a constructor for PE/COFF binaries. This will ensure that the
registration occurs for all modules appropriately. This should resolve
rdar://problem/19045112. The registration should occur prior to
`DllMain` being invoked from `DllMainCRTStartup`.
Restructure the ELF handling to be completely agnostic to the OS.
Rather than usng the loader to query the section information, use the
linker to construct linker tables and synthetic markers for the
beginning and of the table. Save off the values of these pointers and
pass them along through the constructor to the runtime for registration.
This removes the need for the begin/end objects. Remove the special
construction of the begin/end objects through the special assembly
constructs, preferring to do this in C with a bit of inline assembly to
ensure that the section is always allocated.
Remove the special handling for the various targets, the empty object
file can be linked on all the targets.
The new object file has no requirements on the ordering. It needs to
simply be injected into the link.
Name the replacement file `swiftrt.o` mirroring `crt.o` from libc. Merge
the constructor and the definition into a single object file.
This approach is generally more portable, overall simpler to implement,
and more robust.
Thanks to Orlando Bassotto for help analyzing some of the odd behaviours
when switching over.
Rather than storing contextual types in the type witnesses and associated
conformances of NormalProtocolConformance, store only interface types.
@huonw did most of the work here, and @DougGregor patched things up to
complete the change.
@objc enums lower as their raw type, so should go through the same
code path as imported enums.
Fixes <https://bugs.swift.org/browse/SR-5625> and
<rdar://problem/33683103>.
