OpaqueStorageTypeInfo uses iNNN types that don't always have the correct alloc size for an expected
size at the LLVM level. This needs to be fixed before my partial apply closure fixes can hold.
Swift SVN r24551
In order to deal with generic indirect value captures, we need to be able to bind their type metadata in the partial apply forwarder and heap object destructor to have access to the value operations for that type. NecessaryBindings gives us a way to do that and clean up our ad-hoc polymorphic argument forwarding we had before. N(intended)FC yet, aside from some harmless reordering of operations, since we also need to implement NonFixedOffsets for heap objects to be fully operational.
Swift SVN r24526
- A spot fix in SILGen for reabstracting the result of a downcast, which fixes checked casts to function types.
- Associate the layout information in type metadata records with the most abstract representation of the type. This is the correct thing to do in cases where we need the metadata as a tag for an opaque value--if we store a value in an Any, or pass it as an unconstrained generic parameter, we must maximally reabstract it. This fixes the value semantics of existentials containing trivial metatypes.
- To ensure that we get runtime layout of structs and enums correct when they contain reabstractable types, introduce a "metadata for layout" concept, which doesn't need to describe the canonical metadata for the type, but only needs to describe a type with equivalent layout and value semantics. This is a correctness fix that allows us to correctly lay out generic types containing dependent tuples and functions, and although we don't really take advantage of it here, it's also a potential runtime performance win down the road, because we could potentially produce direct metadata for a primitive type that's layout-equivalent with a runtime-instantiated type. To aid in type safety here, push SILType deeper into IRGen in places where we potentially care about specific representations of types.
- Finally, fix an inconsistency between the runtime and IRGen's concept of what spare bits unmanaged references and thick metatypes have.
Together, these fixes address rdar://problem/16406907, rdar://problem/17822208, rdar://problem/18189508, and likely many other related issues, and also fixes crash suite cases 012 and 024.
Swift SVN r21963
IRGen type conversion is invariant with respect to archetypes with the same set of constraints, so instead of redundantly generating a TypeInfo object and IR type for Optional<T> for every T everywhere, key TypeInfo objects using an "exemplar type" that we form using a folding set to collapse together archetypes with the same class-ness, superclass constraint, and protocol constraints.
This is a nice memory and IR size optimization, but will be essential for correctness when lowering interface types, because there is no unique context to ground a dependent type, and we need to lower the same generic parameter with the same context requirements to the same type whenever we instantiate it in order for the IR to type-check.
In this revision, we profile the nested archetypes of each recursively, which I neglected to take into account originally in r12112, causing failures when archetypes that differed by associated type constraints were incorrectly collapsed.
Swift SVN r12116
IRGen type conversion is invariant with respect to archetypes with the same set of constraints, so instead of redundantly generating a TypeInfo object and IR type for Optional<T> for every T everywhere, key TypeInfo objects using an "exemplar type" that we form using a folding set to collapse together archetypes with the same class-ness, superclass constraint, and protocol constraints.
This is a nice memory and IR size optimization, but will be essential for correctness when lowering interface types, because there is no unique context to ground a dependent type, and we need to lower the same generic parameter with the same context requirements to the same type whenever we instantiate it in order for the IR to type-check.
Swift SVN r12112
Emit the deallocObject runtime call inside the deallocating destructor for a heap object, instead of inside swift_release. This will allow for heap objects with known size to directly call fast deallocator entry points and potentially custom deallocators in the future.
Swift SVN r5027
handling non-fixed layouts.
This uncovered a bug where we weren't rounding up the header
size to the element alignment when allocating an array of archetypes.
Writing up a detailed test case for *that* revealed that we were
never initializing the length field of heap arrays. Fixing that
caused a bunch of tests to crash trying to release stuff. So...
I've left this in a workaround state right now because I have to
catch a plane.
Swift SVN r4804
Implement ElementAddrInst for lvalue tuples, and implement the AllocArray, IndexAddr, and IntegerValue insts used to lower variadic tuples. (Actually compiling code that uses variadic tuples still requires support for SpecializeInst and generic functions.)
Swift SVN r3781
The test changes are that we're setting a class body on
some types that we weren't before. For some of these,
this is okay; for others, it's more questionable, but
ultimately not *harmful*.
Swift SVN r3746
The principal difficulty here is that we need accessing the
value witness table for a type to be an efficient operation,
but there (obviously) isn't a VWT field for ObjC classes.
Placing this field after the metatype would tend to bloat
metatypes by quite a bit. Placing it before is best, but
it introduces an unfortunate difference between the address
point of a metatype and the address of the global symbol.
That, however, can be fixed with appropriate linker support.
Still, for now this is rather unfortunately over-subtle.
Swift SVN r3307
The motivations here are that (1) the parametric types
that actually need the 'self' argument don't necessarily
all want to do what tuples do and put the VWT relative
to the metatype at some definable offset and (2)
recovering type parameters from the metatype is much
better defined than also hopping some relationship back.
Plus this allows VWTs to be shared across instances of
generic types. Also, I'm going to need to add a VW
that takes a metatype, and consistency seems right here.
If keeping two values live is actually punishing, I
might have to reconsider this. But the VWT is at least
always recoverable from the metatype, so....
I ended up abstracting the thing that GenHeap was doing
in order to save archetypes for arrays, because I
needed it to save metatypes instead of VWTs and because
it really needed abstractin'.
Swift SVN r3096
This is kindof a pain in a few places where the type system
doesn't propagate canonicality. Also, member initializations
are always direct-initializations and so are allowed to use
explicit constructors, which is a hole in our canonicality
tracking. But overall I like the idea of always working
with canonical types.
Swift SVN r2893
heap allocations it makes, and switch swift_alloc over to pass
that pointer in as well as the alignment. Also, compute
whether a type is POD during its generation and cache that in
the object, and introduce a method on TypeInfo to destroy an
object in memory.
Swift SVN r1356
