External private definitions coming from a -sil-serialize-all
module must be effectively public at the LLVM level too, fixing
a linker error when referencing closures from a deserialized
function body.
Also, add a third [serializable] state for functions whose bodies we
*can* serialize, but only do so if they're referenced from another
serialized function.
This will be used for bodies synthesized for imported definitions,
such as init(rawValue:), etc, and various thunks, but for now this
change is NFC.
This gives big code size wins for unused types and also for types, which are never used in a generic context.
Also it reduces the amount of symbols in the symbol table.
The size wins heavily depend on the project. I have seen binary size reductions from 0 to 20% on real world projects.
rdar://problem/30119960
This is NFC in intent, but I had to restructure the code to emit more
of the lists "inline", which means I inevitably altered some IRGen
emission patterns in ways that are visible to tests:
- GenClass emits property/ivar/whatever descriptors in a somewhat
different order.
- An ext method type list is now emitted as just an array, not a struct
containing only that array.
- Protocol descriptors are no longer emitted as packed structs.
I was sorely tempted to stop using packed structs for all the metadata
emission, but didn't really want to update that many tests in one go.
to correctly handle generalized protocol requirements.
The major missing pieces here are that the conformance search
algorithms in both the AST (type substitution) and IRGen
(witness table reference emission) need to be rewritten to
back-track requirement sources, and the AST needs to actually
represent this stuff in NormalProtocolConformances instead
of just doing ???.
The new generality isn't tested yet; I'm looking into that,
but I wanted to get the abstractions in place first.
Swift's readonly and readnone should not be automatically mapped to LLVM's readonly.
Swift SIL optimizer relies on @effects(readonly) to remove e.g. dead code remaining
from initializers of strings or dictionaries of variables that are not used. But
those initializers are often not really readonly in terms of LLVM IR. For
example, the Dictionary.init() is marked as @effects(readonly) in Swift, but it
does invoke reference-counting operations. As a result, it leads to miscompiles
and runtime crashes.
In the future, we may add the functionality to analyze the body of the function
and if we can prove that is really readonly, we can map it to the LLVM readonly
attribute.
I checked that removal of this mapping does not affect the performance of any of
our existing benchmarks.
Fixes rdar://problem/28830504
Separate formal lowered types from SIL types.
The SIL type of an argument will depend on the SIL module's conventions.
The module conventions are determined by the SIL stage and LangOpts.
Almost NFC, but specialized manglings are broken incidentally as a result of
fixes to the way passes handle book-keeping of aruments. The mangler is fixed in
the subsequent commit.
Otherwise, NFC is intended, but quite possible do to rewriting the logic in many
places.
