This is currently disabled by default. Building the client library can be enabled with the CMake option SWIFT_BUILD_CLIENT_RETAIN_RELEASE, and using the library can be enabled with the flags -Xfrontend -enable-client-retain-release.
To improve retain/release performance, we build a static library containing optimized implementations of the fast paths of swift_retain, swift_release, and the corresponding bridgeObject functions. This avoids going through a stub to make a cross-library call.
IRGen gains awareness of these new functions and emits calls to them when the functionality is enabled and the target supports them. Two options are added to force use of them on or off: -enable-client-retain-release and -disable-client-retain-release. When enabled, the compiler auto-links the static library containing the implementations.
The new calls also use LLVM's preserve_most calling convention. Since retain/release doesn't need a large number of scratch registers, this is mostly harmless for the implementation, while allowing callers to improve code size and performance by spilling fewer registers around refcounting calls. (Experiments with an even more aggressive calling convention preserving x2 and up showed an insignificant savings in code size, so preserve_most seems to be a good middle ground.)
Since the implementations are embedded into client binaries, any change in the runtime's refcounting implementation needs to stay compatible with this new fast path implementation. This is ensured by having the implementation use a runtime-provided mask to check whether it can proceed into its fast path. The mask is provided as the address of the absolute symbol _swift_retainRelease_slowpath_mask_v1. If that mask ANDed with the object's current refcount field is non-zero, then we take the slow path. A future runtime that changes the refcounting implementation can adjust this mask to match, or set the mask to all 1s to disable the old embedded fast path entirely (as long as the new representation never uses 0 as a valid refcount field value).
As part of this work, the overall approach for bridgeObjectRetain is changed slightly. Previously, it would mask off the spare bits from the native pointer and then call through to swift_retain. This either lost the spare bits in the return value (when tail calling swift_retain) which is problematic since it's supposed to return its parameter, or it required pushing a stack frame which is inefficient. Now, swift_retain takes on the responsibility of masking off spare bits from the parameter and preserving them in the return value. This is a trivial addition to the fast path (just a quick mask and an extra register for saving the original value) and makes bridgeObjectRetain quite a bit more efficient when implemented correctly to return the exact value it was passed.
The runtime's implementations of swift_retain/release are now also marked as preserve_most so that they can be tail called from the client library. preserve_most is compatible with callers expecting the standard calling convention so this doesn't break any existing clients. Some ugly tricks were needed to prevent the compiler from creating unnecessary stack frames with the new calling convention. Avert your eyes.
To allow back deployment, the runtime now has aliases for these functions called swift_retain_preservemost and swift_release_preservemost. The client library brings weak definitions of these functions that save the extra registers and call through to swift_retain/release. This allows them to work correctly on older runtimes, with a small performance penalty, while still running at full speed on runtimes that have the new preservemost symbols.
Although this is only supported on Darwin at the moment, it shouldn't be too much work to adapt it to other ARM64 targets. We need to ensure the assembly plays nice with the other platforms' assemblers, and make sure the implementation is correct for the non-ObjC-interop case.
rdar://122595871
The _SwiftifyImport macro is emitted into an unnamed buffer and then
parsed, pretending it was in the header all along. This makes it hard to
add `expected-note` comments for `diag::in_macro_expansion` when they
point here. That's okay, because the macro expansion has already been
pointed out by `expected-expansion` directives. But
-verify-ignore-unrelated is too blunt of a tool, so this adds
-verify-ignore-macro-note to ignore these specific diagnostics.
This pass has been disabled since a very long time (because it's terrible for code size).
It does not work for OSSA. Therefore it cannot be enabled anymore (as is) once we have OSSA throughout the pipeline.
So it's time to completely remove it.
This commit adds -sil-output-path and -ir-output-path frontend options that
allow generating SIL and LLVM IR files as supplementary outputs during normal
compilation.
These options can be useful for debugging and analysis tools
workflows that need access to intermediate compilation artifacts
without requiring separate compiler invocations.
Expected behaviour:
Primary File mode:
- SIL: Generates one .sil file per source file
- IR: Generates one .ll file per source file
Single-threaded WMO mode:
- SIL: Generates one .sil file for the entire module
- IR: Generates one .ll file for the entire module
Multi-threaded WMO mode:
- SIL: Generates one .sil file for the entire module
- IR: Generates separate .ll files per source file
File Maps with WMO:
- Both SIL and IR outputs using first entry's naming, which is
consistent with the behaviour of other supplementary outputs.
rdar://160297898
This adds the implementation required for later changing the default
behaviour of the -verify flag to error when diagnostics are emitted
in buffers other than the main file and files added with
-verify-additional-file. To keep the current behaviour, use the flag
-verify-ignore-unrelated. This flag is added as a no-op so that tests
can start using it before the new behaviour is enabled by default.
This does not enable it by default. Use either of the flags:
```
-enable-copy-propagation
-enable-copy-propagation=always
```
to enable it in -Onone. The previous frontend flag
`-enable-copy-propagation=true` has been renamed to
`-enable-copy-propagation=optimizing`, which is currently default.
rdar://107610971
This flag dumps all imports for each SourceFile after it's gone through
import resolution. It is only intended for testing purposes.
There are other ways to print imports, but they don't correspond 1:1 to
the imports actually resolved, which is a bit problematic when testing
implicit clang module imports.
Set an upper bound on the number of chained lookups we attempt to
avoid spinning while trying to recursively apply the same dynamic
member lookup to itself.
rdar://157288911
Controlled from Swift with '-version-independent-apinotes', which, for the underlying Clang invocation enables '-fswift-version-independent-apinotes', results in PCMs which aggregate all versioned APINotes wrapped in a 'SwiftVersionedAttr', with the intent to have the client pick and apply only those that match its current Swift version, discarding the rest.
This change introduces the configuration flags for this mode as well as the corresponding logic at the beginning of `importDeclImpl` to canonicalize versioned attributes, i.e. select the appropriate attributes for the current target and discard the rest.
Add a new option `-gen-reproducer` that when swift caching is used,
create a standalone reproducer that can be used to reproduce the
`swift-frontend` invocation.
The concrete nesting limit, which defaults to 30, catches
things like A == G<A>. However, with something like
A == (A, A), you end up with an exponential problem size
before you hit the limit.
Add two new limits.
The first is the total size of the concrete type, counting
all leaves, which defaults to 4000. It can be set with the
-requirement-machine-max-concrete-size= frontend flag.
The second avoids an assertion in addTypeDifference() which
can be hit if a certain counter overflows before any other
limit is breached. This also defaults to 4000 and can be set
with the -requirement-machine-max-type-differences= frontend flag.
This was used a long time ago for a design of a scanner which could rely on the client to specify that some modules *will be* present at a given location but are not yet during the scan. We have long ago determined that the scanner must have all modules available to it at the time of scan for soundness. This code has been stale for a couple of years and it is time to simplify things a bit by deleting it.
Create a path that swift-frontend can execute an uncached job from
modules built with CAS based explicit module build. The new flag
-import-module-from-cas will allow an uncached build to load module
from CAS, and combined with source file from real file system to build
the current module. This allows quick iterations that bypasses CAS,
without full dependency scanning every time in between.
rdar://152441866
The LeastValidPointerValue is hard-coded in the runtime.
Therefore this option is only available in embedded swift - which doesn't have a runtime.
rdar://151755654
`-Xfrontend -enable-cond-fail-message-annotation`
LLVM IR produced by the Swift compiler will add the `annotation`
metadata attribute to the branch instruction generated for cond_fail
builtins.
Make `-enable-deterministic-check` a driver option and teach dependency
scanner to propagate the option to explicit module build commmands. This
allows to the option to check every build output from the compiler is
deterministic.
It is possible for a module interface (e.g., ModuleA) to be generated
with C++ interop disabled, and then rebuilt with C++ interop enabled
(e.g., because ModuleB, which imports ModuleA, has C++ interop enabled).
This circumstance can lead to various issues when name lookup behaves
differently depending on whether C++ interop is enabled, e.g., when
a module name is shadowed by a namespace of the same name---this only
happens in C++ because namespaces do not exist in C. Unfortunately,
naming namespaces the same as a module is a common C++ convention,
leading to many textual interfaces whose fully-qualified identifiers
(e.g., c_module.c_member) cannot be correctly resolved when C++ interop
is enabled (because c_module is shadowed by a namespace of the same
name).
This patch does two things. First, it introduces a new frontend flag,
-formal-cxx-interoperability-mode, which records the C++ interop mode
a module interface was originally compiled with. Doing so allows
subsequent consumers of that interface to interpret it according to the
formal C++ interop mode. Note that the actual "versioning" used by this
flag is very crude: "off" means disabled, and "swift-6" means enabled.
This is done to be compatible with C++ interop compat versioning scheme,
which seems to produce some invalid (but unused) version numbers. The
versioning scheme for both the formal and actual C++ interop modes
should be clarified and fixed in a subsequent patch.
The second thing this patch does is fix the module/namespace collision
issue in module interface files. It uses the formal C++ interop mode to
determine whether it should resolve C++-only decls during name lookup.
For now, the fix is very minimal and conservative: it only filters out
C++ namespaces during unqualified name lookup in an interface that was
originally generated without C++ interop. Doing so should fix the issue
while minimizing the chance for collateral breakge. More cases other
than C++ namespaces should be added in subsequent patches, with
sufficient testing and careful consideration.
rdar://144566922
