This reverts commit 4f059033bb. The change
is actually not NFC since previously, there is a cache in the
CompilerInvocation that prevents the same CAS from the same CASOptions
from being initialized multiple times, which was relied upon when
running inside sub invocation. When switching to a non-caching simple
CASOption types, it causes every single sub instance will create its own
CAS, and it can consume too many file descriptors and causing errors
during dependency scanning.
rdar://164903080
Upstream LLVM in llvm/llvm-project#139584 changed `DiagnosticOptions`
from being a referenced counted object to just be a reference, not owned
by the `clang::DiagnosticEngine`.
In 0981b71090 (part of #82243), the usages
of the Swift repository were adapted to the new memory model, but it
introduced at least one use-after-free and a potential one around the
usage of Clang in the Clang Importer.
This commit tries to fix the use-after-free in both cases, by returning
a `unique_ptr` to the `clang::DiagnosticOptions`, which makes the
lifetime of the `DiagnosticOptions` match the lifetime of the variable
that uses it (normally a `CompilerInvocation`).
Other cases in 0981b71090 should be safe
because the lifetime of the `DiagnosticOptions` do not seem to propagate
beyond the scope of the functions where they live (but I am not fully
sure about the one in `IDETool/CompilerInvocation.cpp` completely).
This was causing compiler crashes during the test
`Interop/Cxx/stdlib/unsupported-stdlib.swift` which eventually uses
`createClangDriver` and tries to emit a diagnostic, which in some cases
was reading the memory from `DiagnosticOptions` when it was already out
of scope.
Bring back legacy prefix map option to allow an older swift-driver to
work with newer swift-frontend. For old swift-driver, it will always
send the old style prefix map option, so the new compiler needs to
support that.
rdar://164208526
Currently symbol graphs are always written in files that contain 1 to 2
module names. It's possible for Swift module names to be very long, so
combining 2 of them in file name like `module1@module2...` in the same
path component means the name can be too long for some file systems. The
new option `-symbol-graph-shorten-output-names` changes the symbol graph
output files to use a MD5 hash of the module name(s) as the filename and
outputs an additional JSON file with the original names mapped to the
real filename. The module names JSON can be used to construct a VFS
overlay with the original naming scheme.
fix#83723
I considered using vfsoverlay, which seems like a viable solution, but
the vfsoverlay options don't seem to apply to any of the outputs from
the compiler. When I set an overlay to remap the symbol graph file
outputs, the remapped external paths aren't used so the root problem of
too long file names remains.
This adds the -Rmacro-expansions flag. It provides similar functionality
to -dump-macro-expansions, but instead of dumping the macro expansion to
stderr, it emits it line by line as remarks. This is useful for testing
with -verify, where both macro expansion content and warnings need to be
tested at the same time.
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
This new option allows the Driver to pass the path to a compilation
job's own binary swiftmodule artifact to the frontend. The compiler
then stores this path in the debug info, to allow clients like LLDB to
unambiguously know which binary Swift module belongs to which compile
unit.
rdar://163302154
This experimental feature will be used to force the compiler to treat `Swift`
runtime availability as separate from platform availability when compiling for
targets that have the Swift runtime built-in.
We have long switched to delegate the checking of whether a module is up-to-date to the build system (SwiftDriver) and stopped serializing file dependencies for interface-built binary modules.
Resolves rdar://162881032
_SwiftifyImport assumes types like Swift.Int, Swift.UnsafePointer<T> and
Swift.Span<T> are available. This is not the case when building the
stdlib itself. Disable safe interop in the stdlib to prevent errors.
This currently has no effect, but will when this feature is enabled by
default, which I have manually tested.
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 PR introduces three new instrumentation flags and plumbs them
through to IRGen:
1. `-ir-profile-generate` - enable IR-level instrumentation.
2. `-cs-profile-generate` - enable context-sensitive IR-level
instrumentation.
3. `-ir-profile-use` - IR-level PGO input profdata file to enable
profile-guided optimization (both IRPGO and CSIRPGO)
**Context:**
https://forums.swift.org/t/ir-level-pgo-instrumentation-in-swift/82123
**Swift-driver PR:** https://github.com/swiftlang/swift-driver/pull/1992
**Tests and validation:**
This PR includes ir level verification tests, also checks few edge-cases
when `-ir-profile-use` supplied profile is either missing or is an
invalid IR profile.
However, for argument validation, linking, and generating IR profiles
that can later be consumed by -cs-profile-generate, I’ll need
corresponding swift-driver changes. Those changes are being tracked in
https://github.com/swiftlang/swift-driver/pull/1992
Add support for the `Swift` availability domain, which represents availability
with respect to the Swift runtime. Use of this domain is restricted by the
experimental feature `SwiftRuntimeAvailability`.
We already have -suppress-warnings and -suppress-remarks; this patch
adds support for suppressing notes too. Doing so is useful for -verify
tests where we don't really care about the emitted notes.
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.
With some changes that I am making, we will no longer need this flag at the SIL
level, so we can just make it an IRGen flag (which it really should have been
in the first place).
