This commit fixes a bug where we were not checking that all predecessors had the
cond_fail block as its only successor. This occured since we were bailing early
when we saw a constant. So if we saw a predecessor with a constant before a
predecessor that had multiple successors, we would optimize even though we would
be introducing an extra cond_fail along a path.
I added a new utility pass to test this code since so much is going on in
SimplifyCFG that it is difficult to construct a test case running the full
pass.
Really this code should be in a different pass (properly SIL Code Motion TBH).
But for now, this commit just fixes the bug.
rdar://26904047
Opened archetypes may be used by types of basic block arguments. Therefore, when performing the CSE of open_existential_ref instructions, one should also update the type of any basic block argument that may have used the archetype opened by the open_existential_ref being replaced.
rdar://problem/27386065
The generic specialized would get out of control in certain cases and would not stop generating specializations of generic functions until it runs out of memory after a while.
rdar://21260480
Strict aliasing only applies to memory operations that use strict
addresses. The optimizer needs to be aware of this flag. Uses of raw
addresses should not have their address substituted with a strict
address.
Also add Builtin.LoadRaw which will be used by raw pointer loads.
Exceptions can cause an exit before hitting the unsafeGuaranteedEnd instruction.
Make sure that we check post dominance to find such cases.
rdar://26890751
Introduce bridging of NSError to ErrorProtocol, so an Objective-C API
expressed via an "NSError *" will be imported using ErrorProtocol in
the Swift. For example, the Objective-C method:
- (void)handleError:(NSError *)error userInteractionPermitted:(BOOL)userInteractionPermitted;
will now be imported as:
func handleError(_ error: ErrorProtocol, userInteractionPermitted: Bool)
This is bullet (3) under the proposed solution of SE-0112. Note that
we made one semantic change here: instead of removing the conformance
of NSError to ErrorProtocol, which caused numerous problems both
theoretical and actual because the model expects that an NSError
conforms to ErrorProtocol without requiring wrapping, we instead limit
the ErrorProtocol -> NSError conversion that would be implied by
bridging. This is defensible in the short term because it also
eliminates the implicit conversion, and aligns with SE-0072, which
eliminates implicit bridging conversions altogether.
This ensures that when one is bisecting on pass counts, regardless of whether or
not one removes code in the test case, the pass counts being run remain the
same.
This enables things like redundant load elimination and should fix the
regression after changing ManagedBuffer(Pointer) to use unsafeAddressWithOwner.
rdar://27138023
The option is supposed to be used as follows and takes a comma-seprataed list of SIL pass numbers as input:
-Xllvm -debug-only-pass-number=passnumber1[,passnumber2,..,passnumberN]
It enables the debug printing (i.e. prints inside DEBUG statements ) when one of the mentioned passes is being run.
Now we properly track the dependencies between instructions using opened archetypes and instructions opening those archetypes. No need for any special handling anymore.
Till now there was no way in SIL to explicitly express a dependency of an instruction on any opened archetypes used by it. This was a cause of many errors and correctness issues. In many cases the code was moved around without taking into account these dependencies, which resulted in breaking the invariant that any uses of an opened archetype should be dominated by the definition of this archetype.
This patch does the following:
- Map opened archetypes to the instructions defining them, i.e. to open_existential instructions.
- Introduce a helper class SILOpenedArchetypesTracker for creating and maintaining such mappings.
- Introduce a helper class SILOpenedArchetypesState for providing a read-only API for looking up available opened archetypes.
- Each SIL instruction which uses an opened archetype as a type gets an additional opened archetype operand representing a dependency of the instruction on this archetype. These opened archetypes operands are an in-memory representation. They are not serialized. Instead, they are re-constructed when reading binary or textual SIL files.
- SILVerifier was extended to conduct more thorough checks related to the usage of opened archetypes.
Formerly SILGen would never emit this sequence. In fact in most places
we lower away dynamic Self, replacing it with the concrete Self type
instead.
However, with an upcoming change, I'm using 'metatype $Self' as a handy
way to grab IRGenSILFunction::LocalSelfMetadata, since that's what it
already does.
Note that the tests for this are in the next patch.
Now we properly track the dependencies between instructions using opened archetypes and instructions opening those archetypes. No need for any special handling anymore.
Till now there was no way in SIL to explicitly express a dependency of an instruction on any opened archetypes used by it. This was a cause of many errors and correctness issues. In many cases the code was moved around without taking into account these dependencies, which resulted in breaking the invariant that any uses of an opened archetype should be dominated by the definition of this archetype.
This patch does the following:
- Map opened archetypes to the instructions defining them, i.e. to open_existential instructions.
- Introduce a helper class SILOpenedArchetypesTracker for creating and maintaining such mappings.
- Introduce a helper class SILOpenedArchetypesState for providing a read-only API for looking up available opened archetypes.
- Each SIL instruction which uses an opened archetype as a type gets an additional opened archetype operand representing a dependency of the instruction on this archetype. These opened archetypes operands are an in-memory representation. They are not serialized. Instead, they are re-constructed when reading binary or textual SIL files.
- SILVerifier was extended to conduct more thorough checks related to the usage of opened archetypes.
Now we properly track the dependencies between instructions using opened archetypes and instructions opening those archetypes. No need for any special handling anymore.
Till now there was no way in SIL to explicitly express a dependency of an instruction on any opened archetypes used by it. This was a cause of many errors and correctness issues. In many cases the code was moved around without taking into account these dependencies, which resulted in breaking the invariant that any uses of an opened archetype should be dominated by the definition of this archetype.
This patch does the following:
- Map opened archetypes to the instructions defining them, i.e. to open_existential instructions.
- Introduce a helper class SILOpenedArchetypesTracker for creating and maintaining such mappings.
- Introduce a helper class SILOpenedArchetypesState for providing a read-only API for looking up available opened archetypes.
- Each SIL instruction which uses an opened archetype as a type gets an additional opened archetype operand representing a dependency of the instruction on this archetype. These opened archetypes operands are an in-memory representation. They are not serialized. Instead, they are re-constructed when reading binary or textual SIL files.
- SILVerifier was extended to conduct more thorough checks related to the usage of opened archetypes.
As a first step to allowing the build script to build *only*
static library versions of the stdlib, change `add_swift_library`
such that callers must pass in `SHARED`, `STATIC`, or `OBJECT_LIBRARY`.
Ideally, only these flags would be used to determine whether to
build shared, static, or object libraries, but that is not currently
the case -- `add_swift_library` also checks whether the library
`IS_STDLIB` before performing certain additional actions. This will be
cleaned up in a future commit.
They are mandatory inlined anyway. Therefore the binary code of transparent functions is not used in most cases.
In case the address of a transparent function is taken, the main program (which deserialized the transparent function) generates code for it and treats it as shared function.
This reduces the stdlib code size by 2.8%.
