Commit Graph

2202 Commits

Author SHA1 Message Date
Mikio Takeuchi
488d531846 Enhance -assume-single-threaded option (SR-3945) 2017-02-27 12:17:53 +09:00
Andrew Trick
38d44c2164 AddressLowering: rewrite the call-site lowering logic.
This pass now canonicalizes results before lowering and handles all combinations
of direct and indirect multiple return values. The logic is much less ad-hoc and
more robust.

try_apply still isn't handled, but should be much easier now.

Add visitLoadInst, visitStoreInst, visitDebugValueInst, etc.
2017-02-26 13:52:50 -08:00
Michael Gottesman
784d5d16fc [silgen] Change emitClassConstructorInitializer to use ownership and make calling designated/chaining initializes use proper ownership.
rdar://29791263
2017-02-23 08:48:58 -08:00
Joe Shajrawi
ec1e3ee20e Add support for unconditional checked cast instruction for opaque value types + SILGen support for it 2017-02-22 16:35:46 -08:00
Joe Shajrawi
147f844ce4 Merge pull request #7654 from shajrawi/init_opaque
Add support for Init Existentials for opaque value types
2017-02-20 19:37:01 -08:00
Joe Shajrawi
1e521c453b Add support for Init Existentials for opaque value types 2017-02-20 16:40:02 -08:00
Doug Gregor
042e6510c3 [AST] Drop ProtocolDecl's "inherited protocols" list.
The list of directly inherited protocols of a ProtocolDecl is already
encoded in the requirement signature, as conformance constraints where
the subject is Self. Gather the list from there rather than separately
computing/storing the list of "inherited protocols".
2017-02-20 09:41:00 -08:00
Michael Gottesman
81dcdba0a9 Revert "Make sure we always de-serialize transparent functions."
This reverts commit 821b365054.

It caused a bunch of execution time regressions.
2017-02-18 13:36:34 -08:00
Erik Eckstein
821b365054 Make sure we always de-serialize transparent functions.
We didn’t do that if an optimization looked up a witness table.

Fixes rdar://problem/30544344

I couldn’t come up with an isolated test case, but this should be covered with our existing tests.
(The problem shows up when inlining of generics are enabled)
2017-02-16 17:04:58 -08:00
Andrew Trick
4256112150 AddressLowering: fix handling of result tuples. 2017-02-16 11:04:42 -08:00
Arnold Schwaighofer
876cea81ae SIL: Add an allowed access kind to the opened value of an open_existential_addr instruction
Once we move to a copy-on-write implementation of existential value buffers we
can no longer consume or destroy values of an opened existential unless the
buffer is uniquely owned.

Therefore we need to track the allowed operation on opened values.

Add qualifiers "mutable_access" and "immutable_access" to open_existential_addr
instructions to indicate the allowed access to the opened value.

Once we move to a copy-on-write implementation, an "open_existential_addr
mutable_access" instruction will ensure unique ownership of the value buffer.
2017-02-15 14:23:12 -08:00
practicalswift
65b0219f7b [gardening] Fix typos 2017-02-14 20:04:08 +01:00
practicalswift
1c64f04997 [gardening] Fix header inconsistencies 2017-02-14 20:04:05 +01:00
Roman Levenstein
fc7b9e1081 [sil-linker] Minor clean-ups of function lookup code in SILModule and SIL linker. NFC.
Stop using SILLinkage::Private as a flag for "linkage doesn't matter". Use Optional instead.
2017-02-14 08:15:51 -08:00
Andrew Trick
15699805ac SILModule.h comments 2017-02-13 17:10:02 -08:00
Andrew Trick
10b118dfa9 [Lowering] Make the AddressLowering pass functional. 2017-02-13 17:10:02 -08:00
Andrew Trick
855918c620 [Lowering] Add an AddressLowering pass. 2017-02-13 17:10:02 -08:00
swift-ci
b454c78ea9 Merge pull request #7443 from gottesmm/ownership_unreachable_code 2017-02-13 17:06:46 -08:00
Michael Gottesman
ed67bf9f21 [ownership-verifier] Teach the ownership verifier how to handle unreachable code.
There are a few different use cases here:

1. In Raw SIL, no return folding may not have been run yet implying that a call
to a no-return function /can/ have arbitrary control flow after it (consider
mandatory inlined functions). We need to recognize that the region of code that
is strictly post dominated by the no-return function is "transitively
unreachable" and thus leaking is ok from that point. *Footnote 1*.

2. In Canonical and Raw SIL, we must recognize that unreachables and no-return
functions constitute places where we are allowed to leak.

rdar://29791263

----

*Footnote 1*: The reason why this is done is since we want to emit unreachable
code diagnostics when we run no-return folding. By leaving in the relevant code,
we have preserved all of the SILLocations on that code allowing us to create
really nice diagnostics.
2017-02-13 16:36:01 -08:00
Michael Gottesman
c309a286a3 [sil] Fix a bug that I found on inspection.
rdar://29791263
2017-02-13 16:36:00 -08:00
Jordan Rose
1c60910198 Revert "Merge pull request #6092 from swiftix/wip-generics-inlining-flag-4"
This reverts commit 1b3d29a163, reversing
changes made to b32424953e.

We're seeing a handful of issues from turning on inlining of generics,
so I'm reverting to unblock the bots.
2017-02-13 10:52:17 -08:00
Hugh Bellamy
9f8b91a0a7 Merge pull request #7402 from hughbe/msvc-refresh
Get Swift compiling with MSVC again
2017-02-12 08:56:35 +07:00
Hugh Bellamy
f2f522c789 Merge pull request #6887 from hughbe/msvc-sil-warning
Fix multiple copy constructors warning
2017-02-11 15:00:25 +07:00
Hugh Bellamy
762a5a0aa8 Add some llvm_unreachable annotations for recently introduced MSVC control path warnings 2017-02-11 14:34:35 +07:00
Bob Wilson
c24c9958c0 Fix incorrect SILSuccessorIterator operator++ that broke master-next after llvm r294339.
This operator++ is a post-increment operator but it was implemented with pre-increment behavior.
This bug was exposed by the changes in llvm r294339 and caused the master-next branch to stop
building.
2017-02-10 22:22:29 -08:00
Roman Levenstein
199e76d6d4 Fixes for a lookup of functions by name
Among other things it fixes a bug in the function signature optimization where it would use an already existing mangled name for a different function
2017-02-10 18:07:30 -08:00
swift-ci
ec5a1c2265 Merge pull request #7367 from gottesmm/no_return_refactoring 2017-02-09 13:40:43 -08:00
Joe Shajrawi
df5297d877 Merge pull request #7349 from shajrawi/open_existential
Support for Open Existentials that do no take an address
2017-02-09 12:39:16 -08:00
Michael Gottesman
284a028e61 [sil] Add a new API SILModule::isNoReturnBuiltinOrIntrinsic(Identifier).
This is just refactoring code from DiagnoseUnreachable into a centralized
location.

rdar://29791263
2017-02-09 11:29:41 -08:00
Joe Shajrawi
c478828de7 Support for Open Existentials that do no take an address 2017-02-09 11:25:34 -08:00
Roman Levenstein
163f58e293 Merge pull request #7277 from swiftix/wip-partial-pre-specializations-wip-specialize-attribute-2
Teach EagerSpecializer how to use the new form of the @_specialize attribute
2017-02-08 20:07:54 -08:00
Michael Gottesman
ebf5cc607e [sil] Move PostOrderFunctionInfo from SILOptimizer/Analysis/PostOrderAnalysis => SIL/PostOrder.h.
There is already precedence for doing this via DominanceInfo. The reason I am
doing this is that I need the ownership verifier to be able to be run in Raw SIL
before no return folding has run. This means I need to be able to ignore
unreachable code resulting from SILGen not inserting unreachables after no
return functions.

The reason why SILGen does this is to preserve the source information of the
unreachable code so that we can emit nice errors about unreachable code.

rdar://29791263
2017-02-08 19:15:05 -08:00
Hugh Bellamy
a737a763e3 Fix MSVC "multiple copy constructors specified" warning 2017-02-08 21:25:50 +07:00
Roman Levenstein
2b9444e03a Add mangling support for partial specializations. 2017-02-07 23:16:00 -08:00
Slava Pestov
f2173e9b24 AST: Remove ModuleDecl parameter from GenericEnvironment::getSubstitutionMap()
This method maps interface types to archetypes, which in general
requires a module for performing conformance lookups, if mapping
a member type of a generic parameter which has been made concrete.

However, in practice the types we are mapping here are all canonical
with respect to the generic signature, because they came from
GenericSignature::getAllDependentTypes(), so we actually don't need
to do conformance lookups.

This allows some code to be simplified.
2017-02-06 22:39:26 -08:00
Slava Pestov
3519e0cd25 AST: Introduce new SubstitutionList type to replace ArrayRef<Substitution>
SubstitutionList is going to be a more compact representation of
a SubstitutionMap, suitable for inline allocation inside another
object.

For now, it's just a typedef for ArrayRef<Substitution>.
2017-02-06 21:36:33 -08:00
Slava Pestov
0bfe9d4102 SIL: Remove SILInstruction::getSubstitutionsWithoutSelfSubstitution()
This method was not used anywhere, and furthermore it makes
makes no sense to directly pick apart ArrayRef<Substitution>
like this.
2017-02-06 20:33:39 -08:00
Slava Pestov
1954665a95 AST: Remove more uses of SubstitutionMap::add{Substitution,Conformance}()
A new SubstitutionMap::getProtocolSubstitutions() method handles
the case where we construct a trivial SubstitutionMap to replace
the protocol Self type with a concrete type.

When substituting one opened existential archetype for another,
use the form of Type::subst() that takes two callbacks instead of
building a SubstitutionMap. SubstitutionMaps are intended to be
used with keys that either come from a GenericSignature or a
GenericEnvironment, so using them to replace opened archetypes
doesn't fit the conceptual model we're going for.
2017-02-06 20:33:39 -08:00
Slava Pestov
b22f9ea487 AST: Remove SubstitutionMap::getMap()
We don't want to expose the fact that SubstitutionMaps are
backed by a DenseMap, since that's going to change soon.
2017-02-06 19:43:33 -08:00
Slava Pestov
162b1c0de0 SIL: Remove uncurryLevel parameter from type lowering
There was only one place where we lowered AST types with a non-zero
uncurry level, in SILGenApply.cpp. Add a new overload of the
getSILFunctionType() method that takes an uncurry level. All the
other methods no longer have to thread it through.
2017-02-06 15:24:03 -08:00
Michael Gottesman
15a12a3f34 Merge pull request #7272 from gottesmm/unmanaged_autorelease_value
[semantic-sil] Create unmanaged_autorelease_value.
2017-02-06 13:19:01 -08:00
Michael Gottesman
554feff463 [semantic-sil] Create unmanaged_autorelease_value.
This is an autorelease for use with Builtin.autorelease that does not need to be
balanced as part of the ownership model.

rdar://29791263
2017-02-06 12:11:46 -08:00
Adrian Prantl
4d1ae142c6 Remove the redundant DeclCtx field in SILFunction.
In all cases the DeclCtx field was supposed to be initialized from the
SILLocation of the function, so we can save one pointer per
SILFunction.

There is one test case change where a different (more precise)
diagnostic is being generated after this change.
2017-02-06 11:07:50 -08:00
practicalswift
0193f2b872 [gardening] Fix typos 2017-02-04 09:22:38 +01:00
swift-ci
83fcb32ad5 Merge pull request #7189 from atrick/master 2017-02-01 12:53:51 -08:00
Andrew Trick
a6ebb4e914 Add an option -Xllvm -sil-print-on-error.
Print the SIL function body on an assert. Recovering the SIL code is the
critical path for pretty much any SIL development. The only alternative is
rebuilding the library with string matching or building a debug compiler and
hoping lldb works. The standard library takes a very long time to build with a
debug compiler.
2017-02-01 10:51:53 -08:00
Erik Eckstein
7fb871a7c6 Disable dead conformance elimination for now.
There are still several open issues (e.g. asserts in IRGen) and it seems to be hard to fix them.
Most likely we’ll go for a different approach at all.
2017-02-01 10:30:30 -08:00
Roman Levenstein
a422f8af63 [sil-dead-funciton-elimination] Eliminate unused default methods implementations from default witness tables
We were missing this opportunity, because we forgot to update DeadFunctionElimination after default witness tables were introduced.

rdar://30264699
2017-01-30 11:57:34 -08:00
Andrew Trick
9f141374a6 [NFC] Move SILFunctionConventions into its own header. 2017-01-28 13:13:17 -08:00
Andrew Trick
e9c559b718 [SILGen] Bootstrapping opaque values (#7113)
[NFC] Add -enable-sil-opaque-values frontend option.

This will be used to change the SIL-level calling convention for opaque values,
such as generics and resilient structs, to pass-by-value.  Under this flag,
opaque values have SSA lifetimes, managed by copy_value and destroy_value.

This will make it easier to optimize copies and verify ownership.

* [SILGen] type lowering support for opaque values.

Add OpaqueValueTypeLowering.
Under EnableSILOpaqueValues, lower address-only types as opaque values.

* [SIL] Fix ValueOwnershipKind to support opaque SIL values.

* Test case: SILGen opaque value support for Parameter/ResultConvention.

* [SILGen] opaque value support for function arguments.

* Future Test case: SILGen opaque value specialDest arguments.

* Future Test case: SILGen opaque values: emitOpenExistential.

* Test case: SIL parsing support for EnableSILOpaqueValues.

* SILGen opaque values: prepareArchetypeCallee.

* [SIL Verify] allow copy_value for EnableSILOpaqueValues.

* Test cast: SIL serializer support for opaque values.

* Add a static_assert for ParameterConvention layout.

* Test case: Mandatory SILOpt support for EnableSILOpaqueValues.

* Test case: SILOpt support for EnableSILOpaqueValues.

* SILGen opaque values: TypeLowering emitCopyValue.

* SILBuilder createLoad. Allow loading opaque values.

* SIL Verifier. Allow loading and storing opaque values.

* SILGen emitSemanticStore support for opaque values.

* Test case for SILGen emitSemanticStore.

* Test case for SIL mandatory support for inout assignment.

* Fix SILGen opaque values test case after rebasing.
2017-01-27 18:56:53 -08:00