Specifically for 6.2, we are making optimize hop to executor more conservative
around caller isolation inheriting functions. This means that we are:
1. No longer treating calls to caller isolation inheriting functions as having a
hop in their prologue. In terms of this pass, it means that when determining
dead hop to executors, we no longer think that a caller isolation inheriting
function means that an earlier hop to executor is not required.
2. Treating returns from caller isolation inheriting callees as requiring a
hop. The reason why we are doing this is that we can no longer assume that our
caller will hop after we return.
Post 6.2, there are three main changes we are going to make:
* Forward Dataflow
Caller isolation inheriting functions will no longer be treated as suspension
points meaning that we will be able to propagate hops over them and can assume
that we know the actor that we are on when we enter the function. Practically
this means that trees of calls that involve just nonisolated(nonsending) async
functions will avoid /all/ hop to executor calls since we will be able to
eliminate all of them since the dataflow will just propagate forward from the
entrance that we are already on the actor.
* Backwards Dataflow
A caller isolation inheriting call site will still cause preceding
hop_to_executor functions to be live. This is because we need to ensure that we
are on the caller isolation inheriting actor before we hit the call site. If we
are already on that actor, the hop will be eliminated by the forward pass. But
if the hop has not been eliminated, then the hop must be needed to return us to
the appropriate actor.
We will also keep the behavior that returns from a caller isolation inheriting
function are considered to keep hop to executors alive. If we were able to
propagate to a hop to executor before the return inst with the forward dataflow,
then we know that we are guaranteed to still be on the relevant actor. If the
hop to executor is still there, then we need it to ensure that our caller can
treat the caller isolation inheriting function as a non-suspension point.
rdar://155905383
A call to a `@preconcurrency` function goes through a function conversion
that removes `Sendable` from existentials among other things. Implement
support for this by bitcasting indirect return slots whose type differs
from the formal indirect return type in concurrency markings only.
Fixes rdar://154240007
If we have no substitution map, we still substitute types appearing
in the original function, because we need to remap any local
archetypes, which are always cloned.
However, the conformance lookup callback used for this substitution
was wrong. We should only do mapTypeOutOfContext() if we're going
to callSubstitutionMap::lookupConformance(), otherwise we form a
new abstract conformance with an interface type, and not a primary
archetype as expected.
* add `cloneFunctionBody` without an `entryBlockArguments` argument
* remove the `swift::ClosureSpecializationCloner` from the bridging code and replace it with a more general `SpecializationCloner`
OSSA lifetime canonicalization can take a very long time in certain
cases in which there are large basic blocks. to mitigate this, add logic
to skip walking the liveness boundary for extending liveness to dead
ends when there aren't any dead ends in the function.
Updates `DeadEndBlocks` with a new `isEmpty` method and cache to
determine if there are any dead-end blocks in a given function.
Calling setupLLVMOptimizationRemarks overwrites the MainRemarkStreamer
in the LLVM context. This prevents LLVM from serializing the remark meta
information for the already emitted SIL remarks into the object file.
Without the meta information bitstream remarks don't work correctly.
Instead, emit SIL remarks and LLVM remarks to the same RemarkSerializer,
and keep the file stream alive until after CodeGen.
Remarks emitted by LLVM use the mangled function name. Be consistent
with LLVM when serializing SIL remarks, so external tools can properly
filter for all remarks attributed to a certain function.
LifetimeDependenceInsertion inserts mark_dependence on token result of a begin_apply
when it yields a lifetime dependent value. When such a begin_apply gets inlined,
the inliner can crash because of the remaining uses of the token result.
Fix this by inserting mark_dependence on parameter operands that are lifetime dependence sources
and deleting the mark_dependence on token results in the inliner.
Fixes rdar://151568816
We are going to need to add more flags to the various checked cast
instructions. Generalize the CastingIsolatedConformances bit in all of
these SIL instructions to an "options" struct that's easier to extend.
Precursor to rdar://152335805.
The new function stripAccessAndAccessStorageCasts is analogous to the
existing function stripAccessAndIdentityCasts but differs in that the
latter uses isAccessStorageIdentityCast whereas the new function uses
isAccessStorageCast.
* Move the mutating APIs into Context.swift, because SIL can only be mutated through a MutatingContext
* move the `baseOperand` and `base` properties from the instruction classes to the `MarkDependenceInstruction` protocol
* add `valueOrAddressOperand` and `valueOrAddress` in the `MarkDependenceInstruction` protocol
This prevents simplification and SILCombine passes to remove (alive) `mark_dependence_addr`.
The instruction is conceptually equivalent to
```
%v = load %addr
%d = mark_dependence %v on %base
store %d to %addr
```
Therefore the address operand has to be defined as writing to the address.
It is like `zeroInitializer`, but does not actually initialize the memory.
It only indicates to mandatory passes that the memory is going to be initialized.
Convert a bunch of places where we're dumping to stderr and calling
`abort` over to using `ABORT` such that the message gets printed to
the pretty stack trace. This ensures it gets picked up by
CrashReporter.
Outside of the resilience domain, they have to be treated as opaque and therefore potentially
addressable-for-dependencies, but inside of the resilience domain, we may take advantage of
knowing the type layout to load indirect parameters out of memory and break the (unnecessary)
dependency on a fixed memory location. Fixes rdar://151268401.
We do still however have problems when the type is actually `@_addressableForDependencies`
inside of its resilience domain (rdar://151500074). I'll fix that in a follow up.
This will cause tests today to crash since even though we are placing the
isolation now, to make it easier to read, I left in the old isolation selecting
code. This code uses the witness's isolation instead of the requirement's
isolation which is incorrect since the protocol witness thunk needs to look the
requirement from an ABI perspective since the two must be substitutable. The
crash comes from the ABI verification I added in earlier commits.
Diagnostics only work with `SourceLoc` which is basically a pointer into a buffer of the loaded source file.
But when debug info is de-serialized, the SIL `Location` consists of a filename+line+column.
To "convert" this to a `SourceLoc`, the file must be loaded.
This change adds `DiagnosticEngine.getLocationFromExternalSource` for this purpose.
Also, the new protocol `ProvidingSourceLocation` - to which `SourceLoc` and `Location` conform - help to generalize the helper struct `Diagnostic` and make this "conversion" happen automatically.
