- Introduce an UnownedSerialExecutor type into the concurrency library.
- Create a SerialExecutor protocol which allows an executor type to
change how it executes jobs.
- Add an unownedExecutor requirement to the Actor protocol.
- Change the ABI for ExecutorRef so that it stores a SerialExecutor
witness table pointer in the implementation field. This effectively
makes ExecutorRef an `unowned(unsafe) SerialExecutor`, except that
default actors are represented without a witness table pointer (just
a bit-pattern).
- Synthesize the unownedExecutor method for default actors (i.e. actors
that don't provide an unownedExecutor property).
- Make synthesized unownedExecutor properties `final`, and give them
a semantics attribute specifying that they're for default actors.
- Split `Builtin.buildSerialExecutorRef` into a few more precise
builtins. We're not using the main-actor one yet, though.
Pitch thread:
https://forums.swift.org/t/support-custom-executors-in-swift-concurrency/44425
Tasks shouldn't normally hog the actor context indefinitely after making a call that's bound to
that actor, since that prevents the actor from potentially taking on other jobs it needs to
be able to address. Set up SILGen so that it saves the current executor (using a new runtime
entry point) and hops back to it after every actor call, not only ones where the caller context
is also actor-bound.
The added executor hopping here also exposed a bug in the runtime implementation while processing
DefaultActor jobs, where if an actor job returned to the processing loop having already yielded
the thread back to a generic executor, we would still attempt to make the actor give up the thread
again, corrupting its state.
rdar://71905765
Most of the async runtime functions have been changed to not
expect the task and executor to be passed in. When knowing the
task and executor is necessary, there are runtime functions
available to recover them.
The biggest change I had to make to a runtime function signature
was to swift_task_switch, which has been altered to expect to be
passed the context and resumption function instead of requiring
the caller to park the task. This has the pleasant consequence
of allowing the implementation to very quickly turn around when
it recognizes that the current executor is satisfactory. It does
mean that on arm64e we have to sign the continuation function
pointer as an argument and then potentially resign it when
assigning into the task's resume slot.
rdar://70546948
In their previous form, the non-`_f` variants of these entry points were unused, and IRGen
lowered the `createAsyncTask` builtins to use the `_f` variants with a large amount of caller-side
codegen to manually unpack closure values. Amid all this, it also failed to make anyone responsible
for releasing the closure context after the task completed, causing every task creation to leak.
Redo the `swift_task_create_*` entry points to accept the two words of an async closure value
directly, and unpack the closure to get its invocation entry point and initial context size
inside the runtime. (Also get rid of the non-future `swift_task_create` variant, since it's unused
and it's subtly different in a lot of hairy ways from the future forms. Better to add it later
when it's needed than to have a broken unexercised version now.)
The underlying runtime functions really want to be able to consume the closure being used
to spawn the task, but the implementation was trying to hide this by introducing a retain
at IRGen time, which is not very ARC-optimizer-friendly. Correctly model the builtin operands
as consumed so that the ownership verifier allows them to be taken +1.
Refactor SILGen's ApplyOptions into an OptionSet, add a
DoesNotAwait flag to go with DoesNotThrow, and sink it
all down into SILInstruction.h.
Then, replace the isNonThrowing() flag in ApplyInst and
BeginApplyInst with getApplyOptions(), and plumb it
through to TryApplyInst as well.
Set the flag when SILGen emits a sync call to a reasync
function.
When set, this disables the SIL verifier check against
calling async functions from sync functions.
Finally, this allows us to add end-to-end tests for
rdar://problem/71098795.
It fixes a crash in SILGen if Builtin.createAsyncTaskFuture is used in a non-generic context.
I found this by experiment - we don't use it currently in the stdlib. But it doesn't harm to fix this.
In derivatives of loops, no longer allocate boxes for indirect case payloads. Instead, use a custom pullback context in the runtime which contains a bump-pointer allocator.
When a function contains a differentiated loop, the closure context is a `Builtin.NativeObject`, which contains a `swift::AutoDiffLinearMapContext` and a tail-allocated top-level linear map struct (which represents the linear map struct that was previously directly partial-applied into the pullback). In branching trace enums, the payloads of previously indirect cases will be allocated by `swift::AutoDiffLinearMapContext::allocate` and stored as a `Builtin.RawPointer`.
Implement SIL generation for "async let" constructs, which involves:
1. Creating a child task future at the point of declaration of the "async let",
which runs the initializer in an async closure.
2. Entering a cleanup to destroy the child task.
3. Entering a cleanup to cancel the child task.
4. Waiting for the child task when any of the variables is reference.
5. Decomposing the result of the child task to write the results into the
appropriate variables.
Implements rdar://71123479.
This makes it easier to understand conceptually why a ValueOwnershipKind with
Any ownership is invalid and also allowed me to explicitly document the lattice
that relates ownership constraints/value ownership kinds.
`Builtin.createAsyncTask` takes flags, an optional parent task, and an
async/throwing function to execute, and passes it along to the
`swift_task_create_f` entry point to create a new (potentially child)
task, returning the new task and its initial context.
Implement a new builtin, `cancelAsyncTask()`, to cancel the given
asynchronous task. This lowers down to a call into the runtime
operation `swift_task_cancel()`.
Use this builtin to implement Task.Handle.cancel().
Rather than produce an "unowned" result from `getCurrentAsyncTask()`,
take advantage of the fact that the task is effectively guaranteed in
the scope. Do so be returning it as "unowned", and push an
end_lifetime cleanup to end the lifetime. This eliminates unnecessary
ref-count traffic as well as introducing another use of unowned.
Approach is thanks to Michael Gottesman, bugs are mine.
This introduces a new builtin, `getCurrentAsyncTask()`, that produces a
reference to the current task. This builtin can only be used within
`async` functions, and IR generation merely grabs the task argument
and packages it up.
The type of this function is `() -> Builtin.NativeObject`, because we
don't currently have a Swift-level representation of tasks, and can
probably handle everything through builtins or runtime calls.
`DifferentiableFunctionInst` now stores result indices.
`SILAutoDiffIndices` now stores result indices instead of a source index.
`@differentiable` SIL function types may now have multiple differentiability
result indices and `@noDerivative` resutls.
`@differentiable` AST function types do not have `@noDerivative` results (yet),
so this functionality is not exposed to users.
Resolves TF-689 and TF-1256.
Infrastructural support for TF-983: supporting differentiation of `apply`
instructions with multiple active semantic results.
* Add all [differential operators](https://github.com/apple/swift/blob/master/docs/DifferentiableProgramming.md#list-of-differential-operators).
* Add `withoutDerivative(at:)`, used for efficiently stopping the derivative propagation at a value and causing the derivative at the value to be zero.
* Add utility `differentiableFunction(from:)`, used for creating a `@differentiable` function from an original function and a derivative function.
Mostly work done by @marcrasi and @dan-zheng.
Partially resolves TF-843.
TODO:
* Add `AnyDerivative`.
* Add `Array.differentiableMap(_:)` and `differentiableReduce(_:_:)`.
Define type signatures and SILGen for the following builtins:
```
/// Applies the {jvp|vjp} of `f` to `arg1`, ..., `argN`.
func applyDerivative_arityN_{jvp|vjp}(f, arg1, ..., argN) -> jvp/vjp return type
/// Applies the transpose of `f` to `arg`.
func applyTranspose_arityN(f, arg) -> transpose return type
/// Makes a differentiable function from the given `original`, `jvp`, and
/// `vjp` functions.
func differentiableFunction_arityN(original, jvp, vjp)
/// Makes a linear function from the given `original` and `transpose` functions.
func linearFunction_arityN(original, transpose)
```
Add SILGen FileCheck tests for all builtins.
(BaseT, @inout @unowned(unsafe) T) -> @guaranteed T
The reason for the weird signature is that currently the Builtin infrastructure
does not handle results well. Also, note that we are not actually performing a
call here. We are SILGening directly so we can create a guaranteed result.
The intended semantics is that one passes in a base value that guarantees the
lifetime of the unowned(unsafe) value. The builtin then:
1. Borrows the base.
2. Loads the trivial unowned (unsafe), converts that value to a guaranteed ref
after unsafely unwrapping the optional.
3. Uses mark dependence to tie the lifetimes of the guaranteed base to the
guaranteed ref.
I also updated my small UnsafeValue.swift test to make sure we get the codegen
we expect.
The signature is:
(T, @inout @unowned(unsafe) Optional<T>) -> ()
The reason for the weird signature is that currently the Builtin infrastructure
does not handle results well.
The semantics of this builtin is that it enables one to store the first argument
into an unowned unsafe address without any reference counting operations. It
does this just by SILGening the relevant code. The optimizer chews through this
code well, so we get the expected behavior.
I also included a small proof of concept to validate that this builtin works as
expected.
This ensures that the optimizer has a summary of where the ref_tail_addr will no
longer be used. This is important when analyzing the lifetime of the base of the
ref_tail_addr.
I also cleaned up a little the description around the specification for this in
OperandOwnership. Now all instructions that are "INTERIOR_POINTER_PROJECTIONS"
have their own section/macro as a form of self documenting.
Most of AST, Parse, and Sema deal with FileUnits regularly, but SIL
and IRGen certainly don't. Split FileUnit out into its own header to
cut down on recompilation times when something changes.
No functionality change.
We have to be source compatible to be able to parse old swiftinterface files where the old Builtin.condfail is used in inlineable functions.
rdar://problem/53176692
The SIL generation for this builtin also changes: instead of generating the cond_fail instructions upfront, let the optimizer generate it, if the operand is a static string literal.
In worst case, if the second operand is not a static string literal, the Builtin.condfail is lowered at the end of the optimization pipeline with a default message: "unknown program error".
The SIL generation for this builtin also changes: instead of generating the cond_fail instructions upfront, let the optimizer generate it, if the operand is a static string literal.
In worst case, if the second operand is not a static string literal, the Builtin.condfail is lowered at the end of the optimization pipeline with a default message: "unknown program error".
String -> Builtin.RawPointer that given a string constructed from a
literal, returns the address of the string literal in the global string
table of the compiled binary as a pointer.
For anything else, we can decompose the argument list on the spot.
Note that builtins that are implemented as EarlyEmitters now take a
the argument list as a PreparedArguments instead of a single Expr.
Since the PreparedArguments can still be a scalar with an
ArgumentShuffleExpr, we have to jump through some hoops to turn
it into a list of argument Exprs. This will all go away soon.
Right now we use TupleShuffleExpr for two completely different things:
- Tuple conversions, where elements can be re-ordered and labels can be
introduced/eliminated
- Complex argument lists, involving default arguments or varargs
The first case does not allow default arguments or varargs, and the
second case does not allow re-ordering or introduction/elimination
of labels. Furthermore, the first case has a representation limitation
that prevents us from expressing tuple conversions that change the
type of tuple elements.
For all these reasons, it is better if we use two separate Expr kinds
for these purposes. For now, just make an identical copy of
TupleShuffleExpr and call it ArgumentShuffleExpr. In CSApply, use
ArgumentShuffleExpr when forming the arguments to a call, and keep
using TupleShuffleExpr for tuple conversions. Each usage of
TupleShuffleExpr has been audited to see if it should instead look at
ArgumentShuffleExpr.
In sequent commits I plan on redesigning TupleShuffleExpr to correctly
represent all tuple conversions without any unnecessary baggage.
Longer term, we actually want to change the representation of CallExpr
to directly store an argument list; then instead of a single child
expression that must be a ParenExpr, TupleExpr or ArgumentShuffleExpr,
all CallExprs will have a uniform representation and ArgumentShuffleExpr
will go away altogether. This should reduce memory usage and radically
simplify parts of SILGen.
The ownership kind is Any for trivial types, or Owned otherwise, but
whether a type is trivial or not will soon depend on the resilience
expansion.
This means that a SILModule now uniques two SILUndefs per type instead
of one, and serialization uses two distinct sentinel IDs for this
purpose as well.
For now, the resilience expansion is not actually used here, so this
change is NFC, other than changing the module format.
