Dagger Functions

Dagger.@spawn [option=value]... f(args...; kwargs...) -> DTask

Spawns a Dagger DTask that will call f(args...; kwargs...). This DTask is like a Julia Task, and has many similarities:

• The DTask can be wait'd on and fetch'd from to see its final result
• By default, the DTask will be automatically run on the first available compute resource
• If all dependencies are satisfied, the DTask will be run as soon as possible
• The DTask may be run in parallel with other DTasks, and the scheduler will automatically manage dependencies
• If a DTask throws an exception, it will be propagated to any calls to fetch, but not to calls to wait

However, the DTask also has many key differences from a Task:

• The DTask may run on any thread of any Julia process, and even on a remote machine, in your cluster (see Distributed.addprocs)
• The DTask might automatically utilize GPUs or other accelerators, if available
• If arguments to a DTask are also DTasks, then the scheduler will execute those arguments' DTasks first, before running the "downstream" task
• If an argument to a DTask t2 is a DTask t1, then the result of t1 (gotten via fetch(t1)) will be passed to t2 (no need for t2 to call fetch!)
• DTasks are generally expected to be defined "functionally", meaning that they should not mutate global state, mutate their arguments, or have side effects
• DTasks are function call-focused, meaning that Dagger.@spawn expects a single function call, and not a block of code
• All DTask arguments are expected to be safe to serialize and send to other Julia processes; if not, use the scope option or Dagger.@mutable to control execution location

Options to the DTask can be set before the call to f with key-value syntax, e.g. Dagger.@spawn myopt=2 do_something(1, 3.0), which would set the option myopt to 2 for this task. Multiple options may be provided, which are specified like Dagger.@spawn myopt=2 otheropt=4 do_something(1, 3.0).

These options control a variety of properties of the resulting DTask:

• scope: The execution "scope" of the task, which determines where the task will run. By default, the task will run on the first available compute resource. If you have multiple compute resources, you can specify a scope to run the task on a specific resource. For example, Dagger.@spawn scope=Dagger.scope(worker=2) do_something(1, 3.0) would run do_something(1, 3.0) on worker 2.
• meta: If true, instead of the scheduler automatically fetching values from other tasks, the raw Chunk objects will be passed to f. Useful for doing manual fetching or manipulation of Chunk references. Non-Chunk arguments are still passed as-is.

Other options exist; see Dagger.Sch.ThunkOptions for the full list.

This macro is a semi-thin wrapper around Dagger.spawn - it creates a call to Dagger.spawn on f with arguments args and keyword arguments kwargs, and also passes along any options in an Options struct. For example, Dagger.@spawn myopt=2 do_something(1, 3.0) would essentially become Dagger.spawn(do_something, Dagger.Options(;myopt=2), 1, 3.0).

Dagger.spawn(f, args...; kwargs...) -> DTask

Spawns a DTask that will call f(args...; kwargs...). Also supports passing a Dagger.Options struct as the first argument to set task options. See Dagger.@spawn for more details on DTasks.

with_options(f, options::NamedTuple) -> Any
with_options(f; options...) -> Any

Sets one or more options to the given values, executes f(), resets the options to their previous values, and returns the result of f(). This is the recommended way to set options, as it only affects tasks spawned within its scope. Note that setting an option here will propagate its value across Julia or Dagger tasks spawned by f() or its callees (i.e. the options propagate).

get_options(key::Symbol, default) -> Any
get_options(key::Symbol) -> Any

Returns the value of the option named key. If option does not have a value set, then an error will be thrown, unless default is set, in which case it will be returned instead of erroring.

get_options() -> NamedTuple

Returns a NamedTuple of all option key-value pairs.

@option name myfunc(A, B, C) = value

A convenience macro for defining default_option. For example:

Dagger.@option single mylocalfunc(Int) = 1

The above call will set the single option to 1 for any Dagger task calling mylocalfunc(Int) with an Int argument.

default_option(::Val{name}, Tf, Targs...) where name = value

Defines the default value for option name to value when Dagger is preparing to execute a function with type Tf with the argument types Targs. Users and libraries may override this to set default values for tasks.

An easier way to define these defaults is with @option.

Note that the actual task's argument values are not passed, as it may not always be possible or efficient to gather all Dagger task arguments on one worker.

This function may be executed within the scheduler, so it should generally be made very cheap to execute. If the function throws an error, the scheduler will use whatever the global default value is for that option instead.

tochunk(x, proc::Processor, scope::AbstractScope; device=nothing, kwargs...) -> Chunk

Create a chunk from data x which resides on proc and which has scope scope.

device specifies a MemPool.StorageDevice (which is itself wrapped in a Chunk) which will be used to manage the reference contained in the Chunk generated by this function. If device is nothing (the default), the data will be inspected to determine if it's safe to serialize; if so, the default MemPool storage device will be used; if not, then a MemPool.CPURAMDevice will be used.

All other kwargs are passed directly to MemPool.poolset.

mutable(f::Base.Callable; worker, processor, scope) -> Chunk

Calls f() on the specified worker or processor, returning a Chunk referencing the result with the specified scope scope.

@mutable [worker=1] [processor=OSProc()] [scope=ProcessorScope()] f()

Helper macro for mutable().

Creates a Shard. See Dagger.shard for details.

shard(f; kwargs...) -> Chunk{Shard}

Executes f on all workers in workers, wrapping the result in a process-scoped Chunk, and constructs a Chunk{Shard} containing all of these Chunks on the current worker.

Keyword arguments:

• procs – The list of processors to create pieces on. May be any iterable container of Processors.
• workers – The list of workers to create pieces on. May be any iterable container of Integers.
• per_thread::Bool=false – If true, creates a piece per each thread, rather than a piece per each worker.

spawn_datadeps(f::Base.Callable; traversal::Symbol=:inorder)

Constructs a "datadeps" (data dependencies) region and calls f within it. Dagger tasks launched within f may wrap their arguments with In, Out, or InOut to indicate whether the task will read, write, or read+write that argument, respectively. These argument dependencies will be used to specify which tasks depend on each other based on the following rules:

• Dependencies across different arguments are independent; only dependencies on the same argument synchronize with each other ("same-ness" is determined based on isequal)
• InOut is the same as In and Out applied simultaneously, and synchronizes with the union of the In and Out effects
• Any two or more In dependencies do not synchronize with each other, and may execute in parallel
• An Out dependency synchronizes with any previous In and Out dependencies
• An In dependency synchronizes with any previous Out dependencies
• If unspecified, an In dependency is assumed

In general, the result of executing tasks following the above rules will be equivalent to simply executing tasks sequentially and in order of submission. Of course, if dependencies are incorrectly specified, undefined behavior (and unexpected results) may occur.

Unlike other Dagger tasks, tasks executed within a datadeps region are allowed to write to their arguments when annotated with Out or InOut appropriately.

At the end of executing f, spawn_datadeps will wait for all launched tasks to complete, rethrowing the first error, if any. The result of f will be returned from spawn_datadeps.

The keyword argument traversal controls the order that tasks are launched by the scheduler, and may be set to :bfs or :dfs for Breadth-First Scheduling or Depth-First Scheduling, respectively. All traversal orders respect the dependencies and ordering of the launched tasks, but may provide better or worse performance for a given set of datadeps tasks. This argument is experimental and subject to change.

scope(scs...) -> AbstractScope
scope(;scs...) -> AbstractScope

Constructs an AbstractScope from a set of scope specifiers. Each element in scs is a separate specifier; if scs is empty, an empty UnionScope() is produced; if scs has one element, then exactly one specifier is constructed; if scs has more than one element, a UnionScope of the scopes specified by scs is constructed. A variety of specifiers can be passed to construct a scope:

• :any - Constructs an AnyScope()
• :default - Constructs a DefaultScope()
• (scs...,) - Constructs a UnionScope of scopes, each specified by scs
• thread=tid or threads=[tids...] - Constructs an ExactScope or UnionScope containing all Dagger.ThreadProcs with thread ID tid/tids across all workers.
• worker=wid or workers=[wids...] - Constructs a ProcessScope or UnionScope containing all Dagger.ThreadProcs with worker ID wid/wids across all threads.
• thread=tid/threads=tids and worker=wid/workers=wids - Constructs an ExactScope, ProcessScope, or UnionScope containing all Dagger.ThreadProcs with worker ID wid/wids and threads tid/tids.

Aside from the worker and thread specifiers, it's possible to add custom specifiers for scoping to other kinds of processors (like GPUs) or providing different ways to specify a scope. Specifier selection is determined by a precedence ordering: by default, all specifiers have precedence 0, which can be changed by defining scope_key_precedence(::Val{spec}) = precedence (where spec is the specifier as a Symbol). The specifier with the highest precedence in a set of specifiers is used to determine the scope by calling to_scope(::Val{spec}, sc::NamedTuple) (where sc is the full set of specifiers), which should be overriden for each custom specifier, and which returns an AbstractScope. For example:

# Setup a GPU specifier
Dagger.scope_key_precedence(::Val{:gpu}) = 1
Dagger.to_scope(::Val{:gpu}, sc::NamedTuple) = ExactScope(MyGPUDevice(sc.worker, sc.gpu))

# Generate an `ExactScope` for `MyGPUDevice` on worker 2, device 3
Dagger.scope(gpu=3, worker=2)

constraint(x::AbstractScope, y::AbstractScope) -> ::AbstractScope

Constructs a scope that is the intersection of scopes x and y.

execute!(proc::Processor, f, args...; kwargs...) -> Any

Executes the function f with arguments args and keyword arguments kwargs on processor proc. This function can be overloaded by Processor subtypes to allow executing function calls differently than normal Julia.

iscompatible(proc::Processor, opts, f, Targs...) -> Bool

Indicates whether proc can execute f over Targs given opts. Processor subtypes should overload this function to return true if and only if it is essentially guaranteed that f(::Targs...) is supported. Additionally, iscompatible_func and iscompatible_arg can be overriden to determine compatibility of f and Targs individually. The default implementation returns false.

default_enabled(proc::Processor) -> Bool

Returns whether processor proc is enabled by default. The default value is false, which is an opt-out of the processor from execution when not specifically requested by the user, and true implies opt-in, which causes the processor to always participate in execution when possible.

get_processors(proc::Processor) -> Set{<:Processor}

Returns the set of processors contained in proc, if any. Processor subtypes should overload this function if they can contain sub-processors. The default method will return a Set containing proc itself.

get_parent(proc::Processor) -> Processor

Returns the parent processor for proc. The ultimate parent processor is an OSProc. Processor subtypes should overload this to return their most direct parent.

move(from_proc::Processor, to_proc::Processor, x)

Moves and/or converts x such that it's available and suitable for usage on the to_proc processor. This function can be overloaded by Processor subtypes to transport arguments and convert them to an appropriate form before being used for exection. Subtypes of Processor wishing to implement efficient data movement should provide implementations where x::Chunk.

get_tls()

Gets all Dagger TLS variable as a NamedTuple.

set_tls!(tls)

Sets all Dagger TLS variables from the NamedTuple tls.

addprocs!(ctx::Context, xs)

Add new workers xs to ctx.

Workers will typically be assigned new tasks in the next scheduling iteration if scheduling is ongoing.

Workers can be either Processors or the underlying process IDs as Integers.

rmprocs!(ctx::Context, xs)

Remove the specified workers xs from ctx.

Workers will typically finish all their assigned tasks if scheduling is ongoing but will not be assigned new tasks after removal.

Workers can be either Processors or the underlying process IDs as Integers.

in_task()

Returns true if currently executing in a DTask, else false.

task_processor()

Get the current processor executing the current Dagger task.

Gets the scheduler handle for the currently-executing thunk.

Adds a new Thunk to the DAG.

Base.fetch(c::DArray)

If a DArray tree has a Thunk in it, make the whole thing a big thunk.

Waits on a thunk to complete, and fetches its result.

Waits on a thunk to complete.

Executes an arbitrary function within the scheduler, returning the result.

Commands the scheduler to halt execution immediately.

Returns all Thunks IDs as a Dict, mapping a Thunk to its downstream dependents.