Tasking

Section author: Gustavo A. Hoffmann

Tasks and protected objects allow the implementation of concurrency in Ada. The following sections explain these concepts in more details.

Tasks

A task can be thought as an application that runs concurrently with the main application. In other programming languages, a task can be called a thread, and tasking can be called multithreading.

Tasks may synchronize with the main application but may also process information completely independent from the main application. Here we show how this is accomplished.

Simple task

Tasks are declared using the keyword task. The task implementation is specified in a task body block. For example:

with Ada.Text_IO; use Ada.Text_IO; procedure Show_Simple_Task is task T; task body T is begin Put_Line ("In task T"); end T; begin Put_Line ("In main"); end Show_Simple_Task;

Here, we're declaring and implementing the task T. As soon as the main application starts, task T starts automatically --- it's not necessary to manually start this task. By running the application above, we can see that both calls to Put_Line are performed.

Note that:

  • The main application is itself a task (the main task).

    • In this example, the subprogram Show_Simple_Task is the main task of the application.

  • Task T is a subtask.

    • Each subtask has a master task.
    • Therefore the main task is also the master task of task T.

  • The number of tasks is not limited to one: we could include a task T2 in the example above.

    • This task also starts automatically and runs concurrently with both task T and the main task. For example:

      with Ada.Text_IO; use Ada.Text_IO; procedure Show_Simple_Tasks is task T; task T2; task body T is begin Put_Line ("In task T"); end T; task body T2 is begin Put_Line ("In task T2"); end T2; begin Put_Line ("In main"); end Show_Simple_Tasks;

Simple synchronization

As we've just seen, as soon as the main task starts, its subtasks also start automatically. The main task continues its processing until it has nothing more to do. At that point, however, it will not terminate. Instead, the task waits until its subtasks have finished before it allows itself to terminate. In other words, this waiting process provides synchronization between the main task and its subtasks. After this synchronization, the main task will terminate. For example:

with Ada.Text_IO; use Ada.Text_IO; procedure Show_Simple_Sync is task T; task body T is begin for I in 1 .. 10 loop Put_Line ("hello"); end loop; end T; begin null; -- Will wait here until all tasks have terminated end Show_Simple_Sync;

The same mechanism is used for other subprograms that contain subtasks: the subprogram's master task will wait for its subtasks to finish. So this mechanism is not limited to the main application and also applies to any subprogram called by the main application or its subprograms.

Synchronization also occurs if we move the task to a separate package. In the example below, we declare a task T in the package Simple_Sync_Pkg.

package Simple_Sync_Pkg is task T; end Simple_Sync_Pkg;

This is the corresponding package body:

with Ada.Text_IO; use Ada.Text_IO; package body Simple_Sync_Pkg is task body T is begin for I in 1 .. 10 loop Put_Line ("hello"); end loop; end T; end Simple_Sync_Pkg;

Because the package is with'ed by the main procedure, the task T defined in the package is part of the main task. For example:

with Simple_Sync_Pkg; procedure Test_Simple_Sync_Pkg is begin null; -- Will wait here until all tasks have terminated end Test_Simple_Sync_Pkg;

Again, as soon as the main task reaches its end, it synchronizes with task T from Simple_Sync_Pkg before terminating.

Delay

We can introduce a delay by using the keyword delay. This puts the task to sleep for the length of time (in seconds) specified in the delay statement. For example:

with Ada.Text_IO; use Ada.Text_IO; procedure Show_Delay is task T; task body T is begin for I in 1 .. 5 loop Put_Line ("hello from task T"); delay 1.0; -- ^ Wait 1.0 seconds end loop; end T; begin delay 1.5; Put_Line ("hello from main"); end Show_Delay;

In this example, we're making the task T wait one second after each time it displays the "hello" message. In addition, the main task is waiting 1.5 seconds before displaying its own "hello" message

Synchronization: rendez-vous

The only type of synchronization we've seen so far is the one that happens automatically at the end of the main task. You can also define custom synchronization points using the keyword entry. An entry can be viewed as a special kind of subprogram, which is called by the master task using a similar syntax, as we will see later.

In the task definition, you define which part of the task will accept the entries by using the keyword accept. A task proceeds until it reaches an accept statement and then waits for the master task to synchronize with it. Specifically,

  • The subtask waits at that point (in the accept statement), ready to accept a call to the corresponding entry from the master task.
  • The master task calls the task entry, in a manner similar to a procedure call, to synchronize with the subtask.

This synchronization between tasks is called rendez-vous. Let's see an example:

with Ada.Text_IO; use Ada.Text_IO; procedure Show_Rendezvous is task T is entry Start; end T; task body T is begin accept Start; -- Waiting for somebody to call the entry Put_Line ("In T"); end T; begin Put_Line ("In Main"); T.Start; -- Calling T's entry end Show_Rendezvous;

In this example, we declare an entry Start for task T. In the task body, we implement this entry using accept Start. When task T reaches this point, it waits for the master task. This synchronization occurs in the T.Start statement. After the synchronization completes, the main task and task T again run concurrently until they synchronize one final time when the main task finishes.

An entry may be used to perform more than a simple task synchronization: it also may perform multiple statements during the time both tasks are synchronized. We do this with a do ... end block. For the previous example, we would simply write accept Start do <statements>; end;. We use this kind of block in the next example.

Select loop

There's no limit to the number of times an entry can be accepted. We could even create an infinite loop in the task and accept calls to the same entry over and over again. An infinite loop, however, prevents the subtask from finishing, so it blocks the master task when it reaches the end of its processing. Therefore, a loop containing accept statements in a task body is normally used in conjunction with a select ... or terminate statement. In simple terms, this statement allows the master task to automatically terminate the subtask when the master task finishes. For example:

with Ada.Text_IO; use Ada.Text_IO; procedure Show_Rendezvous_Loop is task T is entry Start; end T; task body T is Cnt : Integer := 0; begin loop select accept Start do Cnt := Cnt + 1; end Start; Put_Line ("In T's loop (" & Integer'Image (Cnt) & ")"); or terminate; end select; end loop; end T; begin Put_Line ("In Main"); for I in 1 .. 4 loop T.Start; -- Calling T's entry multiple times end loop; end Show_Rendezvous_Loop;

In this example, the task body implements an infinite loop that accepts calls to the Start entry. We make the following observations:

  • The accept E do ... end block is used to increment a counter.

    • As long as task T is performing the do ... end block, the main task waits for the block to complete.

  • The main task is calling the Start entry multiple times in the loop from 1 .. 4.

    • Because task T contains an infinite loop, it always accepts calls to the Start entry.
    • When the main task finishes, it checks the status of the T task. Even though task T could accept new calls to the Start entry, the master task is allowed to terminate task T due to the or terminate part of the select statement.

Cycling tasks

In a previous example, we saw how to delay a task a specified time by using the delay keyword. However, using delay statements in a loop is not enough to guarantee regular intervals between those delay statements. For example, we may have a call to a computationally intensive procedure between executions of successive delay statements:

while True loop
   delay 1.0;
   --    ^ Wait 1.0 seconds
   Computational_Intensive_App;
end loop;

In this case, we can't guarantee that exactly 10 seconds have elapsed after 10 calls to the delay statement because a time drift may be introduced by the Computational_Intensive_App procedure. In many cases, this time drift is not relevant, so using the delay keyword is good enough.

However, there are situations where a time drift isn't acceptable. In those cases, we need to use the delay until statement, which accepts a precise time for the end of the delay, allowing us to define a regular interval. This is useful, for example, in real-time applications.

We will soon see an example of how this time drift may be introduced and how the delay until statement circumvents the problem. But before we do that, we look at a package containing a procedure allowing us to measure the elapsed time (Show_Elapsed_Time) and a dummy Computational_Intensive_App procedure which is simulated by using a simple delay. This is the package specification:

with Ada.Real_Time; use Ada.Real_Time; package Delay_Aux_Pkg is function Get_Start_Time return Time with Inline; procedure Show_Elapsed_Time with Inline; procedure Computational_Intensive_App; private Start_Time : Time := Clock; function Get_Start_Time return Time is (Start_Time); end Delay_Aux_Pkg;

And this is the package body:

with Ada.Text_IO; use Ada.Text_IO; package body Delay_Aux_Pkg is procedure Show_Elapsed_Time is Now_Time : Time; Elapsed_Time : Time_Span; begin Now_Time := Clock; Elapsed_Time := Now_Time - Start_Time; Put_Line ("Elapsed time " & Duration'Image (To_Duration (Elapsed_Time)) & " seconds"); end Show_Elapsed_Time; procedure Computational_Intensive_App is begin delay 0.5; end Computational_Intensive_App; end Delay_Aux_Pkg;

Using this auxiliary package, we're now ready to write our time-drifting application:

with Ada.Text_IO; use Ada.Text_IO; with Ada.Real_Time; use Ada.Real_Time; with Delay_Aux_Pkg; procedure Show_Time_Drifting_Task is package Aux renames Delay_Aux_Pkg; task T; task body T is Cnt : Integer := 1; begin for I in 1 .. 5 loop delay 1.0; Aux.Show_Elapsed_Time; Aux.Computational_Intensive_App; Put_Line ("Cycle # " & Integer'Image (Cnt)); Cnt := Cnt + 1; end loop; Put_Line ("Finished time-drifting loop"); end T; begin null; end Show_Time_Drifting_Task;

We can see by running the application that we already have a time difference of about four seconds after three iterations of the loop due to the drift introduced by Computational_Intensive_App. Using the delay until statement, however, we're able to avoid this time drift and have a regular interval of exactly one second:

with Ada.Text_IO; use Ada.Text_IO; with Ada.Real_Time; use Ada.Real_Time; with Delay_Aux_Pkg; procedure Show_Cycling_Task is package Aux renames Delay_Aux_Pkg; task T; task body T is Cycle : constant Time_Span := Milliseconds (1000); Next : Time := Aux.Get_Start_Time + Cycle; Cnt : Integer := 1; begin for I in 1 .. 5 loop delay until Next; Aux.Show_Elapsed_Time; Aux.Computational_Intensive_App; -- Calculate next execution time using a -- cycle of one second Next := Next + Cycle; Put_Line ("Cycle # " & Integer'Image (Cnt)); Cnt := Cnt + 1; end loop; Put_Line ("Finished cycling"); end T; begin null; end Show_Cycling_Task;

Now, as we can see by running the application, the delay until statement ensures that the Computational_Intensive_App doesn't disturb the regular interval of one second between iterations.

Protected objects

When multiple tasks are accessing shared data, corruption of that data may occur. For example, data may be inconsistent if one task overwrites parts of the information that's being read by another task at the same time. In order to avoid these kinds of problems and ensure information is accessed in a coordinated way, we use protected objects.

Protected objects encapsulate data and provide access to that data by means of protected operations, which may be subprograms or protected entries. Using protected objects ensures that data is not corrupted by race conditions or other simultaneous access.

Important

Protected objects can be implemented using Ada tasks. In fact, this was the only possible way of implementing them in Ada 83 (the first version of the Ada language). However, the use of protected objects is much simpler than using similar mechanisms implemented using only tasks. Therefore, you should use protected objects when your main goal is only to protect data.

Simple object

You declare a protected object with the protected keyword. The syntax is similar to that used for packages: you can declare operations (e.g., procedures and functions) in the public part and data in the private part. The corresponding implementation of the operations is included in the protected body of the object. For example:

with Ada.Text_IO; use Ada.Text_IO; procedure Show_Protected_Objects is protected Obj is -- Operations go here (only subprograms) procedure Set (V : Integer); function Get return Integer; private -- Data goes here Local : Integer := 0; end Obj; protected body Obj is -- procedures can modify the data procedure Set (V : Integer) is begin Local := V; end Set; -- functions cannot modify the data function Get return Integer is begin return Local; end Get; end Obj; begin Obj.Set (5); Put_Line ("Number is: " & Integer'Image (Obj.Get)); end Show_Protected_Objects;

In this example, we define two operations for Obj: Set and Get. The implementation of these operations is in the Obj body. The syntax used for writing these operations is the same as that for normal procedures and functions. The implementation of protected objects is straightforward --- we simply access and update Local in these subprograms. To call these operations in the main application, we use prefixed notation, e.g., Obj.Get.

Entries

In addition to protected procedures and functions, you can also define protected entry points. Do this using the entry keyword. Protected entry points allow you to define barriers using the when keyword. Barriers are conditions that must be fulfilled before the entry can start performing its actual processing --- we speak of releasing the barrier when the condition is fulfilled.

The previous example used procedures and functions to define operations on the protected objects. However, doing so permits reading protected information (via Obj.Get) before it's set (via Obj.Set). To allow that to be a defined operation, we specified a default value (0). Instead, by rewriting Obj.Get using an entry instead of a function, we implement a barrier, ensuring no task can read the information before it's been set.

The following example implements the barrier for the Obj.Get operation. It also contains two concurrent subprograms (main task and task T) that try to access the protected object.

with Ada.Text_IO; use Ada.Text_IO; procedure Show_Protected_Objects_Entries is protected Obj is procedure Set (V : Integer); entry Get (V : out Integer); private Local : Integer; Is_Set : Boolean := False; end Obj; protected body Obj is procedure Set (V : Integer) is begin Local := V; Is_Set := True; end Set; entry Get (V : out Integer) when Is_Set is -- Entry is blocked until the condition is true. -- The barrier is evaluated at call of entries and at exits of -- procedures and entries. -- The calling task sleeps until the barrier is released. begin V := Local; Is_Set := False; end Get; end Obj; N : Integer := 0; task T; task body T is begin Put_Line ("Task T will delay for 4 seconds..."); delay 4.0; Put_Line ("Task T will set Obj..."); Obj.Set (5); Put_Line ("Task T has just set Obj..."); end T; begin Put_Line ("Main application will get Obj..."); Obj.Get (N); Put_Line ("Main application has just retrieved Obj..."); Put_Line ("Number is: " & Integer'Image (N)); end Show_Protected_Objects_Entries;

As we see by running it, the main application waits until the protected object is set (by the call to Obj.Set in task T) before it reads the information (via Obj.Get). Because a 4-second delay has been added in task T, the main application is also delayed by 4 seconds. Only after this delay does task T set the object and release the barrier in Obj.Get so that the main application can then resume processing (after the information is retrieved from the protected object).

Task and protected types

In the previous examples, we defined single tasks and protected objects. We can, however, generalize tasks and protected objects using type definitions. This allows us, for example, to create multiple tasks based on just a single task type.

Task types

A task type is a generalization of a task. The declaration is similar to simple tasks: you replace task with task type. The difference between simple tasks and task types is that task types don't create actual tasks that automatically start. Instead, a task declaration is needed. This is exactly the way normal variables and types work: objects are only created by variable definitions, not type definitions.

To illustrate this, we repeat our first example:

with Ada.Text_IO; use Ada.Text_IO; procedure Show_Simple_Task is task T; task body T is begin Put_Line ("In task T"); end T; begin Put_Line ("In main"); end Show_Simple_Task;

We now rewrite it by replacing task T with task type TT. We declare a task (A_Task) based on the task type TT after its definition:

with Ada.Text_IO; use Ada.Text_IO; procedure Show_Simple_Task_Type is task type TT; task body TT is begin Put_Line ("In task type TT"); end TT; A_Task : TT; begin Put_Line ("In main"); end Show_Simple_Task_Type;

We can extend this example and create an array of tasks. Since we're using the same syntax as for variable declarations, we use a similar syntax for task types: array (<>) of Task_Type. Also, we can pass information to the individual tasks by defining a Start entry. Here's the updated example:

with Ada.Text_IO; use Ada.Text_IO; procedure Show_Task_Type_Array is task type TT is entry Start (N : Integer); end TT; task body TT is Task_N : Integer; begin accept Start (N : Integer) do Task_N := N; end Start; Put_Line ("In task T: " & Integer'Image (Task_N)); end TT; My_Tasks : array (1 .. 5) of TT; begin Put_Line ("In main"); for I in My_Tasks'Range loop My_Tasks (I).Start (I); end loop; end Show_Task_Type_Array;

In this example, we're declaring five tasks in the array My_Tasks. We pass the array index to the individual tasks in the entry point (Start). After the synchronization between the individual subtasks and the main task, each subtask calls Put_Line concurrently.

Protected types

A protected type is a generalization of a protected object. The declaration is similar to that for protected objects: you replace protected with protected type. Like task types, protected types require an object declaration to create actual objects. Again, this is similar to variable declarations and allows for creating arrays (or other composite objects) of protected objects.

We can reuse a previous example and rewrite it to use a protected type:

with Ada.Text_IO; use Ada.Text_IO; procedure Show_Protected_Object_Type is protected type Obj_Type is procedure Set (V : Integer); function Get return Integer; private Local : Integer := 0; end Obj_Type; protected body Obj_Type is procedure Set (V : Integer) is begin Local := V; end Set; function Get return Integer is begin return Local; end Get; end Obj_Type; Obj : Obj_Type; begin Obj.Set (5); Put_Line ("Number is: " & Integer'Image (Obj.Get)); end Show_Protected_Object_Type;

In this example, instead of directly defining the protected object Obj, we first define a protected type Obj_Type and then declare Obj as an object of that protected type. Note that the main application hasn't changed: we still use Obj.Set and Obj.Get to access the protected object, just like in the original example.