Running Tasks in Parallel🔗

Contributors: Anirudh Gupta

To know what Tasks are and how to spawn and handle them, check out Spawning Tasks and Worker Threads recipe before.

Running Tasks in Parallel🔗

One of the most powerful uses of tasks is running multiple IO operations at the same time. Since Task α is a primitive for asynchronous computation, you can spawn multiple tasks to perform IO in the background and wait for their results later. This allows you to interleave the outputs of different tasks, demonstrating concurrency.

def heavyWork (name : String) (iters : Nat) : IO Unit := do
  for i in [0:iters] do
    -- We use a small loop to simulate work
    let mut _x := 0
    for _ in [0:1000] do
      _x := _x + 1
    
    -- Printing acts as a synchronization/yield point
    IO.println s!"{name}: step {i}"

def runParallel : IO Unit := do
  IO.println "Starting heavy work..."

  -- Spawn Task A
  let taskA ← IO.asTask (heavyWork "Task A" 5)
  -- Spawn Task B
  let taskB ← IO.asTask (heavyWork "Task B" 2)

  -- Wait for both to finish
  let _ ← IO.wait taskA
  let _ ← IO.wait taskB
  IO.println "Both tasks finished!"

/-
Task A: step 0
Task B: step 0
Task A: step 1
Task B: step 1
Task A: step 2
Task A: step 3
Task A: step 4
Starting heavy work...
Both tasks finished!
-/
-- #eval runParallel

In this example, the outputs of A and B are interleaved, demonstrating that they are running concurrently. Notice that even when flushed, the "Starting heavy work..." is coming later. When printing the output do to the terminal, but to a buffer and so does other worker thread's output too, and hence based on more interrupts and thread behaviour, the order may differ among main thread and worker threads If you have a better explanation, Please share.

However in many cases due to spawning too many threads, you might create a deadlock. Check out the Deadlocking the Task System section for more information on how to avoid deadlocks when spawning too many threads.