Julia High Performance - Second Edition

Using tasks

Consider the following trivial code. How long do you think this will take?

for i in 1:5    sleep(1)end

Well, the answer is obviously five seconds—a sleep of one second at a time, executed five times, as shown in the following code:

julia> @time for i in 1:5        sleep(1)       end  5.026456 seconds (26 allocations: 1.266 KiB)

Now see what happens when we write this code with the @async macro, as follows:

julia> @time for i in 1:5         @async sleep(1)       end  0.000140 seconds (32 allocations: 4.375 KiB)

The @async macro (short for asynchronous) creates and schedules tasks for all code within its scope (in this case, the sleep(1) function call), which means all the sleep calls happen in separate tasks. The for loop, which is running on the REPL task, returns ...