Function timely::communication::initialize
source · pub fn initialize<T, F>(
config: Config,
func: F
) -> Result<WorkerGuards<T>, String>
Expand description
Initializes communication and executes a distributed computation.
This method allocates an allocator::Generic
for each thread, spawns local worker threads,
and invokes the supplied function with the allocator.
The method returns a WorkerGuards<T>
which can be join
ed to retrieve the return values
(or errors) of the workers.
§Examples
use timely_communication::Allocate;
// configure for two threads, just one process.
let config = timely_communication::Config::Process(2);
// initializes communication, spawns workers
let guards = timely_communication::initialize(config, |mut allocator| {
println!("worker {} started", allocator.index());
// allocates a pair of senders list and one receiver.
let (mut senders, mut receiver) = allocator.allocate(0);
// send typed data along each channel
use timely_communication::Message;
senders[0].send(Message::from_typed(format!("hello, {}", 0)));
senders[1].send(Message::from_typed(format!("hello, {}", 1)));
// no support for termination notification,
// we have to count down ourselves.
let mut expecting = 2;
while expecting > 0 {
allocator.receive();
if let Some(message) = receiver.recv() {
use std::ops::Deref;
println!("worker {}: received: <{}>", allocator.index(), message.deref());
expecting -= 1;
}
allocator.release();
}
// optionally, return something
allocator.index()
});
// computation runs until guards are joined or dropped.
if let Ok(guards) = guards {
for guard in guards.join() {
println!("result: {:?}", guard);
}
}
else { println!("error in computation"); }
The should produce output like:
ⓘ
worker 0 started
worker 1 started
worker 0: received: <hello, 0>
worker 1: received: <hello, 1>
worker 0: received: <hello, 0>
worker 1: received: <hello, 1>
result: Ok(0)
result: Ok(1)