|
| 1 | +use timely::dataflow::channels::pact::Pipeline; |
| 2 | +use timely::dataflow::operators::Input; |
| 3 | +use timely::dataflow::InputHandle; |
| 4 | +use timely::Config; |
| 5 | + |
| 6 | +#[test] fn operator_scaling_1() { operator_scaling(1); } |
| 7 | +#[test] fn operator_scaling_10() { operator_scaling(10); } |
| 8 | +#[test] fn operator_scaling_100() { operator_scaling(100); } |
| 9 | +#[test] #[cfg_attr(miri, ignore)] fn operator_scaling_1000() { operator_scaling(1000); } |
| 10 | +#[test] #[cfg_attr(miri, ignore)] fn operator_scaling_10000() { operator_scaling(10000); } |
| 11 | +#[test] #[cfg_attr(miri, ignore)] fn operator_scaling_100000() { operator_scaling(100000); } |
| 12 | + |
| 13 | +fn operator_scaling(scale: u64) { |
| 14 | + timely::execute(Config::thread(), move |worker| { |
| 15 | + let mut input = InputHandle::new(); |
| 16 | + worker.dataflow::<u64, _, _>(|scope| { |
| 17 | + use timely::dataflow::operators::Partition; |
| 18 | + let parts = |
| 19 | + scope |
| 20 | + .input_from(&mut input) |
| 21 | + .partition(scale, |()| (0, ())); |
| 22 | + |
| 23 | + use timely::dataflow::operators::generic::builder_rc::OperatorBuilder; |
| 24 | + let mut builder = OperatorBuilder::new("OpScaling".to_owned(), scope.clone()); |
| 25 | + let mut handles = Vec::with_capacity(parts.len()); |
| 26 | + let mut outputs = Vec::with_capacity(parts.len()); |
| 27 | + for (index, part) in parts.into_iter().enumerate() { |
| 28 | + use timely::container::CapacityContainerBuilder; |
| 29 | + let (output, stream) = builder.new_output_connection::<CapacityContainerBuilder<Vec<()>>,_>([]); |
| 30 | + use timely::progress::Antichain; |
| 31 | + let connectivity = [(index, Antichain::from_elem(Default::default()))]; |
| 32 | + handles.push((builder.new_input_connection(&part, Pipeline, connectivity), output)); |
| 33 | + outputs.push(stream); |
| 34 | + } |
| 35 | + |
| 36 | + builder.build(move |_| { |
| 37 | + move |_frontiers| { |
| 38 | + for (input, output) in handles.iter_mut() { |
| 39 | + let mut output = output.activate(); |
| 40 | + input.for_each(|time, data| { |
| 41 | + let mut output = output.session_with_builder(&time); |
| 42 | + for datum in data.drain(..) { |
| 43 | + output.give(datum); |
| 44 | + } |
| 45 | + }); |
| 46 | + } |
| 47 | + } |
| 48 | + }); |
| 49 | + }); |
| 50 | + }) |
| 51 | + .unwrap(); |
| 52 | +} |
| 53 | + |
| 54 | +#[test] fn subgraph_scaling_1() { subgraph_scaling(1); } |
| 55 | +#[test] fn subgraph_scaling_10() { subgraph_scaling(10); } |
| 56 | +#[test] fn subgraph_scaling_100() { subgraph_scaling(100); } |
| 57 | +#[test] #[cfg_attr(miri, ignore)] fn subgraph_scaling_1000() { subgraph_scaling(1000); } |
| 58 | +#[test] #[cfg_attr(miri, ignore)] fn subgraph_scaling_10000() { subgraph_scaling(10000); } |
| 59 | +#[test] #[cfg_attr(miri, ignore)] fn subgraph_scaling_100000() { subgraph_scaling(100000); } |
| 60 | + |
| 61 | +fn subgraph_scaling(scale: u64) { |
| 62 | + timely::execute(Config::thread(), move |worker| { |
| 63 | + let mut input = InputHandle::new(); |
| 64 | + worker.dataflow::<u64, _, _>(|scope| { |
| 65 | + use timely::dataflow::operators::Partition; |
| 66 | + let parts = |
| 67 | + scope |
| 68 | + .input_from(&mut input) |
| 69 | + .partition(scale, |()| (0, ())); |
| 70 | + |
| 71 | + use timely::dataflow::Scope; |
| 72 | + let _outputs = scope.region(|inner| { |
| 73 | + use timely::dataflow::operators::{Enter, Leave}; |
| 74 | + parts.into_iter().map(|part| part.enter(inner).leave()).collect::<Vec<_>>() |
| 75 | + }); |
| 76 | + }); |
| 77 | + }) |
| 78 | + .unwrap(); |
| 79 | +} |
0 commit comments