async_tbb_reduction_example.cpp File Reference

Example of using uni20 coroutines with the TbbScheduler to parallelize a map–reduce computation. More...

#include <iostream>
#include <numeric>
#include <uni20/async/async.hpp>
#include <uni20/async/buffers.hpp>
#include <uni20/async/tbb_scheduler.hpp>
#include <vector>

Include dependency graph for async_tbb_reduction_example.cpp:

Functions
int	main ()
AsyncTask	square (ReadBuffer< int > in, WriteBuffer< int > out)
AsyncTask	sum (ReadBuffer< int > a, ReadBuffer< int > b, WriteBuffer< int > out)

Detailed Description

Example of using uni20 coroutines with the TbbScheduler to parallelize a map–reduce computation.

This example computes the sum of squares from 1 to N using two coroutine kernels:

• square: squares a single integer
• sum: sums two integers

The computation is structured in two phases:

1. Map stage: a parallel launch of square coroutines over all inputs
2. Reduce stage: a binary-tree reduction using sum coroutines

The scheduler (TbbScheduler) manages coroutine resumption across threads participating in its arena, while Async<int>, ReadBuffer<int>, and WriteBuffer<int> handle dataflow and dependency tracking between tasks.

This demonstrates:

• Expressing parallel computations as coroutines
• Building a DAG of tasks with explicit dependencies
• Executing the DAG in parallel using TBB

Expected output (for N=1000):

Sum of squares 1..N = 333833500

In principle the application thread can reach the last line of main() before any of the scheduled computations begin.

Function Documentation

main()

int main

(

)

square()

AsyncTask square	(	ReadBuffer< int >	in,
		WriteBuffer< int >	out
	)

sum()

AsyncTask sum	(	ReadBuffer< int >	a,
		ReadBuffer< int >	b,
		WriteBuffer< int >	out
	)