多个异步协程以减少 Kotlin 中的执行时间

Question

I had a list and I want to perform 1 heavy task with all elements to minimum executed times. My solution is create a number of coroutines that equals to elements in list with dispatcher Default.

Example:

list.map {
   async {
      // do heavy task
   }
}.awaitAll()

Assume that I had a map with 500 elements and my device has maximum 4 threads that can run parallel maximum 4 heavy tasks. As my understanding that will 4 threads do 4 tasks parallel, when one thread executed done, one task will be allocated to that thread and start. The other solution is create 4 coroutines for 4 threads and sequentially do 125 task per thread. Each threads run sequentially therefore it confined multiple time to allocate task to threads. Therefore it faster than my solution. Is it right? If not, what is the best solution?

Answer 1

I use the following code.

It's for a List of a lot of elements that I need to transform into another object class.

/**
     * Method that converts the [FullStopsInfo] list into [Station] list.
     */
    private fun fastConvert(data: MutableList<FullStopsInfo>): Vector<Station> {
        var data1 = data
        val stations = Vector<Station>()

        var offset = 0
        val range = data1.size / Utils.availableCores()// your 4 cores

        val queue = mutableListOf<Deferred<Int>?>()

        val timeProccessing = measureTimeMillis {
            while (data1.isNotEmpty()) {
                val newRange = min(if(range == 0) 1 else range, data1.size)
                val tempTransfers = data1.subList(0, newRange)

                val preferredTask = addToList(tempTransfers, stations)//the transformation goes here
                queue.add(preferredTask)
                preferredTask.start()

                // Move to the next range
                data1 = data1.subList(newRange, data1.size)
                offset += newRange
            }

            runBlocking {
                queue.forEach {
                    it?.await()//wait until all procceses ended
                }
            }
        }

        Log.d(TAG, "End findSearchAllFullDataStation with $timeProccessing")
        return stations
    }