C# Parallel用法

1、parallel.invoke 主要用于任务的并行
　　这个函数的功能和task有些相似，就是并发执行一系列任务，然后等待所有完成。和task比起来，省略了task.waitall这一步，自然也缺少了task的相关管理功能。它有两种形式:
　　parallel.invoke( params action[] actions);
　　parallel.invoke(action[] actions,taskmanager manager,taskcreationoptions options);

using system;
using system.threading;
using system.threading.tasks;

namespace consoleapp1
{
    class program
    {
        static void main(string[] args)
        {
            var actions = new action[]
            {
                () => actiontest("test 1"),
                () => actiontest("test 2"),
                () => actiontest("test 3"),
                () => actiontest("test 4")
            };

            console.writeline("parallel.invoke 1 test");
            parallel.invoke(actions);

            console.writeline("结束！");
        }

        static void actiontest(object value)
        {
            console.writeline(">>> thread:{0}, value:{1}",
            thread.currentthread.managedthreadid, value);
        }
    }
}

program

2、for方法，主要用于处理针对数组元素的并行操作(数据的并行) 

using system;
using system.threading;
using system.threading.tasks;

namespace consoleapp1
{
    class program
    {
        static void main(string[] args)
        {
            int[] nums = new int[] { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 };
            parallel.for(0, nums.length, (i) =>
            {
                console.writeline("针对数组索引{0}对应的那个元素{1}的一些工作代码……threadid={2}", i, nums[i], thread.currentthread.managedthreadid);
            });
            console.readkey();
        }
    }
}

program

3、foreach方法，主要用于处理泛型集合元素的并行操作(数据的并行)

using system;
using system.collections.generic;
using system.threading;
using system.threading.tasks;

namespace consoleapp1
{
    class program
    {
        static void main(string[] args)
        {
            list<int> nums = new list<int> { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 };
            parallel.foreach(nums, (item) =>
            {
                console.writeline("针对集合元素{0}的一些工作代码……threadid={1}", item, thread.currentthread.managedthreadid);
            });
            console.readkey();
        }
    }
}

program

　　数据的并行的方式二(asparallel()):

using system;
using system.collections.generic;
using system.linq;
using system.threading;

namespace consoleapp1
{
    class program
    {
        static void main(string[] args)
        {
            list<int> nums = new list<int> { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 };
            var evennumbers = nums.asparallel().select(item => calculate(item));
            //注意这里是个延迟加载,也就是不用集合的时候 这个calculate里面的算法 是不会去运行 可以屏蔽下面的代码看效果;
            console.writeline(evennumbers.count());
            //foreach (int item in evennumbers)
            //    console.writeline(item);
            console.readkey();
        }

        static int calculate(int number)
        {
            console.writeline("针对集合元素{0}的一些工作代码……threadid={1}", number, thread.currentthread.managedthreadid);
            return number * 2;
        }
    }
}

program

　　.asordered() 对结果进行排序:

using system;
using system.collections.generic;
using system.linq;
using system.threading;
using system.threading.tasks;

namespace consoleapp
{

    class program
    {
        static void main(string[] args)
        {
            list<int> nums = new list<int> { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 };
            var evennumbers = nums.asparallel().asordered().select(item => calculate(item));
            //注意这里是个延迟加载,也就是不用集合的时候 这个calculate里面的算法 是不会去运行 可以屏蔽下面的代码看效果;
            //console.writeline(evennumbers.count());
            foreach (int item in evennumbers)
                console.writeline(item);
            console.readkey();
        }

        static int calculate(int number)
        {
            console.writeline("针对集合元素{0}的一些工作代码……threadid={1}", number, thread.currentthread.managedthreadid);
            return number * 2;
        }
    }
}

program

　　foreach的独到之处就是可以将数据进行分区，每一个小区内实现串行计算，分区采用partitioner.create实现。

using system;
using system.collections.concurrent;
using system.diagnostics;
using system.threading.tasks;

namespace consoleapp1
{
    class program
    {
        static void main(string[] args)
        {
            for (int j = 1; j < 4; j++)
            {
                concurrentbag<int>  bag = new concurrentbag<int>();
                var watch = stopwatch.startnew();
                watch.start();
                parallel.foreach(partitioner.create(0, 3000000), i =>
                {
                    for (int m = i.item1; m < i.item2; m++)
                    {
                        bag.add(m);
                    }
                });
                console.writeline("并行计算：集合有:{0},总共耗时：{1}", bag.count, watch.elapsedmilliseconds);
                gc.collect();

            }
        }
    }
}

program

　　paralleloptions类
　　paralleloptions options = new paralleloptions();
　　//指定使用的硬件线程数为4
　　options.maxdegreeofparallelism = 4;
　　有时候我们的线程可能会跑遍所有的内核，为了提高其他应用程序的稳定性，就要限制参与的内核，正好paralleloptions提供了maxdegreeofparallelism属性。

using system;
using system.collections.concurrent;
using system.diagnostics;
using system.linq;
using system.threading.tasks;

namespace consoleapp1
{
    public class student
    {
        public int id { get; set; }
        public string name { get; set; }
        public int age { get; set; }
        public datetime createtime { get; set; }
    }

    class program
    {
        static void main(string[] args)
        {
            var dic = loaddata();
            stopwatch watch = new stopwatch();
            watch.start();
            var query2 = (from n in dic.values.asparallel()
                          where n.age > 20 && n.age < 25
                          select n).tolist();
            watch.stop();
            console.writeline("并行计算耗费时间：{0}", watch.elapsedmilliseconds);

            console.read();
        }

        public static concurrentdictionary<int, student> loaddata()
        {
            concurrentdictionary<int, student> dic = new concurrentdictionary<int, student>();
            paralleloptions options = new paralleloptions();
            //指定使用的硬件线程数为4
            options.maxdegreeofparallelism = 4;
            //预加载1500w条记录
            parallel.for(0, 15000000, options, (i) =>
            {
                var single = new student()
                {
                    id = i,
                    name = "hxc" + i,
                    age = i % 151,
                    createtime = datetime.now.addseconds(i)
                };
                dic.tryadd(i, single);
            });

            return dic;
        }
    }
}

program

常见问题的处理

　　<1> 如何中途退出并行循环？
　　是的，在串行代码中我们break一下就搞定了，但是并行就不是这么简单了，不过没关系，在并行循环的委托参数中提供了一个parallelloopstate，该实例提供了break和stop方法来帮我们实现。
　　break: 当然这个是通知并行计算尽快的退出循环，比如并行计算正在迭代100，那么break后程序还会迭代所有小于100的。
　　stop：这个就不一样了，比如正在迭代100突然遇到stop，那它啥也不管了，直接退出。

using system;
using system.collections.concurrent;
using system.threading.tasks;

namespace consoleapp1
{
    class program
    {
        static void main(string[] args)
        {
            concurrentbag<int> bag = new concurrentbag<int>();

            parallel.for(0, 20000000, (i, state) =>
            {
                if (bag.count == 1000)
                {
                    //state.break();
                    state.stop();
                    return;
                }
                bag.add(i);
            });

            console.writeline("当前集合有{0}个元素。", bag.count);

        }
    }
}

program

　　取消(cancel)

using system;
using system.threading;
using system.threading.tasks;

namespace consoleapp1
{
    class program
    {
        public static void main()
        {

            var cts = new cancellationtokensource();
            var ct = cts.token;
            task.factory.startnew(() => fun(ct));
            console.readkey();
            //thread.sleep(3000);
            cts.cancel();
            console.writeline("任务取消了！");

        }

        static void fun(cancellationtoken token)
        {
            parallel.for(0, 100000,
                        new paralleloptions { cancellationtoken = token },
                        (i) =>
                        {
                            console.writeline("针对数组索引{0}的一些工作代码……threadid={1}", i, thread.currentthread.managedthreadid);
                        });
        }
    }
}

program

　　<2> 并行计算中抛出异常怎么处理？
　　首先任务是并行计算的，处理过程中可能会产生n多的异常，那么如何来获取到这些异常呢？普通的exception并不能获取到异常，然而为并行诞生的aggregateexcepation就可以获取到一组异常。

using system;
using system.threading;
using system.threading.tasks;

namespace consoleapp1
{
    class program
    {
        static void main(string[] args)
        {
            try
            {
                parallel.invoke(run1, run2);
            }
            catch (aggregateexception ex)
            {
                foreach (var single in ex.innerexceptions)
                {
                    console.writeline(single.message);
                }
            }
            console.writeline("结束了！");
            //console.read();
        }

        static void run1()
        {
            thread.sleep(3000);
            throw new exception("我是任务1抛出的异常");
        }

        static void run2()
        {
            thread.sleep(5000);
            throw new exception("我是任务2抛出的异常");
        }
    }
}

program

　　注意parallel里面 不建议抛出异常 因为在极端的情况下比如进去的第一批线程先都抛异常了 此时aggregateexcepation就只能捕获到这一批的错误,然后程序就结束了

using system;
using system.collections.generic;
using system.threading.tasks;

namespace consoleapp1
{
    public class testclass
    {
        public static list<int> numberlist = null;
        private static readonly object locker = new object();
        public void test(int number)
        {
            throw new exception("1111");
            //lock (locker)
            //{
            //    if (numberlist == null)
            //    {
            //        console.writeline("执行添加");
            //        numberlist = new list<int>();
            //        numberlist.add(1);
            //        //thread.sleep(1000);
            //    }
            //}
            //if (number == 5 || number == 7) throw new exception(string.format("number{0}boom!", number));
            //console.writeline(number);
        }
    }

    class program
    {
        private static readonly object locker = new object();
        static void main(string[] args)
        {
            list<string> errlist = new list<string>();
            try
            {
                parallel.for(0, 10, (i) =>
                {
                    try
                    {
                        testclass a = new testclass();
                        a.test(i);
                    }
                    catch (exception ex)
                    {
                        lock (locker)
                        {
                            errlist.add(ex.message);
                            throw ex;
                        }
                    }
                });
            }
            catch (aggregateexception ex)
            {
                foreach (var single in ex.innerexceptions)
                {
                    console.writeline(single.message);
                }
            }
            int index = 1;
            foreach (string err in errlist)
            {
                console.writeline("{0}、的错误:{1}", index++, err);
            }
        }
    }
}

program

　　可以向下面这样来处理一下
　　不在aggregateexcepation中来处理 而是在parallel里面的try catch来记录错误,或处理错误

using system;
using system.collections.generic;
using system.threading.tasks;

namespace consoleapp1
{
    public class testclass
    {
        public static list<int> numberlist = null;
        private static readonly object locker = new object();
        public void test(int number)
        {
            throw new exception("1111");
            //lock (locker)
            //{
            //    if (numberlist == null)
            //    {
            //        console.writeline("执行添加");
            //        numberlist = new list<int>();
            //        numberlist.add(1);
            //        //thread.sleep(1000);
            //    }
            //}
            //if (number == 5 || number == 7) throw new exception(string.format("number{0}boom!", number));
            //console.writeline(number);
        }
    }

    class program
    {
        private static readonly object locker = new object();
        static void main(string[] args)
        {
            list<string> errlist = new list<string>();
            parallel.for(0, 10, (i) =>
            {
                try
                {
                    testclass a = new testclass();
                    a.test(i);
                }
                catch (exception ex)
                {
                    lock (locker)
                    {
                        errlist.add(ex.message);
                    }
                    //console.writeline(ex.message);
                    //注:这里不再将错误抛出.....
                    //throw ex;
                }
            });

            int index = 1;
            foreach (string err in errlist)
            {
                console.writeline("{0}、的错误:{1}", index++, err);
            }
        }
    }
}

program