Stanford CS149 -- Assignment 2: Building A Task Execution Library from the Ground Up
作业描述及代码参见：CS149-asst2
Part A
Step 1
只需要实现一个简单的任务系统，在run()的开始生成工作线程，并在run()返回之前从主线程合并这些线程。任务的分配方式采用动态分配，即每个线程每次取一个任务完成，能者多劳。
每个线程的核心实现为：
while (true) {
int taskID = done++;
if (taskID >= num_total_tasks)
return;
runnable->runTask(taskID, num_total_tasks);
}
Step 2
Step 1 实现的任务系统，每次调用run()都会重新生成线程，线程管理开销大。为了减少线程创建开销，引入线程池，即：在任务系统初始化时创建所有线程，无任务时各线程执行忙等待，有任务时，每个线程每次取一个任务完成。
Step 3
Step 2 中，线程忙等待期间也会占用 CPU 核心的执行资源，这会影响性能。Step 3 中要使用环境变量让空闲线程进入睡眠态，直到有任务要执行，在实现时，需要使用两个条件变量：一个工作线程等待有任务到来，另一个用于主线程等待所有任务完成。
Part B
在 Part A 中实现的任务系统是同步的，每次调用run()都会等到所有任务完成才返回。Part B 要求实现异步的任务系统，这样在执行任务期间，主线程可以执行其他操作，此外，在同步任务系统中，调用run()后工作线程只能执行一种任务，在任务数较少或大多数任务完成时，会有一些工作线程处于空闲；异步任务系统，可以多次调用runAsyncWithDeps()添加多种任务，工作线程可以同时执行多种任务，因此可以提高性能。
实现异步任务系统需要考虑任务调度存在依赖的问题，某些任务必须在前面某些任务执行后才能开始执行。分析任务调度的过程，可以发现任务有四种状态：
- 未就绪：需要等待依赖任务完成
- 就绪：依赖的任务都已完成，可以随时被调度执行
- 分配完成：已将任务的
num_total_tasks项工作全部分给各线程 - 已完成：任务的
num_total_tasks项工作全部完成
对应任务的四种状态，定义四种数据结构：
waiting_list：存储未就绪的任务，由于需要在有任务完成后需要遍历所有未就绪的任务，将依赖满足的任务转为就绪态，因此涉及到遍历及频繁的数据添加、删除，可以使用链表std::list实现ready_queue：存储就绪的任务，使用队列std::queue实现即可，工作线程每次从队首的任务中取一项工作完成，直到对任务的工作全部分配给线程后，将队首任务移出，转为分配完成状态working_set：存储分配完成的任务，等到其所有工作完成后，转为已完成态。由于只涉及在将任务加入及在条件满足后将任务移出，可以使用集合std::unordered_set实现completed_set：存储已完成的任务，只用于判断任务是否完成，使用集合std::unordered_set实现即可， 也可以用一个大数组作为哈希表实现
这四种状态的转移条件为：
- 未就绪：为了简化实现，添加的任务无论有无依赖都先添加到
waiting_list中 - 就绪：在添加的任务后，以及有任务完成时，遍历
waiting_list，将依赖满足的任务移出并加入ready_queue - 分配完成：工作线程不断从
ready_queue取工作，队首任务的工作分配完成后将其出队，加入working_set - 已完成：每次工作线程完成一项工作后，都检查该工作所在的任务是否完成，若已完成，将其从
working_set移除，加入completed_set
在具体实现时，为了防止并发访问造成数据冲突，还要为这四种数据结构分别声明一个互斥量，保护访问。
基于runAsyncWithDeps和sync可以实现run：
void TaskSystemParallelThreadPoolSleeping::run(IRunnable* runnable, int num_total_tasks) {
std::vector<TaskID> noDeps;
runAsyncWithDeps(runnable, num_total_tasks, noDeps);
sync();
}
结果
异步任务系统的性能测试结果如下：
================================================================================
Running task system grading harness... (22 total tests)
- Detected CPU with 8 execution contexts
- Task system configured to use at most 8 threads
================================================================================
================================================================================
Executing test: super_super_light...
Reference binary: ./runtasks_ref_linux
Results for: super_super_light
STUDENT REFERENCE PERF?
[Serial] 8.519 12.275 0.69 (OK)
[Parallel + Thread Pool + Sleep] 29.887 46.432 0.64 (OK)
================================================================================
Executing test: super_super_light_async...
Reference binary: ./runtasks_ref_linux
Results for: super_super_light_async
STUDENT REFERENCE PERF?
[Serial] 8.413 12.287 0.68 (OK)
[Parallel + Thread Pool + Sleep] 15.982 24.708 0.65 (OK)
================================================================================
Executing test: super_light...
Reference binary: ./runtasks_ref_linux
Results for: super_light
STUDENT REFERENCE PERF?
[Serial] 79.056 90.503 0.87 (OK)
[Parallel + Thread Pool + Sleep] 60.706 59.286 1.02 (OK)
================================================================================
Executing test: super_light_async...
Reference binary: ./runtasks_ref_linux
Results for: super_light_async
STUDENT REFERENCE PERF?
[Serial] 78.752 93.645 0.84 (OK)
[Parallel + Thread Pool + Sleep] 46.706 48.718 0.96 (OK)
================================================================================
Executing test: ping_pong_equal...
Reference binary: ./runtasks_ref_linux
Results for: ping_pong_equal
STUDENT REFERENCE PERF?
[Serial] 1324.494 1506.428 0.88 (OK)
[Parallel + Thread Pool + Sleep] 398.311 441.353 0.90 (OK)
================================================================================
Executing test: ping_pong_equal_async...
Reference binary: ./runtasks_ref_linux
Results for: ping_pong_equal_async
STUDENT REFERENCE PERF?
[Serial] 1550.31 1726.468 0.90 (OK)
[Parallel + Thread Pool + Sleep] 536.815 615.742 0.87 (OK)
================================================================================
Executing test: ping_pong_unequal...
Reference binary: ./runtasks_ref_linux
Results for: ping_pong_unequal
STUDENT REFERENCE PERF?
[Serial] 2680.887 2732.8 0.98 (OK)
[Parallel + Thread Pool + Sleep] 886.001 910.192 0.97 (OK)
================================================================================
Executing test: ping_pong_unequal_async...
Reference binary: ./runtasks_ref_linux
Results for: ping_pong_unequal_async
STUDENT REFERENCE PERF?
[Serial] 2687.128 2764.332 0.97 (OK)
[Parallel + Thread Pool + Sleep] 875.282 893.719 0.98 (OK)
================================================================================
Executing test: recursive_fibonacci...
Reference binary: ./runtasks_ref_linux
Results for: recursive_fibonacci
STUDENT REFERENCE PERF?
[Serial] 1726.216 2264.539 0.76 (OK)
[Parallel + Thread Pool + Sleep] 583.149 775.575 0.75 (OK)
================================================================================
Executing test: recursive_fibonacci_async...
Reference binary: ./runtasks_ref_linux
Results for: recursive_fibonacci_async
STUDENT REFERENCE PERF?
[Serial] 1717.786 2190.635 0.78 (OK)
[Parallel + Thread Pool + Sleep] 582.04 765.999 0.76 (OK)
================================================================================
Executing test: math_operations_in_tight_for_loop...
Reference binary: ./runtasks_ref_linux
Results for: math_operations_in_tight_for_loop
STUDENT REFERENCE PERF?
[Serial] 1042.031 1008.215 1.03 (OK)
[Parallel + Thread Pool + Sleep] 639.439 719.384 0.89 (OK)
================================================================================
Executing test: math_operations_in_tight_for_loop_async...
Reference binary: ./runtasks_ref_linux
Results for: math_operations_in_tight_for_loop_async
STUDENT REFERENCE PERF?
[Serial] 945.107 989.953 0.95 (OK)
[Parallel + Thread Pool + Sleep] 585.924 602.104 0.97 (OK)
================================================================================
Executing test: math_operations_in_tight_for_loop_fewer_tasks...
Reference binary: ./runtasks_ref_linux
Results for: math_operations_in_tight_for_loop_fewer_tasks
STUDENT REFERENCE PERF?
[Serial] 949.832 993.658 0.96 (OK)
[Parallel + Thread Pool + Sleep] 600.664 673.685 0.89 (OK)
================================================================================
Executing test: math_operations_in_tight_for_loop_fewer_tasks_async...
Reference binary: ./runtasks_ref_linux
Results for: math_operations_in_tight_for_loop_fewer_tasks_async
STUDENT REFERENCE PERF?
[Serial] 973.477 988.745 0.98 (OK)
[Parallel + Thread Pool + Sleep] 299.371 368.859 0.81 (OK)
================================================================================
Executing test: math_operations_in_tight_for_loop_fan_in...
Reference binary: ./runtasks_ref_linux
Results for: math_operations_in_tight_for_loop_fan_in
STUDENT REFERENCE PERF?
[Serial] 513.92 532.803 0.96 (OK)
[Parallel + Thread Pool + Sleep] 194.392 258.495 0.75 (OK)
================================================================================
Executing test: math_operations_in_tight_for_loop_fan_in_async...
Reference binary: ./runtasks_ref_linux
Results for: math_operations_in_tight_for_loop_fan_in_async
STUDENT REFERENCE PERF?
[Serial] 484.185 497.827 0.97 (OK)
[Parallel + Thread Pool + Sleep] 140.66 184.051 0.76 (OK)
================================================================================
Executing test: math_operations_in_tight_for_loop_reduction_tree...
Reference binary: ./runtasks_ref_linux
Results for: math_operations_in_tight_for_loop_reduction_tree
STUDENT REFERENCE PERF?
[Serial] 524.883 524.995 1.00 (OK)
[Parallel + Thread Pool + Sleep] 153.145 201.542 0.76 (OK)
================================================================================
Executing test: math_operations_in_tight_for_loop_reduction_tree_async...
Reference binary: ./runtasks_ref_linux
Results for: math_operations_in_tight_for_loop_reduction_tree_async
STUDENT REFERENCE PERF?
[Serial] 544.436 542.025 1.00 (OK)
[Parallel + Thread Pool + Sleep] 148.916 212.23 0.70 (OK)
================================================================================
Executing test: spin_between_run_calls...
Reference binary: ./runtasks_ref_linux
Results for: spin_between_run_calls
STUDENT REFERENCE PERF?
[Serial] 617.51 775.054 0.80 (OK)
[Parallel + Thread Pool + Sleep] 339.995 484.251 0.70 (OK)
================================================================================
Executing test: spin_between_run_calls_async...
Reference binary: ./runtasks_ref_linux
Results for: spin_between_run_calls_async
STUDENT REFERENCE PERF?
[Serial] 599.196 761.259 0.79 (OK)
[Parallel + Thread Pool + Sleep] 331.343 453.562 0.73 (OK)
================================================================================
Executing test: mandelbrot_chunked...
Reference binary: ./runtasks_ref_linux
Results for: mandelbrot_chunked
STUDENT REFERENCE PERF?
[Serial] 563.182 577.487 0.98 (OK)
[Parallel + Thread Pool + Sleep] 89.698 93.976 0.95 (OK)
================================================================================
Executing test: mandelbrot_chunked_async...
Reference binary: ./runtasks_ref_linux
Results for: mandelbrot_chunked_async
STUDENT REFERENCE PERF?
[Serial] 579.105 565.807 1.02 (OK)
[Parallel + Thread Pool + Sleep] 93.104 94.314 0.99 (OK)
================================================================================
Overall performance results
[Serial] : All passed Perf
[Parallel + Thread Pool + Sleep] : All passed Perf
