并發(fā)算法與理論P(yáng)PT課件

1、并發(fā)算法與理論Concurrency: Algorithms and Theories梁紅瑾Programmers View2C+ Programmers View#include #include void hello() std:coutHello Concurrent Worldn;int main() std:thread t(hello); t.join(); return 0;3C+ Programmers ViewOutputs4Taking a More Abstract View Parallel compositionC1 | C25for(int i=0; i 10; i

2、+) cout“main concurrentn;for(int i=0; i 10; i+) cout“hello concurrentn;Taking a More Abstract View Shared memory & interleaving semantics6. . .MemoryT1T2T3TnTaking a More Abstract View Shared memory & interleaving semantics7for(int i=0; i 10; i+) cout“main concurrentn;for(int i=0; i 10; i+)

3、cout“hello concurrentn;Memory (e.g. cout)T1T2Java Threads8Java Threads9public class MyRunnable implements Runnable Override public void run() for(int i = 0; i 10; i+) System.out.println(i); public class MyRunnableTest public static void main(String args) MyRunnable mr1 = new MyRunnable(); MyRunnable

4、 mr2 = new MyRunnable(); Thread t1 = new Thread(mr1); Thread t2 = new Thread(mr2); t1.start(); t2.start(); Python Threads10Python Threads11Problems with Concurrency Nondeterministic! Recall: interleaving semantics12MemoryT1T2T3TnProblems with Concurrency Nondeterministic! Recall: interleaving semant

5、ics13Possible executionsProgramT1T2|= Difficult to find a bug Difficult to reproduce a bugMore C+14#include #include void thread_function() for (int i = -100; i 0; i+) std:cout thread function: i n;int main()std:thread t(thread_function);for (int i = 0; i 100; i+) std:cout main thread: i n;t.join();

6、return 0;thread function: main thread: 0main thread: 1main thread: 2main thread: 3-100thread function: -99main thread: 4main thread: 5main thread: 6Using Locks 15#include #include #include #include std:mutex mu;void shared_cout(std:string msg, int id)mu.lock();std:cout msg : id std:endl;mu.unlock();

7、Using Locks 16#include #include #include #include std:mutex mu;void shared_cout(std:string msg, int id)mu.lock();std:cout msg : id std:endl;mu.unlock();void thread_function()for (int i = -100; i 0; i-) shared_cout(main thread, i);t.join();return 0;More Locks in C+17#include #include #include #includ

8、e std:mutex mu;void shared_cout(std:string msg, int id)std:lock_guard guard(mu);std:cout msg : id std:endl;More Locks in C+18#include #include #include std:list some_list;std:mutex some_mutex;void add_to_list(int new_value) std:lock_guard guard(some_mutex); some_list.push_back(new_value);bool list_c

9、ontains(int value_to_find) std:lock_guard guard(some_mutex); return std:find(some_list.begin(),some_list.end(),value_to_find) != some_list.end();Taking a More Abstract View Synchronization operations lock/unlock, acq/rel19for (int i=-100; i0; i+) mu.lock();std:cout thread function: i 0; i-) mu.lock(

10、);std:cout main thread: i n; mu.unlock();20But waitHow are lock() and unlock() implemented?Are there interleavings between the threads when executing the implementations of lock()/unlock()?for (int i=-100; i0; i+) mu.lock();std:cout thread function: i 0; i-) mu.lock();std:cout main thread: i n; mu.u

11、nlock();Programmers View A concurrent program = concurrent objects + their clients21mu.lock();mu.unlock();mu.lock();mu.unlock();void lock() void unlock() Client codeConcurrent objectT1T2T3lock() Programmers View A concurrent program = concurrent objects + their clients22s.push(7);x = s.pop();s.push(

12、6);void push(int v) int pop() T1T2T3push() Concurrent objectClient codeConcurrent Objects23memoryobjectobjectthreadsProgrammers View Parallel composition, shared memory & interleaving semantics Locks & synchronization operations Concurrent objects and their clients24Memory Models25https:/en.

13、/wiki/Memory_model_(programming) Memory Models Interleaving semantics Sequential Consistency (SC) model 26MemoryT1T2T3Tn(Weak/Relaxed) Memory Models Java memory model C/C+11 memory model x86-TSO memory model ARMv8 memory model 2728Will the assertion fail?#include #include int x = 0;void

14、 foo() while (true) x = 1; x = 0; assert(x = 0); int main() std:thread t(foo); std:thread t2(foo); t.join(); t2.join();C+ Program 129Will the assertion fail?#include #include int x = 0;void foo() for ( int i = 0; i 10000000; i+ ) x = x + 1; int main() std:thread t1(foo); std:thread t2(foo); t1.join(

15、); t2.join(); assert(x = 20000000);C+ Program 2Course OverviewGoals of the Course Understanding concurrent programs Interleaving semantics Some feelings of weak memory models Understanding concurrent objects (algorithms) Correctness criteria Design ideas Why they are correct Some impossibility resul

16、ts31Preliminary Syllabus Weak memory models Java, C11 Locks Implemented by Read/Write Peterson Lock, Filter Lock, Lamports Bakery Lock Linearizability Synchronization Operations Practical Locks TAS Lock, Queue Locks List-Based Sets Concurrent Queues and Stacks Snapshot32References Maurice Herlihy and Nir Shavit. The Art of Multiprocessor Programming. Morgan Kaufmann. 2008 / 2012.33Lecture notes, homework and notifications will be posted on the course webpage.You can ask questions in QQ group, or email me your questions.Please pay attention t