Threading: Mechanisms for multi-threaded communication, which can refer to Go's chan or Node.js's postMessage.

[winlinvip]

Multi-threaded communication is a series of mechanisms:

1. First is the efficient communication queue.
2. Then there is the mechanism for communication between threads.

Mutex vs CAS Int

We only consider the comparison between locks and CAS, that is, the case of int increment.

• trunk/research/lockless/mutex-int.cpp: Use Mutex, int increment.
• trunk/research/lockless/lockless-int.cpp: Use CAS, int increment.

First, look at single-threaded, so that waiting and conflict can be eliminated, data is as follows:

Number of threads	Synchronization	Number of loops	Actual number of loops	Time consumed
Single thread	None	300M	300M	646ms
Single thread	Mutex	300M	300M	5373ms
Single thread	CAS	300M	300M	1911ms

Then, it's double-threaded, at this time Mutex or CAS must be used, no synchronization will cause data abnormalities and can only be used as a reference, data is as follows:

Number of threads	Synchronization	Number of loops	Actual number of loops	Time consumed
2 threads	None	60M	65M	468ms
2 threads	Mutex	60M	120M	5009ms
2 threads	CAS	60M	120M	3295ms

Note: The actual number of loops should be 120M, because there are 2 threads, each with 60M.

Next, test 3 threads, data is as follows:

Number of threads	Synchronization	Number of loops	Actual number of loops	Time consumed
3 threads	None	50M	60M	677ms
3 threads	Mutex	50M	150M	7909ms
3 threads	CAS	50M	150M	4317ms

Mutex vs CAS Queue

Consider the circular queue.

• trunk/research/lockless/mutex-queue.cpp: Use Mutex.
• trunk/research/lockless/lockless-queue.cpp: Use CAS.

Data is as follows:

Write thread count	Read thread count	Synchronization	Number of loops	Time consumed
1 thread	1 thread	Mutex	30M	2872ms
1 thread	1 thread	CAS	30M	3254ms
2 threads	2 threads	Mutex	30M	6233ms
2 threads	2 threads	CAS	30M	10387ms
3 threads	3 threads	Mutex