Udp receives the data, to carry on the processing, computational cost is larger, how to use multi-core processing? Now I use the threadpool processing can only use a CPU cores

CodePudding user response:

C #, with TPL, namely the tasks below Parallel invocation, such as the Parallel. The For

CodePudding user response:

How do you know a CPU cores were used?

Multi-threading does not necessarily fast, if you are a dual-core, best four threads

CodePudding user response:

reference 1st floor caozhy response:

, c # using the TPL, namely the tasks below Parallel invocation, such as the Parallel. For

Business process play simple say like that, I receive packets around 4000 packages per second, each packet json processing, take the time information, then send the processing, I use the parallel for and foreach all don't know how to write

CodePudding user response:

refer to the second floor of autumn red fruit response:

how do you know a CPU cores were used?

Multi-threading does not necessarily fast, if you are a dual-core, best to open four threads

Computer CPU for six nuclear data backlog, observation is only about 16% of the CPU performance

CodePudding user response:

Will receive the package cache to safety in the queue, for 12 threads (quantity will be subject to test is the best), from the security queue data concurrency

CodePudding user response:

Currently USES is concurrentqueue, use threadpool processing queue data, limitation lies in the processing of large amount of calculation, the problem is how to call more CPU to calculate

CodePudding user response:

Personally, CPU scheduling is the business of operating system
CPU utilization is low, there may be a bottleneck problem, such as memory, when sending the network speed, etc.
You can use the similar loadrunner tools for testing

CodePudding user response:

See # 1,