March 16, 2009
Use Threads Correctly = Isolation + Asynchronous MessagesExample 3: Pipeline Stages
A third classic example of independent work that should run asynchronously is pipelining, and the independent pieces of work are the pipeline stages. In a nutshell, here's the idea: We have a series of pieces of data to be processed in order (e.g., series of packets to prepare for sending). Each piece has to go through several stages of processing (e.g., decorate, compress, encrypt) that need to be applied to a given piece of data in order.
The stages are otherwise independent (e.g., the compressor can run independently of the decorator and encryptor), other than waiting for available work to arrive from the previous stage. As shown in Figure 4, we can express the pipeline with one thread per stage, connected by asynchronous message queues to let the stages run independently and tolerate latency as the different stages typically run at different speeds. Each stage just takes each incoming packet from the previous stage, does its own processing, and throws it over to the next stage.
Figure 4: Using three threads to express three pipeline stages.
SummaryThreads are a low-level tool for expressing asynchronous work. "Uplevel" them by applying discipline: strive to make their data private, and have them communicate and synchronize using asynchronous messages. Each thread that needs to get information from other threads or from people should have a message queue, whether a simple FIFO queue or a priority queue, and organize its work around an event-driven message pump mainline; replacing spaghetti with event-driven logic is a great way to improve the clarity and determinism of your code.
Notes
[1] H. Sutter. "The Pillars of Concurrency" (Dr. Dobb's Journal, August 2007). http://www.ddj.com/architect/200001985. [2] E. Lee. "The Problem with Threads" (EECS Department Technical Report, University of California, Berkeley, 2006). http://www.eecs.berkeley.edu/Pubs/TechRpts/2006/EECS-2006-1.html. [3] The Cilk Project website: http://supertech.csail.mit.edu/cilk/. [4] Intel Threading Building Blocks website: http://www.threadingbuildingblocks.org/. [5] K. Kerr. "Visual C++ 2010 and the Parallel Patterns Library (MSDN Magazine, February 2009). http:// msdn.microsoft.com/en-us/magazine/dd434652.aspx. [6] D. Leijen and J. Hall. "Optimize Managed Code For Multi-Core Machines" (MSDN Magazine, October 2007). http://msdn.microsoft.com/en-us/magazine/cc163340.aspx. [7] J. Duffy and E. Essey. "Running Queries On Multi-Core Processors" (MSDN Magazine, October 2007). http://msdn.microsoft.com/en-us/magazine/cc163329.aspx. [8] D. Lea. "Package jsr166y" (Draft JSR166y documentation). http://gee.cs.oswego.edu/dl/jsr166/dist/jsr166ydocs/. [9] OpenMP Application Program Interface Version 3.0 (OpenMP.org, May 2008). http://www.openmp.org/mp-documents/spec30.pdf. 
Previous Page |
1 Threads: In a Nutshell
|
2 Example 1: GUI
|
3 Example 2: Sockets
|
4 Example 3: Pipeline Stages
|
 |
|
|||||||||||||||||||||||||||||
|
|
|
