Intel Threading Building Blocks

Throughput of pipeline

The throughput of a pipeline is the rate at which tokens flow through it, and it is limited by two constraints. First, if a pipeline is run with n tokens, there obviously cannot be more than n operations running in parallel. Selecting the right value of n may involve some experimentation. Too low a value limits parallelism; too high a value may demand too many resources (for example, more buffers).

Second, the throughput of a pipeline is limited by the throughput of the slowest sequential stage. This is true even for a pipeline with no parallel stages. No matter how fast the other stages are, the slowest sequential stage is the bottleneck. So in general, you should try to keep the sequential stages fast and, when possible, shift work to the parallel stages.

The text-processing example has relatively poor speedup because the serial stages are limited by the I/O speed of the system. Indeed, even when files are on a local disk, you are unlikely to see a speedup of much more than 2X. To really benefit from a pipeline, the parallel stages need to be doing more substantial work compared to the serial stages.

The window size, or subproblem size for each token, can also limit throughput. Making windows too small may cause overheads to dominate the useful work. Making windows too large may cause them to spill out of cache. A good guideline is to try for a large window size that still fits in cache. You may have to experiment a bit to find a good window size.