请教：【微流水线并行机制】是如何执行的

想请教下，介绍中的【微流水线并行机制】是指GraphProcessor间并行，还是在GraphProcessor内并行。通常有这样的场景：当服务收到的数据集比较大时，一般会分batch并行处理，一种是分多个GraphProcessor并行，但目前看example中，在构图阶段，GraphProcessor的数量是固定的，是否可以在运行阶段调整GraphProcessor数量？ 还一种方式在GraphProcessor内并行，研读代码时貌似没有这种支持。

Start with the introduction's description of the micro-pipeline parallel mechanism and the example's graph-construction code. Trace how GraphProcessor instances are created and how data batches are handled, then document whether parallelism is between processors or within one processor and whether the processor count can change at runtime.

Written by the indexing model from the issue text.