Lack of parallelism when using the `Flow` API without partition about gen_stage

[josevalim]

Stages exchange data in batches. For such a small Enumerable, the whole data will be sent in a single batch. Partitioning changes it because the batches are spread based on the integer (although it still batches).

You will see parallelism for large enumerables or if you start a Flow with Flow.new and set :max_demand to 1.

[josevalim]

It is also worth noticing that Flow works better for CPU bound parallelism. After all, we start the same amount of partitions as the number of cores. If you are going to be IO bound, then you can increase the default number of stages via Flow.new as well.

[josevalim]

There is a section called "Performance discussions" in the docs that should answer those questions. :)

[rafbarr]

Ok. I'm officially stupid. I should have read more of the documentation before opening the issue. Sorry about that and thanks for the quick reply.

[josevalim]

Ok. I'm officially stupid.

Not at all. :D GenStage.Flow is quite new so asking questions is part of the process!