Thread by @xpvit

#Java performance Tip: `parallelStream()` is way slower than regular `stream()` on small amount of data. I have created a simple benchmark and run it on my 8-CPU laptop. The difference is 160x for some cases. 😲

Details 🧵 👇

👷The setup is to run the same code for streams of 10, 1000, 10_000, 100_000, and 1000_000 items long in standard and parallel modes:
list.stream()
.filter(i -> i == -1)
.findAny()
.orElse(0);

10 items stream:
✅stream()
❌parallelStream() 161x SLOWER ‼️

1000 items stream:
✅stream()
❌parallelStream() 7x SLOWER

10_000 items stream:
✅stream()
❌parallelStream() 1.5x SLOWER

100_000 items stream (winner change):
❌stream() 1.5x SLOWER
✅parallelStream()

1000_000 items stream (winner changed):
❌stream() 2.4x SLOWER
✅parallelStream()

Are you wondering why it is so?
When running a parallel stream, the code starts multiple threads and splits data across them. Starting threads is not a very cheap operation; therefore, on smaller streams, parallel processing is slower than gains from multi-threading.

Note, that stream sizes where classical/parallel performs better should not be taken as is to your code. These numbers depend a lot on:
- CPU
- complexity of operations your stream performs

As always code could be found in my GitHub:

github.com/xp-vit/java-lo…

That's a wrap!

If you enjoyed this thread:

1. Follow me @xpvit for more of these
2. RT the tweet below to share this thread with your audience

https://twitter.com/421103460/status/1621071984326021120