Experiences/anecdotes for OTP 24 JIT performance?

BEAM Forum BEAM Chat / Discussions
1

Have you deployed OTP 24 apps and can comment on JIT performance?
I’ve read the original blog posts and test reports back when OTP 24 was out. Performance testing example

But I haven’t read much since. My question is purely out of curiosity :slight_smile:

7 Likes
2

We’ve switched to JIT.

Cannot tell that we’ve got 2x speedup.

4 Likes
3

We’ve got an average of 25% - good enough to be noticed by Will:

Actual boost varies. Compute-heavy services can get up to 50-60%, storage gets almost nothing.

8 Likes
4

Thanks for your example…25% is certainly “good enough to be noticed” :slight_smile:

And good to know CPU-bound jobs can benefit up to 60%.

4 Likes
5

Basically it means: everything is getting by a noticeable amount faster and the tasks that can be usefully tackled by Erlang expand (before people avoided compute heavy things before or moved them to NIFs).

Less need for NIFs now!

7 Likes