# Help me Improve my poker hand evaluator

**URL:** <https://discourse.julialang.org/t/help-me-improve-my-poker-hand-evaluator/19360>\
**Category:** Performance\
**Tags:** optimization\
**Created:** [January 7, 2019, 6:27pm UTC](https://discourse.julialang.org/t/help-me-improve-my-poker-hand-evaluator/19360 "2019-01-07T18:27:34Z")\
**Posts on this page:** 1\
**Showing post:** 7

<div class="post-metadata">

**Author:** ![bennedich](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/bennedich/32/4894_2.png) [@bennedich](https://discourse.julialang.org/u/bennedich)\
**Post date:** [January 8, 2019, 11:15pm UTC](https://discourse.julialang.org/t/help-me-improve-my-poker-hand-evaluator/19360/7 "2019-01-08T23:15:12Z")

</div>

> [@ffevotte](#):
>
> are you saying that somewhere between 100 and 1000 hands, the total number of branches to remember (i.e. the total number of branches in all scores computations) gets too large for the branch predictor to keep in memory? At this point, we start seeing benchmarked times which are more accurate, because they are not artificially “helped” by the hot branch predictor?

Yes, exactly. At least on my system (Intel Skylake). You can read more about this in [the topic](https://discourse.julialang.org/t/psa-microbenchmarks-remember-branch-history/17436) I linked above. On AMD Ryzen for example, it appears that even more branches can be remembered.

---

_[View the full topic](https://discourse.julialang.org/t/help-me-improve-my-poker-hand-evaluator/19360)._