Replies: 3 comments
-
I found it interesting that my solution runs almost as fast with 512 BM of heap space as it does with 8 GB. Seems to be very much CPU-limited. |
Beta Was this translation helpful? Give feedback.
-
I suspect that most fast solutions will be fine with pretty small heaps because the cost of regularly touching a lot of auxiliary storage is high. |
Beta Was this translation helpful? Give feedback.
-
In my experience, my solution was always slower when I tried with more heap space. It's not dramatic though, only 50 - 100 ms. But it counts in this challenge for sure. And for smaller sizes, I found serial GC performs best. I didn't see any difference even with Epsilon GC. Maybe Serial GC is also not kicked in since the heap usage is really small. |
Beta Was this translation helpful? Give feedback.
-
Most solutions do not seem to explicitly configure this. On my machine, the default max heap space is 25% of the available RAM.
I seem to get the best result with this heap size, where my solution runs much slower if the heap is larger. With that in mind, it would make sense to configure.
One issue is that if your program runs better the more RAM it has, you can't configure a static value, if the RAM on the evaluation machine differs from your local machine. On the other hand, setting it as a percentage might not be ideal either, as the ratio of free memory to total memory could also differ.
Thoughts on this?
Beta Was this translation helpful? Give feedback.
All reactions