Closed Yaffle closed 2 years ago
I understand that you believe it would be possible to significant boost the performance fo FastBitSet and you find it slow compared to your design.
I invite you to produce a pull request.
I am closing this particular issue. Let us discuss your ideas with actual code in hand.
Thanks again.
@lemire , I think the code change to switch to 31bit word size may affect performance of other operations, so you may not be happy with it. I have no plans to change your lib anyway, I am just trying to implement small BitSet with fast "change" ("xor").
Thanks for your reply.
Hi, I am not using your library, but would like to ask if usage of 32 bit integers affects performance.
In my test of your library it is slower if I set every 30th or 31th bits:
This script outputs something like: default: 2789.47802734375 ms default: 3693.5849609375 ms
On the page https://v8.dev/blog/pointer-compression it is said, that Chrome SMI is a 30 bit signed integer... May be this is why it is slower.