was pointed towards an issue raised here, two possible optimisations that could be made:
using array instead of slice for bucket and fingerprint, sizeof(slice) is 24 bytes. also every slice is a pointer indirection, given bucket and fingerprint sizes don't vary an array would fare better
additionally the original paper proposes usage of 12-bit fingerprints, 8-bit variants have lower memory efficiency for the same false positive probability.
was pointed towards an issue raised here, two possible optimisations that could be made:
array
instead ofslice
forbucket
andfingerprint
,sizeof(slice)
is 24 bytes. also every slice is a pointer indirection, givenbucket
andfingerprint
sizes don't vary anarray
would fare better