Benchmarks

All filters configured for 65536 items at 1% false-positive rate.

2026

Comparison against bits-and-blooms/bloom/v3 and greatroar/blobloom.

Run on Go 1.25, AMD Ryzen AI MAX+ 395, Linux amd64 (March 2026):

go test -bench='BenchmarkPerOp' -benchmem -count=5 -benchtime=2s

Speed (ns/op, lower is better)

Operation	ipfs/bbloom	bits-and-blooms/bloom/v3	greatroar/blobloom
Add	~20.5	~39.3	~4.95
Has/Test	~20.0	~36.1	~4.4
AddTS (locked)	~32.6	n/a	n/a
HasTS (rlocked)	~23.7	n/a	n/a

Fastest: blobloom, but it accepts a pre-hashed uint64, so its numbers do not include hashing time. Comparing only libraries that hash from []byte internally, bbloom is ~2x faster than bits-and-blooms.

Memory (total allocations for 65536 inserts)

Library	Alloc
ipfs/bbloom	133 KiB
bits-and-blooms/bloom/v3	80 KiB
greatroar/blobloom	88 KiB

Smallest: bits-and-blooms uses the least memory. bbloom rounds the bitset up to the next power of two, which uses ~1.7x more memory but produces a much lower false-positive rate as a side effect.

False-positive rate (1M lookups of absent keys)

Library	Measured	Target
ipfs/bbloom	0.07%	1.00%
bits-and-blooms/bloom/v3	1.00%	1.00%
greatroar/blobloom	0.65%	1.00%

Best accuracy: bbloom at 0.07% -- 14x below the 1% target. This is a direct consequence of the power-of-two bitset rounding: more bits means fewer collisions.

Summary

Aspect	Winner
Speed	blobloom (pre-hashed); bbloom (hashing from []byte)
Memory	bits-and-blooms
Accuracy	bbloom (0.07% vs 1% target)
Built-in thread safety	bbloom (AddTS/HasTS with mutex)

2013

Original benchmarks by Andreas Briese on a MacBook Pro 15" (OSX 10.8.5, i7 4-core 2.4 GHz). Bloom filter size 524288 bits, 7 hash locations, 65536 items, 10 repetitions each:

github.com/AndreasBriese/bbloom    Add 65536 items:  6595800 ns (100 ns/op)
github.com/AndreasBriese/bbloom    Has 65536 items:  5986600 ns  (91 ns/op)
github.com/AndreasBriese/bloom     Add 65536 items:  6304684 ns  (96 ns/op)
github.com/AndreasBriese/bloom     Has 65536 items:  6568663 ns (100 ns/op)

github.com/willf/bloom             Add 65536 items: 24367224 ns (371 ns/op)
github.com/willf/bloom            Test 65536 items: 21881142 ns (333 ns/op)
github.com/dataence/bloom/standard Add 65536 items: 23041644 ns (351 ns/op)
github.com/dataence/bloom/standard Check 65536 items: 19153133 ns (292 ns/op)
github.com/cabello/bloom           Add 65536 items: 131921507 ns (2012 ns/op)
github.com/cabello/bloom      Contains 65536 items: 131108962 ns (2000 ns/op)

Note: willf/bloom has since been renamed to bits-and-blooms/bloom and has improved significantly (see 2026 results above: 371 ns/op -> ~39 ns/op).