guava
guava copied to clipboard
Update BloomFilterStrategies.java
Remove imul instruction from murmur operations of 32bit hash functions -- reduces the overall number of instructions by ~1-3 CPU cycles.
Before -- https://godbolt.org/z/ed93nf1sx After -- https://godbolt.org/z/3h1rG399P
I see the difference in bytecode, but is there a difference in benchmarks? I'd tend to expect both versions to get optimized to the same thing.
yeah, running a BM is good idea. By any chance, do we have one for testing the performance of the strategies ?. if so, could you please point me out on how to run them.
thanks