No benefit with new bulk approach for `And` ?

[marniks7]

Hi, based on the release v1.2.0 i have tried to use bulk approach for And operation for 1 + 6 extra bitmaps. Usage: index.And(bms[0], bms[1:]...) I do understand that there is 1 + 3 extra handled in bulk, and all next extra are handled one-by-one, but still

Timings for bitmap.And: 5.31s before vs 5.38s after (though this vary). I did few other tests and I don't see benefits.

Another test:

name              old time/op    new time/op    delta
FindPriceV2_Real    10.4µs ± 3%    10.3µs ± 3%     ~     (p=1.000 n=5+5)

name              old alloc/op   new alloc/op   delta
FindPriceV2_Real      361B ± 0%      537B ± 0%  +48.75%  (p=0.008 n=5+5)

name              old allocs/op  new allocs/op  delta
FindPriceV2_Real      8.00 ± 0%      9.00 ± 0%  +12.50%  (p=0.008 n=5+5)

Non-bulk approach:

Bulk approach:

[kelindar]

How big are your test bitmaps and caches? I assume if your bitmaps are small-medium size or your L1/L2 cache is large enough you won't indeed see much benefit.