Closed kelindar closed 3 years ago
This PR implements an unrolled POPCNT loop to speed things up by around 30%. Note that bits.CountOnes() already uses POPCNT instruction if present, but unrolling that doesn't work very well (I tried).
POPCNT
30%
bits.CountOnes()
This PR implements an unrolled
POPCNTloop to speed things up by around30%. Note thatbits.CountOnes()already usesPOPCNTinstruction if present, but unrolling that doesn't work very well (I tried).