Lzcnt/popcnt functions?

[touisteur]

Hi,

Unless I haven't looked in the right places, there doesn't seem to be any support lane-wise lzcnt or popcount. Is it foreseen in the future or out of scope of this library?

Thanks in advance

[kouchy]

Hi @touisteur,

At this time this feature does not exist. Maybe you can use mipp::sum as a work around until I implement the popcount.

Hope it helps.

[hayguen]

a look at https://github.com/kimwalisch/libpopcnt might be interesting - although i would think, there are different goals:

• libpopcnt is targeted for big data sizes
• mipp IMHO should target inlined code for simple data types

@kouchy : don't see how mipp::sum could be any workaround?!

[yzazik]

@hayguen I guess, you would use LUT to convert each byte/nibble to a number of set bits in it, then shuffle/permute and sum till you reach your lane width. Some libraries even do it recursively and optimizer unwinds it quite nicely.

[hayguen]

@yzazik i had found my non-simd solution at https://graphics.stanford.edu/~seander/bithacks.html specifically https://graphics.stanford.edu/~seander/bithacks.html#CountBitsSetParallel , if i remind right benchmarks showed, that it's very near to CPU intrinsic on x64