MirrorHub/construct
0
0
Fork 0
mirror of https://github.com/matrix-construct/construct synced 2024-09-30 12:48:54 +02:00
Code Issues Releases Wiki Activity

ircd::simd: Add population mask convenience template.

Browse source
This commit is contained in:
Jason Volk 2020-06-29 20:15:11 -07:00
parent e3e6bb8dbf
commit aec21af0e4
1 changed files with 29 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

29
include/ircd/simd/simd.h
View file

@ -18,6 +18,9 @@
namespace ircd::simd
{
template<class T> T popmask(const T) noexcept;
template<class T> size_t popcount(const T) noexcept;
// xmmx shifter workarounds
template<int bits, class T> T shl(const T &a) noexcept;
template<int bits, class T> T shr(const T &a) noexcept;
@ -34,6 +37,32 @@ namespace ircd
using simd::lane_cast;
}
/// Convenience template. Vector compare instructions yield 0xff on equal;
/// sometimes one might need an actual value of 1 for accumulators or maybe
/// some bool-type reason...
template<class T>
inline size_t
ircd::simd::popcount(const T a)
noexcept
{
size_t i(0), ret(0);
while(i < lanes(a))
ret += __builtin_popcountll(a[i++]);
return ret;
}
/// Convenience template. Vector compare instructions yield 0xff on equal;
/// sometimes one might need an actual value of 1 for accumulators or maybe
/// some bool-type reason...
template<class T>
inline T
ircd::simd::popmask(const T a)
noexcept
{
return a & 1;
}
#ifdef HAVE_X86INTRIN_H
template<int b>
[[using gnu: always_inline, gnu_inline, artificial]]