github_mirrors/vectorscan
3
0
Fork 1
mirror of https://github.com/VectorCamp/vectorscan.git synced 2025-06-28 16:41:01 +03:00
Code Issues Packages Projects Releases Wiki Activity

popcount: use intrinsics and restructure defines

Browse Source
This commit is contained in:
Matthew Barr 2016-05-20 15:30:29 +10:00
parent 142e74e8e6
commit 5234639736
1 changed files with 10 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
src/util/popcount.h
View File

@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2015, Intel Corporation * Copyright (c) 2015-2016, Intel Corporation
* *
* Redistribution and use in source and binary forms, with or without * Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met: * modification, are permitted provided that the following conditions are met:
@ -38,21 +38,17 @@
// We have a native popcount where the compiler has defined __POPCNT__. // We have a native popcount where the compiler has defined __POPCNT__.
#if defined(__POPCNT__) #if defined(__POPCNT__)
#define HAVE_POPCOUNT_INSTR #define HAVE_POPCOUNT_INSTR
#endif #elif defined(_WIN32) && defined(__AVX__) // TODO: fix win preproc
#if defined(_WIN32) && defined(__AVX__) // TODO: fix win preproc
#define HAVE_POPCOUNT_INSTR #define HAVE_POPCOUNT_INSTR
#define __builtin_popcount __popcnt
#define __builtin_popcountll __popcnt64
#endif #endif
static really_inline static really_inline
u32 popcount32(u32 x) { u32 popcount32(u32 x) {
#if defined(HAVE_POPCOUNT_INSTR) #if defined(HAVE_POPCOUNT_INSTR)
// Single-instruction builtin. // Single-instruction builtin.
return (u32)__builtin_popcount(x); return _mm_popcnt_u32(x);
#else #else
// Fast branch-free version from bit-twiddling hacks as most Intel // Fast branch-free version from bit-twiddling hacks as older Intel
// processors do not have a POPCNT instruction. // processors do not have a POPCNT instruction.
x -= (x >> 1) & 0x55555555; x -= (x >> 1) & 0x55555555;
x = (x & 0x33333333) + ((x >> 2) & 0x33333333); x = (x & 0x33333333) + ((x >> 2) & 0x33333333);
@ -62,16 +58,18 @@ u32 popcount32(u32 x) {
static really_inline static really_inline
u32 popcount64(u64a x) { u32 popcount64(u64a x) {
#if defined(HAVE_POPCOUNT_INSTR) #if defined(ARCH_X86_64)
# if defined(HAVE_POPCOUNT_INSTR)
// Single-instruction builtin. // Single-instruction builtin.
return (u32)__builtin_popcountll(x); return (u32)_mm_popcnt_u64(x);
#elif defined(ARCH_X86_64) # else
// Fast branch-free version from bit-twiddling hacks as most Intel // Fast branch-free version from bit-twiddling hacks as older Intel
// processors do not have a POPCNT instruction. // processors do not have a POPCNT instruction.
x -= (x >> 1) & 0x5555555555555555; x -= (x >> 1) & 0x5555555555555555;
x = (x & 0x3333333333333333) + ((x >> 2) & 0x3333333333333333); x = (x & 0x3333333333333333) + ((x >> 2) & 0x3333333333333333);
x = (x + (x >> 4)) & 0x0f0f0f0f0f0f0f0f; x = (x + (x >> 4)) & 0x0f0f0f0f0f0f0f0f;
return (x * 0x0101010101010101) >> 56; return (x * 0x0101010101010101) >> 56;
# endif
#else #else
// Synthesise from two 32-bit cases. // Synthesise from two 32-bit cases.
return popcount32(x >> 32) + popcount32(x); return popcount32(x >> 32) + popcount32(x);