hcq/TurboPFor-Integer-Compression
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

BitUtil: c/c++ header

Browse Source
This commit is contained in:
x
2018-01-07 18:36:14 +01:00
parent 4dd49cd55e
commit 62fdd87e7b
1 changed files with 13 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
bitutil.h
View File

@ -80,7 +80,6 @@ static inline uint64_t zigzagdec64(uint64_t x) { return x >> 1 ^
_sv_ = _mm256_add_epi32( _mm256_permute2x128_si256( _mm256_shuffle_epi32(_sv_,_MM_SHUFFLE(3, 3, 3, 3)), _sv_, 0x11), \
_mm256_add_epi32(_v_, _mm256_permute2x128_si256(zv,_mm256_shuffle_epi32(_v_, _MM_SHUFFLE(3, 3, 3, 3)), 0x20)));\
}
#define SCANI256x32(_v_, _sv_, _vi_) SCAN256x32(_v_, _sv_); _sv_ = _mm256_add_epi32(_sv_, _vi_)
#define ZIGZAG256x32(_v_) _mm256_xor_si256(_mm256_slli_epi32(_v_,1), _mm256_srai_epi32(_v_,31))
@ -94,12 +93,25 @@ static inline uint64_t zigzagdec64(uint64_t x) { return x >> 1 ^
#include <emmintrin.h>
#define DELTA128x32(_v_, _sv_) _mm_sub_epi32(_v_, _mm_or_si128(_mm_srli_si128(_sv_, 12), _mm_slli_si128(_v_, 4)))
#ifdef __SSSE3__
#include <tmmintrin.h>
// SIMD Scan ( prefix sum )
#define SCAN128x16( _v_, _sv_) {\
_v_ = _mm_add_epi16(_v_, _mm_slli_si128(_v_, 2));\
_v_ = _mm_add_epi16(_v_, _mm_slli_si128(_v_, 4));\
_sv_ = _mm_add_epi16(_mm_add_epi16(_v_, _mm_slli_si128(_v_, 8)), _mm_shuffle_epi8(_sv_, _mm_set_epi8(15,14,15,14,15,14,15,14,15,14,15,14,15,14,15,14)));\
}
#endif
#define SCAN128x32( _v_, _sv_) _v_ = _mm_add_epi32(_v_, _mm_slli_si128(_v_, 4)); _sv_ = _mm_add_epi32(_mm_shuffle_epi32(_sv_, _MM_SHUFFLE(3, 3, 3, 3)), _mm_add_epi32(_mm_slli_si128(_v_, 8), _v_) )
#define SCANI128x32(_v_, _sv_, _vi_) SCAN128x32(_v_, _sv_); _sv_ = _mm_add_epi32(_sv_, _vi_)
#define ZIGZAG128x16(_v_) _mm_xor_si128(_mm_slli_epi16(_v_,1), _mm_srai_epi16(_v_,15))
#define ZIGZAG128x32(_v_) _mm_xor_si128(_mm_slli_epi32(_v_,1), _mm_srai_epi32(_v_,31))
#define UNZIGZAG128x16(_v_) _mm_xor_si128(_mm_srli_epi16(_v_,1), _mm_srai_epi16(_mm_slli_epi16(_v_,15),15) )
#define UNZIGZAG128x32(_v_) _mm_xor_si128(_mm_srli_epi32(_v_,1), _mm_srai_epi32(_mm_slli_epi32(_v_,31),31) ) //_mm_sub_epi32(cz, _mm_and_si128(iv,c1))
// SIMD Horizontal OR
#define HOR128x32(_v_,_b_) _v_ = _mm_or_si128(_v_, _mm_srli_si128(_v_, 8)); _v_ = _mm_or_si128(_v_, _mm_srli_si128(_v_, 4)); _b_ = (unsigned)_mm_cvtsi128_si32(_v_)
#endif