diff --git a/bitutil.h b/bitutil.h index 282c1ff..7965e8d 100644 --- a/bitutil.h +++ b/bitutil.h @@ -1,5 +1,5 @@ /** - Copyright (C) powturbo 2013-2018 + Copyright (C) powturbo 2013-2019 GPL v2 License This program is free software; you can redistribute it and/or modify @@ -22,11 +22,25 @@ - email : powturbo [_AT_] gmail [_DOT_] com **/ // "Integer Compression: max.bits, delta, zigzag, xor" + #ifdef __AVX2__ +#include + #elif defined(__AVX__) +#include + #elif defined(__SSE4_1__) +#include + #elif defined(__SSSE3__) +#include + #elif defined(__SSE2__) +#include + #elif defined(__ARM_NEON) +#include + #endif #if defined(_MSC_VER) && _MSC_VER < 1600 #include "vs/stdint.h" #else #include #endif +#include "sse_neon.h" #define BITFORSET_(_out_, _n_, _start_, _mindelta_) do { unsigned _i;\ for(_i = 0; _i != (_n_&~3); _i+=4) {\ @@ -51,26 +65,24 @@ #define BITSIZE16(_in_, _n_, _b_) BITSIZE_(_in_, _n_, _b_, 16) #define BITSIZE64(_in_, _n_, _b_) BITSIZE_(_in_, _n_, _b_, 64) -static inline unsigned char zigzagenc8( char x) { return x << 1 ^ x >> 7; } -static inline char zigzagdec8( unsigned char x) { return x >> 1 ^ -(x & 1); } +static inline unsigned char zigzagenc8( signed char x) { return x << 1 ^ x >> 7; } +static inline char zigzagdec8( unsigned char x) { return x >> 1 ^ -(x & 1); } -static inline unsigned short zigzagenc16(short x) { return x << 1 ^ x >> 15; } -static inline short zigzagdec16(unsigned short x) { return x >> 1 ^ -(x & 1); } +static inline unsigned short zigzagenc16(short x) { return x << 1 ^ x >> 15; } +static inline short zigzagdec16(unsigned short x) { return x >> 1 ^ -(x & 1); } -static inline unsigned zigzagenc31(int x) { x = (x << 2 | ((x>>30)& 2)) ^ x >> 31; return x; } // for signed x -static inline unsigned zigzagdec31(unsigned x) { return (x >> 2 | (x& 2)<<30 ) ^ -(x & 1); } +static inline unsigned zigzagenc31(int x) { x = (x << 2 | ((x>>30)& 2)) ^ x >> 31; return x; } // for signed x +static inline unsigned zigzagdec31(unsigned x) { return (x >> 2 | (x& 2)<<30 ) ^ -(x & 1); } static inline unsigned long long zigzagenc63(long long x) { x = (x << 2 | ((x>>62)& 2)) ^ x >> 63; return x; } // for signed x static inline long long zigzagdec63(unsigned long long x) { return (x >> 2 | (x& 2)<<62 ) ^ -(x & 1); } -static inline unsigned zigzagenc32(int x) { return x << 1 ^ x >> 31; } -static inline int zigzagdec32(unsigned x) { return x >> 1 ^ -(x & 1); } +static inline unsigned zigzagenc32(int x) { return x << 1 ^ x >> 31; } +static inline int zigzagdec32(unsigned x) { return x >> 1 ^ -(x & 1); } -static inline uint64_t zigzagenc64(int64_t x) { return x << 1 ^ x >> 63; } -static inline int64_t zigzagdec64(uint64_t x) { return x >> 1 ^ -(x & 1); } +static inline uint64_t zigzagenc64(int64_t x) { return x << 1 ^ x >> 63; } +static inline int64_t zigzagdec64(uint64_t x) { return x >> 1 ^ -(x & 1); } #ifdef __AVX2__ -#include - #define DELTA256x32(_v_, _sv_) _mm256_sub_epi32(_v_, _mm256_alignr_epi8(_v_, _mm256_permute2f128_si256(_sv_, _v_, _MM_SHUFFLE(0, 2, 0, 1)), 12)) #define SCAN256x32( _v_, _sv_) {\ @@ -87,25 +99,28 @@ static inline int64_t zigzagdec64(uint64_t x) { return x >> 1 ^ #define HOR256x32(_v_,_b_) _v_ = _mm256_or_si256(_v_, _mm256_srli_si256(_v_, 8)); _v_ = _mm256_or_si256(_v_, _mm256_srli_si256(_v_, 4)); _b_ = _mm256_extract_epi32(_v_,0) | _mm256_extract_epi32(_v_, 4) #endif - #ifdef __SSE2__ -#include - - #ifdef __SSSE3__ -#include + #if defined(__SSE2__) || defined(__ARM_NEON) + #if defined(__SSSE3__) || defined(__ARM_NEON) #define DELTA128x16(_v_, _sv_) _mm_sub_epi16(_v_, _mm_alignr_epi8(_v_, _sv_, 14)) -#define DELTA128x32(_v_, _sv_) _mm_sub_epi32(_v_, _mm_alignr_epi8(_v_, _sv_, 12)) +#define DELTA128x32(_v_, _sv_) _mm_sub_epi32(_v_, _mm_alignr_epi8(_v_, _sv_, 12)) // 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)));\ + _sv_ = _mm_add_epi16(_mm_add_epi16(_v_, _mm_slli_si128(_v_, 8)), _mm_shuffle_epi8(_sv_, _mm_set1_epi16(0x0f0e)/*_mm_set_epi8(15,14,15,14,15,14,15,14,15,14,15,14,15,14,15,14)*/));\ } #else #define DELTA128x16(_v_, _sv_) _mm_sub_epi16(_v_, _mm_or_si128(_mm_srli_si128(_sv_, 14), _mm_slli_si128(_v_, 2))) #define DELTA128x32(_v_, _sv_) _mm_sub_epi32(_v_, _mm_or_si128(_mm_srli_si128(_sv_, 12), _mm_slli_si128(_v_, 4))) #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 SUBI128x16(_v_, _sv_) _mm_sub_epi16(_v_, _sv_) +#define SUBI128x32(_v_, _sv_) _mm_sub_epi32(_v_, _sv_) +#define ADDI128x16(_v_, _sv_, _vi_) _sv_ = _mm_add_epi16(_mm_add_epi16(_sv_, _vi_),_v_) +#define ADDI128x32(_v_, _sv_, _vi_) _sv_ = _mm_add_epi32(_mm_add_epi32(_sv_, _vi_),_v_) + +#define SCAN128x32( _v_, _sv_) _v_ = _mm_add_epi32(_v_, _mm_slli_si128(_v_, 4)); _sv_ = _mm_add_epi32(mm_shuffle_nnnn_epi32(_sv_, 3), _mm_add_epi32(_mm_slli_si128(_v_, 8), _v_) ) // _mm_shuffle_epi32(_sv_, _MM_SHUFFLE(3, 3, 3, 3)) #define SCANI128x16(_v_, _sv_, _vi_) SCAN128x16(_v_, _sv_); _sv_ = _mm_add_epi16(_sv_, _vi_) #define SCANI128x32(_v_, _sv_, _vi_) SCAN128x32(_v_, _sv_); _sv_ = _mm_add_epi32(_sv_, _vi_) @@ -146,7 +161,7 @@ static inline int64_t zigzagdec64(uint64_t x) { return x >> 1 ^ _sv = _mm256_add_epi32(_sv, _cv); _cv = _mm256_set1_epi32(4*_mindelta_); do { _mm256_storeu_si256(_ov++, _sv), _sv = _mm256_add_epi32(_sv, _cv); } while(_ov < _ove);\ } while(0) - #elif defined(__SSE2__) && defined(USE_SSE) // ------------- + #elif (defined(__SSE2__) || defined(__ARM_NEON)) && defined(USE_SSE) // ------------- #define BITSIZE32(_in_, _n_, _b_) { unsigned *_ip; __m128i _v = _mm_setzero_si128();\ for(_ip = _in_; _ip != _in_+(_n_&~(4-1)); _ip+=4)\ _v = _mm_or_si128(_v, _mm_loadu_si128((__m128i*)_ip));\ @@ -209,6 +224,9 @@ unsigned bitd18( uint8_t *in, unsigned n, uint8_t start); unsigned bitd116( uint16_t *in, unsigned n, uint16_t start); unsigned bitd132( uint32_t *in, unsigned n, uint32_t start); unsigned bitd164( uint64_t *in, unsigned n, uint64_t start); + +unsigned bits128v16( uint16_t *in, unsigned n, uint16_t start); +unsigned bits128v32( uint32_t *in, unsigned n, uint32_t start); //-- in-place reverse delta of delta (encoded w/ bitdiencNN and parameter mindelta = 1) void bitd1dec8( uint8_t *p, unsigned n, uint8_t start); // non strictly decreasing (out[i] = in[i] - in[i-1] - 1) void bitd1dec16( uint16_t *p, unsigned n, uint16_t start); @@ -309,3 +327,4 @@ void padfloat64(double *in, size_t n, double *out, double e); #ifdef __cplusplus } #endif +