SSE4 1和SSE4 2 Intrinsics各函数介绍
Posted xkiwnchwhd
tags:
篇首语：本文由小常识网(cha138.com)小编为大家整理，主要介绍了SSE4 1和SSE4 2 Intrinsics各函数介绍相关的知识，希望对你有一定的参考价值。
SIMD相关头文件包括：
//#include <ivec.h>//MMX
//#include <fvec.h>//SSE(also include ivec.h)
//#include <dvec.h>//SSE2(also include fvec.h)
#include <mmintrin.h> //MMX
#include <xmmintrin.h> //SSE(include mmintrin.h)
#include <emmintrin.h> //SSE2(include xmmintrin.h)
#include <pmmintrin.h> //SSE3(include emmintrin.h)
#include <tmmintrin.h>//SSSE3(include pmmintrin.h)
#include <smmintrin.h>//SSE4.1(include tmmintrin.h)
#include <nmmintrin.h>//SSE4.2(include smmintrin.h)
#include <wmmintrin.h>//AES(include nmmintrin.h)
#include <immintrin.h>//AVX(include wmmintrin.h)
#include <intrin.h>//(include immintrin.h)mmintrin.h为MMX 头文件，其中__m64的定义为：
typedef union __declspec(intrin_type) _CRT_ALIGN(8) __m64
{
unsigned __int64 m64_u64;
float m64_f32[2];
__int8 m64_i8[8];
__int16 m64_i16[4];
__int32 m64_i32[2];
__int64 m64_i64;
unsigned __int8 m64_u8[8];
unsigned __int16 m64_u16[4];
unsigned __int32 m64_u32[2];
} __m64;xmmintrin.h为SSE 头文件，此头文件里包含MMX头文件，其中__m128的定义为：
typedef union __declspec(intrin_type) _CRT_ALIGN(16) __m128 {
float m128_f32[4];
unsigned __int64 m128_u64[2];
__int8 m128_i8[16];
__int16 m128_i16[8];
__int32 m128_i32[4];
__int64 m128_i64[2];
unsigned __int8 m128_u8[16];
unsigned __int16 m128_u16[8];
unsigned __int32 m128_u32[4];
} __m128;emmintrin.h为SSE2头文件，此头文件里包含SSE头文件，其中__m128i和__m128d的定义为：
typedef union __declspec(intrin_type) _CRT_ALIGN(16) __m128i {
__int8 m128i_i8[16];
__int16 m128i_i16[8];
__int32 m128i_i32[4];
__int64 m128i_i64[2];
unsigned __int8 m128i_u8[16];
unsigned __int16 m128i_u16[8];
unsigned __int32 m128i_u32[4];
unsigned __int64 m128i_u64[2];
} __m128i;
typedef struct __declspec(intrin_type) _CRT_ALIGN(16) __m128d {
double m128d_f64[2];
} __m128d;smmintrin.h为SSE4.1头文件，其文件中各函数的介绍：
/*Integer blend instructions - select data from 2 sources
using constant/variable mask*/
//v1=(v10, v11, ..., v17), v2=(v20, v21, ..., v27)
//mask:If the corresponding flag bit is 0, the value is selected from parameter v1.
//Otherwise the value is from parameter v2.
//则r0=(mask0 == 0) ? v10 : v20,...,r7= (mask7 == 0) ? v17 : v27
extern __m128i _mm_blend_epi16 (__m128i v1, __m128i v2, const int mask);
//v1=(v10, v11, ..., v115), v2=(v20, v21, ..., v215), mask=(mask1, ..., mask15)
//则r0=(mask0 & 0x80) ? v20 : v10, ..., r15=(mask15 & 0x80) ? v215 : v115
extern __m128i _mm_blendv_epi8 (__m128i v1, __m128i v2, __m128i mask);
/*Float single precision blend instructions - select data
from 2 sources using constant/variable mask */
//v1=(v10, v11, v12, v13), v2=(v20, v21, v22, v23)
//则r0=(mask0 == 0) ? v10 : v20,..., r3= (mask3 == 0) ? v13 : v23
extern __m128 _mm_blend_ps (__m128 v1, __m128 v2, const int mask);
//v1=(v10, v11, v12, v13), v2=(v20, v21, v22, v23)
//则r0= (v30 & 0x80000000) ? v20 : v10,...,r3= (v33 & 0x80000000) ? v23 : v13
extern __m128 _mm_blendv_ps(__m128 v1, __m128 v2, __m128 v3);
/*Float double precision blend instructions - select data
from 2 sources using constant/variable mask*/
//v1=(v10, v11), v2=(v20, v21)
//则r0 = (mask0 == 0) ? v10 : v20, r1 = (mask1 == 0) ? v11 : v21
extern __m128d _mm_blend_pd (__m128d v1, __m128d v2, const int mask);
//v1=(v10, v11), v2=(v20, v21)
//则r0 = (v30 & 0x8000000000000000) ? v20 : v10,
//r1 = (v31 & 0x8000000000000000) ? v21 : v11
extern __m128d _mm_blendv_pd(__m128d v1, __m128d v2, __m128d v3);
/*Dot product instructions with mask-defined summing and zeroing
of result‘s parts*/
//val1=(val10, ..., val13), val2=(val20,...,val23)
/*则tmp0 := (mask4 == 1) ? (val10 * val20) : +0.0
tmp1 := (mask5 == 1) ? (val11 * val21) : +0.0
tmp2 := (mask6 == 1) ? (val12 * val22) : +0.0
tmp3 := (mask7 == 1) ? (val13 * val23) : +0.0
tmp4 := tmp0 + tmp1 + tmp2 + tmp3
r0 := (mask0 == 1) ? tmp4 : +0.0
r1 := (mask1 == 1) ? tmp4 : +0.0
r2 := (mask2 == 1) ? tmp4 : +0.0
r3 := (mask3 == 1) ? tmp4 : +0.0 */
extern __m128 _mm_dp_ps(__m128 val1, __m128 val2, const int mask);
//val1=(val10, val11), val2=(val20, val21)
/*则tmp0 := (mask4 == 1) ? (val10 * val20) : +0.0
tmp1 := (mask5 == 1) ? (val11 * val21) : +0.0
tmp2 := tmp0 + tmp1
r0 := (mask0 == 1) ? tmp2 : +0.0
r1 := (mask1 == 1) ? tmp2 : +0.0 */
extern __m128d _mm_dp_pd(__m128d val1, __m128d val2, const int mask);
/*Packed integer 64-bit comparison, zeroing or filling with ones
corresponding parts of result */
//val1=(val10, val11), val2=(val20, val21)
//则r0 = (val10 == val20) ? 0xffffffffffffffff : 0,
//r1 = (val11 == val21) ? 0xffffffffffffffff : 0
extern __m128i _mm_cmpeq_epi64(__m128i val1, __m128i val2);
/* Min/max packed integer instructions*/
//val1=(val10,...,val115), val2=(val20,...,val215)
//则r0 = (val10 < val20) ? val10 : val20, ...,
//r15 = (val115 < val215) ? val115 : val215
extern __m128i _mm_min_epi8 (__m128i val1, __m128i val2);
//val1=(val10,...,val115), val2=(val20,...,val215)
//则r0 = (val10 > val20) ? val10 : val20, ...,
//r15 = (val115 > val215) ? val115 : val215
extern __m128i _mm_max_epi8 (__m128i val1, __m128i val2);
//val1=(val10,...,val17), val2=(val20,...,val27), eight 16-bit unsigned integers
//则r0 = (val10 < val20) ? val10 : val20, ...,
//r7 = (val17 < val27) ? val17 : val27
extern __m128i _mm_min_epu16(__m128i val1, __m128i val2);
//val1=(val10,...,val17), val2=(val20,...,val27),eight 16-bit unsigned integers
//则r0 = (val10 > val20) ? val10 : val20, ...,
//r7 = (val17 > val27) ? val17 : val27
extern __m128i _mm_max_epu16(__m128i val1, __m128i val2);
//val1=(val10,...,val13), val2=(val20,...,val23)
//则r0 = (val10 < val20) ? val10 : val20, ...,
//r3 = (val13 < val23) ? val13 : val23
extern __m128i _mm_min_epi32(__m128i val1, __m128i val2);
//val1=(val10,...,val13), val2=(val20,...,val23)
//则r0 = (val10 > val20) ? val10 : val20, ...,
//r3 = (val13 > val23) ? val13 : val23
extern __m128i _mm_max_epi32(__m128i val1, __m128i val2);
//val1=(val10,...,val13), val2=(val20,...,val23), four 32-bit unsigned integers
//则r0 = (val10 < val20) ? val10 : val20, ...,
//r3 = (val13 < val23) ? val13 : val23
extern __m128i _mm_min_epu32(__m128i val1, __m128i val2);
//val1=(val10,...,val13), val2=(val20,...,val23), four 32-bit unsigned integers
//则r0 = (val10 > val20) ? val10 : val20, ...,
//r3 = (val13 > val23) ? val13 : val23
extern __m128i _mm_max_epu32(__m128i val1, __m128i val2);
/*Packed integer 32-bit multiplication with truncation
of upper halves of results*/
//a=(a0,...,a3), b=(b0,...,b3), 则r0=a0 * b0, ..., r3=a3 * b3
//Only the lower 32-bits of each product are saved
extern __m128i _mm_mullo_epi32(__m128i a, __m128i b);
/*Packed integer 32-bit multiplication of 2 pairs of operands
producing two 64-bit results */
//a=(a0,a1,a2,a3), b=(b0,b1,b2,b3)
//r0=low_half(a0*b0), r1=high_half(a0*b0),r2=low_half(a2*b2), r3=high_half(a2*b2)
//The upper 32-bits of each quadword of the input parameters are not used
extern __m128i _mm_mul_epi32(__m128i a, __m128i b);
/*Packed integer 128-bit bitwise comparison.
return 1 if (val ‘and‘ mask) == 0*/
//则r = (mask & val) == 0, Generates a return value of 0 or 1
extern int _mm_testz_si128(__m128i mask, __m128i val);
/*Packed integer 128-bit bitwise comparison.
return 1 if (val ‘and_not‘ mask) == 0 */
//则r=1 if all the bits set in val are set in mask; otherwise 0
//Generates a return value of 0 or 1
extern int _mm_testc_si128(__m128i mask, __m128i val);
/*Packed integer 128-bit bitwise comparison
ZF = ((val ‘and‘ mask) == 0) CF = ((val ‘and_not‘ mask) == 0)
return 1 if both ZF and CF are 0 */
//则 ZF := (mask & s2) == 0，CF := (~mask & s2) == 0, r = ~ZF & ~CF
//Generates a return value of 0 or 1
extern int _mm_testnzc_si128(__m128i mask, __m128i s2);
/*Insert single precision float into packed single precision
array element selected by index.
The bits [7-6] of the 3d parameter define src index,
the bits [5-4] define dst index, and bits [3-0] define zeroing
mask for dst */
/* sx := ndx6-7
sval := (sx == 0) ? src0 : ((sx == 1) ? src1 : ((sx == 2) ? src2 : src3))
dx := ndx4-5
r0 := (dx == 0) ? sval : dst0
r1 := (dx == 1) ? sval : dst1
r2 := (dx == 2) ? sval : dst2
r3 := (dx == 3) ? sval : dst3
zmask := ndx0-3
r0 := (zmask0 == 1) ? +0.0 : r0
r1 := (zmask1 == 1) ? +0.0 : r1
r2 := (zmask2 == 1) ? +0.0 : r2
r3 := (zmask3 == 1) ? +0.0 : r3 */
extern __m128 _mm_insert_ps(__m128 dst, __m128 src, const int ndx);
/*Extract binary representation of single precision float from
packed single precision array element selected by index */
//src=(src0, src1, src2, src3)
//则r = (ndx == 0) ? src0 : ((ndx == 1) ? src1 : ((ndx == 2) ? src2 : src3))
//Only the least significant two bits of ndx are used
extern int _mm_extract_ps(__m128 src, const int ndx);
/*Insert integer into packed integer array element
selected by index */
//则r0=(ndx == 0) ? s : dst0, ..., r15=(ndx == 15) ? s : dst15
//Only the lowest 8 bits of s are used,
//Only the least significant 4 bits of ndx are used
extern __m128i _mm_insert_epi8 (__m128i dst, int s, const int ndx);
//则r0=(ndx == 0) ? s : dst0, ..., r3=(ndx == 3) ? s : dst3
//Only the least significant 2 bits of ndx are interpreted
extern __m128i _mm_insert_epi32(__m128i dst, int s, const int ndx);
//则r0=(ndx == 0) ? s : dst0, r1=(ndx == 1) ? s : dst1
//Only the least significant bit of ndx is interpreted
extern __m128i _mm_insert_epi64(__m128i dst, __int64 s, const int ndx);
/*Extract integer from packed integer array element
selected by index */
//则r=(ndx == 0) ? src0 : ((ndx == 1) ? src1 : ...((ndx == 14) ? src14 : src15))
//Only the least significant four bits of ndx are used
//注意：The result is the unsigned equivalent of the appropriate 8-bits in parameter src
extern int _mm_extract_epi8 (__m128i src, const int ndx);
//则r=(ndx == 0) ? src0 : ((ndx == 1) ? src1 : ((ndx == 2) ? src2 : src3))
//Only the least significant two bits of ndx are used.
extern int _mm_extract_epi32(__m128i src, const int ndx);
//则r = (ndx == 0) ? src0 : src1
//Only the least significant bit of parameter ndx is used
extern __int64 _mm_extract_epi64(__m128i src, const int ndx);
/*Horizontal packed word minimum and its index in
result[15:0] and result[18:16] respectively */
//The lowest order 16 bits are the minimum value found in parameter shortValues.
//The second-lowest order 16 bits are the index of the minimum value
//found in parameter shortValues.
extern __m128i _mm_minpos_epu16(__m128i shortValues);
/* Packed/single float double precision rounding */
//则r0=RND(val0), r1=RND(val1),详见参考文献1
extern __m128d _mm_round_pd(__m128d val, int iRoundMode);
//则r0=RND(val0), r1=dst1, 详见参考文献1
// The lowest 64 bits are the result of the rounding function on val.
//The higher order 64 bits are copied directly from input parameter dst
extern __m128d _mm_round_sd(__m128d dst, __m128d val, int iRoundMode);
/*Packed/single float single precision rounding */
//则r0=RND(val0), r1=RND(val1), r2=RND(val2), r3=RND(val3),详见参考文献1
extern __m128 _mm_round_ps(__m128 val, int iRoundMode);
//则r0=RND(val0), r1=dst1, r2=dst2, r3=dst3,
//The lowest 32 bits are the result of the rounding function on val.
//The higher order 96 bits are copied directly from input parameter dst
extern __m128 _mm_round_ss(__m128 dst, __m128 val, int iRoundMode);
/*Packed integer sign-extension */
//byteValues: A 128-bit parameter that contains four signed 8-bit integers
//in the lower 32 bits, byteValues=(a0, a1, ..., a15)
/*则r0 := a0
r1 := (a0 < 0) ? 0xff : 0
r2 := (a0 < 0) ? 0xff : 0
r3 := (a0 < 0) ? 0xff : 0
r4 := a1
r5 := (a1 < 0) ? 0xff : 0
r6 := (a1 < 0) ? 0xff : 0
r7 := (a1 < 0) ? 0xff : 0
r8 := a2
r9 := (a2 < 0) ? 0xff : 0
r10 := (a2 < 0) ? 0xff : 0
r11 := (a2 < 0) ? 0xff : 0
r12 := a3
r13 := (a3 < 0) ? 0xff : 0
r14 := (a3 < 0) ? 0xff : 0
r15 := (a3 < 0) ? 0xff : 0 */
extern __m128i _mm_cvtepi8_epi32 (__m128i byteValues);
//shortValues: A 128-bit parameter that contains four signed 16-bit integers
//in the lower 64 bits, shortValues=(a0, a1, ..., a7)
/*则r0 := a0
r1 := (a0 < 0) ? 0xffff : 0
r2 := a1
r3 := (a1 < 0) ? 0xffff : 0
r4 := a2
r5 := (a2 < 0) ? 0xffff : 0
r6 := a3
r7 := (a3 < 0) ? 0xffff : 0 */
extern __m128i _mm_cvtepi16_epi32(__m128i shortValues);
//byteValues: A 128-bit parameter that contains two signed 8-bit integers
//in the lower 16 bits, byteValues=(a0, a1, ... , a15)
/*则r0 := a0
r1 := (a0 < 0) ? 0xff : 0
r2 := (a0 < 0) ? 0xff : 0
r3 := (a0 < 0) ? 0xff : 0
r4 := (a0 < 0) ? 0xff : 0
r5 := (a0 < 0) ? 0xff : 0
r6 := (a0 < 0) ? 0xff : 0
r7 := (a0 < 0) ? 0xff : 0
r8 := a1
r9 := (a1 < 0) ? 0xff : 0
r10 := (a1 < 0) ? 0xff : 0
r11 := (a1 < 0) ? 0xff : 0
r12 := (a1 < 0) ? 0xff : 0
r13 := (a1 < 0) ? 0xff : 0
r14 := (a1 < 0) ? 0xff : 0
r15 := (a1 < 0) ? 0xff : 0 */
extern __m128i _mm_cvtepi8_epi64 (__m128i byteValues);
//intValues: A 128-bit parameter that contains two signed 32-bit
//integers in the lower 64 bits, intValues=(a0, a1, a2, a3)
/*则r0 := a0
r1 := (a0 < 0) ? 0xffffffff : 0
r2 := a1
r3 := (a1 < 0) ? 0xffffffff : 0*/
extern __m128i _mm_cvtepi32_epi64(__m128i intValues);
//shortValues:A 128-bit parameter that contains two signed 16-bit integers
//in the lower 32 bits, shortValues=(a0, a1, ..., a7)
/*则r0 := a0
r1 := (a0 < 0) ? 0xffff : 0
r2 := (a0 < 0) ? 0xffff : 0
r3 := (a0 < 0) ? 0xffff : 0
r4 := a1
r5 := (a1 < 0) ? 0xffff : 0
r6 := (a1 < 0) ? 0xffff : 0
r7 := (a1 < 0) ? 0xffff : 0*/
extern __m128i _mm_cvtepi16_epi64(__m128i shortValues);
//byteValues:A 128-bit parameter that contains eight signed 8-bit integers
//in the lower 64 bits, byteValues=(a0, a1, ..., a15)
/*则r0 := a0
r1 := (a0 < 0) ? 0xff : 0
r2 := a1
r3 := (a1 < 0) ? 0xff : 0
...
r14 := a7
r15 := (a7 < 0) ? 0xff : 0*/
extern __m128i _mm_cvtepi8_epi16 (__m128i byteValues);
/*Packed integer zero-extension*/
//byteValues:A 128-bit parameter that contains four unsigned 8-bit integers
//in the lower 32 bits, byteValues=(a0, a1, ... , a15)
/*则r0 := a0
r1 := 0
r2 := 0
r3 := 0
r4 := a1
r5 := 0
r6 := 0
r7 := 0
r8 := a2
r9 := 0
r10 := 0
r11 := 0
r12 := a3
r13 := 0
r14 := 0
r15 := 0*/
extern __m128i _mm_cvtepu8_epi32 (__m128i byteValues);
//shortValues:A 128-bit parameter that contains four unsigned 16-bit integers
//in the lower 64 bits, shortValues=(a0, a1, ... , a7)
/*则r0 := a0
r1 := 0
r2 := a1
r3 := 0
r4 := a2
r5 := 0
r6 := a3
r7 := 0*/
extern __m128i _mm_cvtepu16_epi32(__m128i shortValues);
//shortValues:A 128-bit parameter that contains two unsigned 8-bit integers
//in the lower 16 bits, shortValues=(a0, a1, ..., a15)
/*则r0 := a0
r1 := 0
r2 := 0
r3 := 0
r4 := 0
r5 := 0
r6 := 0
r7 := 0
r8 := a1
r9 := 0
r10 := 0
r11 := 0
r12 := 0
r13 := 0
r14 := 0
r15 := 0*/
extern __m128i _mm_cvtepu8_epi64 (__m128i shortValues);
//intValues:A 128-bit parameter that contains two unsigned 32-bit integers
//in the lower 64 bits, intValues=(a0, a1, a2, a3)
/*则r0 = a0
r1 = 0
r2 = a1
r3 = 0*/
extern __m128i _mm_cvtepu32_epi64(__m128i intValues);
//shortValues:A 128-bit parameter that contains two unsigned 16-bit integers
//in the lower 32 bits, shortValues=(a0, a1, ... , a7)
/*则r0 := a0
r1 := 0
r2 := 0
r3 := 0
r4 := a1
r5 := 0
r6 := 0
r7 := 0*/
extern __m128i _mm_cvtepu16_epi64(__m128i shortValues);
//byteValues:A 128-bit parameter that contains eight unsigned 8-bit integers
//in the lower 64 bits, byteValues=(a0, a1, ... , a15)
/*则r0 := a0
r1 := 0
r2 := a1
r3 := 0
...
r14 := a7
r15 := 0*/
extern __m128i _mm_cvtepu8_epi16 (__m128i byteValues);
/*Pack 8 double words from 2 operands into 8 words of result
with unsigned saturation */
//val1=(val10,...,vall3), val2=(val20, ..., val23)
/*则r0 := (val10 < 0) ? 0 : ((val10 > 0xffff) ? 0xffff : val10)
r1 := (val11 < 0) ? 0 : ((val11 > 0xffff) ? 0xffff : val11)
r2 := (val12 < 0) ? 0 : ((val12 > 0xffff) ? 0xffff : val12)
r3 := (val13 < 0) ? 0 : ((val13 > 0xffff) ? 0xffff : val13)
r4 := (val20 < 0) ? 0 : ((val20 > 0xffff) ? 0xffff : val20)
r5 := (val21 < 0) ? 0 : ((val21 > 0xffff) ? 0xffff : val21)
r6 := (val22 < 0) ? 0 : ((val22 > 0xffff) ? 0xffff : val22)
r7 := (val23 < 0) ? 0 : ((val23 > 0xffff) ? 0xffff : val23)*/
extern __m128i _mm_packus_epi32(__m128i val1, __m128i val2);
/*Sum absolute 8-bit integer difference of adjacent groups of 4 byte
integers in operands. Starting offsets within operands are
determined by mask */
//s1, s2: sixteen 8-bit unsigned integers
// msk0, msk1, and msk2 are the three least significant bits of parameter msk
/*则i = msk2 * 4
j = msk0-1 * 4
for (k = 0; k < 8; k = k + 1) {
t0 = abs(s1[i + k + 0] - s2[j + 0])
t1 = abs(s1[i + k + 1] - s2[j + 1])
t2 = abs(s1[i + k + 2] - s2[j + 2])
t3 = abs(s1[i + k + 3] - s2[j + 3])
r[k] = t0 + t1 + t2 + t3
}*/
extern __m128i _mm_mpsadbw_epu8(__m128i s1, __m128i s2, const int msk);
/*
* Load double quadword using non-temporal aligned hint
*/
//This instruction loads data from a specified address.The memory source must be
//16-byte aligned because the return value consists of sixteen bytes.则r=*v1
extern __m128i _mm_stream_load_si128(__m128i* v1);nmmintrin.h为SSE4.2头文件，其文件中各函数的介绍：
/*
* Intrinsics for text/string processing.
*/
//Either the computed mask of MaxSize bits or its expansion to a 128-bit parameter.
//If the return value is expanded, each bit of the result mask is expanded to a
//byte or a word.详见参考文献2
extern __m128i _mm_cmpistrm (__m128i a, __m128i b, const int mode);
//An integer between 0 and Maxsize. MaxSize when the computed mask equals 0.
//Otherwise, the index of the leftmost or rightmost bit set to 1 in this mask.
//详见参考文献2
extern int _mm_cmpistri (__m128i a, __m128i b, const int mode);
//Either the computed mask of MaxSize bits or its expansion to a 128-bit parameter.
//If the return value is expanded, each bit of the result mask is expanded to
//a byte or a word.详见参考文献3
extern __m128i _mm_cmpestrm (__m128i a, int la, __m128i b, int lb, const int mode);
//An integer that ranges between 0 and MaxSize. Maxsize is returned when the
//resulting bitmask is equal to 0. Otherwise, the index of either the leftmost
//or rightmost bit set to 1 in this mask.详见参考文献3
extern int _mm_cmpestri (__m128i a, int la, __m128i b, int lb, const int mode);
/*
* Intrinsics for text/string processing and reading values of EFlags.
*/
//Returns one if the null character occurs in b. Otherwise, zero. When one is
//returned, it means that b contains the ending fragment of the string that is
//being compared.详见参考文献2
extern int _mm_cmpistrz (__m128i a, __m128i b, const int mode);
//Zero if the resulting mask is equal to zero. Otherwise, one.
//详见参考文献2
extern int _mm_cmpistrc (__m128i a, __m128i b, const int mode);
//One if the null character occurs in a. Otherwise, zero. When one is returned,
//it means that a contains the ending fragment of the string that is being compared.
//详见参考文献2
extern int _mm_cmpistrs (__m128i a, __m128i b, const int mode);
//bit0 of the resulting bitmask.详见参考文献2
extern int _mm_cmpistro (__m128i a, __m128i b, const int mode);
//One if b is does not contain the null character and the resulting mask is
//equal to zero. Otherwise, zero. 详见参考文献2
extern int _mm_cmpistra (__m128i a, __m128i b, const int mode);
//One if the absolute value of lb is less than MaxSize. Otherwise, zero.详见参考文献3
extern int _mm_cmpestrz (__m128i a, int la, __m128i b, int lb, const int mode);
//Zero if the resulting mask is equal to zero. Otherwise, one.详见参考文献3
extern int _mm_cmpestrc (__m128i a, int la, __m128i b, int lb, const int mode);
//One if the absolute value of la is less than MaxSize. Otherwise, zero.详见参考文献3
extern int _mm_cmpestrs (__m128i a, int la, __m128i b, int lb, const int mode);
//bit0 of the resulting bitmask. 详见参考文献3
extern int _mm_cmpestro (__m128i a, int la, __m128i b, int lb, const int mode);
//One if the absolute value of lb is larger than or equal to MaxSize and the
//resulting mask is equal to zero. Otherwise, zero.详见参考文献3
extern int _mm_cmpestra (__m128i a, int la, __m128i b, int lb, const int mode);
/*
* Packed integer 64-bit comparison, zeroing or filling with ones
* corresponding parts of result
*/
//val1=(val10, val11), val2=(val20, val21)
//则,r0 = (val10 > val20) ? 0xffffffffffffffff : 0x0
// r1 = (val11 > val21) ? 0xffffffffffffffff : 0x0
extern __m128i _mm_cmpgt_epi64(__m128i val1, __m128i val2);
/*
* Calculate a number of bits set to 1
*/
//The number of bits set to one in v
extern int _mm_popcnt_u32(unsigned int v);
//The number of bits set to one in v
extern __int64 _mm_popcnt_u64(unsigned __int64 v);
/*
* Accumulate CRC32 (polynomial 0x11EDC6F41) value
*/
//crc：循环冗余校验码，CRC32-C algorithm is based on polynomial 0x1EDC6F41,
//r = crc + CRC-32C(v)
extern unsigned int _mm_crc32_u8 (unsigned int crc, unsigned char v);
//crc：循环冗余校验码，CRC32-C algorithm is based on polynomial 0x1EDC6F41,
//r = crc + CRC-32C(v)
extern unsigned int _mm_crc32_u16(unsigned int crc, unsigned short v);
//crc：循环冗余校验码，CRC32-C algorithm is based on polynomial 0x1EDC6F41,
//r = crc + CRC-32C(v)
extern unsigned int _mm_crc32_u32(unsigned int crc, unsigned int v);
//crc：循环冗余校验码，CRC32-C algorithm is based on polynomial 0x1EDC6F41,
//r = crc + CRC-32C(v)
extern unsigned __int64 _mm_crc32_u64(unsigned __int64 crc, unsigned __int64 v);参考文献：
1、http://msdn.microsoft.com/zh-cn/library/bb514044(v=vs.100).aspx
2、http://msdn.microsoft.com/zh-cn/library/bb513993(v=vs.100).aspx
3、http://msdn.microsoft.com/zh-cn/library/bb514048(v=vs.100).aspx
再分享一下我老师大神的人工智能教程吧。零基础！通俗易懂！风趣幽默！还带黄段子！希望你也加入到我们人工智能的队伍中来！https://blog.csdn.net/jiangjunshow
以上是关于SSE4 1和SSE4 2 Intrinsics各函数介绍的主要内容，如果未能解决你的问题，请参考以下文章
使用 SSE4.2 编译 TensorFlow，AVX 因 bazel 失败
是否所有支持 AVX2 的 CPU 也支持 SSE4.2 和 AVX？
如何使用 SSE4.2 和 AVX 指令编译 Tensorflow？
如何使用 SSE4.2 和 AVX 指令编译 Tensorflow？
如何使用 SSE4.2 和 AVX 指令编译 Tensorflow？
在win10里 为啥 Bazel 编译 tensorflow.dll 提示 /arch:SSE4.2 无效命令？要编译SSE4.2的参数怎么配置