X86架构上SIMD指令集使用

SIMD说明

Single Instruction Multiple Data

Intel Processor上对应的指令集include文件

// AVX      immintrin.h
// MMX		mmintrin.h
// SSE		xmmintrin.h
// SSE2		emmintrin.h
// SSE3		pmmintrin.h

// suffix
// si256 – signed 256-bit integer
// si128 – signed 128-bit integer
// epi8, epi32, epi64 — an vector of signed 8-bit integers (32 in a __m256 and 16 in a __m128) or signed 32-bit integers or signed 64-bit integers
// epu8 — an vecotr of unsigned 8-bit integers (when there is a difference between what an operation would do with signed and unsigned numbers, such as with conversion to a larger integer or multiplication)
// epu16, epu32 — an array of unsigned 16-bit integers or 8 unsigned 32-bit integers (when the operation would be different than signed)
// ps — “packed single” — 8 single-precision floats
// pd — “packed double” — 4 doubles
// ss — one float (only 32-bits of a 256-bit or 128-bit value are used)
// sd — one double (only 64-bits of a 256-bit or 256-bit value are used)

以下例子中，输入的数据是否需要对齐取决于load指令的要求，在这里查找不同load指令的要求：
Intel Intrinsic指令含义查询链接

并行4个32 bit integer相加例子 – 输入无须对齐，__m128i赋值时间较长

#include<emmintrin.h>
#include<immintrin.h>

void print128_num_epi32(__m128i var)
{
   
    uint32_t val[4];
    memcpy(val, &var, sizeof(val))