_mm_set1_epi8的干了什么？

最新推荐文章于 2023-05-08 12:56:48 发布
原创 最新推荐文章于 2023-05-08 12:56:48 发布 · 8.5k 阅读 · 2 ·
CC 4.0 BY-SA版权
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。

本文深入探讨了Intel SSE2 intrinsics中的_mm_set1_epi8函数，解释了其用途和工作原理，以及如何通过单指令多数据流思想实现高效的数据操作。

在qemu中有#define SPLAT(p) _mm_set1_epi8(*(p))

那么，_mm_set1_epi8到底是干什么用的呢？下面这个说明文档说的很详细，算是个记录吧。

Set Intrinsics

The Intel® Streaming SIMD Extensions 2 (Intel® SSE2) intrinsics for integer set operations are listed in this topic. These intrinsics are composite intrinsics because they require more than one instruction to implement them. The prototypes for the Intel® SSE2 intrinsics are in the emmintrin.h header file.

The results of each intrinsic operation are placed in registers. The information about what is placed in each register appears in the tables below, in the detailed explanation of each intrinsic. R, R0, R1...R15 represent the registers in which results are placed.（其实就是通过这些指令给寄存器赋值，利用了单指令多数据流的思想）

Intrinsic Name

Operation

Corresponding
Intel® SSE2 Instruction

_mm_set_epi64

Set two integer values

Composite

_mm_set_epi64x

Set two integer values

Composite

_mm_set_epi32

Set four integer values

Composite

_mm_set_epi16

Set eight integer values

Composite

_mm_set_epi8

Set sixteen integer values

Composite

_mm_set1_epi64

Set two integer values

Composite

_mm_set1_epi64x

Set two integer values

Composite

_mm_set1_epi32

Set four integer values

Composite

_mm_set1_epi16

Set eight integer values

Composite

_mm_set1_epi8

Set sixteen integer values

Composite

_mm_setr_epi64

Set two integer values in reverse order

Composite

_mm_setr_epi32

Set four integer values in reverse order

Composite

_mm_setr_epi16

Set eight integer values in reverse order

Composite

_mm_setr_epi8

Set sixteen integer values in reverse order

Composite

_mm_setzero_si128

Set to zero

Composite

__m128i _mm_set_epi64(__m64 q1, __m64 q0)

Sets the 2 64-bit integer values.

R0

R1

q0

q1

__m128i _mm_set_epi64x(__int64 b, __int64 a)

Sets the 2 64-bit integer values.

R0

R1

a

b

__m128i _mm_set_epi32(int i3, int i2, int i1, int i0)

Sets the 4 signed 32-bit integer values.

R0

R1

R2

R3

i0

i1

i2

i3

__m128i _mm_set_epi16(short w7, short w6, short w5, short w4, short w3, short w2, short w1, short w0)

Sets the 8 signed 16-bit integer values.

R0

R1

...

R7

w0

w1

...

w7

__m128i _mm_set_epi8(char b15, char b14, char b13, char b12, char b11, char b10, char b9, char b8, char b7, char b6, char b5, char b4, char b3, char b2, char b1, char b0)

Sets the 16 signed 8-bit integer values.

R0

R1

...

R15

b0

b1

...

b15

__m128i _mm_set1_epi64(__m64 q)

Sets the 2 64-bit integer values to q.

R0

R1

q

q

__m128i _mm_set1_epi64x(__int64 a)

Sets the 2 64-bit integer values to a.

R0

R1

a

a

__m128i _mm_set1_epi32(int i)

Sets the 4 signed 32-bit integer values to i.

R0

R1

R2

R3

i

i

i

i

__m128i _mm_set1_epi16(short w)

Sets the 8 signed 16-bit integer values to w.

R0

R1

...

R7

w

w

w

w

__m128i _mm_set1_epi8(char b)

Sets the 16 signed 8-bit integer values to b.

R0

R1

...

R15

b

b

b

b

__m128i _mm_setr_epi64(__m64 q0, __m64 q1)

Sets the 2 64-bit integer values in reverse order.

R0

R1

q0

q1

__m128i _mm_setr_epi32(int i0, int i1, int i2, int i3)

Sets the 4 signed 32-bit integer values in reverse order.

R0

R1

R2

R3

i0

i1

i2

i3

__m128i _mm_setr_epi16(short w0, short w1, short w2, short w3, short w4, short w5, short w6, short w7)

Sets the 8 signed 16-bit integer values in reverse order.

R0

R1

...

R7

w0

w1

...

w7

__m128i _mm_setr_epi8(char b15, char b14, char b13, char b12, char b11, char b10, char b9, char b8, char b7, char b6, char b5, char b4, char b3, char b2, char b1, char b0)

Sets the 16 signed 8-bit integer values in reverse order.

R0

R1

...

R15

b0

b1

...

b15

__m128i _mm_setzero_si128()

Sets the 128-bit value to zero.

R

0x0

SIMD指令集分析(C/C++)
AAAA202012的博客
04-06 1万+
阅读代码的时候遇到了__m128i、_mm_set1_epi8、_mm_loadu_si128、_mm_max_epu8、_mm_min_epu8、_mm_store_si128、_mm_unpackhi_epi8、_mm_adds_epi16、_mm_srli_si128等SIMD指令集，所以想着作一个总结。 0. SIMD基础知识 SIMD是单指令多数据技术，目前Intel处理器支持的SIMD技术包括MMX、SSE以及AVX。 MMX是MultiMedia eXtensions(多媒体扩展)的缩写，是
_mm_movemask_epi8解析
kupePoem的专栏
03-17 605
通过 _mm_movemask_epi8，开发者可以将 SIMD 并行计算结果快速压缩为标量掩码，从而简化分支处理和后续逻辑，是 SIMD 编程中的关键工具之一。_mm_movemask_epi8 是 Intel SIMD 指令集（SSE/AVX）中的一个重要函数，主要用于从 128 位向量寄存器中提取字节级别的掩码信息。AVX2 提供 _mm256_movemask_epi8，支持 256 位向量（生成 32 位掩码）。
SSE2的一些常用指令集介绍
热门推荐
小白_努力
03-14 1万+
开门见山，前段时间学习OpenCV的FAST算法，中间有很多SSE2的指令集，深受其惑。下面我把学习过程中学到的一些指令集介绍给大家，希望能对大家有所帮助！ __m128i被称为128bits的整数，对其进行赋值时，可以调用__m128i_mm_set1_epi8或者__m128i_mm_set1_epi816等，前者就是把128bits设置为16个8bits的整数值，后一个例子就是把128bi
SSE指令
fyire的博客
10-15 2748
load类 __m128i_mm_loadl_epi64(__m128i const*mem_addr)
AVX-512指令_mm512_shuffle_epi8分析
zzokest的博客
06-12 4452
从Intel网站https://software.intel.com/sites/landingpage/IntrinsicsGuide/#expand=5159,5640,5159&text=_mm512_shuffle_epi8可以得到
SSE指令学习：_mm_shuffle_epi8
Stone_石头
01-14 6949
_mm_setr_epi8：Integer, sets中的函数，初始化赋值。 _mm_shuffle_epi8：第一个参数为输入数组；第二个参数为掩模，对应输入数组的第几个值。 功能：对数据进行重组，非常灵活。当掩模值大于1时，该值除以16求余，如17即为1。当掩模值为-1时对应值为0。 __m128i a= _mm_setr_epi8(1, 2, 3, 4, 5, 6, 7, 8, 9,...
_mm256_set1_epi32和_mm256_set_epi32区别
11-16
`_mm256_set1_epi32`和`_mm256_set_epi32`都是AVX2指令集中的函数，用于设置256位整型向量的值。它们的区别在于： `_mm256_set1_epi32`将一个整数复制到256位向量的所有位置，即所有位置的值都相同。 `_mm256_set_...
_mm256_set_epi32
08-17
`_mm256_set_epi32` 是 Intel Intrinsics 库中的一个函数，用于创建一个包含 8 个 32 位整数的 AVX 寄存器。 它的函数原型如下： ```c __m256i _mm256_set_epi32(int e7, int e6, int e5, int e4, int e3, int e2,...
_mm_cmpeq_epi8
最新发布
06-19
__m128i b = _mm_set_epi8(15, 14, 13, 11, 11, 10, 8, 8, 7, 5, 5, 4, 3, 1, 1, 0); // 使用 _mm_cmpeq_epi8 进行逐元素比较 __m128i result = _mm_cmpeq_epi8(a, b); // 打印结果 for (int i = 0; i ; ++i)...
void YUVToRGB(unsigned char *Y, unsigned char *U, unsigned char *V, unsigned char *RGB, int Width, int Height, int Stride){ const int Shift = 13; const int HalfV = 1 << (Shift - 1); const int B_Y_WT = 1 << Shift, B_U_WT = 2.03211f * (1 << Shift), B_V_WT = 0; const int G_Y_WT = 1 << Shift, G_U_WT = -0.39465f * (1 << Shift), G_V_WT = -0.58060f * (1 << Shift); const int R_Y_WT = 1 << Shift, R_U_WT = 0, R_V_WT = 1.13983 * (1 << Shift); for (int YY = 0; YY < Height; YY++) { unsigned char *LinePD = RGB + YY * Stride; unsigned char *LinePY = Y + YY * Width; unsigned char *LinePU = U + YY * Width; unsigned char *LinePV = V + YY * Width; for (int XX = 0; XX < Width; XX++, LinePD += 3) { int YV = LinePY[XX], UV = LinePU[XX] - 128, VV = LinePV[XX] - 128; LinePD[0] = ClampToByte(YV + ((B_U_WT * UV + HalfV) >> Shift)); LinePD[1] = ClampToByte(YV + ((G_U_WT * UV + G_V_WT * VV + HalfV) >> Shift)); LinePD[2] = ClampToByte(YV + ((R_V_WT * VV + HalfV) >> Shift)); } }}12345678910111213141516171819202122YUV2RGB的SSE初级实现和上面一样，直接翻译为SSE代码，没什么技巧：void YUVToRGBSSE_1(unsigned char *Y, unsigned char *U, unsigned char *V, unsigned char *RGB, int Width, int Height, int Stride) { const int Shift = 13; const int HalfV = 1 << (Shift - 1); const int B_Y_WT = 1 << Shift, B_U_WT = 2.03211f * (1 << Shift), B_V_WT = 0; const int G_Y_WT = 1 << Shift, G_U_WT = -0.39465f * (1 << Shift), G_V_WT = -0.58060f * (1 << Shift); const int R_Y_WT = 1 << Shift, R_U_WT = 0, R_V_WT = 1.13983 * (1 << Shift); __m128i Weight_B_Y = _mm_set1_epi32(B_Y_WT), Weight_B_U = _mm_set1_epi32(B_U_WT), Weight_B_V = _mm_set1_epi32(B_V_WT); __m128i Weight_G_Y = _mm_set1_epi32(G_Y_WT), Weight_G_U = _mm_set1_epi32(G_U_WT), Weight_G_V = _mm_set1_epi32(G_V_WT); __m128i Weight_R_Y = _mm_set1_epi32(R_Y_WT), Weight_R_U = _mm_set1_epi32(R_U_WT), Weight_R_V = _mm_set1_epi32(R_V_WT); __m128i Half = _mm_set1_epi32(HalfV); __m128i C128 = _mm_set1_epi32(128); __m128i Zero = _mm_setzero_si128(); const int BlockSize = 16, Block = Width / BlockSize; for (int YY = 0; YY < Height; YY++) { unsigned char *LinePD = RGB + YY * Stride; unsigned char *LinePY = Y + YY * Width; unsigned char *LinePU = U + YY * Width; unsigned char *LinePV = V + YY * Width; for (int XX = 0; XX < Block * BlockSize; XX += BlockSize, LinePY += BlockSize, LinePU += BlockSize, LinePV += BlockSize) { __m128i Blue, Green, Red, YV, UV, VV, Dest1, Dest2, Dest3; YV = _mm_loadu_si128((__m128i *)(LinePY + 0)); UV = _mm_loadu_si128((__m128i *)(LinePU + 0)); VV = _mm_loadu_si128((__m128i *)(LinePV + 0)); //UV = _mm_sub_epi32(UV, C128); //VV = _mm_sub_epi32(VV, C128); __m128i YV16L = _mm_unpacklo_epi8(YV, Zero); __m128i YV16H = _mm_unpackhi_epi8(YV, Zero); __m128i YV32LL = _mm_unpacklo_epi16(YV16L, Zero); __m128i YV32LH = _mm_unpackhi_epi16(YV16L, Zero); __m128i YV32HL = _mm_unpacklo_epi16(YV16H, Zero); __m128i YV32HH = _mm_unpackhi_epi16(YV16H, Zero); __m128i UV16L = _mm_unpacklo_epi8(UV, Zero); __m128i UV16H = _mm_unpackhi_epi8(UV, Zero); __m128i UV32LL = _mm_unpacklo_epi16(UV16L, Zero); __m128i UV32LH = _mm_unpackhi_epi16(UV16L, Zero); __m128i UV32HL = _mm_unpacklo_epi16(UV16H, Zero); __m128i UV32HH = _mm_unpackhi_epi16(UV16H, Zero); UV32LL = _mm_sub_epi32(UV32LL, C128); UV32LH = _mm_sub_epi32(UV32LH, C128); UV32HL = _mm_sub_epi32(UV32HL, C128); UV32HH = _mm_sub_epi32(UV32HH, C128); __m128i VV16L = _mm_unpacklo_epi8(VV, Zero); __m128i VV16H = _mm_unpackhi_epi8(VV, Zero); __m128i VV32LL = _mm_unpacklo_epi16(VV16L, Zero); __m128i VV32LH = _mm_unpackhi_epi16(VV16L, Zero); __m128i VV32HL = _mm_unpacklo_epi16(VV16H, Zero); __m128i VV32HH = _mm_unpackhi_epi16(VV16H, Zero); VV32LL = _mm_sub_epi32(VV32LL, C128); VV32LH = _mm_sub_epi32(VV32LH, C128); VV32HL = _mm_sub_epi32(VV32HL, C128); VV32HH = _mm_sub_epi32(VV32HH, C128); __m128i LL_B = _mm_add_epi32(YV32LL, _mm_srai_epi32(_mm_add_epi32(Half, _mm_mullo_epi32(UV32LL, Weight_B_U)), Shift)); __m128i LH_B = _mm_add_epi32(YV32LH, _mm_srai_epi32(_mm_add_epi32(Half, _mm_mullo_epi32(UV32LH, Weight_B_U)), Shift)); __m128i HL_B = _mm_add_epi32(YV32HL, _mm_srai_epi32(_mm_add_epi32(Half, _mm_mullo_epi32(UV32HL, Weight_B_U)), Shift)); __m128i HH_B = _mm_add_epi32(YV32HH, _mm_srai_epi32(_mm_add_epi32(Half, _mm_mullo_epi32(UV32HH, Weight_B_U)), Shift)); Blue = _mm_packus_epi16(_mm_packus_epi32(LL_B, LH_B), _mm_packus_epi32(HL_B, HH_B)); __m128i LL_G = _mm_add_epi32(YV32LL, _mm_srai_epi32(_mm_add_epi32(Half, _mm_add_epi32(_mm_mullo_epi32(Weight_G_U, UV32LL), _mm_mullo_epi32(Weight_G_V, VV32LL))), Shift)); __m128i LH_G = _mm_add_epi32(YV32LH, _mm_srai_epi32(_mm_add_epi32(Half, _mm_add_epi32(_mm_mullo_epi32(Weight_G_U, UV32LH), _mm_mullo_epi32(Weight_G_V, VV32LH))), Shift)); __m128i HL_G = _mm_add_epi32(YV32HL, _mm_srai_epi32(_mm_add_epi32(Half, _mm_add_epi32(_mm_mullo_epi32(Weight_G_U, UV32HL), _mm_mullo_epi32(Weight_G_V, VV32HL))), Shift)); __m128i HH_G = _mm_add_epi32(YV32HH, _mm_srai_epi32(_mm_add_epi32(Half, _mm_add_epi32(_mm_mullo_epi32(Weight_G_U, UV32HH), _mm_mullo_epi32(Weight_G_V, VV32HH))), Shift)); Green = _mm_packus_epi16(_mm_packus_epi32(LL_G, LH_G), _mm_packus_epi32(HL_G, HH_G)); __m128i LL_R = _mm_add_epi32(YV32LL, _mm_srai_epi32(_mm_add_epi32(Half, _mm_mullo_epi32(VV32LL, Weight_R_V)), Shift)); __m128i LH_R = _mm_add_epi32(YV32LH, _mm_srai_epi32(_mm_add_epi32(Half, _mm_mullo_epi32(VV32LH, Weight_R_V)), Shift)); __m128i HL_R = _mm_add_epi32(YV32HL, _mm_srai_epi32(_mm_add_epi32(Half, _mm_mullo_epi32(VV32HL, Weight_R_V)), Shift)); __m128i HH_R = _mm_add_epi32(YV32HH, _mm_srai_epi32(_mm_add_epi32(Half, _mm_mullo_epi32(VV32HH, Weight_R_V)), Shift)); Red = _mm_packus_epi16(_mm_packus_epi32(LL_R, LH_R), _mm_packus_epi32(HL_R, HH_R)); Dest1 = _mm_shuffle_epi8(Blue, _mm_setr_epi8(0, -1, -1, 1, -1, -1, 2, -1, -1, 3, -1, -1, 4, -1, -1, 5)); Dest1 = _mm_or_si128(Dest1, _mm_shuffle_epi8(Green, _mm_setr_epi8(-1, 0, -1, -1, 1, -1, -1, 2, -1, -1, 3, -1, -1, 4, -1, -1))); Dest1 = _mm_or_si128(Dest1, _mm_shuffle_epi8(Red, _mm_setr_epi8(-1, -1, 0, -1, -1, 1, -1, -1, 2, -1, -1, 3, -1, -1, 4, -1))); Dest2 = _mm_shuffle_epi8(Blue, _mm_setr_epi8(-1, -1, 6, -1, -1, 7, -1, -1, 8, -1, -1, 9, -1, -1, 10, -1)); Dest2 = _mm_or_si128(Dest2, _mm_shuffle_epi8(Green, _mm_setr_epi8(5, -1, -1, 6, -1, -1, 7, -1, -1, 8, -1, -1, 9, -1, -1, 10))); Dest2 = _mm_or_si128(Dest2, _mm_shuffle_epi8(Red, _mm_setr_epi8(-1, 5, -1, -1, 6, -1, -1, 7, -1, -1, 8, -1, -1, 9, -1, -1))); Dest3 = _mm_shuffle_epi8(Blue, _mm_setr_epi8(-1, 11, -1, -1, 12, -1, -1, 13, -1, -1, 14, -1, -1, 15, -1, -1)); Dest3 = _mm_or_si128(Dest3, _mm_shuffle_epi8(Green, _mm_setr_epi8(-1, -1, 11, -1, -1, 12, -1, -1, 13, -1, -1, 14, -1, -1, 15, -1))); Dest3 = _mm_or_si128(Dest3, _mm_shuffle_epi8(Red, _mm_setr_epi8(10, -1, -1, 11, -1, -1, 12, -1, -1, 13, -1, -1, 14, -1, -1, 15))); _mm_storeu_si128((__m128i*)(LinePD + (XX / BlockSize) * BlockSize * 3), Dest1); _mm_storeu_si128((__m128i*)(LinePD + (XX / BlockSize) * BlockSize * 3 + BlockSize), Dest2); _mm_storeu_si128((__m128i*)(LinePD + (XX / BlockSize) * BlockSize * 3 + BlockSize * 2), Dest3); } for (int XX = Block * BlockSize; XX < Width; XX++, LinePU++, LinePV++, LinePY++) { int YV = LinePY[XX], UV = LinePU[XX] - 128, VV = LinePV[XX] - 128; LinePD[XX + 0] = ClampToByte(YV + ((B_U_WT * UV + HalfV) >> Shift)); LinePD[XX + 1] = ClampToByte(YV + ((G_U_WT * UV + G_V_WT * VV + HalfV) >> Shift)); LinePD[XX + 2] = ClampToByte(YV + ((R_V_WT * VV + HalfV) >> Shift)); } }}
03-10
__m128i v_V = _mm_set1_epi16(src_uv[col + 1]); // Convert from signed shorts back to bytes after processing... // ...and perform the actual color space transformation here... // Store results ...
SIMD and Avx2
TH_NUM的博客
06-07 622
SIMD 一条指令可以执行多个数据group的计算和输出。对于SIMD相对应的SISD. intel SSE2 , AVX2, AVX-512 假设有一个任务是统计字符串中每一个字符出现的次数，我们可以用128bit 的SISD指令进行统计。每8个bit代表一个字符，所以只需要两个SIMD指令（move mask、pop count）。 详细测试: #include <stdio.h> #include <thread> #define INC_TO 1000000 // o
C++基础-------一些基本类型概述
weixin_39609623的博客
08-02 445
开门见山，前段时间学习OpenCV的FAST算法，中间有很多SSE2的指令集，深受其惑。下面我把学习过程中学到的一些指令集介绍给大家，希望能对大家有所帮助！ #include <xmmintrin.h> __m128i被称为128bits的整数，对其进行赋值时，可以调用__m128i_mm_set1_epi8或者__m128i_mm_set1_epi816等，前者就是把12...
SSE一些指令集
huashaopan的博客
09-14 4112
0、变量类型 01. __m128i（以下内容来源于 此处） // data_types__m128i.cpp #include <emmintrin.h> int main() { __m128i x; } SSE2中的指令集指令，定义在 emmintrin.h 中。__m128i 类型变量映射到XMM[0-7]寄存器，它的地址会自动对齐到16字节的边界, 即__m128i类型的变量的起始地址总是16的整数倍。 1、函数接口 01. _mm_load_si128 __m128i _mm
使用AVX-512指令来实现一个比特位反转
zzokest的博客
06-13 2650
项目里面遇到需要把一个45位的比特流按照比特位来反转一下，也就是需要把第45位和第1位互换，第44位和第2位互换，依次类推。数据是用一个u64的整数来存储的，也就是需要把一个u64整数的低45位做一个bi ...
向量化代码实践与思考：如何借助向量化技术给代码提速
阿里技术
08-16 5071
本文将深入向量化计算技术，为大家讲解SIMD指令，以及如何写出规范的可向量化的代码。
AVX/AVX2、内存对齐学习
Vicczyq
05-08 4501
AVX指令函数查询——Linux系统判断自己电脑CPU是否支持AVX和AVX2，可以用lscpu命令查询Windows系统可以用CPU-Z工具进行查询cpp文件需要加上头文件在编译时需要添加后缀-mavx -mavx2。
SSE2指令--使用SSE2的整型内存操作和初始化：设置(Set)
bwl111的专栏
05-27 7776
SSE2指令使用下面三种数据类型： __m128, __m128i, 和 __m128d ，但是安腾处理器家族( Itanium Processor Family,IPF)处理器并不支持这三种数据类型；       x64的处理器不支持任何使用 __m64 数据类型的SSE2指令。      所有的SSE2指令的声明都在emmintrin.h头文件中可以找得到。     1.__m128
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值