AVX / AVX2 Intrinsics Example Code

• AVX / AVX2 Intrinsics Example Code
  • Quick Start
    • Compile
    • Run
    • Clean
  • Initialization Intrinsics
    • Initialization with Scalar Values
    • Loading Data to Memory
  • Arithmetic Intrinsics
    • Addition and Subtraction
    • Multiplication and Division
    • Fused Multiply and Add (FMA)
  • Permuting and Shuffling
    • Permuting
  • Copyright

Quick Start

Compile

$ make

All the source files in src/ will be compiled and generate binary files to the bin/ in each subdirectory.

Run

Fast compile & run at one time!

Execute this command:

$ make run

At the project root directory, then you can see all the program output on your terminal :tada:

Clean

It's easy to clean all the output files, just enter the following command at the project root directory:

$ make clean

Then you would find out all the output files are gone away!

Initialization Intrinsics

Initialization with Scalar Values

• setzero

  • _mm256_setzero_ps
  • _mm256_setzero_pd
  • _mm256_setzero_si256

• set1

  • _mm256_set1_ps
  • _mm256_set1_pd
  • _mm256_set1_epi32
  • _mm256_set1_epi64x
  • _mm256_set1_epi16
  • _mm256_set1_epi8

• set

  • _mm256_set_ps
  • _mm256_set_pd
  • _mm256_set_epi32
  • _mm256_set_epi64x
  • _mm256_set_epi16
  • _mm256_set_epi8
  • _mm256_set_m128
  • _mm256_set_m128d
  • _mm256_set_m128i

• setr

  • _mm256_setr_ps
  • _mm256_setr_pd
  • _mm256_setr_epi32
  • _mm256_setr_epi64x
  • _mm256_setr_epi16
  • _mm256_setr_epi8
  • _mm256_setr_m128
  • _mm256_setr_m128d
  • _mm256_setr_m128i

Loading Data to Memory

• load

  • _mm256_load_ps
  • _mm256_load_pd
  • _mm256_load_si256

• loadu

  • _mm256_loadu_ps
  • _mm256_loadu_pd
  • _mm256_loadu_si256

• maskload

  • _mm_maskload_ps
  • _mm_maskload_pd
  • _mm256_maskload_ps
  • _mm256_maskload_pd
  • _mm_maskload_epi32 AVX2
  • _mm_maskload_epi64 AVX2
  • _mm256_maskload_epi32 AVX2
  • _mm256_maskload_epi64 AVX2

Arithmetic Intrinsics

Addition and Subtraction

• add

  • _mm256_add_ps
  • _mm256_add_pd
  • _mm256_add_epi8 AVX2
  • _mm256_add_epi16 AVX2
  • _mm256_add_epi32 AVX2
  • _mm256_add_epi64 AVX2

• sub

  • _mm256_sub_ps
  • _mm256_sub_pd
  • _mm256_sub_epi8 AVX2
  • _mm256_sub_epi16 AVX2
  • _mm256_sub_epi32 AVX2
  • _mm256_sub_epi64 AVX2

• adds

  • _mm256_adds_epi8 AVX2
  • _mm256_adds_epi16 AVX2
  • _mm256_adds_epu8 AVX2
  • _mm256_adds_epu16 AVX2

• subs

  • _mm256_subs_epi8 AVX2
  • _mm256_subs_epi16 AVX2
  • _mm256_subs_epu8 AVX2
  • _mm256_subs_epu16 AVX2

• hadd

  • _mm256_hadd_ps
  • _mm256_hadd_pd
  • _mm256_hadd_epi16 AVX2
  • _mm256_hadd_epi32 AVX2

• hsub

  • _mm256_hadd_ps
  • _mm256_hadd_pd
  • _mm256_hadd_epi16 AVX2
  • _mm256_hadd_epi32 AVX2

• hadds

  • _mm256_hadds_epi16 AVX2

• hsubs

  • _mm256_hsubs_epi16 AVX2

• addsub

  • _mm256_addsub_ps
  • _mm256_addsub_pd

Multiplication and Division

• mul

  • _mm256_mul_ps
  • _mm256_mul_pd
  • _mm256_mul_epi32 AVX2
  • _mm256_mul_epu32 AVX2

• mullo

  • _mm256_mullo_epi16 AVX2
  • _mm256_mullo_epi32 AVX2

• mulhi

  • _mm256_mulhi_epi16 AVX2
  • _mm256_mulhi_epu16 AVX2

• mulhrs

  • _mm256_mulhrs_epi16 AVX2

• div

  • _mm256_div_ps
  • _mm256_div_pd

Fused Multiply and Add (FMA)

• fmadd

  • _mm_fmadd_ps FMA
  • _mm_fmadd_pd FMA
  • _mm256_fmadd_ps FMA
  • _mm256_fmadd_pd FMA
  • _mm_fmadd_ss FMA
  • _mm_fmadd_sd FMA

• fmsub

  • _mm_fmsub_ps FMA
  • _mm_fmsub_pd FMA
  • _mm256_fmsub_ps FMA
  • _mm256_fmsub_pd FMA
  • _mm_fmsub_ss FMA
  • _mm_fmsub_sd FMA

• fnmadd

  • _mm_fnmadd_ps FMA
  • _mm_fnmadd_pd FMA
  • _mm256_fnmadd_ps FMA
  • _mm256_fnmadd_pd FMA
  • _mm_fnmadd_ss FMA
  • _mm_fnmadd_sd FMA

• fnmsub

  • _mm_fnmsub_ps FMA
  • _mm_fnmsub_pd FMA
  • _mm256_fnmsub_ps FMA
  • _mm256_fnmsub_pd FMA
  • _mm_fnmsub_ss FMA
  • _mm_fnmsub_sd FMA

• fmaddsub

  • _mm_fmaddsub_ps FMA
  • _mm_fmaddsub_pd FMA
  • _mm256_fmaddsub_ps FMA
  • _mm256_fmaddsub_pd FMA

• fmsubadd

  • _mm_fmsubadd_ps FMA
  • _mm_fmsubadd_pd FMA
  • _mm256_fmsubadd_ps FMA
  • _mm256_fmsubadd_pd FMA

Permuting and Shuffling

Permuting

• permute

  • _mm_permute_ps
  • _mm_permute_pd
  • _mm256_permute_ps
  • _mm256_permute_pd

• permute4x64

  • _mm256_permute4x64_pd AVX2
  • _mm256_permute4x64_epi64 AVX2

• permute2f128

  • _mm256_permute2f128_ps
  • _mm256_permute2f128_pd
  • _mm256_permute2f128_si256

• permutevar

  • _mm_permutevar_ps
  • _mm_permutevar_pd
  • _mm256_permutevar_ps
  • _mm256_permutevar_pd

• permutevar8x32

  • _mm256_permutevar8x32_ps AVX2
  • _mm256_permutevar8x32_epi32 AVX2

• shuffle

  • _mm256_shuffle_ps
  • _mm256_shuffle_pd
  • _mm256_shuffle_epi32
  • _mm256_shuffle_epi8

• shufflehi

  • _mm256_shufflehi_epi16 AVX2

• shufflelo

  • _mm256_shufflelo_epi16 AVX2

Copyright

This project is licensed under the BSD 3-Clause license.