xsimd

C++ wrappers for SIMD intrinsics

Introduction

SIMD (Single Instruction, Multiple Data) is a feature of microprocessors that has been available for many years. SIMD instructions perform a single operation on a batch of values at once, and thus provide a way to significantly accelerate code execution. However, these instructions differ between microprocessor vendors and compilers.

xsimd provides a unified means for using these features for library authors. Namely, it enables manipulation of batches of numbers with the same arithmetic operators as for single values. It also provides accelerated implementation of common mathematical functions operating on batches.

CI status

Workflow	Status
android
cross-rvv
cross-sve
cross
cxx-no-exceptions
cxx-versions
emscripten
linux
macos
windows

Adoption

Beyond Xtensor, Xsimd has been adopted by major open-source projects, such as:

• Mozilla Firefox - web browser,
• Apache Arrow - cross-language framework for tabular data management,
• Pandas - Python data analysis library,
• KDE Krita - digital painting and illustration application,
• Meta Velox - a composable execution engine,
• Milvus - vector database for AI applications,
• Pythran - Python-to-C++ compiler for numerical computing.

History

The XSimd project started with a series of blog articles by Johan Mabille on how to implement wrappers for SIMD intrinsicts. The archives of the blog can be found here: The C++ Scientist. The design described in the articles remained close to the actual architecture of XSimd up until Version 8.0.

The mathematical functions are a lightweight implementation of the algorithms originally implemented in the now deprecated boost.SIMD project.

Requirements

xsimd builds with any C++17 compliant compiler. Official support covers the compilers tested in the CI:

Compiler	Version
Microsoft Visual Studio	MSVC 2022 and above
g++	10 and above
clang	16 and above

The following SIMD instruction set extensions are supported:

Architecture	Instruction set extensions
x86	SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, AVX, AVX2, FMA3+SSE, FMA3+AVX, FMA3+AVX2
x86	AVX512BW, AVX512CD, AVX512DQ, AVX512F, AVX512VL
x86 AMD	FMA4
ARM	NEON, NEON64, SVE128/256/512 (fixed vector size)
WebAssembly	WASM
powerpc64	VSX
RISC-V	RISC-V128/256/512 (fixed vector size)
LoongArch64	LSX, LASX
IBM Z (s390x)	VXE (IBM z14)

Installation

Install from conda-forge

A package for xsimd is available on the mamba (or conda) package manager.

mamba install -c conda-forge xsimd

Install with Spack

A package for xsimd is available on the Spack package manager.

spack install xsimd
spack load xsimd

Install from sources

You can directly install it from the sources with cmake:

cmake -D CMAKE_INSTALL_PREFIX=your_install_prefix .
make install

Documentation

To get started with using xsimd, check out the full documentation

http://xsimd.readthedocs.io/

Dependencies

xsimd has an optional dependency on the xtl library:

xsimd	xtl (optional)
master	^0.7.0
12.x	^0.7.0
11.x	^0.7.0
10.x	^0.7.0
9.x	^0.7.0
8.x	^0.7.0

The dependency on xtl is required if you want to support vectorization for xtl::xcomplex.

Usage

The version 8 of the library is a complete rewrite and there are some slight differences with 7.x versions. A migration guide will be available soon. In the meanwhile, the following examples show how to use both versions 7 and 8 of the library?

Explicit use of an instruction set extension

Here is an example that computes the mean of two sets of 4 double floating point values, assuming AVX extension is supported:

#include <iostream>
#include "xsimd/xsimd.hpp"

namespace xs = xsimd;

int main(int argc, char* argv[])
{
    xs::batch<double, xs::avx2> a = {1.5, 2.5, 3.5, 4.5};
    xs::batch<double, xs::avx2> b = {2.5, 3.5, 4.5, 5.5};
    auto mean = (a + b) / 2;
    std::cout << mean << std::endl;
    return 0;
}

Do not forget to enable AVX extension when building the example. With gcc or clang, this is done with the -mavx flag, on MSVC you have to pass the /arch:AVX option.

This example outputs:

(2.0, 3.0, 4.0, 5.0)

Auto detection of the instruction set extension to be used

The same computation operating on vectors and using the most performant instruction set available at compile time, based on the provided compiler flags (e.g. -mavx2 for GCC and Clang to target AVX2):

#include <cstddef>
#include <vector>
#include "xsimd/xsimd.hpp"

namespace xs = xsimd;
using vector_type = std::vector<double, xsimd::aligned_allocator<double>>;

void mean(const vector_type& a, const vector_type& b, vector_type& res)
{
    std::size_t size = a.size();
    constexpr std::size_t simd_size = xsimd::simd_type<double>::size;
    std::size_t vec_size = size - size % simd_size;

    for(std::size_t i = 0; i < vec_size; i += simd_size)
    {
        auto ba = xs::load_aligned(&a[i]);
        auto bb = xs::load_aligned(&b[i]);
        auto bres = (ba + bb) / 2.;
        bres.store_aligned(&res[i]);
    }
    for(std::size_t i = vec_size; i < size; ++i)
    {
        res[i] = (a[i] + b[i]) / 2.;
    }
}

Building and Running the Tests

Building the tests requires cmake.

cmake is available as a package for most linux distributions. Besides, they can also be installed with the conda package manager (even on windows):

conda install -c conda-forge cmake

Once cmake is installed, you can build and run the tests:

mkdir build
cd build
cmake ../ -DBUILD_TESTS=ON
make xtest

In the context of continuous integration with Travis CI, tests are run in a conda environment, which can be activated with

cd test
conda env create -f ./test-environment.yml
source activate test-xsimd
cd ..
cmake . -DBUILD_TESTS=ON
make xtest

Building the HTML Documentation

xsimd’s documentation is built with three tools

• doxygen
• sphinx
• breathe

While doxygen must be installed separately, you can install breathe by typing

pip install breathe

Breathe can also be installed with conda

conda install -c conda-forge breathe

Finally, build the documentation with

make html

from the docs subdirectory.

License

We use a shared copyright model that enables all contributors to maintain the copyright on their contributions.

This software is licensed under the BSD-3-Clause license. See the LICENSE file for details.