Eigen RVV

Eigen 简介

Eigen 是一个高性能的 C++ 模板库，专用于线性代数运算，涵盖矩阵、向量、数值求解及相关算法。它被广泛应用于机器人、计算机视觉、数值仿真等领域，是 ROS、OpenCV、PCL 等主流框架的核心依赖之一。

主要特点：

• 纯头文件：无需编译，直接 #include 即可使用
• 高性能：通过表达式模板（expression templates）消除临时变量，自动向量化（SIMD）
• 功能丰富：支持稠密矩阵、稀疏矩阵、几何变换（旋转矩阵、四元数）、线性求解器等
• 跨平台：支持 x86、ARM、RISC-V 等主流架构

RVV 加速

RVV（RISC-V Vector Extension）是 RISC-V 架构的向量扩展指令集，可对矩阵运算进行 SIMD 加速。K3 搭载的 SpacemiT X100 + A100 16 核处理器原生支持 RVV 1.0（含 zve64d、zvfh 等向量子扩展）。

针对 K3 平台，我们提供了专为 RVV 优化的定制 Eigen 库 eigen-spacemit，相比系统默认的 Eigen 有显著差异：

对比项	系统 Eigen（libeigen3-dev）	eigen-spacemit
安装路径	/usr/include/eigen3	/opt/eigen-spacemit
RVV 加速	未启用	针对 RVV 1.0 优化
适用平台	通用（x86/ARM/RISC-V）	专为 SpacemiT X100 定制
安装方式	apt install libeigen3-dev	apt install eigen-spacemit

在 K3 上进行高性能线性代数运算（如点云处理、SLAM、卡尔曼滤波）时，推荐优先使用 eigen-spacemit 以充分发挥 RVV 硬件能力。

使用示例

以下示例对比系统 Eigen 与 eigen-spacemit 在 512×512 矩阵乘法上的性能差异。

安装必要依赖

sudo apt update
sudo apt install eigen-spacemit libeigen3-dev

目录结构：

eigen_bench/
├── CMakeLists.txt
└── bench.cpp

CMakeLists.txt：

cmake_minimum_required(VERSION 3.16)
project(eigen_bench LANGUAGES CXX)

# OpenMP 让 Eigen 支持的单个算子（主要是大矩阵乘法）可以使用多线程。
option(EIGEN_BENCH_ENABLE_OPENMP "Enable OpenMP so Eigen can parallelize supported operators" ON)

# RVV 编译参数。
set(RVV_MARCH "rv64gcv_zvl256b" CACHE STRING "RISC-V architecture flags for RVV builds")
set(RVV_MABI  "lp64d"           CACHE STRING "RISC-V ABI flags for RVV builds")

# 系统 Eigen 对照组保持无 RVV，需显式指定标量目标架构和 ABI。
set(SCALAR_MARCH "rv64gc" CACHE STRING "RISC-V architecture flags for scalar non-RVV builds")
set(SCALAR_MABI  "lp64d"  CACHE STRING "RISC-V ABI flags for scalar non-RVV builds")

if(EIGEN_BENCH_ENABLE_OPENMP)
  find_package(OpenMP)
endif()

# 统一创建 benchmark 目标：传入不同 Eigen include 路径即可对比不同实现。
function(add_bench target eigen_include)
  add_executable(${target} bench.cpp)
  target_include_directories(${target} PRIVATE "${eigen_include}")
  target_compile_options(${target} PRIVATE -O3 -DNDEBUG -std=c++17)
  if(EIGEN_BENCH_ENABLE_OPENMP AND OpenMP_CXX_FOUND)
    target_link_libraries(${target} PRIVATE OpenMP::OpenMP_CXX)
  endif()
endfunction()

# SpacemiT Eigen（RVV 加速）
add_bench(bench_spacemit /opt/eigen-spacemit/include/eigen3)
target_compile_options(bench_spacemit PRIVATE -march=${RVV_MARCH} -mabi=${RVV_MABI} -mrvv-vector-bits=zvl)
target_compile_definitions(bench_spacemit PRIVATE EIGEN_RISCV64_USE_RVV10)

# 系统 Eigen（无 RVV）
add_bench(bench_system /usr/include/eigen3)
target_compile_options(bench_system PRIVATE -march=${SCALAR_MARCH} -mabi=${SCALAR_MABI})

bench.cpp：

#include <Eigen/Dense>
#include <chrono>
#include <cstdlib>
#include <iostream>
#include <stdexcept>
#include <string>

namespace {

int parsePositiveInt(const char* value, const std::string& name) {
    const int parsed = std::stoi(value);
    if (parsed <= 0) {
        throw std::runtime_error(name + " must be positive");
    }
    return parsed;
}

std::string envOrUnset(const char* name) {
    const char* value = std::getenv(name);
    return value ? value : "not_set";
}

}  // namespace

int main(int argc, char** argv) {
    try {
    const int N = argc > 1 ? parsePositiveInt(argv[1], "matrix_size") : 512;
    const int iters = argc > 2 ? parsePositiveInt(argv[2], "iterations") : 10;
    const char* threads_env = std::getenv("EIGEN_BENCH_THREADS");
    const int threads = argc > 3 ? parsePositiveInt(argv[3], "threads") : (threads_env ? parsePositiveInt(threads_env, "EIGEN_BENCH_THREADS") : 8);
    Eigen::setNbThreads(threads);

    Eigen::MatrixXf A = Eigen::MatrixXf::Random(N, N);
    Eigen::MatrixXf B = Eigen::MatrixXf::Random(N, N);
    Eigen::MatrixXf C(N, N);

    C.noalias() = A * B; // warmup

    auto t0 = std::chrono::high_resolution_clock::now();
    for (int i = 0; i < iters; ++i)
        C.noalias() = A * B;
    auto t1 = std::chrono::high_resolution_clock::now();

    double ms = std::chrono::duration<double, std::milli>(t1 - t0).count() / iters;
    std::cout << "Eigen version: " << EIGEN_WORLD_VERSION << '.' << EIGEN_MAJOR_VERSION << '.' << EIGEN_MINOR_VERSION << '\n';
    std::cout << "SimdInstructionSetsInUse(): " << Eigen::SimdInstructionSetsInUse() << '\n';
#ifdef EIGEN_HAS_OPENMP
    std::cout << "EIGEN_HAS_OPENMP: defined\n";
#else
    std::cout << "EIGEN_HAS_OPENMP: not_defined\n";
#endif
    std::cout << "matrix_size: " << N << '\n';
    std::cout << "iterations: " << iters << '\n';
    std::cout << "requested_threads: " << threads << '\n';
    std::cout << "Eigen nbThreads: " << Eigen::nbThreads() << '\n';
    std::cout << "OMP_NUM_THREADS: " << envOrUnset("OMP_NUM_THREADS") << '\n';
    std::cout << "EIGEN_BENCH_THREADS: " << envOrUnset("EIGEN_BENCH_THREADS") << '\n';
    std::cout << N << "x" << N << " matmul avg: " << ms << " ms\n";
    return 0;
    } catch (const std::exception& error) {
        std::cerr << "error: " << error.what() << '\n';
        return 1;
    }
}

编译与运行：

mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make -j$(nproc)

./bench_system    # 系统 Eigen
./bench_spacemit  # SpacemiT Eigen（RVV）

单核对比

可以看出，RVV加速后的矩阵乘法性能有明显提升（77 ms -> 22 ms）

八核对比

使用 openblas-spacemit 作为后端

Eigen 默认使用自身的模板实现完成矩阵运算，而 OpenBLAS 是一个经过高度优化的 BLAS（Basic Linear Algebra Subprograms）数值库。通过在编译时添加 -DEIGEN_USE_BLAS 宏，Eigen 会将底层的矩阵乘法（gemm）等关键运算委托给 OpenBLAS 执行，从而利用其针对特定硬件深度调优的实现。

openblas-spacemit 是专为 SpacemiT K3 平台定制的 OpenBLAS 版本，充分利用了 RVV 1.0 向量扩展及平台特性。与单独使用 eigen-spacemit（RVV 向量化）相比，通过 openblas-spacemit 后端可以进一步将大规模矩阵乘法性能提升。

安装必要依赖

sudo apt install openblas-spacemit

编译与运行

仍使用上一节的bench.cpp

export LD_LIBRARY_PATH=/opt/openblas-spacemit/lib:$LD_LIBRARY_PATH
g++ -O3 -DNDEBUG -std=c++17 \
	-fopenmp \
    -march=rv64gcv_zvl256b -mabi=lp64d -mrvv-vector-bits=zvl \
    -DEIGEN_RISCV64_USE_RVV10 -DEIGEN_USE_BLAS \
    -I"/opt/eigen-spacemit/include/eigen3" \
    "bench.cpp" \
    "/opt/openblas-spacemit/lib/libopenblas.so" -lpthread \
    -o /tmp/eigen_bench_blas

/tmp/eigen_bench_blas

单核性能：

使用 openblas-spacemit 作为 eigen 后端，矩阵乘法的性能还有进一步的提升（22 ms -> 9.48 ms）

八核性能：

更多性能测试数据

测试环境

• SpacemiT RISCV64 X100 CPU（2.4Ghz）
• Bianbu 4.0.1 操作系统
• 内存：32GB

多核测试说明

• 多核测试编译需要 target_link_libraries(${target_name} PRIVATE OpenMP::OpenMP_CXX)
• 多核测试代码需要 Eigen::setNbThreads(threads)，threads 建议与测试时限制的核数一致
• 以下测试 threads 与核数均取一致

性能下降情况说明

后续表格中 speedup 小于 1 或 improvement 为负数表示对比方案性能低于基准方案。需要注意的是，这类结果主要集中在以下几类场景：

• 极小规模几何运算：如 Quaternion::inverse()、Quaternion::normalize()、Quaternion::toRotationMatrix()、cross() 等输入规模仅为 3x1 或 3x3，单次耗时约 0.0001 ms，已经接近计时分辨率与函数调用开销，少量负收益更多体现为测试抖动或额外分支/向量化开销，不代表大规模计算性能退化。
• 低计算密度或访存主导算子：如 trace()、部分 Identity()、block()、transpose()、rowwise().sum() 等，计算量较小或内存访问占主导，RVV/BLAS 后端的调度、打包、线程同步等开销可能抵消向量化收益。
• 使用 BLAS 后端时的非 GEMM 算子：EIGEN_USE_BLAS 主要加速大矩阵乘法等 BLAS Level-3 操作；对于大量数组逐元素运算、标量运算、归约或小矩阵操作，调用 BLAS 后端并不会带来直接收益，部分场景会出现 1%～3% 左右的小幅下降，通常属于开销差异或测试波动。
• 多核小算子并行开销：在四核、八核测试中，部分单次计算量较小的算子会受到线程调度、同步和缓存竞争影响，出现比单核更明显的负收益。例如 BLAS 后端对比中，matrix_gemv、matrix_identity 等在多核下存在下降，但大矩阵乘法仍保持显著提升。

因此，性能下降项应结合算子类型、输入规模与标准差一起判断。实际应用中，建议优先将 eigen-spacemit 或 openblas-spacemit 用于矩阵乘法、矩阵分解、求逆、归约等计算密集型任务；对于小尺寸几何操作或访存主导算子，可根据端到端业务耗时进行取舍。

eigen-spacemit VS libeigen3-dev

• 使用 taskset -c 限制核数
• 单次测试的计时策略是每个算子跑 50 次取平均值，预热10次。本测试重复10次单次测试，取其指标的平均值并计算标准差
• mean 表示 10 次测试的均值，sd 为标准差

单核对比

taskset -c 0

threads：1

module	function	api	dtype	input_size	output_size	libeigen3-dev avg_ms mean±sd	eigen-spacemit avg_ms mean±sd	libeigen3-dev avg_gflops mean±sd	eigen-spacemit avg_gflops mean±sd	speedup mean±sd	improvement mean±sd
array	vector_abs	abs(a - scalar)	float32	262144x1	262144x1	0.6099 ± 0.0158	0.2440 ± 0.0056	0.4301 ± 0.0111	1.0751 ± 0.0245	2.50x ± 0.01x	149.99% ± 1.37%
array	vector_abs	abs(a - scalar)	float64	262144x1	262144x1	0.8122 ± 0.0206	0.6826 ± 0.0123	0.3229 ± 0.0079	0.3842 ± 0.0069	1.19x ± 0.03x	19.00% ± 2.58%
array	vector_add_scalar	a + scalar	float32	262144x1	262144x1	0.6101 ± 0.0149	0.1996 ± 0.0070	0.4299 ± 0.0104	1.3147 ± 0.0447	3.06x ± 0.05x	205.77% ± 5.15%
array	vector_add_scalar	a + scalar	float64	262144x1	262144x1	0.7155 ± 0.0168	0.6065 ± 0.0099	0.3666 ± 0.0085	0.4324 ± 0.0070	1.18x ± 0.02x	17.99% ± 2.41%
array	vector_add_vector	a + b	float32	262144x1	262144x1	0.7388 ± 0.0191	0.2521 ± 0.0173	0.3550 ± 0.0091	1.0443 ± 0.0719	2.94x ± 0.21x	194.30% ± 21.45%
array	vector_add_vector	a + b	float64	262144x1	262144x1	1.0307 ± 0.0347	0.8972 ± 0.0215	0.2546 ± 0.0086	0.2923 ± 0.0069	1.15x ± 0.04x	14.91% ± 4.05%
array	vector_clamp	cwiseMax + cwiseMin	float32	262144x1	262144x1	0.7360 ± 0.0175	0.4418 ± 0.0107	0.7127 ± 0.0168	1.1872 ± 0.0284	1.67x ± 0.00x	66.58% ± 0.49%
array	vector_clamp	cwiseMax + cwiseMin	float64	262144x1	262144x1	0.8870 ± 0.0165	1.0334 ± 0.0201	0.5913 ± 0.0110	0.5075 ± 0.0098	0.86x ± 0.01x	-14.17% ± 0.62%
array	vector_cos	cos(a)	float32	262144x1	262144x1	5.6887 ± 0.1369	2.0954 ± 0.0505	0.0461 ± 0.0011	0.1252 ± 0.0030	2.71x ± 0.01x	171.49% ± 0.85%
array	vector_cos	cos(a)	float64	262144x1	262144x1	10.9408 ± 0.2610	10.7751 ± 0.2532	0.0240 ± 0.0006	0.0244 ± 0.0006	1.02x ± 0.00x	1.54% ± 0.16%
array	vector_div_scalar	a / scalar	float32	262144x1	262144x1	1.5860 ± 0.0382	0.6609 ± 0.0243	0.1654 ± 0.0039	0.3972 ± 0.0142	2.40x ± 0.04x	140.11% ± 4.25%
array	vector_div_scalar	a / scalar	float64	262144x1	262144x1	2.6190 ± 0.0821	2.3817 ± 0.0522	0.1002 ± 0.0031	0.1101 ± 0.0024	1.10x ± 0.02x	9.95% ± 1.50%
array	vector_div_vector	a / (b + scalar)	float32	262144x1	262144x1	1.5903 ± 0.0384	0.7656 ± 0.0279	0.3299 ± 0.0079	0.6856 ± 0.0243	2.08x ± 0.05x	107.85% ± 5.38%
array	vector_div_vector	a / (b + scalar)	float64	262144x1	262144x1	2.6453 ± 0.0850	2.6033 ± 0.0547	0.1984 ± 0.0063	0.2015 ± 0.0042	1.02x ± 0.02x	1.60% ± 1.93%
array	vector_dot	dot()	float32	262144x1	1x1	0.3717 ± 0.0184	0.0981 ± 0.0095	1.4133 ± 0.0669	5.3871 ± 0.5120	3.82x ± 0.37x	281.66% ± 37.21%
array	vector_dot	dot()	float64	262144x1	1x1	0.5054 ± 0.0402	0.4009 ± 0.0175	1.0430 ± 0.0786	1.3098 ± 0.0542	1.26x ± 0.10x	26.15% ± 9.90%
array	vector_exp	exp(a * scalar)	float32	262144x1	262144x1	6.7104 ± 0.1583	2.1390 ± 0.0716	0.0782 ± 0.0018	0.2454 ± 0.0080	3.14x ± 0.05x	213.83% ± 4.63%
array	vector_exp	exp(a * scalar)	float64	262144x1	262144x1	7.0105 ± 0.1669	6.8168 ± 0.1582	0.0748 ± 0.0018	0.0770 ± 0.0018	1.03x ± 0.00x	2.84% ± 0.25%
array	vector_fma	a * b + c	float32	262144x1	262144x1	1.2195 ± 0.0269	0.4835 ± 0.0119	0.4301 ± 0.0094	1.0849 ± 0.0266	2.52x ± 0.08x	152.33% ± 7.97%
array	vector_fma	a * b + c	float64	262144x1	262144x1	1.4908 ± 0.0232	1.2728 ± 0.0207	0.3518 ± 0.0056	0.4120 ± 0.0066	1.17x ± 0.03x	17.16% ± 2.95%
array	vector_log	log(a + scalar)	float32	262144x1	262144x1	6.4867 ± 0.2051	2.0757 ± 0.0702	0.0809 ± 0.0025	0.2529 ± 0.0083	3.13x ± 0.06x	212.58% ± 6.31%
array	vector_log	log(a + scalar)	float64	262144x1	262144x1	7.3730 ± 0.2169	6.1661 ± 0.1434	0.0712 ± 0.0020	0.0851 ± 0.0020	1.20x ± 0.01x	19.56% ± 1.13%
array	vector_max	max(a, b)	float32	262144x1	262144x1	0.7613 ± 0.0424	0.3161 ± 0.0143	0.3452 ± 0.0184	0.8309 ± 0.0382	2.41x ± 0.15x	141.16% ± 15.07%
array	vector_max	max(a, b)	float64	262144x1	262144x1	1.0221 ± 0.0252	0.9173 ± 0.0252	0.2566 ± 0.0061	0.2860 ± 0.0077	1.11x ± 0.04x	11.49% ± 3.59%
array	vector_min	min(a, b)	float32	262144x1	262144x1	0.8580 ± 0.0196	0.3150 ± 0.0145	0.3057 ± 0.0070	0.8340 ± 0.0387	2.73x ± 0.10x	172.79% ± 9.95%
array	vector_min	min(a, b)	float64	262144x1	262144x1	1.0048 ± 0.0185	0.9050 ± 0.0261	0.2610 ± 0.0048	0.2899 ± 0.0081	1.11x ± 0.03x	11.08% ± 2.72%
array	vector_mul_scalar	a * scalar	float32	262144x1	262144x1	0.7347 ± 0.0213	0.2149 ± 0.0076	0.3571 ± 0.0101	1.2210 ± 0.0418	3.42x ± 0.07x	241.96% ± 6.92%
array	vector_mul_scalar	a * scalar	float64	262144x1	262144x1	0.7132 ± 0.0163	0.6299 ± 0.0115	0.3678 ± 0.0083	0.4163 ± 0.0076	1.13x ± 0.02x	13.22% ± 1.93%
array	vector_mul_vector	a * b	float32	262144x1	262144x1	0.7392 ± 0.0186	0.2642 ± 0.0190	0.3548 ± 0.0089	0.9968 ± 0.0721	2.81x ± 0.21x	181.07% ± 21.30%
array	vector_mul_vector	a * b	float64	262144x1	262144x1	1.0304 ± 0.0253	0.9050 ± 0.0334	0.2546 ± 0.0063	0.2900 ± 0.0106	1.14x ± 0.05x	13.99% ± 4.89%
array	vector_norm	norm()	float32	262144x1	1x1	0.3660 ± 0.0096	0.0841 ± 0.0023	1.4333 ± 0.0373	6.2364 ± 0.1713	4.35x ± 0.03x	335.21% ± 2.91%
array	vector_norm	norm()	float64	262144x1	1x1	0.3664 ± 0.0091	0.1884 ± 0.0090	1.4316 ± 0.0351	2.7879 ± 0.1301	1.95x ± 0.06x	94.68% ± 6.04%
array	vector_normalize	normalize()	float32	262144x1	262144x1	2.1937 ± 0.0534	0.8286 ± 0.0203	0.3587 ± 0.0086	0.9496 ± 0.0231	2.65x ± 0.01x	164.76% ± 0.75%
array	vector_normalize	normalize()	float64	262144x1	262144x1	3.2807 ± 0.0731	2.7954 ± 0.1001	0.2398 ± 0.0053	0.2816 ± 0.0098	1.17x ± 0.02x	17.42% ± 1.96%
array	vector_segment	segment()	float32	262144x1	131072x1	0.3051 ± 0.0075	0.0634 ± 0.0016	0.0000 ± 0.0000	0.0000 ± 0.0000	4.81x ± 0.04x	381.18% ± 3.64%
array	vector_segment	segment()	float64	262144x1	131072x1	0.3085 ± 0.0087	0.1298 ± 0.0076	0.0000 ± 0.0000	0.0000 ± 0.0000	2.38x ± 0.08x	138.05% ± 7.98%
array	vector_select	mask.select(a, b)	float32	262144x1	262144x1	0.8723 ± 0.0534	0.4859 ± 0.0146	0.3016 ± 0.0184	0.5400 ± 0.0162	1.80x ± 0.10x	79.55% ± 9.87%
array	vector_select	mask.select(a, b)	float64	262144x1	262144x1	0.8716 ± 0.0160	1.1196 ± 0.0231	0.3009 ± 0.0056	0.2342 ± 0.0048	0.78x ± 0.01x	-22.14% ± 1.45%
array	vector_sin	sin(a)	float32	262144x1	262144x1	5.6152 ± 0.1374	2.1254 ± 0.0533	0.0467 ± 0.0011	0.1234 ± 0.0030	2.64x ± 0.01x	164.20% ± 0.94%
array	vector_sin	sin(a)	float64	262144x1	262144x1	11.1028 ± 0.2664	10.9457 ± 0.3047	0.0236 ± 0.0006	0.0240 ± 0.0007	1.01x ± 0.01x	1.45% ± 0.77%
array	vector_sqrt	(a + scalar).sqrt()	float32	262144x1	262144x1	5.3624 ± 0.1308	0.6281 ± 0.0189	0.0978 ± 0.0024	0.8354 ± 0.0247	8.54x ± 0.09x	753.91% ± 8.62%
array	vector_sqrt	(a + scalar).sqrt()	float64	262144x1	262144x1	6.4080 ± 0.1798	2.3226 ± 0.0505	0.0819 ± 0.0023	0.2258 ± 0.0049	2.76x ± 0.03x	175.88% ± 2.78%
array	vector_sub_scalar	a - scalar	float32	262144x1	262144x1	0.6099 ± 0.0147	0.1990 ± 0.0070	0.4300 ± 0.0103	1.3190 ± 0.0450	3.07x ± 0.05x	206.68% ± 4.71%
array	vector_sub_scalar	a - scalar	float64	262144x1	262144x1	0.7140 ± 0.0164	0.6069 ± 0.0120	0.3673 ± 0.0083	0.4321 ± 0.0085	1.18x ± 0.02x	17.66% ± 2.20%
array	vector_sub_vector	a - b	float32	262144x1	262144x1	0.7392 ± 0.0189	0.2508 ± 0.0181	0.3549 ± 0.0091	1.0504 ± 0.0765	2.96x ± 0.22x	196.14% ± 22.40%
array	vector_sub_vector	a - b	float64	262144x1	262144x1	1.0259 ± 0.0285	0.9138 ± 0.0488	0.2557 ± 0.0071	0.2876 ± 0.0147	1.13x ± 0.06x	12.52% ± 6.01%
array	vector_sum	sum()	float32	262144x1	1x1	0.3661 ± 0.0085	0.0450 ± 0.0013	0.7163 ± 0.0166	5.8288 ± 0.1619	8.13x ± 0.10x	713.41% ± 10.38%
array	vector_sum	sum()	float64	262144x1	1x1	0.3671 ± 0.0091	0.0918 ± 0.0040	0.7145 ± 0.0175	2.8612 ± 0.1207	4.00x ± 0.16x	300.47% ± 15.59%
array	vector_tanh	tanh(a * scalar)	float32	262144x1	262144x1	2.8000 ± 0.0678	1.2313 ± 0.0309	0.1873 ± 0.0045	0.4260 ± 0.0106	2.27x ± 0.00x	127.40% ± 0.48%
array	vector_tanh	tanh(a * scalar)	float64	262144x1	262144x1	15.6625 ± 0.4161	15.4361 ± 0.3628	0.0335 ± 0.0009	0.0340 ± 0.0008	1.01x ± 0.00x	1.46% ± 0.48%
geometry	quaternion_inverse	Quaternion::inverse()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1083 ± 0.0078	0.0989 ± 0.0024	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_inverse	Quaternion::inverse()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.0972 ± 0.0060	0.0699 ± 0.0053	0.85x ± 0.24x	-15.00% ± 24.15%
geometry	quaternion_normalize	Quaternion::normalize()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1261 ± 0.0047	0.1134 ± 0.0023	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_normalize	Quaternion::normalize()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1074 ± 0.0040	0.0821 ± 0.0027	0.90x ± 0.21x	-10.00% ± 21.08%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float32	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.3592 ± 0.0117	0.2732 ± 0.0086	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float64	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.3604 ± 0.0119	0.2669 ± 0.0238	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1288 ± 0.0043	0.1125 ± 0.0029	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1290 ± 0.0053	0.1104 ± 0.0047	1.00x ± 0.00x	0.00% ± 0.00%
matrix	matrix_add	A + B	float32	512x512	512x512	0.7408 ± 0.0198	0.2449 ± 0.0092	0.3541 ± 0.0093	1.0717 ± 0.0393	3.03x ± 0.13x	202.83% ± 13.27%
matrix	matrix_add	A + B	float64	512x512	512x512	1.0272 ± 0.0284	0.9091 ± 0.0211	0.2554 ± 0.0070	0.2885 ± 0.0066	1.13x ± 0.05x	13.07% ± 5.01%
matrix	matrix_block	block()	float32	512x512	256x256	0.1435 ± 0.0035	0.0338 ± 0.0009	0.0000 ± 0.0000	0.0000 ± 0.0000	4.24x ± 0.03x	324.39% ± 2.75%
matrix	matrix_block	block()	float64	512x512	256x256	0.1593 ± 0.0039	0.0640 ± 0.0013	0.0000 ± 0.0000	0.0000 ± 0.0000	2.49x ± 0.05x	149.02% ± 4.70%
matrix	matrix_colwise_sum	colwise().sum()	float32	512x512	1x512	0.3701 ± 0.0092	0.0473 ± 0.0014	0.7074 ± 0.0174	5.5389 ± 0.1637	7.83x ± 0.09x	683.04% ± 9.44%
matrix	matrix_colwise_sum	colwise().sum()	float64	512x512	1x512	0.3733 ± 0.0092	0.0945 ± 0.0048	0.7012 ± 0.0172	2.7750 ± 0.1313	3.96x ± 0.18x	295.81% ± 18.29%
matrix	matrix_cwise_product	cwiseProduct	float32	512x512	512x512	0.6404 ± 0.0145	0.2580 ± 0.0107	0.4095 ± 0.0092	1.0177 ± 0.0413	2.49x ± 0.10x	148.58% ± 10.46%
matrix	matrix_cwise_product	cwiseProduct	float64	512x512	512x512	1.0405 ± 0.0209	0.9080 ± 0.0274	0.2520 ± 0.0051	0.2889 ± 0.0085	1.15x ± 0.04x	14.68% ± 4.11%
matrix	matrix_determinant	determinant()	float32	512x512	1x1	42.2178 ± 0.9951	14.7710 ± 0.4316	2.1205 ± 0.0496	6.0623 ± 0.1745	2.86x ± 0.02x	185.86% ± 2.47%
matrix	matrix_determinant	determinant()	float64	512x512	1x1	53.0025 ± 1.5707	25.6416 ± 0.6506	1.6895 ± 0.0501	3.4916 ± 0.0877	2.07x ± 0.04x	106.72% ± 4.05%
matrix	matrix_gemv	A * x	float32	512x512	512x1	0.1724 ± 0.0041	0.0818 ± 0.0017	3.0390 ± 0.0709	6.4019 ± 0.1328	2.11x ± 0.03x	110.71% ± 2.56%
matrix	matrix_gemv	A * x	float64	512x512	512x1	0.3434 ± 0.0233	0.1834 ± 0.0070	1.5312 ± 0.0975	2.8592 ± 0.1058	1.87x ± 0.12x	87.32% ± 12.38%
matrix	matrix_identity	Identity()	float32	1x1	512x512	0.9450 ± 0.0187	0.3809 ± 0.0093	0.0000 ± 0.0000	0.0000 ± 0.0000	2.48x ± 0.05x	148.18% ± 5.28%
matrix	matrix_identity	Identity()	float64	1x1	512x512	0.6247 ± 0.0140	0.4294 ± 0.0115	0.0000 ± 0.0000	0.0000 ± 0.0000	1.46x ± 0.03x	45.52% ± 3.29%
matrix	matrix_inverse	inverse()	float32	512x512	512x512	217.4607 ± 5.4640	58.4411 ± 1.3855	1.2351 ± 0.0307	4.5956 ± 0.1079	3.72x ± 0.01x	272.09% ± 1.43%
matrix	matrix_inverse	inverse()	float64	512x512	512x512	224.3264 ± 6.5301	100.5545 ± 2.4741	1.1976 ± 0.0351	2.6710 ± 0.0651	2.23x ± 0.05x	123.11% ± 5.04%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float32	512x512	512x1	0.2000 ± 0.0083	0.1439 ± 0.0033	2.6252 ± 0.1054	3.6446 ± 0.0837	1.39x ± 0.04x	38.95% ± 3.52%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float64	512x512	512x1	0.3129 ± 0.0116	0.2903 ± 0.0069	1.6778 ± 0.0605	1.8069 ± 0.0424	1.08x ± 0.03x	7.77% ± 2.86%
matrix	matrix_llt_solve	LLT::solve(x)	float32	512x512	512x1	0.2019 ± 0.0070	0.1494 ± 0.0036	2.5991 ± 0.0870	3.5117 ± 0.0847	1.35x ± 0.03x	35.18% ± 2.66%
matrix	matrix_llt_solve	LLT::solve(x)	float64	512x512	512x1	0.3126 ± 0.0078	0.2897 ± 0.0061	1.6782 ± 0.0421	1.8108 ± 0.0379	1.08x ± 0.02x	7.93% ± 1.98%
matrix	matrix_matmul	A * B	float32	512x512	512x512	71.7630 ± 1.8402	22.4028 ± 0.5166	3.7391 ± 0.0949	11.9762 ± 0.2737	3.20x ± 0.01x	220.31% ± 1.10%
matrix	matrix_matmul	A * B	float64	512x512	512x512	93.7697 ± 2.3706	46.0085 ± 1.0569	2.8616 ± 0.0723	5.8315 ± 0.1328	2.04x ± 0.03x	103.82% ± 2.80%
matrix	matrix_rowwise_sum	rowwise().sum()	float32	512x512	512x1	1.8259 ± 0.0486	0.2395 ± 0.0056	0.1434 ± 0.0038	1.0928 ± 0.0254	7.62x ± 0.14x	662.34% ± 13.72%
matrix	matrix_rowwise_sum	rowwise().sum()	float64	512x512	512x1	2.5977 ± 0.0687	0.6699 ± 0.0711	0.1008 ± 0.0027	0.3939 ± 0.0351	3.91x ± 0.35x	291.00% ± 35.02%
matrix	matrix_sub	A - B	float32	512x512	512x512	0.6398 ± 0.0157	0.2459 ± 0.0093	0.4100 ± 0.0100	1.0674 ± 0.0389	2.60x ± 0.10x	160.45% ± 10.38%
matrix	matrix_sub	A - B	float64	512x512	512x512	1.0246 ± 0.0291	0.9139 ± 0.0376	0.2560 ± 0.0071	0.2873 ± 0.0114	1.12x ± 0.06x	12.31% ± 6.26%
matrix	matrix_trace	trace()	float32	512x512	1x1	0.0008 ± 0.0001	0.0009 ± 0.0001	0.6096 ± 0.0363	0.5975 ± 0.0523	0.97x ± 0.08x	-2.93% ± 8.27%
matrix	matrix_trace	trace()	float64	512x512	1x1	0.0011 ± 0.0000	0.0011 ± 0.0000	0.4747 ± 0.0121	0.4709 ± 0.0135	0.96x ± 0.05x	-3.56% ± 4.60%
matrix	matrix_transpose	transpose()	float32	512x512	512x512	2.2251 ± 0.0531	1.8858 ± 0.0435	0.0000 ± 0.0000	0.0000 ± 0.0000	1.18x ± 0.01x	17.99% ± 0.73%
matrix	matrix_transpose	transpose()	float64	512x512	512x512	3.4085 ± 0.1070	3.5317 ± 0.1711	0.0000 ± 0.0000	0.0000 ± 0.0000	0.97x ± 0.04x	-3.34% ± 4.33%
matrix	matrix_zero	Zero()	float32	1x1	512x512	0.4170 ± 0.0109	0.0926 ± 0.0020	0.0000 ± 0.0000	0.0000 ± 0.0000	4.50x ± 0.04x	350.19% ± 3.56%
matrix	matrix_zero	Zero()	float64	1x1	512x512	0.4684 ± 0.0105	0.1874 ± 0.0065	0.0000 ± 0.0000	0.0000 ± 0.0000	2.50x ± 0.08x	150.17% ± 8.28%

四核对比

taskset -c 0-3

threads：4

module	function	api	dtype	input_size	output_size	libeigen3-dev avg_ms mean±sd	eigen-spacemit avg_ms mean±sd	libeigen3-dev avg_gflops mean±sd	eigen-spacemit avg_gflops mean±sd	speedup mean±sd	improvement mean±sd
array	vector_abs	abs(a - scalar)	float32	262144x1	262144x1	0.6274 ± 0.0143	0.2503 ± 0.0061	0.4180 ± 0.0096	1.0479 ± 0.0264	2.51x ± 0.06x	150.71% ± 5.71%
array	vector_abs	abs(a - scalar)	float64	262144x1	262144x1	0.8665 ± 0.0251	0.6945 ± 0.0091	0.3028 ± 0.0089	0.3775 ± 0.0049	1.25x ± 0.04x	24.79% ± 4.23%
array	vector_add_scalar	a + scalar	float32	262144x1	262144x1	0.6428 ± 0.0214	0.2105 ± 0.0101	0.4082 ± 0.0140	1.2479 ± 0.0613	3.06x ± 0.11x	205.69% ± 10.72%
array	vector_add_scalar	a + scalar	float64	262144x1	262144x1	0.7743 ± 0.0215	0.6206 ± 0.0153	0.3388 ± 0.0096	0.4226 ± 0.0104	1.25x ± 0.04x	24.82% ± 4.01%
array	vector_add_vector	a + b	float32	262144x1	262144x1	0.7733 ± 0.0248	0.2626 ± 0.0146	0.3393 ± 0.0111	1.0011 ± 0.0571	2.95x ± 0.11x	194.91% ± 10.77%
array	vector_add_vector	a + b	float64	262144x1	262144x1	1.0735 ± 0.0306	0.8903 ± 0.0272	0.2444 ± 0.0071	0.2947 ± 0.0085	1.21x ± 0.04x	20.64% ± 4.12%
array	vector_clamp	cwiseMax + cwiseMin	float32	262144x1	262144x1	0.7595 ± 0.0188	0.4512 ± 0.0095	0.6907 ± 0.0170	1.1626 ± 0.0251	1.68x ± 0.04x	68.37% ± 4.04%
array	vector_clamp	cwiseMax + cwiseMin	float64	262144x1	262144x1	0.9385 ± 0.0169	1.0677 ± 0.0210	0.5588 ± 0.0100	0.4913 ± 0.0097	0.88x ± 0.02x	-12.08% ± 1.81%
array	vector_cos	cos(a)	float32	262144x1	262144x1	5.8698 ± 0.1365	2.1296 ± 0.0422	0.0447 ± 0.0010	0.1231 ± 0.0025	2.76x ± 0.05x	175.65% ± 5.05%
array	vector_cos	cos(a)	float64	262144x1	262144x1	11.2248 ± 0.1838	11.1351 ± 0.2361	0.0234 ± 0.0004	0.0235 ± 0.0005	1.01x ± 0.01x	0.82% ± 1.18%
array	vector_div_scalar	a / scalar	float32	262144x1	262144x1	1.6620 ± 0.0552	0.6891 ± 0.0273	0.1579 ± 0.0053	0.3810 ± 0.0154	2.41x ± 0.07x	141.31% ± 6.84%
array	vector_div_scalar	a / scalar	float64	262144x1	262144x1	2.7798 ± 0.0835	2.4699 ± 0.0762	0.0944 ± 0.0029	0.1062 ± 0.0033	1.13x ± 0.03x	12.58% ± 3.03%
array	vector_div_vector	a / (b + scalar)	float32	262144x1	262144x1	1.6439 ± 0.0429	0.7931 ± 0.0330	0.3191 ± 0.0083	0.6621 ± 0.0278	2.07x ± 0.06x	107.47% ± 6.28%
array	vector_div_vector	a / (b + scalar)	float64	262144x1	262144x1	2.7681 ± 0.1037	2.6653 ± 0.0562	0.1896 ± 0.0071	0.1968 ± 0.0041	1.04x ± 0.04x	3.88% ± 4.09%
array	vector_dot	dot()	float32	262144x1	1x1	0.3929 ± 0.0334	0.1007 ± 0.0122	1.3422 ± 0.1031	5.2688 ± 0.5900	3.96x ± 0.65x	296.11% ± 65.05%
array	vector_dot	dot()	float64	262144x1	1x1	0.5018 ± 0.0491	0.4140 ± 0.0218	1.0526 ± 0.0875	1.2696 ± 0.0666	1.22x ± 0.15x	21.61% ± 15.03%
array	vector_exp	exp(a * scalar)	float32	262144x1	262144x1	6.9051 ± 0.1454	2.2167 ± 0.0852	0.0760 ± 0.0016	0.2368 ± 0.0093	3.12x ± 0.11x	211.81% ± 10.69%
array	vector_exp	exp(a * scalar)	float64	262144x1	262144x1	7.2381 ± 0.0942	7.0569 ± 0.1752	0.0725 ± 0.0010	0.0743 ± 0.0019	1.03x ± 0.02x	2.61% ± 1.94%
array	vector_fma	a * b + c	float32	262144x1	262144x1	1.2551 ± 0.0285	0.4861 ± 0.0110	0.4179 ± 0.0093	1.0791 ± 0.0245	2.58x ± 0.07x	158.32% ± 7.42%
array	vector_fma	a * b + c	float64	262144x1	262144x1	1.4918 ± 0.0234	1.2963 ± 0.0273	0.3515 ± 0.0055	0.4046 ± 0.0084	1.15x ± 0.03x	15.14% ± 3.41%
array	vector_log	log(a + scalar)	float32	262144x1	262144x1	6.6297 ± 0.1274	2.1498 ± 0.0804	0.0791 ± 0.0015	0.2442 ± 0.0093	3.09x ± 0.11x	208.72% ± 11.28%
array	vector_log	log(a + scalar)	float64	262144x1	262144x1	7.5689 ± 0.1399	6.3410 ± 0.1317	0.0693 ± 0.0013	0.0827 ± 0.0017	1.19x ± 0.02x	19.38% ± 1.84%
array	vector_max	max(a, b)	float32	262144x1	262144x1	0.7821 ± 0.0690	0.3281 ± 0.0172	0.3372 ± 0.0262	0.8011 ± 0.0422	2.39x ± 0.24x	138.94% ± 24.04%
array	vector_max	max(a, b)	float64	262144x1	262144x1	1.0944 ± 0.0531	0.9409 ± 0.0279	0.2400 ± 0.0111	0.2788 ± 0.0084	1.16x ± 0.07x	16.44% ± 7.38%
array	vector_min	min(a, b)	float32	262144x1	262144x1	0.8819 ± 0.0216	0.3255 ± 0.0157	0.2974 ± 0.0073	0.8069 ± 0.0378	2.71x ± 0.11x	171.33% ± 11.42%
array	vector_min	min(a, b)	float64	262144x1	262144x1	1.0774 ± 0.0491	0.9156 ± 0.0224	0.2438 ± 0.0108	0.2864 ± 0.0070	1.18x ± 0.07x	17.75% ± 6.54%
array	vector_mul_scalar	a * scalar	float32	262144x1	262144x1	0.7686 ± 0.0273	0.2254 ± 0.0107	0.3415 ± 0.0123	1.1652 ± 0.0570	3.41x ± 0.12x	241.25% ± 12.24%
array	vector_mul_scalar	a * scalar	float64	262144x1	262144x1	0.7651 ± 0.0184	0.6495 ± 0.0214	0.3428 ± 0.0084	0.4040 ± 0.0130	1.18x ± 0.04x	17.89% ± 4.20%
array	vector_mul_vector	a * b	float32	262144x1	262144x1	0.7690 ± 0.0255	0.2754 ± 0.0139	0.3412 ± 0.0113	0.9540 ± 0.0499	2.80x ± 0.14x	179.67% ± 13.87%
array	vector_mul_vector	a * b	float64	262144x1	262144x1	1.0740 ± 0.0396	0.8982 ± 0.0283	0.2444 ± 0.0090	0.2921 ± 0.0089	1.20x ± 0.07x	19.72% ± 6.77%
array	vector_norm	norm()	float32	262144x1	1x1	0.3798 ± 0.0118	0.0855 ± 0.0017	1.3816 ± 0.0428	6.1338 ± 0.1220	4.44x ± 0.12x	344.27% ± 11.66%
array	vector_norm	norm()	float64	262144x1	1x1	0.3772 ± 0.0092	0.1904 ± 0.0055	1.3907 ± 0.0336	2.7558 ± 0.0801	1.98x ± 0.05x	98.19% ± 5.16%
array	vector_normalize	normalize()	float32	262144x1	262144x1	2.2772 ± 0.0695	0.8451 ± 0.0177	0.3456 ± 0.0106	0.9309 ± 0.0200	2.69x ± 0.07x	169.48% ± 6.68%
array	vector_normalize	normalize()	float64	262144x1	262144x1	3.3798 ± 0.0886	2.8385 ± 0.0706	0.2328 ± 0.0061	0.2772 ± 0.0069	1.19x ± 0.03x	19.09% ± 2.70%
array	vector_segment	segment()	float32	262144x1	131072x1	0.3164 ± 0.0098	0.0646 ± 0.0011	0.0000 ± 0.0000	0.0000 ± 0.0000	4.90x ± 0.11x	389.82% ± 11.47%
array	vector_segment	segment()	float64	262144x1	131072x1	0.3174 ± 0.0101	0.1354 ± 0.0055	0.0000 ± 0.0000	0.0000 ± 0.0000	2.35x ± 0.09x	134.71% ± 8.73%
array	vector_select	mask.select(a, b)	float32	262144x1	262144x1	0.9031 ± 0.0361	0.4990 ± 0.0163	0.2907 ± 0.0119	0.5258 ± 0.0173	1.81x ± 0.09x	81.12% ± 8.87%
array	vector_select	mask.select(a, b)	float64	262144x1	262144x1	0.9579 ± 0.0377	1.1470 ± 0.0242	0.2741 ± 0.0107	0.2286 ± 0.0048	0.84x ± 0.03x	-16.46% ± 3.46%
array	vector_sin	sin(a)	float32	262144x1	262144x1	5.7507 ± 0.0914	2.1754 ± 0.0579	0.0456 ± 0.0007	0.1206 ± 0.0032	2.64x ± 0.05x	164.45% ± 5.10%
array	vector_sin	sin(a)	float64	262144x1	262144x1	11.3619 ± 0.1629	11.3429 ± 0.2903	0.0231 ± 0.0003	0.0231 ± 0.0006	1.00x ± 0.02x	0.20% ± 1.75%
array	vector_sqrt	(a + scalar).sqrt()	float32	262144x1	262144x1	5.4976 ± 0.0881	0.6506 ± 0.0249	0.0954 ± 0.0016	0.8069 ± 0.0313	8.46x ± 0.27x	745.83% ± 27.12%
array	vector_sqrt	(a + scalar).sqrt()	float64	262144x1	262144x1	6.6503 ± 0.1665	2.4066 ± 0.0649	0.0789 ± 0.0020	0.2180 ± 0.0059	2.76x ± 0.09x	176.46% ± 8.55%
array	vector_sub_scalar	a - scalar	float32	262144x1	262144x1	0.6421 ± 0.0208	0.2097 ± 0.0100	0.4087 ± 0.0136	1.2528 ± 0.0616	3.07x ± 0.11x	206.55% ± 10.72%
array	vector_sub_scalar	a - scalar	float64	262144x1	262144x1	0.7746 ± 0.0235	0.6244 ± 0.0198	0.3387 ± 0.0104	0.4202 ± 0.0133	1.24x ± 0.05x	24.14% ± 4.85%
array	vector_sub_vector	a - b	float32	262144x1	262144x1	0.7705 ± 0.0244	0.2614 ± 0.0130	0.3406 ± 0.0108	1.0054 ± 0.0512	2.95x ± 0.13x	195.25% ± 12.72%
array	vector_sub_vector	a - b	float64	262144x1	262144x1	1.0788 ± 0.0456	0.8893 ± 0.0266	0.2434 ± 0.0102	0.2950 ± 0.0084	1.21x ± 0.07x	21.42% ± 6.67%
array	vector_sum	sum()	float32	262144x1	1x1	0.3786 ± 0.0105	0.0458 ± 0.0010	0.6929 ± 0.0193	5.7290 ± 0.1326	8.27x ± 0.22x	726.95% ± 22.18%
array	vector_sum	sum()	float64	262144x1	1x1	0.3775 ± 0.0091	0.0937 ± 0.0041	0.6947 ± 0.0167	2.8015 ± 0.1179	4.03x ± 0.17x	303.33% ± 17.47%
array	vector_tanh	tanh(a * scalar)	float32	262144x1	262144x1	2.9032 ± 0.0848	1.2529 ± 0.0247	0.1807 ± 0.0053	0.4186 ± 0.0085	2.32x ± 0.06x	131.73% ± 5.53%
array	vector_tanh	tanh(a * scalar)	float64	262144x1	262144x1	16.0917 ± 0.2985	15.9251 ± 0.3440	0.0326 ± 0.0006	0.0329 ± 0.0007	1.01x ± 0.01x	1.06% ± 1.09%
geometry	quaternion_inverse	Quaternion::inverse()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1071 ± 0.0036	0.0947 ± 0.0057	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_inverse	Quaternion::inverse()	float64	3x1	3x1	0.0001 ± 0.0001	0.0001 ± 0.0000	0.0864 ± 0.0193	0.0685 ± 0.0020	1.10x ± 0.70x	10.00% ± 69.92%
geometry	quaternion_normalize	Quaternion::normalize()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1231 ± 0.0097	0.1115 ± 0.0045	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_normalize	Quaternion::normalize()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1029 ± 0.0037	0.0815 ± 0.0054	0.85x ± 0.24x	-15.00% ± 24.15%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float32	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.3439 ± 0.0103	0.2628 ± 0.0115	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float64	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.3574 ± 0.0300	0.2560 ± 0.0085	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1241 ± 0.0037	0.1094 ± 0.0030	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1224 ± 0.0051	0.1076 ± 0.0031	1.00x ± 0.00x	0.00% ± 0.00%
matrix	matrix_add	A + B	float32	512x512	512x512	0.7672 ± 0.0245	0.2550 ± 0.0146	0.3420 ± 0.0108	1.0312 ± 0.0587	3.02x ± 0.23x	201.97% ± 22.84%
matrix	matrix_add	A + B	float64	512x512	512x512	1.0685 ± 0.0329	0.8896 ± 0.0194	0.2455 ± 0.0073	0.2948 ± 0.0062	1.20x ± 0.05x	20.17% ± 4.96%
matrix	matrix_block	block()	float32	512x512	256x256	0.1484 ± 0.0040	0.0352 ± 0.0011	0.0000 ± 0.0000	0.0000 ± 0.0000	4.22x ± 0.06x	321.95% ± 6.38%
matrix	matrix_block	block()	float64	512x512	256x256	0.1623 ± 0.0037	0.0664 ± 0.0019	0.0000 ± 0.0000	0.0000 ± 0.0000	2.45x ± 0.04x	144.58% ± 4.16%
matrix	matrix_colwise_sum	colwise().sum()	float32	512x512	1x512	0.3820 ± 0.0103	0.0490 ± 0.0019	0.6855 ± 0.0185	5.3448 ± 0.2062	7.80x ± 0.29x	679.92% ± 29.15%
matrix	matrix_colwise_sum	colwise().sum()	float64	512x512	1x512	0.3858 ± 0.0127	0.0969 ± 0.0038	0.6788 ± 0.0225	2.7027 ± 0.1067	3.98x ± 0.19x	298.46% ± 18.86%
matrix	matrix_cwise_product	cwiseProduct	float32	512x512	512x512	0.6632 ± 0.0275	0.2684 ± 0.0128	0.3959 ± 0.0160	0.9784 ± 0.0436	2.47x ± 0.14x	147.45% ± 13.65%
matrix	matrix_cwise_product	cwiseProduct	float64	512x512	512x512	1.0739 ± 0.0355	0.9244 ± 0.0324	0.2444 ± 0.0080	0.2839 ± 0.0100	1.16x ± 0.06x	16.32% ± 6.20%
matrix	matrix_determinant	determinant()	float32	512x512	1x1	25.0644 ± 0.6742	11.0655 ± 0.3305	3.5724 ± 0.0997	8.0929 ± 0.2473	2.27x ± 0.06x	126.60% ± 6.36%
matrix	matrix_determinant	determinant()	float64	512x512	1x1	30.3570 ± 0.7799	17.9423 ± 0.4328	2.9494 ± 0.0787	4.9897 ± 0.1230	1.69x ± 0.03x	69.21% ± 3.09%
matrix	matrix_gemv	A * x	float32	512x512	512x1	0.1794 ± 0.0058	0.0856 ± 0.0037	2.9220 ± 0.0949	6.1306 ± 0.2670	2.10x ± 0.08x	109.90% ± 8.43%
matrix	matrix_gemv	A * x	float64	512x512	512x1	0.3457 ± 0.0083	0.1925 ± 0.0065	1.5160 ± 0.0364	2.7236 ± 0.0946	1.80x ± 0.06x	79.70% ± 6.35%
matrix	matrix_identity	Identity()	float32	1x1	512x512	0.9681 ± 0.0481	0.3927 ± 0.0122	0.0000 ± 0.0000	0.0000 ± 0.0000	2.46x ± 0.07x	146.44% ± 7.27%
matrix	matrix_identity	Identity()	float64	1x1	512x512	0.6387 ± 0.0150	0.4434 ± 0.0126	0.0000 ± 0.0000	0.0000 ± 0.0000	1.44x ± 0.02x	44.07% ± 2.35%
matrix	matrix_inverse	inverse()	float32	512x512	512x512	205.4377 ± 3.8438	53.1489 ± 1.3958	1.3071 ± 0.0252	5.0538 ± 0.1357	3.87x ± 0.07x	286.64% ± 6.58%
matrix	matrix_inverse	inverse()	float64	512x512	512x512	195.1338 ± 4.9092	91.9283 ± 2.1262	1.3764 ± 0.0348	2.9215 ± 0.0682	2.12x ± 0.05x	112.30% ± 4.64%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float32	512x512	512x1	0.2127 ± 0.0114	0.1515 ± 0.0057	2.4712 ± 0.1338	3.4650 ± 0.1339	1.40x ± 0.06x	40.44% ± 6.05%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float64	512x512	512x1	0.3276 ± 0.0107	0.3058 ± 0.0103	1.6019 ± 0.0553	1.7160 ± 0.0600	1.07x ± 0.03x	7.17% ± 3.26%
matrix	matrix_llt_solve	LLT::solve(x)	float32	512x512	512x1	0.2124 ± 0.0080	0.1570 ± 0.0053	2.4712 ± 0.0964	3.3415 ± 0.1162	1.35x ± 0.05x	35.32% ± 4.64%
matrix	matrix_llt_solve	LLT::solve(x)	float64	512x512	512x1	0.3287 ± 0.0110	0.3079 ± 0.0114	1.5969 ± 0.0564	1.7051 ± 0.0659	1.07x ± 0.04x	6.82% ± 3.72%
matrix	matrix_matmul	A * B	float32	512x512	512x512	21.2244 ± 0.6392	6.0132 ± 0.1926	12.6459 ± 0.3981	44.6392 ± 1.4428	3.53x ± 0.08x	253.07% ± 8.25%
matrix	matrix_matmul	A * B	float64	512x512	512x512	35.5351 ± 1.1246	12.5752 ± 0.3502	7.5539 ± 0.2536	21.3411 ± 0.6207	2.83x ± 0.05x	182.58% ± 4.51%
matrix	matrix_rowwise_sum	rowwise().sum()	float32	512x512	512x1	1.8989 ± 0.0616	0.2482 ± 0.0106	0.1379 ± 0.0045	1.0559 ± 0.0452	7.66x ± 0.33x	666.04% ± 32.85%
matrix	matrix_rowwise_sum	rowwise().sum()	float64	512x512	512x1	2.6737 ± 0.0848	0.6785 ± 0.0265	0.0979 ± 0.0031	0.3862 ± 0.0154	3.94x ± 0.17x	294.49% ± 17.24%
matrix	matrix_sub	A - B	float32	512x512	512x512	0.6607 ± 0.0310	0.2564 ± 0.0127	0.3975 ± 0.0176	1.0244 ± 0.0482	2.58x ± 0.21x	158.42% ± 20.88%
matrix	matrix_sub	A - B	float64	512x512	512x512	1.0674 ± 0.0574	0.9165 ± 0.0418	0.2462 ± 0.0128	0.2866 ± 0.0129	1.17x ± 0.09x	16.74% ± 9.15%
matrix	matrix_trace	trace()	float32	512x512	1x1	0.0008 ± 0.0000	0.0009 ± 0.0002	0.6024 ± 0.0180	0.5690 ± 0.0877	0.93x ± 0.17x	-7.08% ± 16.75%
matrix	matrix_trace	trace()	float64	512x512	1x1	0.0011 ± 0.0001	0.0011 ± 0.0001	0.4578 ± 0.0164	0.4521 ± 0.0171	0.98x ± 0.07x	-2.42% ± 7.20%
matrix	matrix_transpose	transpose()	float32	512x512	512x512	2.2959 ± 0.0572	1.9306 ± 0.0522	0.0000 ± 0.0000	0.0000 ± 0.0000	1.19x ± 0.02x	18.94% ± 2.22%
matrix	matrix_transpose	transpose()	float64	512x512	512x512	3.4579 ± 0.0786	3.5339 ± 0.0795	0.0000 ± 0.0000	0.0000 ± 0.0000	0.98x ± 0.02x	-2.13% ± 2.14%
matrix	matrix_zero	Zero()	float32	1x1	512x512	0.4422 ± 0.0122	0.0951 ± 0.0030	0.0000 ± 0.0000	0.0000 ± 0.0000	4.65x ± 0.07x	364.90% ± 7.12%
matrix	matrix_zero	Zero()	float64	1x1	512x512	0.4998 ± 0.0105	0.1928 ± 0.0068	0.0000 ± 0.0000	0.0000 ± 0.0000	2.59x ± 0.07x	159.39% ± 7.43%

八核对比

taskset -c 0-7

threads：8

module	function	api	dtype	input_size	output_size	libeigen3-dev avg_ms mean±sd	eigen-spacemit avg_ms mean±sd	libeigen3-dev avg_gflops mean±sd	eigen-spacemit avg_gflops mean±sd	speedup mean±sd	improvement mean±sd
array	vector_abs	abs(a - scalar)	float32	262144x1	262144x1	0.6158 ± 0.0202	0.2466 ± 0.0046	0.4261 ± 0.0135	1.0633 ± 0.0200	2.50x ± 0.07x	149.72% ± 6.83%
array	vector_abs	abs(a - scalar)	float64	262144x1	262144x1	0.8182 ± 0.0314	0.6886 ± 0.0119	0.3208 ± 0.0118	0.3808 ± 0.0066	1.19x ± 0.04x	18.80% ± 3.51%
array	vector_add_scalar	a + scalar	float32	262144x1	262144x1	0.6149 ± 0.0160	0.2031 ± 0.0050	0.4266 ± 0.0110	1.2917 ± 0.0317	3.03x ± 0.08x	202.85% ± 7.68%
array	vector_add_scalar	a + scalar	float64	262144x1	262144x1	0.7339 ± 0.0323	0.6193 ± 0.0195	0.3578 ± 0.0150	0.4237 ± 0.0133	1.19x ± 0.05x	18.55% ± 5.13%
array	vector_add_vector	a + b	float32	262144x1	262144x1	0.7493 ± 0.0242	0.2465 ± 0.0151	0.3502 ± 0.0112	1.0671 ± 0.0696	3.05x ± 0.26x	205.27% ± 26.00%
array	vector_add_vector	a + b	float64	262144x1	262144x1	1.0344 ± 0.0228	0.8399 ± 0.0970	0.2535 ± 0.0055	0.3163 ± 0.0404	1.25x ± 0.17x	24.93% ± 17.30%
array	vector_clamp	cwiseMax + cwiseMin	float32	262144x1	262144x1	0.7434 ± 0.0248	0.4459 ± 0.0092	0.7060 ± 0.0229	1.1763 ± 0.0245	1.67x ± 0.04x	66.71% ± 3.76%
array	vector_clamp	cwiseMax + cwiseMin	float64	262144x1	262144x1	0.8968 ± 0.0320	1.0468 ± 0.0310	0.5853 ± 0.0202	0.5012 ± 0.0147	0.86x ± 0.02x	-14.32% ± 2.06%
array	vector_cos	cos(a)	float32	262144x1	262144x1	5.7194 ± 0.1173	2.1087 ± 0.0474	0.0458 ± 0.0009	0.1244 ± 0.0028	2.71x ± 0.03x	171.26% ± 3.44%
array	vector_cos	cos(a)	float64	262144x1	262144x1	11.0917 ± 0.3198	10.8393 ± 0.3460	0.0237 ± 0.0007	0.0242 ± 0.0007	1.02x ± 0.02x	2.36% ± 2.10%
array	vector_div_scalar	a / scalar	float32	262144x1	262144x1	1.6055 ± 0.0489	0.6647 ± 0.0141	0.1634 ± 0.0049	0.3945 ± 0.0084	2.42x ± 0.06x	141.54% ± 6.24%
array	vector_div_scalar	a / scalar	float64	262144x1	262144x1	2.6389 ± 0.0598	2.4164 ± 0.0988	0.0994 ± 0.0023	0.1086 ± 0.0044	1.09x ± 0.04x	9.31% ± 3.52%
array	vector_div_vector	a / (b + scalar)	float32	262144x1	262144x1	1.5999 ± 0.0351	0.7725 ± 0.0135	0.3278 ± 0.0072	0.6789 ± 0.0120	2.07x ± 0.03x	107.11% ± 3.21%
array	vector_div_vector	a / (b + scalar)	float64	262144x1	262144x1	2.6680 ± 0.0827	2.6169 ± 0.0749	0.1967 ± 0.0063	0.2005 ± 0.0057	1.02x ± 0.04x	2.02% ± 3.99%
array	vector_dot	dot()	float32	262144x1	1x1	0.3711 ± 0.0122	0.1016 ± 0.0096	1.4142 ± 0.0453	5.2031 ± 0.4840	3.68x ± 0.35x	268.14% ± 34.66%
array	vector_dot	dot()	float64	262144x1	1x1	0.4870 ± 0.0246	0.3998 ± 0.0169	1.0788 ± 0.0512	1.3134 ± 0.0532	1.22x ± 0.08x	22.04% ± 8.39%
array	vector_exp	exp(a * scalar)	float32	262144x1	262144x1	6.7447 ± 0.1368	2.1485 ± 0.0460	0.0778 ± 0.0016	0.2441 ± 0.0053	3.14x ± 0.06x	213.98% ± 5.78%
array	vector_exp	exp(a * scalar)	float64	262144x1	262144x1	7.1180 ± 0.1497	6.8649 ± 0.1892	0.0737 ± 0.0016	0.0764 ± 0.0021	1.04x ± 0.02x	3.72% ± 1.87%
array	vector_fma	a * b + c	float32	262144x1	262144x1	1.2301 ± 0.0239	0.4826 ± 0.0121	0.4264 ± 0.0083	1.0871 ± 0.0281	2.55x ± 0.06x	155.00% ± 6.02%
array	vector_fma	a * b + c	float64	262144x1	262144x1	1.5160 ± 0.0387	1.2732 ± 0.0245	0.3460 ± 0.0089	0.4119 ± 0.0079	1.19x ± 0.04x	19.11% ± 3.82%
array	vector_log	log(a + scalar)	float32	262144x1	262144x1	6.4831 ± 0.1375	2.0885 ± 0.0423	0.0809 ± 0.0017	0.2511 ± 0.0051	3.10x ± 0.05x	210.45% ± 4.70%
array	vector_log	log(a + scalar)	float64	262144x1	262144x1	7.4521 ± 0.2067	6.2224 ± 0.2027	0.0704 ± 0.0020	0.0843 ± 0.0027	1.20x ± 0.04x	19.82% ± 3.55%
array	vector_max	max(a, b)	float32	262144x1	262144x1	0.7674 ± 0.0624	0.3113 ± 0.0095	0.3434 ± 0.0248	0.8428 ± 0.0262	2.47x ± 0.20x	146.62% ± 20.31%
array	vector_max	max(a, b)	float64	262144x1	262144x1	1.0518 ± 0.0681	0.9136 ± 0.0286	0.2503 ± 0.0176	0.2872 ± 0.0088	1.15x ± 0.08x	15.19% ± 7.85%
array	vector_min	min(a, b)	float32	262144x1	262144x1	0.8658 ± 0.0262	0.3097 ± 0.0129	0.3030 ± 0.0089	0.8477 ± 0.0370	2.80x ± 0.10x	179.79% ± 9.59%
array	vector_min	min(a, b)	float64	262144x1	262144x1	1.0125 ± 0.0643	0.9064 ± 0.0185	0.2599 ± 0.0175	0.2893 ± 0.0058	1.12x ± 0.06x	11.68% ± 6.48%
array	vector_mul_scalar	a * scalar	float32	262144x1	262144x1	0.7404 ± 0.0242	0.2193 ± 0.0042	0.3544 ± 0.0114	1.1957 ± 0.0221	3.38x ± 0.11x	237.68% ± 11.15%
array	vector_mul_scalar	a * scalar	float64	262144x1	262144x1	0.7241 ± 0.0168	0.6447 ± 0.0230	0.3622 ± 0.0084	0.4071 ± 0.0142	1.12x ± 0.04x	12.42% ± 3.86%
array	vector_mul_vector	a * b	float32	262144x1	262144x1	0.7539 ± 0.0156	0.2589 ± 0.0156	0.3479 ± 0.0073	1.0159 ± 0.0656	2.92x ± 0.23x	192.35% ± 22.93%
array	vector_mul_vector	a * b	float64	262144x1	262144x1	1.0405 ± 0.0291	0.8708 ± 0.1008	0.2521 ± 0.0071	0.3055 ± 0.0435	1.21x ± 0.19x	21.42% ± 18.95%
array	vector_norm	norm()	float32	262144x1	1x1	0.3704 ± 0.0110	0.0847 ± 0.0019	1.4167 ± 0.0412	6.1915 ± 0.1378	4.37x ± 0.11x	337.10% ± 10.94%
array	vector_norm	norm()	float64	262144x1	1x1	0.3744 ± 0.0163	0.1925 ± 0.0121	1.4027 ± 0.0601	2.7332 ± 0.1691	1.95x ± 0.08x	94.85% ± 8.40%
array	vector_normalize	normalize()	float32	262144x1	262144x1	2.2142 ± 0.0516	0.8405 ± 0.0129	0.3553 ± 0.0082	0.9359 ± 0.0144	2.63x ± 0.04x	163.44% ± 4.08%
array	vector_normalize	normalize()	float64	262144x1	262144x1	3.3215 ± 0.1106	2.7936 ± 0.0623	0.2370 ± 0.0078	0.2816 ± 0.0063	1.19x ± 0.02x	18.89% ± 2.44%
array	vector_segment	segment()	float32	262144x1	131072x1	0.3072 ± 0.0098	0.0636 ± 0.0016	0.0000 ± 0.0000	0.0000 ± 0.0000	4.83x ± 0.10x	382.57% ± 10.03%
array	vector_segment	segment()	float64	262144x1	131072x1	0.3107 ± 0.0105	0.1337 ± 0.0066	0.0000 ± 0.0000	0.0000 ± 0.0000	2.33x ± 0.09x	132.68% ± 9.05%
array	vector_select	mask.select(a, b)	float32	262144x1	262144x1	0.8753 ± 0.0384	0.4883 ± 0.0100	0.3001 ± 0.0142	0.5371 ± 0.0111	1.79x ± 0.08x	79.29% ± 8.17%
array	vector_select	mask.select(a, b)	float64	262144x1	262144x1	0.8822 ± 0.0347	1.1391 ± 0.0298	0.2975 ± 0.0119	0.2303 ± 0.0059	0.77x ± 0.03x	-22.54% ± 2.65%
array	vector_sin	sin(a)	float32	262144x1	262144x1	5.6462 ± 0.1128	2.1422 ± 0.0447	0.0464 ± 0.0009	0.1224 ± 0.0026	2.64x ± 0.03x	163.59% ± 2.99%
array	vector_sin	sin(a)	float64	262144x1	262144x1	11.2308 ± 0.2559	10.9707 ± 0.2745	0.0233 ± 0.0005	0.0239 ± 0.0006	1.02x ± 0.02x	2.39% ± 2.12%
array	vector_sqrt	(a + scalar).sqrt()	float32	262144x1	262144x1	5.3944 ± 0.1229	0.6349 ± 0.0106	0.0973 ± 0.0022	0.8260 ± 0.0139	8.50x ± 0.16x	749.68% ± 16.11%
array	vector_sqrt	(a + scalar).sqrt()	float64	262144x1	262144x1	6.4811 ± 0.1675	2.3448 ± 0.0704	0.0809 ± 0.0021	0.2238 ± 0.0066	2.77x ± 0.07x	176.50% ± 6.77%
array	vector_sub_scalar	a - scalar	float32	262144x1	262144x1	0.6170 ± 0.0197	0.2059 ± 0.0032	0.4252 ± 0.0133	1.2733 ± 0.0196	3.00x ± 0.11x	199.72% ± 11.04%
array	vector_sub_scalar	a - scalar	float64	262144x1	262144x1	0.7278 ± 0.0239	0.6184 ± 0.0201	0.3605 ± 0.0116	0.4243 ± 0.0137	1.18x ± 0.03x	17.73% ± 3.23%
array	vector_sub_vector	a - b	float32	262144x1	262144x1	0.7498 ± 0.0230	0.2463 ± 0.0153	0.3499 ± 0.0107	1.0683 ± 0.0723	3.06x ± 0.25x	205.68% ± 24.89%
array	vector_sub_vector	a - b	float64	262144x1	262144x1	1.0351 ± 0.0236	0.8779 ± 0.0879	0.2534 ± 0.0058	0.3017 ± 0.0344	1.19x ± 0.13x	19.08% ± 13.33%
array	vector_sum	sum()	float32	262144x1	1x1	0.3722 ± 0.0147	0.0454 ± 0.0012	0.7052 ± 0.0270	5.7789 ± 0.1521	8.20x ± 0.23x	719.87% ± 22.77%
array	vector_sum	sum()	float64	262144x1	1x1	0.3743 ± 0.0162	0.0946 ± 0.0074	0.7015 ± 0.0299	2.7866 ± 0.2052	3.97x ± 0.27x	297.27% ± 26.68%
array	vector_tanh	tanh(a * scalar)	float32	262144x1	262144x1	2.8239 ± 0.0774	1.2404 ± 0.0258	0.1858 ± 0.0050	0.4229 ± 0.0088	2.28x ± 0.04x	127.66% ± 4.03%
array	vector_tanh	tanh(a * scalar)	float64	262144x1	262144x1	15.8625 ± 0.5494	15.4945 ± 0.3613	0.0331 ± 0.0011	0.0339 ± 0.0008	1.02x ± 0.02x	2.36% ± 1.73%
geometry	quaternion_inverse	Quaternion::inverse()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1102 ± 0.0039	0.0972 ± 0.0022	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_inverse	Quaternion::inverse()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.0943 ± 0.0020	0.0729 ± 0.0058	0.95x ± 0.16x	-5.00% ± 15.81%
geometry	quaternion_normalize	Quaternion::normalize()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1243 ± 0.0117	0.1137 ± 0.0030	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_normalize	Quaternion::normalize()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1063 ± 0.0029	0.0827 ± 0.0030	0.95x ± 0.16x	-5.00% ± 15.81%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float32	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.3538 ± 0.0140	0.2728 ± 0.0183	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float64	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.3566 ± 0.0139	0.2691 ± 0.0212	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1278 ± 0.0037	0.1104 ± 0.0025	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1268 ± 0.0028	0.1098 ± 0.0042	1.00x ± 0.00x	0.00% ± 0.00%
matrix	matrix_add	A + B	float32	512x512	512x512	0.7502 ± 0.0202	0.2510 ± 0.0129	0.3496 ± 0.0093	1.0465 ± 0.0501	2.99x ± 0.10x	199.25% ± 10.20%
matrix	matrix_add	A + B	float64	512x512	512x512	1.0258 ± 0.0212	0.8888 ± 0.0727	0.2556 ± 0.0053	0.2970 ± 0.0281	1.16x ± 0.13x	16.30% ± 12.53%
matrix	matrix_block	block()	float32	512x512	256x256	0.1442 ± 0.0034	0.0348 ± 0.0014	0.0000 ± 0.0000	0.0000 ± 0.0000	4.14x ± 0.10x	314.48% ± 9.75%
matrix	matrix_block	block()	float64	512x512	256x256	0.1601 ± 0.0062	0.0645 ± 0.0021	0.0000 ± 0.0000	0.0000 ± 0.0000	2.48x ± 0.04x	148.12% ± 4.44%
matrix	matrix_colwise_sum	colwise().sum()	float32	512x512	1x512	0.3769 ± 0.0137	0.0493 ± 0.0039	0.6950 ± 0.0247	5.3323 ± 0.3836	7.68x ± 0.52x	667.54% ± 51.54%
matrix	matrix_colwise_sum	colwise().sum()	float64	512x512	1x512	0.3780 ± 0.0163	0.1002 ± 0.0069	0.6933 ± 0.0293	2.6234 ± 0.1838	3.79x ± 0.32x	279.00% ± 31.64%
matrix	matrix_cwise_product	cwiseProduct	float32	512x512	512x512	0.6582 ± 0.0279	0.2629 ± 0.0111	0.3990 ± 0.0166	0.9989 ± 0.0407	2.51x ± 0.16x	150.84% ± 15.94%
matrix	matrix_cwise_product	cwiseProduct	float64	512x512	512x512	1.0043 ± 0.0663	0.8983 ± 0.0840	0.2622 ± 0.0191	0.2945 ± 0.0324	1.13x ± 0.14x	12.75% ± 13.71%
matrix	matrix_determinant	determinant()	float32	512x512	1x1	25.0008 ± 1.7309	15.2198 ± 0.5286	3.5940 ± 0.2402	5.8855 ± 0.2042	1.64x ± 0.11x	64.31% ± 10.50%
matrix	matrix_determinant	determinant()	float64	512x512	1x1	32.1621 ± 4.2668	21.2876 ± 0.6420	2.8195 ± 0.3165	4.2067 ± 0.1255	1.51x ± 0.19x	51.04% ± 18.80%
matrix	matrix_gemv	A * x	float32	512x512	512x1	0.1753 ± 0.0062	0.0889 ± 0.0215	2.9914 ± 0.1035	6.0938 ± 0.9241	2.04x ± 0.31x	103.93% ± 31.24%
matrix	matrix_gemv	A * x	float64	512x512	512x1	0.3356 ± 0.0152	0.1887 ± 0.0093	1.5633 ± 0.0693	2.7823 ± 0.1371	1.78x ± 0.11x	78.21% ± 10.87%
matrix	matrix_identity	Identity()	float32	1x1	512x512	0.9374 ± 0.0348	0.3871 ± 0.0155	0.0000 ± 0.0000	0.0000 ± 0.0000	2.42x ± 0.07x	142.31% ± 7.00%
matrix	matrix_identity	Identity()	float64	1x1	512x512	0.6304 ± 0.0250	0.4298 ± 0.0144	0.0000 ± 0.0000	0.0000 ± 0.0000	1.47x ± 0.03x	46.68% ± 3.37%
matrix	matrix_inverse	inverse()	float32	512x512	512x512	205.2442 ± 8.2741	57.1382 ± 2.0209	1.3098 ± 0.0519	4.7033 ± 0.1650	3.59x ± 0.08x	259.22% ± 7.83%
matrix	matrix_inverse	inverse()	float64	512x512	512x512	198.7883 ± 13.7298	93.3451 ± 2.3236	1.3559 ± 0.0900	2.8773 ± 0.0711	2.13x ± 0.13x	112.90% ± 12.73%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float32	512x512	512x1	0.2052 ± 0.0077	0.1477 ± 0.0067	2.5588 ± 0.0955	3.5560 ± 0.1592	1.39x ± 0.06x	39.04% ± 6.14%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float64	512x512	512x1	0.3569 ± 0.0235	0.3023 ± 0.0191	1.4748 ± 0.1014	1.7408 ± 0.1093	1.18x ± 0.09x	18.38% ± 8.92%
matrix	matrix_llt_solve	LLT::solve(x)	float32	512x512	512x1	0.2087 ± 0.0077	0.1535 ± 0.0073	2.5145 ± 0.0937	3.4225 ± 0.1610	1.36x ± 0.06x	36.18% ± 6.38%
matrix	matrix_llt_solve	LLT::solve(x)	float64	512x512	512x1	0.3555 ± 0.0237	0.3027 ± 0.0185	1.4810 ± 0.1024	1.7374 ± 0.1033	1.18x ± 0.06x	17.50% ± 5.90%
matrix	matrix_matmul	A * B	float32	512x512	512x512	11.9369 ± 0.6003	3.6781 ± 0.6593	22.5172 ± 1.1346	74.5816 ± 10.4920	3.32x ± 0.48x	231.97% ± 47.54%
matrix	matrix_matmul	A * B	float64	512x512	512x512	17.4131 ± 1.2841	6.8375 ± 0.6139	15.4721 ± 1.0804	39.4869 ± 3.2992	2.55x ± 0.17x	155.47% ± 17.31%
matrix	matrix_rowwise_sum	rowwise().sum()	float32	512x512	512x1	1.8579 ± 0.0771	0.2662 ± 0.0801	0.1411 ± 0.0057	1.0312 ± 0.1791	7.32x ± 1.28x	632.25% ± 128.15%
matrix	matrix_rowwise_sum	rowwise().sum()	float64	512x512	512x1	2.6229 ± 0.1116	0.7331 ± 0.1011	0.0999 ± 0.0041	0.3626 ± 0.0462	3.64x ± 0.53x	263.98% ± 53.09%
matrix	matrix_sub	A - B	float32	512x512	512x512	0.6553 ± 0.0188	0.2529 ± 0.0099	0.4003 ± 0.0116	1.0379 ± 0.0391	2.59x ± 0.09x	159.33% ± 9.39%
matrix	matrix_sub	A - B	float64	512x512	512x512	1.0116 ± 0.0419	0.8950 ± 0.0626	0.2596 ± 0.0117	0.2943 ± 0.0227	1.14x ± 0.10x	13.60% ± 10.48%
matrix	matrix_trace	trace()	float32	512x512	1x1	0.0008 ± 0.0001	0.0009 ± 0.0001	0.6028 ± 0.0239	0.6019 ± 0.0262	0.99x ± 0.09x	-0.83% ± 8.67%
matrix	matrix_trace	trace()	float64	512x512	1x1	0.0011 ± 0.0001	0.0011 ± 0.0001	0.4644 ± 0.0189	0.4643 ± 0.0249	0.99x ± 0.07x	-0.63% ± 7.31%
matrix	matrix_transpose	transpose()	float32	512x512	512x512	2.2642 ± 0.1053	1.9411 ± 0.0744	0.0000 ± 0.0000	0.0000 ± 0.0000	1.17x ± 0.04x	16.67% ± 3.86%
matrix	matrix_transpose	transpose()	float64	512x512	512x512	3.4664 ± 0.0721	3.6390 ± 0.3019	0.0000 ± 0.0000	0.0000 ± 0.0000	0.96x ± 0.08x	-4.17% ± 8.09%
matrix	matrix_zero	Zero()	float32	1x1	512x512	0.4281 ± 0.0088	0.0940 ± 0.0036	0.0000 ± 0.0000	0.0000 ± 0.0000	4.56x ± 0.10x	355.52% ± 10.11%
matrix	matrix_zero	Zero()	float64	1x1	512x512	0.4941 ± 0.0179	0.1869 ± 0.0064	0.0000 ± 0.0000	0.0000 ± 0.0000	2.64x ± 0.07x	164.44% ± 6.96%

eigen-sapcemit USE BLAS VS Without BLAS

• 对比 eigen-sapcemit 使用 openblas-spacemit 作为后端和不使用之间的性能差异
• 使用 taskset -c 限制核数
• 单次测试的计时策略是每个算子跑 50 次取平均值，预热10次。本测试重复10次单次测试，取其指标的平均值并计算标准差
• no_blas 表示直接使用 eigen-spacemit 的 API
• blas 表示 eigen-spacemit 使用 openblas-spacemit 作为后端
• mean 表示 10 次测试的均值，sd 为标准差

单核对比

taskset -c 0

threads：1

module	function	api	dtype	input_size	output_size	no_blas avg_ms mean±sd	blas avg_ms mean±sd	no_blas avg_gflops mean±sd	blas avg_gflops mean±sd	speedup mean±sd	improvement mean±sd
array	vector_abs	abs(a - scalar)	float32	262144x1	262144x1	0.2426 ± 0.0060	0.2457 ± 0.0088	1.0810 ± 0.0263	1.0679 ± 0.0374	0.99x ± 0.03x	-1.21% ± 2.74%
array	vector_abs	abs(a - scalar)	float64	262144x1	262144x1	0.6786 ± 0.0092	0.6801 ± 0.0148	0.3864 ± 0.0053	0.3856 ± 0.0083	1.00x ± 0.02x	-0.19% ± 1.58%
array	vector_add_scalar	a + scalar	float32	262144x1	262144x1	0.1986 ± 0.0074	0.2023 ± 0.0097	1.3213 ± 0.0472	1.2982 ± 0.0608	0.98x ± 0.03x	-1.74% ± 3.14%
array	vector_add_scalar	a + scalar	float64	262144x1	262144x1	0.6039 ± 0.0079	0.6084 ± 0.0252	0.4342 ± 0.0056	0.4315 ± 0.0174	0.99x ± 0.03x	-0.63% ± 3.43%
array	vector_add_vector	a + b	float32	262144x1	262144x1	0.2458 ± 0.0150	0.2509 ± 0.0205	1.0700 ± 0.0640	1.0515 ± 0.0928	0.99x ± 0.11x	-1.31% ± 11.42%
array	vector_add_vector	a + b	float64	262144x1	262144x1	0.9074 ± 0.0300	0.9024 ± 0.0319	0.2892 ± 0.0093	0.2908 ± 0.0100	1.01x ± 0.04x	0.63% ± 3.81%
array	vector_clamp	cwiseMax + cwiseMin	float32	262144x1	262144x1	0.4406 ± 0.0104	0.4427 ± 0.0148	1.1905 ± 0.0277	1.1855 ± 0.0378	1.00x ± 0.02x	-0.41% ± 2.16%
array	vector_clamp	cwiseMax + cwiseMin	float64	262144x1	262144x1	1.0264 ± 0.0182	1.0284 ± 0.0240	0.5110 ± 0.0090	0.5101 ± 0.0118	1.00x ± 0.01x	-0.18% ± 0.95%
array	vector_cos	cos(a)	float32	262144x1	262144x1	2.0988 ± 0.0478	2.0865 ± 0.0696	0.1250 ± 0.0028	0.1258 ± 0.0040	1.01x ± 0.02x	0.63% ± 2.16%
array	vector_cos	cos(a)	float64	262144x1	262144x1	10.7241 ± 0.2396	10.7488 ± 0.3031	0.0244 ± 0.0005	0.0244 ± 0.0007	1.00x ± 0.01x	-0.21% ± 0.79%
array	vector_div_scalar	a / scalar	float32	262144x1	262144x1	0.6508 ± 0.0212	0.6610 ± 0.0262	0.4032 ± 0.0126	0.3972 ± 0.0150	0.99x ± 0.02x	-1.50% ± 2.19%
array	vector_div_scalar	a / scalar	float64	262144x1	262144x1	2.3885 ± 0.0865	2.3830 ± 0.0969	0.1099 ± 0.0038	0.1102 ± 0.0043	1.00x ± 0.01x	0.25% ± 0.67%
array	vector_div_vector	a / (b + scalar)	float32	262144x1	262144x1	0.7589 ± 0.0231	0.7706 ± 0.0325	0.6914 ± 0.0204	0.6815 ± 0.0280	0.99x ± 0.03x	-1.44% ± 3.00%
array	vector_div_vector	a / (b + scalar)	float64	262144x1	262144x1	2.5917 ± 0.0838	2.5770 ± 0.0674	0.2025 ± 0.0063	0.2036 ± 0.0052	1.01x ± 0.01x	0.56% ± 1.03%
array	vector_dot	dot()	float32	262144x1	1x1	0.0960 ± 0.0101	0.0929 ± 0.0073	5.5139 ± 0.5420	5.6720 ± 0.4122	1.04x ± 0.11x	3.61% ± 11.22%
array	vector_dot	dot()	float64	262144x1	1x1	0.4065 ± 0.0338	0.4005 ± 0.0166	1.2974 ± 0.1011	1.3110 ± 0.0539	1.02x ± 0.09x	1.63% ± 9.43%
array	vector_exp	exp(a * scalar)	float32	262144x1	262144x1	2.1141 ± 0.0482	2.1223 ± 0.0715	0.2481 ± 0.0055	0.2473 ± 0.0078	1.00x ± 0.02x	-0.33% ± 2.39%
array	vector_exp	exp(a * scalar)	float64	262144x1	262144x1	6.8590 ± 0.1686	6.8094 ± 0.1905	0.0765 ± 0.0018	0.0770 ± 0.0021	1.01x ± 0.01x	0.74% ± 0.86%
array	vector_fma	a * b + c	float32	262144x1	262144x1	0.4823 ± 0.0119	0.4822 ± 0.0209	1.0876 ± 0.0257	1.0891 ± 0.0451	1.00x ± 0.05x	0.20% ± 5.18%
array	vector_fma	a * b + c	float64	262144x1	262144x1	1.2767 ± 0.0174	1.2782 ± 0.0266	0.4107 ± 0.0056	0.4103 ± 0.0084	1.00x ± 0.02x	-0.09% ± 2.23%
array	vector_log	log(a + scalar)	float32	262144x1	262144x1	2.0563 ± 0.0479	2.0593 ± 0.0679	0.2551 ± 0.0058	0.2548 ± 0.0079	1.00x ± 0.02x	-0.10% ± 2.24%
array	vector_log	log(a + scalar)	float64	262144x1	262144x1	6.1495 ± 0.1527	6.1545 ± 0.1703	0.0853 ± 0.0021	0.0852 ± 0.0023	1.00x ± 0.01x	-0.07% ± 1.07%
array	vector_max	max(a, b)	float32	262144x1	262144x1	0.3110 ± 0.0129	0.3166 ± 0.0235	0.8443 ± 0.0353	0.8321 ± 0.0613	0.99x ± 0.09x	-1.25% ± 8.76%
array	vector_max	max(a, b)	float64	262144x1	262144x1	0.9116 ± 0.0393	0.9189 ± 0.0363	0.2880 ± 0.0118	0.2857 ± 0.0110	0.99x ± 0.07x	-0.61% ± 6.75%
array	vector_min	min(a, b)	float32	262144x1	262144x1	0.3086 ± 0.0119	0.3159 ± 0.0221	0.8506 ± 0.0325	0.8336 ± 0.0581	0.98x ± 0.09x	-1.81% ± 8.52%
array	vector_min	min(a, b)	float64	262144x1	262144x1	0.9000 ± 0.0211	0.8979 ± 0.0178	0.2914 ± 0.0066	0.2920 ± 0.0057	1.00x ± 0.03x	0.27% ± 3.25%
array	vector_mul_scalar	a * scalar	float32	262144x1	262144x1	0.2132 ± 0.0075	0.2175 ± 0.0091	1.2307 ± 0.0409	1.2070 ± 0.0500	0.98x ± 0.03x	-1.91% ± 3.00%
array	vector_mul_scalar	a * scalar	float64	262144x1	262144x1	0.6311 ± 0.0114	0.6329 ± 0.0253	0.4155 ± 0.0074	0.4148 ± 0.0161	1.00x ± 0.03x	-0.19% ± 3.03%
array	vector_mul_vector	a * b	float32	262144x1	262144x1	0.2604 ± 0.0170	0.2647 ± 0.0201	1.0104 ± 0.0653	0.9960 ± 0.0817	0.99x ± 0.10x	-1.07% ± 10.06%
array	vector_mul_vector	a * b	float64	262144x1	262144x1	0.9002 ± 0.0382	0.9071 ± 0.0341	0.2916 ± 0.0117	0.2893 ± 0.0105	0.99x ± 0.06x	-0.62% ± 5.91%
array	vector_norm	norm()	float32	262144x1	1x1	0.0841 ± 0.0029	0.0837 ± 0.0027	6.2425 ± 0.2041	6.2732 ± 0.1949	1.01x ± 0.03x	0.54% ± 2.81%
array	vector_norm	norm()	float64	262144x1	1x1	0.1858 ± 0.0072	0.1859 ± 0.0083	2.8251 ± 0.1038	2.8258 ± 0.1227	1.00x ± 0.03x	0.05% ± 3.12%
array	vector_normalize	normalize()	float32	262144x1	262144x1	0.8263 ± 0.0301	0.8253 ± 0.0242	0.9529 ± 0.0331	0.9536 ± 0.0270	1.00x ± 0.03x	0.12% ± 2.56%
array	vector_normalize	normalize()	float64	262144x1	262144x1	2.7610 ± 0.0519	2.7621 ± 0.0844	0.2849 ± 0.0053	0.2850 ± 0.0084	1.00x ± 0.01x	-0.01% ± 1.45%
array	vector_segment	segment()	float32	262144x1	131072x1	0.0636 ± 0.0020	0.0626 ± 0.0013	0.0000 ± 0.0000	0.0000 ± 0.0000	1.02x ± 0.02x	1.65% ± 1.87%
array	vector_segment	segment()	float64	262144x1	131072x1	0.1294 ± 0.0067	0.1276 ± 0.0039	0.0000 ± 0.0000	0.0000 ± 0.0000	1.01x ± 0.06x	1.45% ± 6.00%
array	vector_select	mask.select(a, b)	float32	262144x1	262144x1	0.4872 ± 0.0168	0.4854 ± 0.0210	0.5387 ± 0.0184	0.5409 ± 0.0232	1.01x ± 0.05x	0.51% ± 5.00%
array	vector_select	mask.select(a, b)	float64	262144x1	262144x1	1.1199 ± 0.0173	1.1256 ± 0.0231	0.2341 ± 0.0036	0.2330 ± 0.0047	1.00x ± 0.01x	-0.49% ± 1.32%
array	vector_sin	sin(a)	float32	262144x1	262144x1	2.0844 ± 0.0453	2.1151 ± 0.0676	0.1258 ± 0.0027	0.1240 ± 0.0037	0.99x ± 0.02x	-1.41% ± 2.09%
array	vector_sin	sin(a)	float64	262144x1	262144x1	10.9153 ± 0.3224	10.9188 ± 0.3461	0.0240 ± 0.0007	0.0240 ± 0.0007	1.00x ± 0.01x	-0.02% ± 0.91%
array	vector_sqrt	(a + scalar).sqrt()	float32	262144x1	262144x1	0.6227 ± 0.0137	0.6263 ± 0.0216	0.8423 ± 0.0182	0.8380 ± 0.0275	0.99x ± 0.02x	-0.52% ± 2.16%
array	vector_sqrt	(a + scalar).sqrt()	float64	262144x1	262144x1	2.3245 ± 0.0733	2.3158 ± 0.0565	0.2258 ± 0.0068	0.2265 ± 0.0054	1.00x ± 0.02x	0.37% ± 1.53%
array	vector_sub_scalar	a - scalar	float32	262144x1	262144x1	0.1983 ± 0.0064	0.2024 ± 0.0087	1.3230 ± 0.0413	1.2975 ± 0.0546	0.98x ± 0.03x	-1.92% ± 2.97%
array	vector_sub_scalar	a - scalar	float64	262144x1	262144x1	0.6066 ± 0.0098	0.6093 ± 0.0258	0.4323 ± 0.0069	0.4309 ± 0.0178	1.00x ± 0.04x	-0.30% ± 3.76%
array	vector_sub_vector	a - b	float32	262144x1	262144x1	0.2453 ± 0.0149	0.2508 ± 0.0207	1.0722 ± 0.0640	1.0523 ± 0.0942	0.99x ± 0.12x	-1.47% ± 11.50%
array	vector_sub_vector	a - b	float64	262144x1	262144x1	0.8995 ± 0.0200	0.9162 ± 0.0359	0.2916 ± 0.0064	0.2866 ± 0.0112	0.98x ± 0.03x	-1.73% ± 3.19%
array	vector_sum	sum()	float32	262144x1	1x1	0.0454 ± 0.0017	0.0446 ± 0.0016	5.7800 ± 0.2011	5.8800 ± 0.1962	1.02x ± 0.03x	1.78% ± 2.88%
array	vector_sum	sum()	float64	262144x1	1x1	0.0920 ± 0.0054	0.0924 ± 0.0053	2.8561 ± 0.1551	2.8450 ± 0.1572	1.00x ± 0.06x	-0.22% ± 5.88%
array	vector_tanh	tanh(a * scalar)	float32	262144x1	262144x1	1.2219 ± 0.0283	1.2256 ± 0.0403	0.4293 ± 0.0098	0.4282 ± 0.0134	1.00x ± 0.02x	-0.26% ± 2.17%
array	vector_tanh	tanh(a * scalar)	float64	262144x1	262144x1	15.4090 ± 0.4466	15.4376 ± 0.4939	0.0340 ± 0.0010	0.0340 ± 0.0010	1.00x ± 0.01x	-0.18% ± 0.64%
geometry	quaternion_inverse	Quaternion::inverse()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.0990 ± 0.0044	0.1104 ± 0.0038	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_inverse	Quaternion::inverse()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.0717 ± 0.0025	0.0735 ± 0.0034	0.95x ± 0.16x	-5.00% ± 15.81%
geometry	quaternion_normalize	Quaternion::normalize()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1138 ± 0.0048	0.1290 ± 0.0054	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_normalize	Quaternion::normalize()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.0833 ± 0.0025	0.0868 ± 0.0031	1.10x ± 0.32x	10.00% ± 31.62%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float32	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.2749 ± 0.0120	0.2995 ± 0.0140	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float64	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.2697 ± 0.0137	0.2985 ± 0.0120	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1120 ± 0.0051	0.1296 ± 0.0033	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1140 ± 0.0036	0.1295 ± 0.0051	1.00x ± 0.00x	0.00% ± 0.00%
matrix	matrix_add	A + B	float32	512x512	512x512	0.2503 ± 0.0110	0.2512 ± 0.0131	1.0492 ± 0.0451	1.0459 ± 0.0529	1.00x ± 0.06x	-0.20% ± 5.60%
matrix	matrix_add	A + B	float64	512x512	512x512	0.9127 ± 0.0427	0.9066 ± 0.0351	0.2877 ± 0.0130	0.2895 ± 0.0109	1.01x ± 0.06x	0.80% ± 5.88%
matrix	matrix_block	block()	float32	512x512	256x256	0.0337 ± 0.0011	0.0341 ± 0.0015	0.0000 ± 0.0000	0.0000 ± 0.0000	0.99x ± 0.02x	-1.09% ± 1.86%
matrix	matrix_block	block()	float64	512x512	256x256	0.0642 ± 0.0020	0.0646 ± 0.0025	0.0000 ± 0.0000	0.0000 ± 0.0000	0.99x ± 0.02x	-0.59% ± 2.04%
matrix	matrix_colwise_sum	colwise().sum()	float32	512x512	1x512	0.0472 ± 0.0016	0.0469 ± 0.0012	5.5489 ± 0.1786	5.5816 ± 0.1347	1.01x ± 0.02x	0.63% ± 1.58%
matrix	matrix_colwise_sum	colwise().sum()	float64	512x512	1x512	0.0940 ± 0.0045	0.0938 ± 0.0048	2.7882 ± 0.1299	2.7949 ± 0.1357	1.00x ± 0.07x	0.44% ± 6.69%
matrix	matrix_cwise_product	cwiseProduct	float32	512x512	512x512	0.2626 ± 0.0111	0.2622 ± 0.0105	1.0001 ± 0.0423	1.0013 ± 0.0398	1.00x ± 0.05x	0.23% ± 4.62%
matrix	matrix_cwise_product	cwiseProduct	float64	512x512	512x512	0.8996 ± 0.0318	0.9034 ± 0.0385	0.2917 ± 0.0102	0.2906 ± 0.0119	1.00x ± 0.06x	-0.27% ± 5.56%
matrix	matrix_determinant	determinant()	float32	512x512	1x1	14.6256 ± 0.4043	8.9489 ± 0.2659	6.1220 ± 0.1651	10.0065 ± 0.2866	1.63x ± 0.02x	63.45% ± 1.83%
matrix	matrix_determinant	determinant()	float64	512x512	1x1	25.5483 ± 0.7610	14.4550 ± 0.5234	3.5051 ± 0.1018	6.1972 ± 0.2169	1.77x ± 0.02x	76.79% ± 1.89%
matrix	matrix_gemv	A * x	float32	512x512	512x1	0.0796 ± 0.0027	0.1093 ± 0.0024	6.5897 ± 0.2140	4.7949 ± 0.1043	0.73x ± 0.01x	-27.21% ± 1.24%
matrix	matrix_gemv	A * x	float64	512x512	512x1	0.1840 ± 0.0083	0.2257 ± 0.0158	2.8515 ± 0.1235	2.3304 ± 0.1577	0.82x ± 0.05x	-18.24% ± 4.86%
matrix	matrix_identity	Identity()	float32	1x1	512x512	0.3504 ± 0.0119	0.3808 ± 0.0123	0.0000 ± 0.0000	0.0000 ± 0.0000	0.92x ± 0.02x	-7.98% ± 2.21%
matrix	matrix_identity	Identity()	float64	1x1	512x512	0.3988 ± 0.0148	0.4305 ± 0.0188	0.0000 ± 0.0000	0.0000 ± 0.0000	0.93x ± 0.02x	-7.33% ± 1.66%
matrix	matrix_inverse	inverse()	float32	512x512	512x512	58.6058 ± 1.6097	35.6504 ± 0.8409	4.5834 ± 0.1225	7.5333 ± 0.1717	1.64x ± 0.02x	64.39% ± 2.29%
matrix	matrix_inverse	inverse()	float64	512x512	512x512	102.4444 ± 2.8414	52.7110 ± 1.6649	2.6221 ± 0.0707	5.0970 ± 0.1557	1.94x ± 0.01x	94.37% ± 1.36%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float32	512x512	512x1	0.1428 ± 0.0041	0.1392 ± 0.0067	3.6731 ± 0.1037	3.7733 ± 0.1742	1.03x ± 0.02x	2.70% ± 2.44%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float64	512x512	512x1	0.2893 ± 0.0104	0.2291 ± 0.0048	1.8146 ± 0.0635	2.2898 ± 0.0469	1.26x ± 0.03x	26.26% ± 2.50%
matrix	matrix_llt_solve	LLT::solve(x)	float32	512x512	512x1	0.1477 ± 0.0032	0.1441 ± 0.0071	3.5506 ± 0.0759	3.6474 ± 0.1738	1.03x ± 0.03x	2.67% ± 2.82%
matrix	matrix_llt_solve	LLT::solve(x)	float64	512x512	512x1	0.2913 ± 0.0111	0.2313 ± 0.0051	1.8022 ± 0.0653	2.2674 ± 0.0498	1.26x ± 0.03x	25.90% ± 3.10%
matrix	matrix_matmul	A * B	float32	512x512	512x512	22.2456 ± 0.5236	9.3863 ± 0.3394	12.0610 ± 0.2787	28.6032 ± 1.0011	2.37x ± 0.03x	137.10% ± 3.41%
matrix	matrix_matmul	A * B	float64	512x512	512x512	48.6433 ± 1.3853	24.6871 ± 0.6813	5.5170 ± 0.1528	10.8702 ± 0.2933	1.97x ± 0.01x	97.04% ± 1.34%
matrix	matrix_rowwise_sum	rowwise().sum()	float32	512x512	512x1	0.2364 ± 0.0076	0.2372 ± 0.0062	1.1077 ± 0.0343	1.1035 ± 0.0283	1.00x ± 0.02x	-0.35% ± 1.93%
matrix	matrix_rowwise_sum	rowwise().sum()	float64	512x512	512x1	0.6596 ± 0.0482	0.6753 ± 0.0792	0.3983 ± 0.0264	0.3914 ± 0.0376	0.99x ± 0.12x	-1.27% ± 12.09%
matrix	matrix_sub	A - B	float32	512x512	512x512	0.2532 ± 0.0114	0.2513 ± 0.0103	1.0372 ± 0.0454	1.0447 ± 0.0425	1.01x ± 0.05x	0.86% ± 5.46%
matrix	matrix_sub	A - B	float64	512x512	512x512	0.9192 ± 0.0440	0.9067 ± 0.0432	0.2858 ± 0.0136	0.2897 ± 0.0135	1.02x ± 0.09x	1.70% ± 8.57%
matrix	matrix_trace	trace()	float32	512x512	1x1	0.0008 ± 0.0000	0.0008 ± 0.0000	0.6114 ± 0.0214	0.6266 ± 0.0149	1.04x ± 0.06x	3.75% ± 6.04%
matrix	matrix_trace	trace()	float64	512x512	1x1	0.0011 ± 0.0000	0.0011 ± 0.0001	0.4692 ± 0.0192	0.4796 ± 0.0116	1.06x ± 0.05x	5.82% ± 5.02%
matrix	matrix_transpose	transpose()	float32	512x512	512x512	1.9027 ± 0.0676	1.8776 ± 0.0508	0.0000 ± 0.0000	0.0000 ± 0.0000	1.01x ± 0.02x	1.33% ± 1.91%
matrix	matrix_transpose	transpose()	float64	512x512	512x512	3.4747 ± 0.1341	3.5373 ± 0.2039	0.0000 ± 0.0000	0.0000 ± 0.0000	0.98x ± 0.06x	-1.52% ± 5.97%
matrix	matrix_zero	Zero()	float32	1x1	512x512	0.0925 ± 0.0032	0.0923 ± 0.0031	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.03x	0.25% ± 2.73%
matrix	matrix_zero	Zero()	float64	1x1	512x512	0.1882 ± 0.0095	0.1882 ± 0.0106	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.05x	0.14% ± 5.10%

四核对比

taskset -c 0-3

threads：4

module	function	api	dtype	input_size	output_size	no_blas avg_ms mean±sd	blas avg_ms mean±sd	no_blas avg_gflops mean±sd	blas avg_gflops mean±sd	speedup mean±sd	improvement mean±sd
array	vector_abs	abs(a - scalar)	float32	262144x1	262144x1	0.2416 ± 0.0051	0.2392 ± 0.0011	1.0857 ± 0.0224	1.0958 ± 0.0050	1.01x ± 0.02x	0.98% ± 2.31%
array	vector_abs	abs(a - scalar)	float64	262144x1	262144x1	0.6711 ± 0.0086	0.6737 ± 0.0087	0.3907 ± 0.0050	0.3891 ± 0.0049	1.00x ± 0.01x	-0.39% ± 0.93%
array	vector_add_scalar	a + scalar	float32	262144x1	262144x1	0.1981 ± 0.0060	0.1955 ± 0.0039	1.3242 ± 0.0387	1.3413 ± 0.0258	1.01x ± 0.04x	1.37% ± 3.89%
array	vector_add_scalar	a + scalar	float64	262144x1	262144x1	0.5992 ± 0.0114	0.6036 ± 0.0129	0.4376 ± 0.0083	0.4345 ± 0.0092	0.99x ± 0.02x	-0.71% ± 2.05%
array	vector_add_vector	a + b	float32	262144x1	262144x1	0.2482 ± 0.0162	0.2472 ± 0.0172	1.0601 ± 0.0695	1.0652 ± 0.0735	1.01x ± 0.09x	0.81% ± 8.91%
array	vector_add_vector	a + b	float64	262144x1	262144x1	0.8884 ± 0.0304	0.9132 ± 0.0334	0.2954 ± 0.0097	0.2874 ± 0.0106	0.97x ± 0.05x	-2.60% ± 4.79%
array	vector_clamp	cwiseMax + cwiseMin	float32	262144x1	262144x1	0.4353 ± 0.0044	0.4329 ± 0.0011	1.2044 ± 0.0122	1.2110 ± 0.0030	1.01x ± 0.01x	0.55% ± 1.06%
array	vector_clamp	cwiseMax + cwiseMin	float64	262144x1	262144x1	1.0154 ± 0.0089	1.0145 ± 0.0101	0.5164 ± 0.0045	0.5168 ± 0.0051	1.00x ± 0.01x	0.10% ± 0.88%
array	vector_cos	cos(a)	float32	262144x1	262144x1	2.0710 ± 0.0057	2.0510 ± 0.0047	0.1266 ± 0.0004	0.1278 ± 0.0003	1.01x ± 0.00x	0.98% ± 0.32%
array	vector_cos	cos(a)	float64	262144x1	262144x1	10.5539 ± 0.0189	10.5530 ± 0.0136	0.0248 ± 0.0000	0.0248 ± 0.0000	1.00x ± 0.00x	0.01% ± 0.19%
array	vector_div_scalar	a / scalar	float32	262144x1	262144x1	0.6404 ± 0.0043	0.6442 ± 0.0084	0.4094 ± 0.0027	0.4070 ± 0.0052	0.99x ± 0.02x	-0.57% ± 1.66%
array	vector_div_scalar	a / scalar	float64	262144x1	262144x1	2.3229 ± 0.0110	2.3314 ± 0.0146	0.1129 ± 0.0005	0.1124 ± 0.0007	1.00x ± 0.00x	-0.36% ± 0.48%
array	vector_div_vector	a / (b + scalar)	float32	262144x1	262144x1	0.7443 ± 0.0095	0.7498 ± 0.0110	0.7045 ± 0.0090	0.6994 ± 0.0103	0.99x ± 0.02x	-0.71% ± 1.78%
array	vector_div_vector	a / (b + scalar)	float64	262144x1	262144x1	2.5419 ± 0.0166	2.5347 ± 0.0094	0.2063 ± 0.0014	0.2069 ± 0.0008	1.00x ± 0.01x	0.29% ± 0.52%
array	vector_dot	dot()	float32	262144x1	1x1	0.0937 ± 0.0090	0.0959 ± 0.0109	5.6377 ± 0.5052	5.5242 ± 0.5756	0.99x ± 0.14x	-1.20% ± 14.36%
array	vector_dot	dot()	float64	262144x1	1x1	0.4170 ± 0.0202	0.3983 ± 0.0125	1.2601 ± 0.0617	1.3175 ± 0.0411	1.05x ± 0.08x	4.88% ± 7.62%
array	vector_exp	exp(a * scalar)	float32	262144x1	262144x1	2.0859 ± 0.0049	2.0883 ± 0.0097	0.2513 ± 0.0006	0.2511 ± 0.0012	1.00x ± 0.01x	-0.11% ± 0.64%
array	vector_exp	exp(a * scalar)	float64	262144x1	262144x1	6.7441 ± 0.0091	6.6802 ± 0.0097	0.0777 ± 0.0001	0.0785 ± 0.0001	1.01x ± 0.00x	0.96% ± 0.16%
array	vector_fma	a * b + c	float32	262144x1	262144x1	0.4809 ± 0.0120	0.4727 ± 0.0191	1.0909 ± 0.0276	1.1106 ± 0.0442	1.02x ± 0.05x	1.86% ± 4.59%
array	vector_fma	a * b + c	float64	262144x1	262144x1	1.3205 ± 0.0267	1.2779 ± 0.0266	0.3972 ± 0.0082	0.4104 ± 0.0083	1.03x ± 0.03x	3.39% ± 3.49%
array	vector_log	log(a + scalar)	float32	262144x1	262144x1	2.0275 ± 0.0035	2.0242 ± 0.0018	0.2586 ± 0.0004	0.2590 ± 0.0002	1.00x ± 0.00x	0.16% ± 0.24%
array	vector_log	log(a + scalar)	float64	262144x1	262144x1	6.0402 ± 0.0110	6.0428 ± 0.0042	0.0868 ± 0.0002	0.0867 ± 0.0001	1.00x ± 0.00x	-0.04% ± 0.18%
array	vector_max	max(a, b)	float32	262144x1	262144x1	0.3140 ± 0.0156	0.3132 ± 0.0126	0.8368 ± 0.0413	0.8381 ± 0.0327	1.00x ± 0.06x	0.38% ± 6.43%
array	vector_max	max(a, b)	float64	262144x1	262144x1	0.9226 ± 0.0392	0.9106 ± 0.0327	0.2846 ± 0.0119	0.2882 ± 0.0102	1.02x ± 0.07x	1.52% ± 7.27%
array	vector_min	min(a, b)	float32	262144x1	262144x1	0.3115 ± 0.0134	0.3140 ± 0.0160	0.8430 ± 0.0359	0.8369 ± 0.0422	0.99x ± 0.06x	-0.60% ± 5.86%
array	vector_min	min(a, b)	float64	262144x1	262144x1	0.9007 ± 0.0235	0.8977 ± 0.0200	0.2912 ± 0.0073	0.2921 ± 0.0063	1.00x ± 0.04x	0.38% ± 3.61%
array	vector_mul_scalar	a * scalar	float32	262144x1	262144x1	0.2125 ± 0.0057	0.2087 ± 0.0005	1.2342 ± 0.0324	1.2560 ± 0.0032	1.02x ± 0.03x	1.84% ± 2.85%
array	vector_mul_scalar	a * scalar	float64	262144x1	262144x1	0.6229 ± 0.0096	0.6244 ± 0.0117	0.4210 ± 0.0065	0.4200 ± 0.0078	1.00x ± 0.02x	-0.22% ± 1.89%
array	vector_mul_vector	a * b	float32	262144x1	262144x1	0.2597 ± 0.0143	0.2584 ± 0.0164	1.0121 ± 0.0562	1.0180 ± 0.0640	1.01x ± 0.08x	0.83% ± 7.96%
array	vector_mul_vector	a * b	float64	262144x1	262144x1	0.9140 ± 0.0364	0.9077 ± 0.0335	0.2872 ± 0.0114	0.2892 ± 0.0106	1.01x ± 0.06x	0.83% ± 5.81%
array	vector_norm	norm()	float32	262144x1	1x1	0.0824 ± 0.0002	0.0823 ± 0.0001	6.3630 ± 0.0172	6.3693 ± 0.0096	1.00x ± 0.00x	0.07% ± 0.33%
array	vector_norm	norm()	float64	262144x1	1x1	0.1824 ± 0.0056	0.1816 ± 0.0046	2.8773 ± 0.0837	2.8883 ± 0.0691	1.00x ± 0.04x	0.48% ± 4.17%
array	vector_normalize	normalize()	float32	262144x1	262144x1	0.8117 ± 0.0062	0.8099 ± 0.0019	0.9690 ± 0.0073	0.9711 ± 0.0022	1.00x ± 0.01x	0.22% ± 0.89%
array	vector_normalize	normalize()	float64	262144x1	262144x1	2.7192 ± 0.0094	2.7161 ± 0.0083	0.2892 ± 0.0010	0.2895 ± 0.0009	1.00x ± 0.00x	0.11% ± 0.45%
array	vector_segment	segment()	float32	262144x1	131072x1	0.0624 ± 0.0004	0.0623 ± 0.0002	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.01x	0.13% ± 0.75%
array	vector_segment	segment()	float64	262144x1	131072x1	0.1268 ± 0.0050	0.1264 ± 0.0047	0.0000 ± 0.0000	0.0000 ± 0.0000	1.01x ± 0.06x	0.50% ± 5.83%
array	vector_select	mask.select(a, b)	float32	262144x1	262144x1	0.4819 ± 0.0190	0.4808 ± 0.0130	0.5447 ± 0.0205	0.5456 ± 0.0145	1.00x ± 0.05x	0.29% ± 4.98%
array	vector_select	mask.select(a, b)	float64	262144x1	262144x1	1.1194 ± 0.0152	1.1066 ± 0.0102	0.2342 ± 0.0032	0.2369 ± 0.0021	1.01x ± 0.02x	1.17% ± 2.05%
array	vector_sin	sin(a)	float32	262144x1	262144x1	2.0531 ± 0.0053	2.0818 ± 0.0052	0.1277 ± 0.0003	0.1259 ± 0.0003	0.99x ± 0.00x	-1.38% ± 0.41%
array	vector_sin	sin(a)	float64	262144x1	262144x1	10.7043 ± 0.0138	10.7054 ± 0.0090	0.0245 ± 0.0000	0.0245 ± 0.0000	1.00x ± 0.00x	-0.01% ± 0.15%
array	vector_sqrt	(a + scalar).sqrt()	float32	262144x1	262144x1	0.6149 ± 0.0045	0.6135 ± 0.0018	0.8527 ± 0.0061	0.8546 ± 0.0025	1.00x ± 0.01x	0.23% ± 0.85%
array	vector_sqrt	(a + scalar).sqrt()	float64	262144x1	262144x1	2.2786 ± 0.0086	2.2822 ± 0.0073	0.2301 ± 0.0009	0.2298 ± 0.0007	1.00x ± 0.00x	-0.16% ± 0.41%
array	vector_sub_scalar	a - scalar	float32	262144x1	262144x1	0.1981 ± 0.0055	0.1943 ± 0.0013	1.3245 ± 0.0359	1.3489 ± 0.0090	1.02x ± 0.03x	1.92% ± 3.18%
array	vector_sub_scalar	a - scalar	float64	262144x1	262144x1	0.6015 ± 0.0122	0.6018 ± 0.0137	0.4360 ± 0.0088	0.4358 ± 0.0097	1.00x ± 0.02x	-0.02% ± 2.36%
array	vector_sub_vector	a - b	float32	262144x1	262144x1	0.2478 ± 0.0146	0.2471 ± 0.0170	1.0612 ± 0.0633	1.0653 ± 0.0734	1.01x ± 0.09x	0.67% ± 8.65%
array	vector_sub_vector	a - b	float64	262144x1	262144x1	0.9075 ± 0.0381	0.9062 ± 0.0361	0.2893 ± 0.0120	0.2897 ± 0.0115	1.00x ± 0.07x	0.34% ± 6.70%
array	vector_sum	sum()	float32	262144x1	1x1	0.0440 ± 0.0003	0.0440 ± 0.0004	5.9621 ± 0.0391	5.9556 ± 0.0476	1.00x ± 0.01x	-0.16% ± 0.78%
array	vector_sum	sum()	float64	262144x1	1x1	0.0901 ± 0.0054	0.0896 ± 0.0043	2.9172 ± 0.1565	2.9326 ± 0.1269	1.01x ± 0.08x	0.85% ± 7.77%
array	vector_tanh	tanh(a * scalar)	float32	262144x1	262144x1	1.2058 ± 0.0058	1.1980 ± 0.0020	0.4348 ± 0.0021	0.4376 ± 0.0007	1.01x ± 0.01x	0.65% ± 0.61%
array	vector_tanh	tanh(a * scalar)	float64	262144x1	262144x1	15.1154 ± 0.0124	15.1210 ± 0.0172	0.0347 ± 0.0000	0.0347 ± 0.0000	1.00x ± 0.00x	-0.04% ± 0.12%
geometry	quaternion_inverse	Quaternion::inverse()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1016 ± 0.0019	0.1128 ± 0.0026	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_inverse	Quaternion::inverse()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.0733 ± 0.0007	0.0749 ± 0.0005	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_normalize	Quaternion::normalize()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1156 ± 0.0015	0.1277 ± 0.0041	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_normalize	Quaternion::normalize()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.0843 ± 0.0020	0.0888 ± 0.0013	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float32	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.2782 ± 0.0096	0.3108 ± 0.0062	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float64	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.2822 ± 0.0259	0.3023 ± 0.0051	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1125 ± 0.0021	0.1302 ± 0.0022	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1142 ± 0.0035	0.1315 ± 0.0035	1.00x ± 0.00x	0.00% ± 0.00%
matrix	matrix_add	A + B	float32	512x512	512x512	0.2496 ± 0.0168	0.2487 ± 0.0152	1.0548 ± 0.0753	1.0573 ± 0.0631	1.01x ± 0.08x	0.56% ± 7.64%
matrix	matrix_add	A + B	float64	512x512	512x512	0.8835 ± 0.0266	0.8924 ± 0.0257	0.2969 ± 0.0087	0.2940 ± 0.0082	0.99x ± 0.04x	-0.94% ± 3.62%
matrix	matrix_block	block()	float32	512x512	256x256	0.0334 ± 0.0003	0.0333 ± 0.0004	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.01x	0.49% ± 1.17%
matrix	matrix_block	block()	float64	512x512	256x256	0.0630 ± 0.0002	0.0631 ± 0.0005	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.01x	-0.08% ± 0.65%
matrix	matrix_colwise_sum	colwise().sum()	float32	512x512	1x512	0.0463 ± 0.0002	0.0462 ± 0.0002	5.6552 ± 0.0232	5.6621 ± 0.0187	1.00x ± 0.00x	0.15% ± 0.45%
matrix	matrix_colwise_sum	colwise().sum()	float64	512x512	1x512	0.0928 ± 0.0032	0.0929 ± 0.0041	2.8215 ± 0.0953	2.8198 ± 0.1170	1.00x ± 0.06x	0.10% ± 6.48%
matrix	matrix_cwise_product	cwiseProduct	float32	512x512	512x512	0.2625 ± 0.0148	0.2615 ± 0.0144	1.0016 ± 0.0584	1.0053 ± 0.0537	1.01x ± 0.06x	0.55% ± 6.04%
matrix	matrix_cwise_product	cwiseProduct	float64	512x512	512x512	0.8918 ± 0.0423	0.8939 ± 0.0253	0.2945 ± 0.0135	0.2935 ± 0.0080	1.00x ± 0.05x	-0.20% ± 4.65%
matrix	matrix_determinant	determinant()	float32	512x512	1x1	10.5614 ± 0.0747	5.5022 ± 0.0131	8.4726 ± 0.0593	16.2622 ± 0.0386	1.92x ± 0.01x	91.95% ± 1.39%
matrix	matrix_determinant	determinant()	float64	512x512	1x1	17.0516 ± 0.1539	9.4279 ± 0.1019	5.2479 ± 0.0474	9.4918 ± 0.1023	1.81x ± 0.03x	80.88% ± 2.54%
matrix	matrix_gemv	A * x	float32	512x512	512x1	0.0784 ± 0.0007	0.1082 ± 0.0004	6.6805 ± 0.0594	4.8415 ± 0.0170	0.72x ± 0.01x	-27.52% ± 0.55%
matrix	matrix_gemv	A * x	float64	512x512	512x1	0.1802 ± 0.0033	0.2180 ± 0.0106	2.9076 ± 0.0515	2.4073 ± 0.1092	0.83x ± 0.04x	-17.17% ± 4.31%
matrix	matrix_identity	Identity()	float32	1x1	512x512	0.3432 ± 0.0005	0.3741 ± 0.0006	0.0000 ± 0.0000	0.0000 ± 0.0000	0.92x ± 0.00x	-8.25% ± 0.23%
matrix	matrix_identity	Identity()	float64	1x1	512x512	0.3904 ± 0.0060	0.4230 ± 0.0047	0.0000 ± 0.0000	0.0000 ± 0.0000	0.92x ± 0.02x	-7.71% ± 1.63%
matrix	matrix_inverse	inverse()	float32	512x512	512x512	51.1701 ± 0.0814	17.7917 ± 0.0713	5.2460 ± 0.0083	15.0879 ± 0.0603	2.88x ± 0.01x	187.61% ± 1.33%
matrix	matrix_inverse	inverse()	float64	512x512	512x512	89.4104 ± 0.1895	28.1555 ± 0.1274	3.0023 ± 0.0064	9.5342 ± 0.0431	3.18x ± 0.01x	217.56% ± 1.30%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float32	512x512	512x1	0.1406 ± 0.0007	0.1352 ± 0.0010	3.7295 ± 0.0172	3.8787 ± 0.0288	1.04x ± 0.01x	4.01% ± 1.09%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float64	512x512	512x1	0.2820 ± 0.0019	0.2273 ± 0.0010	1.8591 ± 0.0122	2.3067 ± 0.0103	1.24x ± 0.01x	24.09% ± 0.95%
matrix	matrix_llt_solve	LLT::solve(x)	float32	512x512	512x1	0.1459 ± 0.0014	0.1401 ± 0.0009	3.5931 ± 0.0331	3.7411 ± 0.0233	1.04x ± 0.01x	4.12% ± 1.13%
matrix	matrix_llt_solve	LLT::solve(x)	float64	512x512	512x1	0.2869 ± 0.0022	0.2304 ± 0.0022	1.8277 ± 0.0137	2.2758 ± 0.0218	1.25x ± 0.02x	24.52% ± 1.68%
matrix	matrix_matmul	A * B	float32	512x512	512x512	5.6547 ± 0.0307	3.1467 ± 0.0362	47.4266 ± 0.2580	85.2327 ± 0.9693	1.80x ± 0.02x	79.72% ± 2.27%
matrix	matrix_matmul	A * B	float64	512x512	512x512	12.3670 ± 0.0774	7.0604 ± 0.0411	21.6854 ± 0.1353	37.9837 ± 0.2203	1.75x ± 0.01x	75.16% ± 1.40%
matrix	matrix_rowwise_sum	rowwise().sum()	float32	512x512	512x1	0.2312 ± 0.0013	0.2334 ± 0.0012	1.1316 ± 0.0062	1.1208 ± 0.0056	0.99x ± 0.01x	-0.94% ± 0.85%
matrix	matrix_rowwise_sum	rowwise().sum()	float64	512x512	512x1	0.6573 ± 0.0552	0.6698 ± 0.0789	0.4003 ± 0.0300	0.3948 ± 0.0396	0.99x ± 0.15x	-0.58% ± 14.87%
matrix	matrix_sub	A - B	float32	512x512	512x512	0.2495 ± 0.0144	0.2498 ± 0.0153	1.0540 ± 0.0636	1.0526 ± 0.0625	1.00x ± 0.06x	0.05% ± 6.31%
matrix	matrix_sub	A - B	float64	512x512	512x512	0.8966 ± 0.0476	0.8820 ± 0.0110	0.2931 ± 0.0146	0.2972 ± 0.0037	1.02x ± 0.06x	1.67% ± 5.66%
matrix	matrix_trace	trace()	float32	512x512	1x1	0.0008 ± 0.0000	0.0008 ± 0.0000	0.6198 ± 0.0058	0.6326 ± 0.0075	1.00x ± 0.00x	0.00% ± 0.00%
matrix	matrix_trace	trace()	float64	512x512	1x1	0.0011 ± 0.0000	0.0011 ± 0.0001	0.4794 ± 0.0054	0.4864 ± 0.0032	1.04x ± 0.05x	4.00% ± 5.16%
matrix	matrix_transpose	transpose()	float32	512x512	512x512	1.8550 ± 0.0161	1.8488 ± 0.0118	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.01x	0.34% ± 1.21%
matrix	matrix_transpose	transpose()	float64	512x512	512x512	3.4719 ± 0.1458	3.4733 ± 0.1662	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.07x	0.21% ± 7.01%
matrix	matrix_zero	Zero()	float32	1x1	512x512	0.0907 ± 0.0002	0.0910 ± 0.0003	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.00x	-0.26% ± 0.36%
matrix	matrix_zero	Zero()	float64	1x1	512x512	0.1859 ± 0.0072	0.1858 ± 0.0051	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.04x	0.07% ± 4.06%

八核对比

taskset -c 0-7

threads：8

module	function	api	dtype	input_size	output_size	no_blas avg_ms mean±sd	blas avg_ms mean±sd	no_blas avg_gflops mean±sd	blas avg_gflops mean±sd	speedup mean±sd	improvement mean±sd
array	vector_abs	abs(a - scalar)	float32	262144x1	262144x1	0.2215 ± 0.0044	0.2218 ± 0.0042	1.1841 ± 0.0226	1.1821 ± 0.0220	1.00x ± 0.03x	-0.13% ± 3.27%
array	vector_abs	abs(a - scalar)	float64	262144x1	262144x1	0.6470 ± 0.0127	0.6327 ± 0.0088	0.4053 ± 0.0079	0.4144 ± 0.0057	1.02x ± 0.02x	2.26% ± 1.68%
array	vector_add_scalar	a + scalar	float32	262144x1	262144x1	0.1839 ± 0.0062	0.1838 ± 0.0072	1.4268 ± 0.0476	1.4283 ± 0.0548	1.00x ± 0.05x	0.21% ± 5.11%
array	vector_add_scalar	a + scalar	float64	262144x1	262144x1	0.5746 ± 0.0176	0.5733 ± 0.0165	0.4566 ± 0.0134	0.4576 ± 0.0126	1.00x ± 0.04x	0.30% ± 4.11%
array	vector_add_vector	a + b	float32	262144x1	262144x1	0.2411 ± 0.0153	0.2343 ± 0.0145	1.0912 ± 0.0691	1.1227 ± 0.0703	1.03x ± 0.07x	3.09% ± 6.68%
array	vector_add_vector	a + b	float64	262144x1	262144x1	0.8765 ± 0.0190	0.8773 ± 0.0327	0.2992 ± 0.0063	0.2992 ± 0.0106	1.00x ± 0.05x	0.04% ± 4.65%
array	vector_clamp	cwiseMax + cwiseMin	float32	262144x1	262144x1	0.4023 ± 0.0092	0.4031 ± 0.0106	1.3038 ± 0.0285	1.3013 ± 0.0328	1.00x ± 0.03x	-0.16% ± 3.15%
array	vector_clamp	cwiseMax + cwiseMin	float64	262144x1	262144x1	0.9576 ± 0.0140	0.9420 ± 0.0096	0.5476 ± 0.0080	0.5566 ± 0.0057	1.02x ± 0.02x	1.66% ± 1.52%
array	vector_cos	cos(a)	float32	262144x1	262144x1	1.9118 ± 0.0227	1.8944 ± 0.0224	0.1371 ± 0.0016	0.1384 ± 0.0016	1.01x ± 0.01x	0.93% ± 1.28%
array	vector_cos	cos(a)	float64	262144x1	262144x1	9.7176 ± 0.0652	9.6950 ± 0.0125	0.0270 ± 0.0002	0.0270 ± 0.0000	1.00x ± 0.01x	0.23% ± 0.67%
array	vector_div_scalar	a / scalar	float32	262144x1	262144x1	0.5916 ± 0.0084	0.5913 ± 0.0035	0.4432 ± 0.0063	0.4433 ± 0.0026	1.00x ± 0.01x	0.04% ± 1.16%
array	vector_div_scalar	a / scalar	float64	262144x1	262144x1	2.1528 ± 0.0141	2.1794 ± 0.0446	0.1217 ± 0.0008	0.1203 ± 0.0024	0.99x ± 0.02x	-1.17% ± 2.47%
array	vector_div_vector	a / (b + scalar)	float32	262144x1	262144x1	0.6959 ± 0.0173	0.6939 ± 0.0165	0.7538 ± 0.0185	0.7559 ± 0.0175	1.00x ± 0.03x	0.32% ± 2.56%
array	vector_div_vector	a / (b + scalar)	float64	262144x1	262144x1	2.3395 ± 0.0203	2.3372 ± 0.0172	0.2241 ± 0.0019	0.2243 ± 0.0016	1.00x ± 0.01x	0.10% ± 1.34%
array	vector_dot	dot()	float32	262144x1	1x1	0.0888 ± 0.0098	0.0872 ± 0.0112	5.9631 ± 0.6088	6.0899 ± 0.6829	1.03x ± 0.16x	3.11% ± 16.47%
array	vector_dot	dot()	float64	262144x1	1x1	0.4038 ± 0.0222	0.3922 ± 0.0146	1.3019 ± 0.0724	1.3384 ± 0.0478	1.03x ± 0.04x	2.97% ± 4.46%
array	vector_exp	exp(a * scalar)	float32	262144x1	262144x1	1.9252 ± 0.0191	1.9539 ± 0.0824	0.2724 ± 0.0027	0.2687 ± 0.0104	0.99x ± 0.03x	-1.35% ± 3.20%
array	vector_exp	exp(a * scalar)	float64	262144x1	262144x1	6.2123 ± 0.0463	6.1377 ± 0.0068	0.0844 ± 0.0006	0.0854 ± 0.0001	1.01x ± 0.01x	1.22% ± 0.79%
array	vector_fma	a * b + c	float32	262144x1	262144x1	0.4734 ± 0.0156	0.4852 ± 0.0254	1.1085 ± 0.0364	1.0831 ± 0.0564	0.98x ± 0.05x	-2.26% ± 4.70%
array	vector_fma	a * b + c	float64	262144x1	262144x1	1.2533 ± 0.0297	1.2412 ± 0.0196	0.4185 ± 0.0099	0.4225 ± 0.0065	1.01x ± 0.03x	1.01% ± 3.19%
array	vector_log	log(a + scalar)	float32	262144x1	262144x1	1.8750 ± 0.0289	1.8875 ± 0.0561	0.2797 ± 0.0042	0.2780 ± 0.0078	0.99x ± 0.01x	-0.62% ± 1.47%
array	vector_log	log(a + scalar)	float64	262144x1	262144x1	5.5634 ± 0.0324	5.5517 ± 0.0101	0.0942 ± 0.0005	0.0944 ± 0.0002	1.00x ± 0.01x	0.21% ± 0.64%
array	vector_max	max(a, b)	float32	262144x1	262144x1	0.2948 ± 0.0096	0.2913 ± 0.0125	0.8900 ± 0.0296	0.9015 ± 0.0391	1.01x ± 0.06x	1.40% ± 5.62%
array	vector_max	max(a, b)	float64	262144x1	262144x1	0.8995 ± 0.0250	0.8867 ± 0.0369	0.2916 ± 0.0081	0.2961 ± 0.0117	1.02x ± 0.04x	1.57% ± 4.16%
array	vector_min	min(a, b)	float32	262144x1	262144x1	0.2958 ± 0.0119	0.2900 ± 0.0112	0.8876 ± 0.0355	0.9052 ± 0.0353	1.02x ± 0.05x	2.10% ± 5.13%
array	vector_min	min(a, b)	float64	262144x1	262144x1	0.8940 ± 0.0165	0.8842 ± 0.0235	0.2933 ± 0.0054	0.2967 ± 0.0076	1.01x ± 0.02x	1.16% ± 2.47%
array	vector_mul_scalar	a * scalar	float32	262144x1	262144x1	0.1970 ± 0.0055	0.1955 ± 0.0044	1.3317 ± 0.0367	1.3414 ± 0.0294	1.01x ± 0.04x	0.80% ± 3.68%
array	vector_mul_scalar	a * scalar	float64	262144x1	262144x1	0.5933 ± 0.0114	0.6234 ± 0.0583	0.4420 ± 0.0085	0.4234 ± 0.0347	0.96x ± 0.07x	-4.23% ± 7.32%
array	vector_mul_vector	a * b	float32	262144x1	262144x1	0.2523 ± 0.0169	0.2513 ± 0.0180	1.0431 ± 0.0683	1.0479 ± 0.0761	1.01x ± 0.06x	0.58% ± 6.05%
array	vector_mul_vector	a * b	float64	262144x1	262144x1	0.8744 ± 0.0260	0.8814 ± 0.0277	0.3000 ± 0.0085	0.2977 ± 0.0092	0.99x ± 0.04x	-0.71% ± 4.15%
array	vector_norm	norm()	float32	262144x1	1x1	0.0766 ± 0.0025	0.0759 ± 0.0008	6.8540 ± 0.2127	6.9066 ± 0.0737	1.01x ± 0.03x	0.88% ± 3.41%
array	vector_norm	norm()	float64	262144x1	1x1	0.1679 ± 0.0066	0.1707 ± 0.0089	3.1263 ± 0.1158	3.0791 ± 0.1523	0.99x ± 0.07x	-1.32% ± 7.21%
array	vector_normalize	normalize()	float32	262144x1	262144x1	0.7610 ± 0.0260	0.7496 ± 0.0074	1.0345 ± 0.0338	1.0492 ± 0.0103	1.02x ± 0.04x	1.53% ± 3.65%
array	vector_normalize	normalize()	float64	262144x1	262144x1	2.5184 ± 0.0235	2.5160 ± 0.0116	0.3123 ± 0.0029	0.3126 ± 0.0014	1.00x ± 0.01x	0.10% ± 1.01%
array	vector_segment	segment()	float32	262144x1	131072x1	0.0586 ± 0.0026	0.0588 ± 0.0030	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.07x	-0.20% ± 6.67%
array	vector_segment	segment()	float64	262144x1	131072x1	0.1173 ± 0.0058	0.1183 ± 0.0072	0.0000 ± 0.0000	0.0000 ± 0.0000	0.99x ± 0.07x	-0.54% ± 7.38%
array	vector_select	mask.select(a, b)	float32	262144x1	262144x1	0.4471 ± 0.0103	0.4443 ± 0.0165	0.5865 ± 0.0133	0.5908 ± 0.0215	1.01x ± 0.05x	0.78% ± 4.58%
array	vector_select	mask.select(a, b)	float64	262144x1	262144x1	1.0560 ± 0.0177	1.0460 ± 0.0093	0.2483 ± 0.0041	0.2506 ± 0.0022	1.01x ± 0.02x	0.96% ± 1.68%
array	vector_sin	sin(a)	float32	262144x1	262144x1	1.9014 ± 0.0312	1.9291 ± 0.0312	0.1379 ± 0.0022	0.1359 ± 0.0022	0.99x ± 0.01x	-1.43% ± 0.74%
array	vector_sin	sin(a)	float64	262144x1	262144x1	9.8594 ± 0.0947	9.8342 ± 0.0129	0.0266 ± 0.0003	0.0267 ± 0.0001	1.00x ± 0.01x	0.26% ± 0.94%
array	vector_sqrt	(a + scalar).sqrt()	float32	262144x1	262144x1	0.5683 ± 0.0099	0.5801 ± 0.0299	0.9228 ± 0.0157	0.9057 ± 0.0432	0.98x ± 0.05x	-1.83% ± 4.73%
array	vector_sqrt	(a + scalar).sqrt()	float64	262144x1	262144x1	2.1125 ± 0.0147	2.1021 ± 0.0069	0.2482 ± 0.0017	0.2494 ± 0.0008	1.00x ± 0.01x	0.50% ± 0.83%
array	vector_sub_scalar	a - scalar	float32	262144x1	262144x1	0.1826 ± 0.0062	0.1828 ± 0.0076	1.4372 ± 0.0472	1.4364 ± 0.0567	1.00x ± 0.04x	-0.02% ± 3.94%
array	vector_sub_scalar	a - scalar	float64	262144x1	262144x1	0.5716 ± 0.0095	0.6237 ± 0.1619	0.4587 ± 0.0077	0.4365 ± 0.0695	0.95x ± 0.15x	-4.94% ± 14.69%
array	vector_sub_vector	a - b	float32	262144x1	262144x1	0.2433 ± 0.0161	0.2384 ± 0.0223	1.0818 ± 0.0718	1.1080 ± 0.0982	1.03x ± 0.09x	2.63% ± 9.26%
array	vector_sub_vector	a - b	float64	262144x1	262144x1	0.8753 ± 0.0231	0.8923 ± 0.0464	0.2997 ± 0.0077	0.2944 ± 0.0146	0.98x ± 0.06x	-1.64% ± 6.23%
array	vector_sum	sum()	float32	262144x1	1x1	0.0405 ± 0.0006	0.0402 ± 0.0001	6.4745 ± 0.1026	6.5170 ± 0.0213	1.01x ± 0.01x	0.67% ± 1.41%
array	vector_sum	sum()	float64	262144x1	1x1	0.0823 ± 0.0038	0.0875 ± 0.0116	3.1913 ± 0.1329	3.0371 ± 0.3500	0.95x ± 0.13x	-4.54% ± 12.79%
array	vector_tanh	tanh(a * scalar)	float32	262144x1	262144x1	1.1129 ± 0.0115	1.1068 ± 0.0162	0.4712 ± 0.0048	0.4738 ± 0.0068	1.01x ± 0.01x	0.56% ± 1.10%
array	vector_tanh	tanh(a * scalar)	float64	262144x1	262144x1	13.9063 ± 0.1157	13.8740 ± 0.0113	0.0377 ± 0.0003	0.0378 ± 0.0000	1.00x ± 0.01x	0.23% ± 0.81%
geometry	quaternion_inverse	Quaternion::inverse()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1102 ± 0.0012	0.1240 ± 0.0111	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_inverse	Quaternion::inverse()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.0798 ± 0.0004	0.0816 ± 0.0010	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_normalize	Quaternion::normalize()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1261 ± 0.0021	0.1372 ± 0.0055	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_normalize	Quaternion::normalize()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.0926 ± 0.0010	0.0963 ± 0.0016	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float32	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.3043 ± 0.0025	0.3390 ± 0.0056	1.00x ± 0.00x	0.00% ± 0.00%
geometry	quaternion_to_rotation_matrix	Quaternion::toRotationMatrix()	float64	3x1	3x3	0.0001 ± 0.0000	0.0001 ± 0.0000	0.2987 ± 0.0048	0.3307 ± 0.0036	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float32	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1232 ± 0.0018	0.1422 ± 0.0043	1.00x ± 0.00x	0.00% ± 0.00%
geometry	vector_cross	cross()	float64	3x1	3x1	0.0001 ± 0.0000	0.0001 ± 0.0000	0.1248 ± 0.0019	0.1533 ± 0.0148	1.00x ± 0.00x	0.00% ± 0.00%
matrix	matrix_add	A + B	float32	512x512	512x512	0.2444 ± 0.0162	0.2363 ± 0.0151	1.0768 ± 0.0690	1.1134 ± 0.0731	1.04x ± 0.11x	3.90% ± 10.64%
matrix	matrix_add	A + B	float64	512x512	512x512	0.8806 ± 0.0214	0.8841 ± 0.0331	0.2979 ± 0.0072	0.2969 ± 0.0108	1.00x ± 0.04x	-0.30% ± 3.90%
matrix	matrix_block	block()	float32	512x512	256x256	0.0307 ± 0.0004	0.0306 ± 0.0003	0.0000 ± 0.0000	0.0000 ± 0.0000	1.01x ± 0.02x	0.60% ± 2.13%
matrix	matrix_block	block()	float64	512x512	256x256	0.0577 ± 0.0003	0.0579 ± 0.0010	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.02x	-0.42% ± 1.83%
matrix	matrix_colwise_sum	colwise().sum()	float32	512x512	1x512	0.0428 ± 0.0008	0.0426 ± 0.0007	6.1163 ± 0.1147	6.1366 ± 0.0896	1.00x ± 0.03x	0.40% ± 2.56%
matrix	matrix_colwise_sum	colwise().sum()	float64	512x512	1x512	0.0878 ± 0.0077	0.0848 ± 0.0036	2.9994 ± 0.2339	3.0920 ± 0.1214	1.04x ± 0.09x	3.65% ± 8.98%
matrix	matrix_cwise_product	cwiseProduct	float32	512x512	512x512	0.2589 ± 0.0170	0.2466 ± 0.0136	1.0166 ± 0.0655	1.0661 ± 0.0618	1.05x ± 0.09x	5.23% ± 8.60%
matrix	matrix_cwise_product	cwiseProduct	float64	512x512	512x512	0.8840 ± 0.0300	0.8962 ± 0.0369	0.2968 ± 0.0097	0.2929 ± 0.0121	0.99x ± 0.03x	-1.29% ± 3.10%
matrix	matrix_determinant	determinant()	float32	512x512	1x1	12.1405 ± 1.0030	5.2100 ± 0.1221	7.4167 ± 0.6275	17.1831 ± 0.4104	2.33x ± 0.21x	133.20% ± 20.93%
matrix	matrix_determinant	determinant()	float64	512x512	1x1	17.5771 ± 1.1407	9.4609 ± 1.0266	5.1098 ± 0.3295	9.5376 ± 0.8177	1.87x ± 0.19x	87.22% ± 19.18%
matrix	matrix_gemv	A * x	float32	512x512	512x1	0.0728 ± 0.0024	0.0998 ± 0.0011	7.2020 ± 0.2387	5.2476 ± 0.0586	0.73x ± 0.02x	-27.08% ± 2.31%
matrix	matrix_gemv	A * x	float64	512x512	512x1	0.1675 ± 0.0050	0.2008 ± 0.0099	3.1296 ± 0.0911	2.6139 ± 0.1215	0.84x ± 0.05x	-16.39% ± 4.98%
matrix	matrix_identity	Identity()	float32	1x1	512x512	0.3157 ± 0.0035	0.3437 ± 0.0031	0.0000 ± 0.0000	0.0000 ± 0.0000	0.92x ± 0.01x	-8.14% ± 1.38%
matrix	matrix_identity	Identity()	float64	1x1	512x512	0.3610 ± 0.0094	0.3946 ± 0.0177	0.0000 ± 0.0000	0.0000 ± 0.0000	0.92x ± 0.04x	-8.38% ± 4.02%
matrix	matrix_inverse	inverse()	float32	512x512	512x512	50.5584 ± 0.9010	15.5381 ± 1.3623	5.3109 ± 0.0952	17.3835 ± 1.3633	3.27x ± 0.24x	227.24% ± 24.16%
matrix	matrix_inverse	inverse()	float64	512x512	512x512	85.6748 ± 0.6326	25.2014 ± 1.8696	3.1333 ± 0.0230	10.7039 ± 0.7830	3.42x ± 0.26x	241.66% ± 25.55%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float32	512x512	512x1	0.1301 ± 0.0041	0.1256 ± 0.0023	4.0339 ± 0.1187	4.1760 ± 0.0738	1.04x ± 0.05x	3.64% ± 4.56%
matrix	matrix_ldlt_solve	LDLT::solve(x)	float64	512x512	512x1	0.2601 ± 0.0030	0.2116 ± 0.0029	2.0160 ± 0.0226	2.4787 ± 0.0328	1.23x ± 0.02x	22.96% ± 1.98%
matrix	matrix_llt_solve	LLT::solve(x)	float32	512x512	512x1	0.1339 ± 0.0007	0.1297 ± 0.0010	3.9156 ± 0.0201	4.0419 ± 0.0303	1.03x ± 0.01x	3.24% ± 1.20%
matrix	matrix_llt_solve	LLT::solve(x)	float64	512x512	512x1	0.2642 ± 0.0040	0.2143 ± 0.0028	1.9848 ± 0.0296	2.4463 ± 0.0314	1.23x ± 0.03x	23.29% ± 3.13%
matrix	matrix_matmul	A * B	float32	512x512	512x512	3.0865 ± 0.4786	1.7980 ± 0.5509	88.2614 ± 9.8672	158.5934 ± 34.6302	1.84x ± 0.55x	83.85% ± 54.82%
matrix	matrix_matmul	A * B	float64	512x512	512x512	6.3781 ± 0.7170	3.9874 ± 1.1349	42.4325 ± 3.8308	70.3070 ± 12.0051	1.67x ± 0.35x	67.29% ± 35.34%
matrix	matrix_rowwise_sum	rowwise().sum()	float32	512x512	512x1	0.2161 ± 0.0075	0.2155 ± 0.0026	1.2117 ± 0.0397	1.2145 ± 0.0146	1.00x ± 0.04x	0.34% ± 4.24%
matrix	matrix_rowwise_sum	rowwise().sum()	float64	512x512	512x1	0.6470 ± 0.1152	0.6061 ± 0.0617	0.4141 ± 0.0609	0.4350 ± 0.0369	1.08x ± 0.22x	7.55% ± 21.61%
matrix	matrix_sub	A - B	float32	512x512	512x512	0.2458 ± 0.0175	0.2370 ± 0.0148	1.0712 ± 0.0730	1.1103 ± 0.0735	1.04x ± 0.11x	4.21% ± 11.15%
matrix	matrix_sub	A - B	float64	512x512	512x512	0.8784 ± 0.0195	0.8945 ± 0.0402	0.2986 ± 0.0065	0.2936 ± 0.0130	0.98x ± 0.04x	-1.65% ± 4.48%
matrix	matrix_trace	trace()	float32	512x512	1x1	0.0008 ± 0.0000	0.0007 ± 0.0000	0.6784 ± 0.0045	0.6900 ± 0.0081	1.10x ± 0.07x	10.00% ± 6.90%
matrix	matrix_trace	trace()	float64	512x512	1x1	0.0010 ± 0.0000	0.0011 ± 0.0004	0.5202 ± 0.0068	0.4994 ± 0.0966	0.95x ± 0.19x	-4.54% ± 18.59%
matrix	matrix_transpose	transpose()	float32	512x512	512x512	1.7316 ± 0.0269	1.7152 ± 0.0236	0.0000 ± 0.0000	0.0000 ± 0.0000	1.01x ± 0.01x	0.96% ± 1.26%
matrix	matrix_transpose	transpose()	float64	512x512	512x512	3.3745 ± 0.2731	3.2961 ± 0.1231	0.0000 ± 0.0000	0.0000 ± 0.0000	1.02x ± 0.08x	2.41% ± 7.99%
matrix	matrix_zero	Zero()	float32	1x1	512x512	0.0841 ± 0.0018	0.0838 ± 0.0009	0.0000 ± 0.0000	0.0000 ± 0.0000	1.00x ± 0.03x	0.41% ± 2.68%
matrix	matrix_zero	Zero()	float64	1x1	512x512	0.1706 ± 0.0086	0.1716 ± 0.0048	0.0000 ± 0.0000	0.0000 ± 0.0000	0.99x ± 0.05x	-0.56% ± 4.75%