Rindow Math Matrix

Overview

“Rindow Math Matrix” is a library that provides a vector operation environment.

It is designed to be similar to Python’s NumPy, helping many people save learning time.

It has the following features:

  • Implements a common array object interface, “NDArray”.
  • Provides a flexible N-dimensional array operation library.
  • Can use OpenBLAS as an external backend library (optional).
  • Enables the use of hardware like GPUs via OpenCL (optional).

Module Structure

“Rindow Math Matrix” is modularized to support various operating environments. Each module is replaceable, and new functionalities can be added.

Structure Diagram:

Structure Diagram

Requirements

  • PHP 8.1, 8.2, 8.3, 8.4
    • (If using in a PHP 7.2 to 8.0 environment, please use release 1.1.)
  • Windows 10, 11, or Linux (when using OpenBLAS)

Recommends

Installation

Installing Rindow Math Matrix

Please set up using composer.

$ composer require rindow/rindow-math-matrix

If graph display is needed, set up rindow-math-plot.

For Windows, enable the GD extension in php.ini.

extension=gd

For Linux, install the GD extension.

$ sudo apt install phpX.X-gd

Install rindow-math-plot.

$ composer require rindow/rindow-math-plot

For Linux, image viewer configuration is required for rindow-math-plot.

$ RINDOW_MATH_PLOT_VIEWER=/some/bin/dir/png-file-viewer
$ export RINDOW_MATH_PLOT_VIEWER

Note: Specify something like “viewnior” for RINDOW_MATH_PLOT_VIEWER.

Accelerate with CPU only

First, enable FFI and install external libraries.

For Windows:

Enable FFI. Enable the FFI extension in php.ini.

extension=ffi

Confirm that FFI is enabled.

work> php -m | findstr "FFI"

Download and extract the Windows binaries for OpenBLAS from the following site:

  • https://github.com/OpenMathLib/OpenBLAS/releases/ Download and extract the Windows binaries for Rindow-Matlib from the following site:
  • https://github.com/rindow/rindow-matlib/releases/

Set the respective bin directories in your PATH.

work> SET PATH=%PATH%;C:\path\to\OpenBLAS\bin;C:\path\to\Matlib\bin

For Linux:

Don’t forget to install the FFI extension, OpenBLAS, and LAPACKe using the apt command.

$ sudo apt install phpX.X-ffi libopenblas0 liblapacke

Download and install the Linux binaries for Rindow-Matlib from the following site:

  • https://github.com/rindow/rindow-matlib/releases/ 
    $ wget https://github.com/rindow/rindow-matlib/releases/download/vX.X.X/rindow-matlib_X.X.X_amd64.deb
$ sudo apt install ./rindow-matlib_X.X.X_amd64.deb

Install the service for acceleration. Common for Windows and Linux

Please install rindow-math-matrix-matlibffi.

$ composer require rindow/rindow-math-matrix-matlibffi

Confirm the installation.

$ vendor/bin/rindow-math-matrix
Service Level   : Advanced
Buffer Factory  : Rindow\Math\Buffer\FFI\BufferFactory
BLAS Driver     : Rindow\OpenBLAS\FFI\Blas
LAPACK Driver   : Rindow\OpenBLAS\FFI\Lapack
Math Driver     : Rindow\Matlib\FFI\Matlib

If the service level is Basic, external libraries are not available for some reason. Check with the command’s verbose option.

$ vendor/bin/rindow-math-matrix -v

For detailed installation instructions, see here.

Accelerate with GPU

First, configure the libraries as described in the previous section “Accelerate with CPU only”. Then, install the GPU acceleration library.

For Windows:

Download and extract the Windows binaries for CLBlast from the following site:

  • https://github.com/CNugteren/CLBlast/releases

Set the bin directory in your PATH.

work> SET PATH=%PATH%;C:\path\to\CLBlast\bin

For Linux: Install OpenCL and CLBlast. OpenCL drivers vary for each hardware. Standard Ubuntu provides the following:

  • For Intel iGPU: intel-opencl-icd
  • For AMD APU: mesa-opencl-icd Other GPUs require drivers specific to their manufacturers.
$ sudo apt install clinfo
$ sudo apt install XXX-XXX-icd  # Replace XXX-XXX-icd with your specific OpenCL ICD loader
$ sudo apt install libclblast0

Confirm that the service level becomes “Accelerated” using the command. Common for Windows and Linux

$ vendor/bin/rindow-math-matrix
Service Level   : Accelerated
Buffer Factory  : Rindow\Math\Buffer\FFI\BufferFactory
BLAS Driver     : Rindow\OpenBLAS\FFI\Blas
LAPACK Driver   : Rindow\OpenBLAS\FFI\Lapack
Math Driver     : Rindow\Matlib\FFI\Matlib
OpenCL Factory  : Rindow\OpenCL\FFI\OpenCLFactory
CLBlast Factory : Rindow\CLBlast\FFI\CLBlastFactory

If the service level is Basic or Advanced, external libraries are not available for some reason. Check with the command’s verbose option.

$ vendor/bin/rindow-math-matrix -v

For detailed installation instructions, see here

How to use Rindow Math Matrix

<?php
include 'vendor/autoload.php';

$mo = new Rindow\Math\Matrix\MatrixOperator();

$a = $mo->array([1.0, 2.0]);
$b = $mo->array([3.0, 4.0]);

$c = $mo->add($a,$b);

echo $mo->toString($c)."\n";

### To create a graph like this:

$plt = new Rindow\Math\Plot\Plot();

$plt->bar(['x','y'],$c);
$plt->show();

When using a linear algebra library:

<?php
include 'vendor/autoload.php';

$mo = new Rindow\Math\Matrix\MatrixOperator();
$la = $mo->la();

$a = $mo->array([[1.0, 2.0],[3.0, 4.0]]);
$b = $mo->array([[3.0, 4.0],[5.0, 6.0]]);

$c = $la->gemm($a,$b);

echo $mo->toString($c)."\n";

### To create a graph for this:

$plt = new Rindow\Math\Plot\Plot();

$plt->bar(['x','y'],$c);
$plt->show();

When using the GPU version of the linear algebra library:

<?php
include 'vendor/autoload.php';

use Interop\Polite\Math\Matrix\OpenCL;

$mo = new Rindow\Math\Matrix\MatrixOperator();
$la = $mo->laAccelerated('clblast',['deviceType'=>OpenCL::CL_DEVICE_TYPE_GPU]);
$la->blocking(true);

$a = $mo->array([[1.0, 2.0],[3.0, 4.0]]);
$b = $mo->array([[3.0, 4.0],[5.0, 6.0]]);

// Move arrays to GPU memory
$a_gpu = $la->array($a);
$b_gpu = $la->array($b);

// Perform operation on GPU
$c_gpu = $la->gemm($a_gpu,$b_gpu);

// Move result back to host NDArray
$c = $la->toNDArray($c_gpu);

echo $mo->toString($c)."\n";

### To create a graph for this:

$plt = new Rindow\Math\Plot\Plot();

$plt->bar(['x','y'],$c);
$plt->show();