Compiling from source
Compiling NumPower from the source allows you to customize the installation and optimize it for your specific environment. This method requires more manual steps but provides flexibility and control over the installation process.
Before installing NumPower, ensure that your system meets the following requirements:
- PHP development files and header files. This may vary based on your operating system. Consult your system's package manager or PHP documentation for specific instructions.
- C++ compiler with C++11 support (such as GCC or Clang).
- CBLAS: CBLAS library (for CPU-based linear algebra operations)
- LAPACKE: LAPACKE library (for advanced linear algebra operations)
Optional requirements (for enhanced performance and additional functionalities):
- CUBLAS: NVIDIA CUDA Toolkit with CUBLAS library (for GPU-accelerated linear algebra operations)
- CUDA development toolkit (NVCC) and the necessary CUDA header files
Clone the NumPower source code repository from the official GitHub repository to your local machine.
Build and compile the PHP extension by running the appropriate commands specific to your environment.
$ make install
For now, when using the build with GPU support, AVX2 will be disabled even when available (due to a limitation of NVCC itself) so some CPU operations may be slower.
We are studying a way to mix NVCC and GCC to use both technologies simultaneously.
$ ./configure --with-cuda
$ make install-cuda
Once the compilation is complete, you will obtain a compiled PHP extension file. If necessary, copy the compiled extension file to the appropriate location in your PHP extensions directory.
Update your PHP configuration file (php.ini) to include the extension entry for NumPower. Add the following line to the