# MatrixOperationsCuda

MatrixOperationsCuda is a Rust crate for performing matrix operations using CUDA

## Installation

• MatrixOperationsCuda

  Add the following to your Cargo.toml file:

  toml [dependencies] matrix_operations_cuda = "0.1.0"

• Matrix_Operations

  To works with matrix_operations_cuda, you need to install the core matrix_operations from matrix_operations

  Add the following to your Cargo.toml file:

  toml [dependencies] matrix_operations = "0.1.3"

• CUDA

  This crate does NOT include CUDA itself. You need to install on your own. the official installer

  Usage

  This crate allow to use common operations using cuda:

  ```rust use matrixoperations::matrix; use matrixoperationscuda::{addmatrices, addscalar, CudaEnv, dot, submatrices};

  let cudaenv; unsafe { cudaenv = CudaEnv::new(0, 0).unwrap(); }

  let m1 = matrix![[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]]; let m2 = matrix![[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]];

  let m3; unsafe { m3 = dot(&m1, &m2, &cuda_env).unwrap(); }

  asserteq!(m3[0], [22.0, 28.0]); asserteq!(m3[1], [49.0, 64.0]);

  let m4; unsafe { m4 = addscalar(&m3, 10.0, &cudaenv).unwrap(); }

  asserteq!(m4[0], [32.0, 38.0]); asserteq!(m4[1], [59.0, 74.0]);

  let m5; unsafe { m5 = submatrices(&m4, &m3, &cudaenv).unwrap(); }

  asserteq!(m5[0], [10.0, 10.0]); asserteq!(m5[1], [10.0, 10.0]); ```

  You also can import your own module from a .ptx file or from a module data as Vec<u8>

  ```rust use cudadriversys::*; use matrixoperationscuda::cudaenv::CudaEnv; use matrixoperationscuda::cudamodule::CudaModule; use matrixoperations::{Matrix, matrix}; use matrixoperationscuda::matrixapply::applyfunctionmatrix;

  unsafe { let mut cuda_env = CudaEnv::new(0, 0).unwrap();

  let module = CudaModule::new(b"resources/kerneltest.ptx\0").unwrap(); let function = module.loadfunction(b"mulby2\0").unwrap();

  let matrix = matrix![[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]];

  let result = applyfunctionmatrix(&matrix, &cuda_env, function).unwrap();

  asserteq!(result[0], [2.0, 4.0, 6.0]); asserteq!(result[1], [8.0, 10.0, 12.0]);

  module.free().unwrap(); } ```

  Features

• Initialize a cuda environment

• Launch common operations on matrices
• Import and use custom kernel to perform custom operations on matrices
• Allocate and Free memory in GPU
• Copy data between Host and Device