matrix_sigmoid_derivative.cu

Go to the documentation of this file.

/**
 * @file matrix_sigmoid_derivative.cu
 * @brief Implementation of the sigmoid derivative function for matrices.
 */
 
#include "matrix.h"
#include <cuda_runtime.h>
 
/**
 * @brief CUDA kernel for applying the sigmoid derivative function element-wise.
 * @param input Pointer to the input matrix data.
 * @param output Pointer to the output matrix data.
 * @param size Total number of elements in the matrix.
 */
__global__ void sigmoidDerivativeKernel(const double* input, double* output, int size) {
    // Calculate the global thread index
    int idx = blockIdx.x * blockDim.x + threadIdx.x;
    
    // Check if the thread is within the matrix bounds
    if (idx < size) {
        // Compute sigmoid activation
        double sigmoid_x = 1.0 / (1.0 + exp(-input[idx]));
        
        // Apply sigmoid derivative: sigmoid(x) * (1 - sigmoid(x))
        output[idx] = sigmoid_x * (1.0 - sigmoid_x);
    }
}
 
/**
 * @brief Applies the sigmoid derivative function to the matrix.
 * @return A new Matrix object with sigmoid derivative applied.
 */
Matrix Matrix::sigmoid_derivative() const {
    // Create a new matrix with the same dimensions
    Matrix result(rows, cols);
    
    // Calculate the total number of elements
    int size = rows * cols;
    
    // Define the number of threads per block
    int threadsPerBlock = 256;
    
    // Calculate the number of blocks needed
    int blocksPerGrid = (size + threadsPerBlock - 1) / threadsPerBlock;
    
    // Launch the CUDA kernel
    sigmoidDerivativeKernel<<<blocksPerGrid, threadsPerBlock>>>(d_data, result.d_data, size);
    
    // Synchronize to ensure the kernel execution is complete
    cudaDeviceSynchronize();
    
    return result;
}