Add documentation comments

include/kernels/activations.cuh

@@ -3,9 +3,23 @@
 
 namespace Kernels {
 
+/**
+ * @brief Sigmoid activation function kernel
+ * 
+ * @param src Pointer to the source array
+ * @param dst Pointer to the destination array
+ * @param len Length of the arrays
+ */
 __global__ void
 sigmoid(const float* __restrict__ src, float* __restrict__ dst, int len);
 
+/**
+ * @brief Relu activation function kernel
+ * 
+ * @param src Pointer to the source array
+ * @param dst Pointer to the destination array
+ * @param len Length of the arrays
+ */
 __global__ void
 relu(const float* __restrict__ src, float* __restrict__ dst, int len);
 

include/kernels/convolution.cuh

@@ -3,6 +3,16 @@
 
 namespace Kernels {
 
+/**
+ * @brief Kernel that pads the input matrix with zeros
+ * 
+ * @param d_input Device pointer to the input matrix (as vector)
+ * @param d_padded Device pointer to the padded matrix (as vector)
+ * @param w Width of the input matrix
+ * @param h Height of the input matrix
+ * @param n Number of input channels
+ * @param p Padding size
+ */
 __global__ void padding(
     const float* d_input,
     float*       d_padded,
@@ -12,6 +22,19 @@ __global__ void padding(
     int          p
 );
 
+/**
+ * @brief Convolution kernel
+ * 
+ * @param d_input Device pointer to the input matrix
+ * @param d_kernel Device pointer to the convolution kernel
+ * @param d_output Device pointer to the output matrix
+ * @param inputSize Width and height of the input matrix
+ * @param nChannels Number of channels in the input matrix
+ * @param kernelSize Width and height of the convolution kernel
+ * @param stride Convolution stride
+ * @param nFilters Number of output filters
+ * @param outputSize Width and height of the output matrix
+ */
 __global__ void convolution(
     const float* d_input,
     const float* d_kernel,

include/kernels/matmul.cuh

@@ -3,6 +3,15 @@
 
 namespace Kernels {
 
+/**
+ * @brief Matrix vector multiplication kernel
+ * 
+ * @param d_matrix Device pointer to matrix
+ * @param d_vector Device pointer to vector
+ * @param d_output Device pointer to output vector
+ * @param w Width of the matrix
+ * @param h Height of the matrix
+ */
 __global__ void mat_vec_mul(
     const float* d_matrix,
     const float* d_vector,
@@ -11,6 +20,14 @@ __global__ void mat_vec_mul(
     int          h
 );
 
+/**
+ * @brief Vector vector addition kernel
+ * 
+ * @param d_vector1 Device pointer to first vector
+ * @param d_vector2 Device pointer to second vector
+ * @param d_output Device pointer to output vector
+ * @param w Length of the vectors
+ */
 __global__ void vec_vec_add(
     const float* d_vector1,
     const float* d_vector2,