Add support for non square matrices

This commit is contained in:
2024-05-20 15:20:43 +02:00
parent 6f8b5f4081
commit 74098b24e3
21 changed files with 314 additions and 299 deletions

View File

@@ -2,27 +2,28 @@
#define CUDANET_POOLING_H
#include <cuda_runtime.h>
#include "layer.cuh"
namespace CUDANet::Kernels {
__global__ void max_pooling(
const float* __restrict__ d_input,
float* __restrict__ d_output,
const int inputSize,
const int outputSize,
const dim2d inputSize,
const dim2d outputSize,
const int nChannels,
const int poolingSize,
const int stride
const dim2d poolingSize,
const dim2d stride
);
__global__ void avg_pooling(
const float* __restrict__ d_input,
float* __restrict__ d_output,
const int inputSize,
const int outputSize,
const dim2d inputSize,
const dim2d outputSize,
const int nChannels,
const int poolingSize,
const int stride
const dim2d poolingSize,
const dim2d stride
);
} // namespace CUDANet::Kernels