#pragma once

void convolution(
    float *_input, float *_output, float *_weight, 
    int _N, int _C, int _H, int _W,
    int _K, int _R, int _S,
    int _pad, int _dilation, int _stride);

void convolution_init( 
    int _N, int _C, int _H, int _W,
    int _K, int _R, int _S,
    int _pad, int _dilation, int _stride);

void convolution_final(
    int _N, int _C, int _H, int _W,
    int _K, int _R, int _S,
    int _pad, int _dilation, int _stride);

void cuda_device_init(void);
void cuda_device_malloc(void);
void cuda_memcpy_host_to_device(void);
void cuda_memcpy_device_to_host(void);
void cuda_kernel_call(void);

void print_filter(float *filter, int K, int C, int R, int S);
void print_input(float *input, int N, int C, int H, int W);