Simple implemenetation of 2D convolution using HIP on AMD GPU

This code repository pertains to Bangtian's introductory task, focusing on the implementation of 2D convolution using HIP on AMD GPU.

Setup

1. Build: make
2. Run: ./conv2d_gpu

parameters for convolution can be changed in conv_config.hpp.

Bangtian's Convolution Implementations

ID	Kernel Method	Comment
1	direct 2D convolution	shared memory is used, NCHW format (kernels/1_*.hpp)

(Additional implementations will be included, with the im2col implementation currently in progress.)