search

melsonGit / serial_cpp_vs_parallel_cuda

Code work for completion of MSc ICT. The original version of this project was submitted alongside a 15,000 word thesis paper and put into practice the theoretical CPU and GPGPU/GPU elements explored in the write-up. I continued to develop this project post-graduation.
0 stars 1 forks source link

GPU twoConv container change; native arrays -> 2D vectors #5

Open melsonGit opened 2 years ago

melsonGit commented 2 years ago

Issue

GPU twoConv containers need changing from native arrays to 2D vectors. This will resemble its CPU counterpart.

Additional Info

GPU twoConv currently uses a native array approach (as used in the original implementation), however, this should be moved to a 2D vector.

melsonGit commented 2 years ago

Such implementation requires significant changes in GPU twoConvFunc() and the overall core twoConv operation logic. This would also affect how data is copied from device to host and vice versa. With this, I may revisit this 'enhancement' in the future.

Closing.

melsonGit commented 2 years ago

With constant memory problems observed in issue #8, I will reopen this just in case I'm unable to find a resolution.

melsonGit commented 2 years ago

Issue #8 closed, however I believe it's worth keeping this open as I plan to implement 2D vectors in the near future.