Skip to content

Latest commit

 

History

History
13 lines (7 loc) · 618 Bytes

README.md

File metadata and controls

13 lines (7 loc) · 618 Bytes

Working GPU parallelized linear algebra library in CUDA C++!

The Linear Algebra library that powers all of this (Matrix.cu) is fully working. Please check it out!

XOR.cu is a demonstration of the autodiff library successfully solving the XOR problem. MNIST.cu will be implemented soon, with a demonstration of a simple feed-forward network on the MNIST dataset

TODOs:

-figure out the cuda linker errors so that i don't need to copy the matrix and autodiff library components into new main files for each demonstration

-add mnist

-add cache tiling to the matmuls, get rid of the superfluous matrix fill kernel