LeetGPU

This repository contains my CUDA C++ solutions to the LeetGPU challenges.

I aim to use optimized methods for each problem and document my thought process.

Additionally, I have created test cases for every problem. You can verify the correctness and robustness of your solution by running:

./run.sh <PROBLEM_PATH> [kernel.cu]

For example, if you want to test the solution for "Easy/Vector_Addition":

./run.sh ./CUDA/Easy/Vector_Addition 

If you do not provide a kernel name, the default is kernel.cu.

---

TODO List (Updated 2025/09/20)

• Easy
  • Vector Addition
  • Matrix Multiplication
  • Matrix Transpose
  • Color Inversion
  • 1D Convolution
  • Reverse Array
  • ReLU
  • Leaky ReLU
  • Rainbow Table
  • Matrix Copy
  • Count Array Element
  • Count 2D Array Element
  • Sigmoid Linear Unit
  • Swish-Gated Linear Unit
• Medium
  • Reduction
  • Softmax
  • Softmax Attention
  • 2D Convolution
  • Histogramming
  • Sorting
  • Prefix Sum
  • Dot Product
  • Sparse Matrix-Vector Multiplication
  • General Matrix Multiplication
  • Categorical Cross Entropy Loss
  • Mean Square Error
  • Gaussian Blur
  • Ordinary Least Squares
  • Logistic Regression
  • Monte Carlo Integration
  • Radix Sort
  • Matrix Power
  • Nearest Neighbor
  • Batch Normalization
  • 2D Max Pooling
  • Count 3D Array Element
  • BFS Shortest Path
  • Subarray Sum
  • 2D Subarray Sum
  • 3D Subarray Sum
  • RMS Normalization
  • Max Subarray Sum
  • Attention with Linear Biases
• Hard
  • 3D Convolution
  • Multi-Head Attention
  • Multi-Agent Simulation
  • K-Means Clustering
  • Fast Fourier Transform
  • Causal Self-Attention
  • Linear Self-Attention