Quantum Based Cuda Kernel

This project demonstrates advanced CUDA kernels for deep learning workloads and provides:

• High-performance GEMM (classical and Tensor Core)
• Mixed precision (FP16) support
• Fused GEMM+ReLU kernel
• Quantum-inspired search kernel
• Auto-tuning logic
• CUDA Graphs for reduced launch overhead
• Comprehensive unit tests (GoogleTest)
• Performance logging and plotting (CSV and matplotlib)

Requirements

• NVIDIA GPU with CUDA support (Compute Capability >= 7.0 for Tensor Cores)
• CUDA Toolkit installed
• CMake >= 3.10
• Python 3 and matplotlib (pip install matplotlib)

Usage

./run_all.sh