[ARCHIVED] Cooperative primitives for CUDA C++. See https://github.com/NVIDIA/cccl

study of cutlass

A CUDA concurrency library designed to simplify concurrency programming, offering C++-style wrappers for selected CUDA Runtime APIs

A GPU-Accelerated Ray Tracer with Monte Carlo Rendering

Mean Shift C++17 implementations: Sequential, OpenMP and CUDA

A managed resource for your CUDA augmented NumPy needs in C++ 17.

Parallel Heterogeneous CPU/GPU computing

GPU Programming Specialisation Capstone Project submission : A custom CUDA based real time video processor with varied filters to generate new and updated videos in minimal time. Used CUDA architecture to harness the power of Nvidia developed GPUs, and Kernels

Cuda mini-project to benchmark vector operations on some datatypes.

From-scratch LLM inference engine in C++17/CUDA. Custom kernels, GGUF model loading, quantized inference (Q4/Q8). Runs SmolLM2-135M and Llama 3.2 1B on a 6 GB GPU.

Custom CUDA kernels for accelerating 1.58-bit ternary LLM inference with 2:4 structured sparsity on consumer Ampere GPUs. Exploits both ternary arithmetic (no multiplies) and hardware sparse tensor cores to maximize throughput on RTX 3060. Based on the Sparse-BitNet paper (Zhang et al., 2026).

Fun with CUDA C++