Fast, reproducible, and portable software development environments
Remote development on HPC clusters with VSCode
CLI wrapper and Claude Code / Codex skill for NVIDIA Nsight Graphics 2026.1+ — capture GPU frames, export GPU Trace, and drill into bottlenecks via compact JSON.
Accelerate and optimize existing C/C++ CPU-only applications using the most essential CUDA tools and techniques.
Matrix multiplication example performed with OpenMP, OpenACC, BLAS, cuBLABS, and CUDA
High-performance Sobel edge detection using CUDA with CPU vs GPU benchmarking, roofline analysis, and Nsight profiling.
Reproducible GPU performance workflow with Nsight CSV parsing, Roofline analysis and portable backend seams.
Hand-written CUDA kernels + Qwen3-VL ViT TensorRT deployment. LLM/VLM inference optimization
A simple and understandable CUDA kernel for batch-matmul operation
16-step CUDA optimization of FlashAttention-2 achieving 99.2% of official performance on A100 — Ampere architecture
Repository for Architecture of computers and parallel systems course on VŠB
Reading Nsight Systems traces to find why a CUDA kernel is slow — transpose bandwidth on A100, a bank-conflict diagnosis, and a tool that extracts kernel timings from trace databases
Reproducible, profiler-driven optimization studies covering CUDA kernels, deep-learning inference, multi-GPU communication and accelerated computing.
Reproducible CUDA kernel optimization lab: warp reduction, shared-memory transpose, WMMA, tests and profiling.
C++23 benchmarking framework with 6 profiler backends, CUDA GPU support, statistical regression detection, cross-compilation for 5 architectures, and CLI tools for analysis and visualization.
The MNIST classification problem is a fundamental machine learning task that involves recognizing handwritten digits (0- 9) from a dataset of 70,000 grayscale images (28x28 pixels each). It serves as a benchmark for evaluating machine learning models, particularly neural networks.
Custom PyTorch CUDA kernel implementing optimized ReLU activation with vectorization, performance profiling, and memory analysis on Tesla T4 GPU achieving 75% bandwidth efficiency.
High-Performance Computing (HPC) & Optimization studies using CUDA C++. Includes Grid-Stride Loops, Shared Memory tiling, and Nsight Compute profiling analysis.
To associate your repository with the nsight topic, visit your repo's landing page and select "manage topics."