A CUDA implementation of the transpose-free Quasi-Minimal Residual method

The definitive Strix Halo LLM guide — 65 t/s on a $2,999 mini PC. Live benchmarks, tested optimizations, and everything that doesn't work.

Unified Memory Abstraction Layer for AI Inference on AMD APUs and Intel iGPUs

Fundamentals of Accelerated Computing C/C++ is a course provided by NVIDIA.

Performance comparison of two different forms of memory management in CUDA

NVML unified memory shim for NVIDIA DGX Spark Grace Blackwell GB10 - enables MAX Engine, PyTorch, and GPU monitoring

Talos-O (Omni): A sovereign, embodied agentic organism forged on AMD Strix Halo. Integrating the Chimera Kernel (Linux 7.0), Zero-Copy Introspection, and the Phronesis Engine. Built from First Principles.

gpu thrashingNVIDIA GPU Unified Memory diagnostic tool — architecture-aware, measurement-based, PCIe/coherent transport detection

3D U-Net with tf.keras for Large-Model-Support or Unified Memory

Apple Silicon Unified Memory for GPU-Accelerated Analytics — TPC-H benchmarks across DuckDB, NumPy, and MLX

Reproducible Pascal GPU Unified Memory benchmark with Nsight and nvprof profiling

Extended the UVM Benchmark such that we can test for huge data workloads(16GiB and more). Needed to make it overflow save and add dataset creation logic for some Applications.

AI-native OS kernel written from scratch in C and x86_64/aarch64 assembly — kernel-level tensor compute, capability-based security, SMP, TCP/IP, and 95 userland programs

NVIDIA GPU validation: PCIe transport, Unified Memory prefetch, SGEMM compute, drift detection.

Local inference server for Apple Silicon — hot-swaps MLX models (LLM, vision, embeddings, TTS, STT) via OpenAI API

Run LLMs larger than your RAM — native GGUF inference engine with SSD streaming, no GPU required

Cycle-accurate UMA fault latency and bandwidth measurement for NVIDIA GPUs. C and PTX. No Python. Pascal (SM 6.0) through Blackwell GB10 (SM 12.1).

3-bit Lloyd-Max KV Cache Compression for LLM Inference on NVIDIA DGX Spark GB10 — 5.12x compression, 0.983 cosine similarity, pure numpy on ARM unified memory

This project will provide an overview on how to programm a GPU. How can we exploit Unified Memory and is it an actual competition to pinned memory.

Unlock fast, local LLM inference on AMD-powered mini PCs delivering 65-87 t/s for large models without cloud or subscription costs