A simple profiler to count Nvidia PTX assembly instructions of OpenCL/SYCL/CUDA kernels for roofline model analysis.
A Systems View of Efficient Machine Learning, from Silicon to Agents. Fourteen parts, from roofline analysis and vector units up through kernels, compilers, quantisation, compression, vision, on-device language models, robotics, profiling, serving and agents.
A GPU performance prediction toolkit for CUDA programs
A Python script for plotting roofline analyses. Intel Advisor style.
A command-line tool to plot roofline models for NVIDIA GPU kernels from profiler output.
High-performance Sobel edge detection using CUDA with CPU vs GPU benchmarking, roofline analysis, and Nsight profiling.
Code Comprehension Assistance for Evidence-Based performance Tuning
Fork of the CS Roofline Toolkit from Berkeley Lab
Silicon workbench for chip blueprints, roofline analysis, tile-memory flow, and AI accelerator co-design.
What a latency promise costs in seats and dollars on a rented GPU — a slice of an LLM inference operator's platform, with every number traceable to the run that produced it.
CLI tool for estimating compute, memory bandwidth, and operational intensity of transformer models from Hugging Face configuration files. Ideal for performance and hardware deployment analysis.
Design Space Exploration of bandwidth and latency constraints in SIMD neural accelerators using the Roofline model. Analyzes how memory bottlenecks impact EDP, GOPS, stall cycles, and hardware utilization through automated RTL synthesis with Python and industry-standard EDA tools.
Repository for the research paper, "District-scale life cycle costing of climate-neutral retrofits, based on automatic envelope detection workflows and LOD3 3D city model".
High-performance discrete-event simulator (C++20/Python) for modeling agentic LLM traffic, KV cache dynamics, Prefill-Decode Disaggregation (PDD), and scheduling policies. Features roofline model analysis, K-Means request clustering, and a real-time web dashboard.
Why more compute doesn't mean faster AI : The Roofline Model explained, with an interactive calculator and Python workload analyzer
Physics-based LLM capacity planner - size a deployment before provisioning hardware. Roofline model → replica count, cost envelope, and confidence band. Benchmark harness validates estimates against real endpoints and compares configs on latency, cost, and quality.
Simulator for brain-inspired many-core dataflow + GPU large-model inference. Built to test whether a PD+A split really delivers 2x throughput and 2x energy efficiency.
FlashAttention + Multi-Head Latent Attention, from scratch: Triton kernel, checkpointed PyTorch attention, MLA, and a reproduced roofline argument on real PTB-XL multimodal data.
Hand-written CUDA convolution & classifier kernels for a VGG-style CNN, profiled on an NVIDIA Titan V, with a CuDNN comparison and a roofline performance model.
Cross-platform system benchmark (Windows 11 / Astra Linux). Roofline-model scoring: CPU, GPU (OpenCL), RAM, disk, stress testing, thermal monitoring. CLI + Web UI.
To associate your repository with the roofline-model topic, visit your repo's landing page and select "manage topics."