@ariG23498: Made a simple utility to share your GPU profile traces. `uvx trace-tuil <local traces> -b <hf bucket name>`
Summary
A simple utility to share GPU profile traces using a command-line tool, designed for developers working with GPU profiling.
View Cached Full Text
Cached at: 05/15/26, 11:09 PM
Made a simple utility to share your GPU profile traces.
uvx trace-tuil <local traces> -b <hf bucket name> https://t.co/tLYc2JegJW
Similar Articles
We built an open-source GPU profiler you point an AI agent at, instead of reading traces yourself
This article introduces an open-source GPU profiler that serves metrics in JSON format, allowing AI agents to automate performance tuning for CUDA-based AI engines like vLLM and SGLang without manual trace analysis.
gpuvis: GPU Trace Visualizer
gpuvis is a Valve-authored, MIT-licensed GPU trace visualizer built with Dear ImGui, SDL, and gl3w for inspecting graphics traces. The project is documented via its GitHub repo and wiki, embedding a range of third-party libraries.
I built an open-source tool that reduces TensorBoard trace sizes by 90%+ for JAX/XLA (XProf Cubism Reducer)
An open-source tool called XProf Cubism Reducer reduces TensorBoard trace sizes by over 90% for JAX/XLA, making performance profiling more efficient.
3 days benchmarking most llama.cpp flags on my weird 40gb vram laptop + tb4 egpu setup. Got +70% generation, +40% prefill, 60k more context, and filed a bug in llama around MTP. What I learned.
The author benchmarked llama.cpp flags on a hybrid GPU setup with an RTX 4090 laptop and AMD XTX 7900 eGPU, achieving 70% faster generation, 40% faster prefill, and discovering a bug related to MTP in multi-GPU configurations.
janestreet/magic-trace
magic-trace is a high-resolution tracing tool using Intel Processor Trace to capture full control flow snapshots of Linux processes with low overhead, useful for debugging and performance analysis.