Profile, analyze, and optimize CUDA kernels in LibuIPC using Nsight Compute.
npx skills add https://github.com/spiriMirror/libuipc --skill gpu-optimization
Please help me install this Agent Skill. Skill: gpu-optimization Source: https://github.com/spiriMirror/libuipc/tree/main/.cursor/skills/gpu-optimization Command: npx skills add https://github.com/spiriMirror/libuipc --skill gpu-optimization
GPU optimization workflow automates profiling, analysis, and optimization of CUDA kernels within LibuIPC to achieve faster simulation times and more efficient GPU usage.
Run the gpu-optimization workflow to profile and optimize CUDA kernels in libuipc using uipc.profile, uipc.profile.nsight, and Nsight Compute CLI.
High-intent search queries and answers about installing and using this skill.
Profiling LibuIPC CUDA kernels involves using uipc.profile and Nsight Compute to capture kernel metrics and timer frames. The workflow enforces Release builds for accurate benchmarking and guides optimization based on the generated reports.
Benchmarking large-scale rigid and soft body simulations requires enforcing a Release build type to measure accurate kernel duration. The workflow uses timer outputs and Nsight Compute reports to yield meaningful frame-time improvements.
Yes, Nsight Compute is required to identify bottlenecks in LibuIPC CUDA kernels. The workflow relies on Nsight Compute reports and timer frames to guide optimization across CPU-GPU and GPU-bound stages.
A Release build type is required for benchmarking because it removes debug overhead, ensuring timer outputs and kernel metrics reflect real-world GPU performance. This provides accurate data for optimizing CUDA kernels.
Yes, you can use uipc.profile to tune kernel parameters for both CPU-GPU and GPU-bound stages. The guided optimization workflow analyzes kernel metrics across build variants to improve simulation times.
Relying solely on timer outputs limits optimization to duration analysis without kernel-level execution details. The workflow addresses this by combining timer frames with Nsight Compute reports for comprehensive GPU bottleneck identification.
igamenovoerConfigure CUDA build environments for Pixi projects with CMake and nvcc.
NVIDIAAnalyze Nsight Compute .ncu-rep reports to classify CUDA kernel bottlenecks.
mit-han-labProfile CUDA kernels with Nsight Compute and generate optimization reports for Blackwell B200 GPUs.
michaeldubuParse and analyze CUDA, CPU, and memory benchmark outputs across CSV, JSON, and text formats.
