Analyze CUDA kernel code and produce performance improvements.
npx skills add https://github.com/jarmak-personal/vibeSpatial --skill cuda-optimizer
Please help me install this Agent Skill. Skill: cuda-optimizer Source: https://github.com/jarmak-personal/vibeSpatial/tree/main/.agents/skills/cuda-optimizer Command: npx skills add https://github.com/jarmak-personal/vibeSpatial --skill cuda-optimizer
This Skill enables users to optimize existing CUDA and GPU kernel code, improving runtime efficiency and performance.
Use the cuda-optimizer Skill to review and optimize your existing CUDA kernel files in vibeSpatial.
High-intent search queries and answers about installing and using this skill.
To optimize existing CUDA kernel code, you can analyze kernel source files and dispatch logic to identify memory access improvements, synchronization elimination, and targeted rewrites that align with NVIDIA best practices.
Unnecessary synchronization patterns in GPU modules often slow down execution; analyzing and eliminating these host-device boundary interactions helps produce concrete performance improvements and enhanced execution speed.
Yes, you can automate GPU code review by reading existing CUDA kernel source files to identify optimization opportunities and produce concrete, applicable code modifications for enhanced runtime efficiency.
This CUDA optimization approach specifically targets host-device boundary interactions and kernel source files to ensure code adheres to NVIDIA best practices through automated rewrites and suggestions.
The best way to reduce manual trial-and-error is to systematically analyze your CUDA code to implement performance best practices through targeted rewrites and concrete suggestions for memory access and synchronization.
Harmeet10000Explain CUDA-free Mojo GPU programming with LayoutTensor and shared memory patterns.
spacecat2002Add and validate lightweight CUDA JIT kernels in SGLang's jit_kernel module.
linkedinProfile, generate variants, and validate Liger Kernel Triton kernels for NVIDIA GPUs.
hkqaiProfile CUDA/GPU simulations with Nsight Systems and analyze bottlenecks statistically.
