II.
SkillArea overview
Reference · liveskill-area:profiling-cuda
CUDA Profiling overview
Profiling CUDA workloads with Nsight Systems / Compute — kernel timing, occupancy, memory throughput, and roofline analysis.
Attributes
displayName
CUDA Profiling
description
Profiling CUDA workloads with Nsight Systems / Compute — kernel
timing, occupancy, memory throughput, and roofline analysis.
domains
expertiseLevels
- expert
Outgoing edges
applies_to1
- specialization:gpu-programming·Specialization
Incoming edges
lib_requires_skill_area9
- lib-agent:gpu-programming--cuda-kernel-expert·LibraryAgentcuda-kernel-expert
- lib-agent:gpu-programming--multi-gpu-systems-expert·LibraryAgentmulti-gpu-systems-expert
- lib-process:gpu-programming--cuda-kernel-development·LibraryProcessspecializations/gpu-programming/cuda-kernel-development
- lib-process:gpu-programming--custom-cuda-operator-development·LibraryProcessspecializations/gpu-programming/custom-cuda-operator-development
- lib-process:gpu-programming--dynamic-parallelism-implementation·LibraryProcessspecializations/gpu-programming/dynamic-parallelism-implementation
- lib-skill:gpu-programming--cuda-debugging·LibrarySkillcuda-debugging
- lib-skill:gpu-programming--cuda-graphs·LibrarySkillcuda-graphs
- lib-skill:gpu-programming--cuda-toolkit·LibrarySkillcuda-toolkit
- lib-skill:gpu-programming--nvenc-nvdec·LibrarySkillnvenc-nvdec
prerequisite_for_learning1
- skill-area:shader-programming·SkillAreaShader Programming