II.
LibraryProcess overview
Reference · livelib-process:gpu-programming--occupancy-optimization
specializations/gpu-programming/occupancy-optimization overview
Occupancy Optimization - Process for optimizing SM occupancy by balancing resource usage (registers, shared memory, thread block size) to maximize parallelism.
Attributes
displayName
specializations/gpu-programming/occupancy-optimization
description
Occupancy Optimization - Process for optimizing SM occupancy by balancing resource usage
(registers, shared memory, thread block size) to maximize parallelism.
libraryPath
library/specializations/gpu-programming/occupancy-optimization.js
specialization
gpu-programming
references
- - CUDA Occupancy Calculator: https://docs.nvidia.com/cuda/cuda-occupancy-calculator/
- - Occupancy Best Practices: https://docs.nvidia.com/cuda/cuda-c-best-practices-guide/
example
const result = await orchestrate('specializations/gpu-programming/occupancy-optimization', {
projectName: 'convolution_kernels',
targetKernels: ['conv2d_forward', 'conv2d_backward'],
targetOccupancy: 75,
targetArch: 'sm_86'
});
usesAgents
- gpu-performance-engineer
Outgoing edges
lib_applies_to_domain1
- domain:scientific-computing·DomainScientific Computing
lib_belongs_to_specialization1
- specialization:gpu-programming·Specialization
lib_implements_workflow1
- workflow:performance-profiling-cycle·WorkflowPerformance Profiling Cycle
lib_involves_role2
- role:computational-scientist·RoleComputational Scientist
- role:ml-engineer·RoleMachine Learning Engineer
lib_requires_skill_area2
- skill-area:cuda-kernels·SkillAreaCUDA Kernel Programming
- skill-area:compute-shaders·SkillAreaCompute Shaders
uses_agent1
- lib-agent:gpu-programming--gpu-performance-engineer·LibraryAgentgpu-performance-engineer
Incoming edges
None.