II.
LibraryProcess overview
Reference · livelib-process:gpu-programming--gpu-performance-regression-testing
specializations/gpu-programming/gpu-performance-regression-testing overview
GPU Performance Regression Testing - Workflow for establishing and maintaining GPU performance benchmarks to detect regressions across code changes and hardware.
Attributes
displayName
specializations/gpu-programming/gpu-performance-regression-testing
description
GPU Performance Regression Testing - Workflow for establishing and maintaining GPU performance
benchmarks to detect regressions across code changes and hardware.
libraryPath
library/specializations/gpu-programming/gpu-performance-regression-testing.js
specialization
gpu-programming
references
- - Google Benchmark: https://github.com/google/benchmark
- - NVIDIA/cuda-samples benchmarks: https://github.com/NVIDIA/cuda-samples
- - Performance Testing Best Practices: https://developer.nvidia.com/blog/
example
const result = await orchestrate('specializations/gpu-programming/gpu-performance-regression-testing', {
projectName: 'cuda_library',
benchmarkSuite: ['gemm', 'convolution', 'reduction'],
thresholds: { latencyRegression: 5, throughputRegression: 5 },
ciIntegration: true
});
usesAgents
- gpu-performance-engineer
- devops-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.