How nucleonautomation/pyceler works
PyCeler analyzes Python source, identifies eligible numeric loops, generates readable Cython or native extensions, then validates and measures candidate implementations to select guarded runtime behavior with Python fallback. The graph preserves the documented optimization families and separates analysis, lowering, build, and selection. Source excerpts expose several internal dependencies; other subsystem links reflect README-level workflow and capabilities rather than inspection of unsampled implementations.
Main entry points: Native_Optimizer.py Candidate optimization · Common_Analysis.py Shared loop analysis · Fast_Math.py Numeric expression lowering · __init__.py Public API
Components 24
User API and analysis
- Public API
PyCeler/__init__.py - Compile workflow
Compile.py - Source analysis
PyCeler/Source_Analyzer.py
Analysis and lowering
- Shared loop analysis
PyCeler/Common_Analysis.py - Math contracts
PyCeler/Math_Registry.py - Math maps
PyCeler/Math_Maps.py - Numeric regions
PyCeler/Numeric_Regions.py - Array regions
PyCeler/Array_Regions.py - Numeric expression lowering
PyCeler/Fast_Math.py - Region code emitter
PyCeler/Numeric_Region_Emitter.py - Array kernel generation
PyCeler/Array_Region_Kernel.py - Numeric kernel generation
PyCeler/Numeric_Region_Kernel.py - Pure helper calls
PyCeler/Helper_Kernel.py - Iterator kernels
PyCeler/Iterator_Kernel.py - Numeric list kernels
PyCeler/List_Grid_Kernel.py - Indexed distance filters
PyCeler/Indexed_Filters.py - Direct Cython bindings
PyCeler/Direct_Source.py
Native build
- Cython native build
PyCeler/Native_Build.py
Validation and selection
- Candidate optimization
PyCeler/Native_Optimizer.py - Behavior and timing worker
PyCeler/Selection_Worker.py - Backend selection policy
PyCeler/Selection_Policy.py - Runtime guards and fallback
PyCeler/Runtime_Resources.py
Other components
- Python developer
- Cython toolchain
Connections 23
- Python developer → Public API: calls
README.md - Public API → Source analysis: analyzes source
README.md - Public API → Compile workflow: converts or compiles
README.md - Compile workflow → Shared loop analysis: plans loops
- Shared loop analysis → Math contracts: checks calls
PyCeler/Common_Analysis.py - Array regions → Numeric regions: plans regions
PyCeler/Array_Regions.py - Array regions → Shared loop analysis: uses analysis
PyCeler/Array_Regions.py - Numeric regions → Shared loop analysis: uses analysis
PyCeler/Numeric_Regions.py - Numeric regions → Numeric expression lowering: plans expressions
PyCeler/Numeric_Regions.py - Math maps → Math contracts: inspects math calls
PyCeler/Math_Maps.py - Math maps → Numeric expression lowering: plans fast math
PyCeler/Math_Maps.py - Array kernel generation → Region code emitter: emits regions
PyCeler/Array_Region_Kernel.py - Array kernel generation → Numeric expression lowering: lowers expressions
PyCeler/Array_Region_Kernel.py - Numeric kernel generation → Region code emitter: emits regions
PyCeler/Numeric_Region_Kernel.py - Numeric kernel generation → Numeric expression lowering: lowers expressions
PyCeler/Numeric_Region_Kernel.py - Compile workflow → Cython native build: builds candidates
- Cython native build → Cython toolchain: invokes compiler
README.md - Candidate optimization → Cython native build: builds variants
PyCeler/Native_Optimizer.py - Candidate optimization → Behavior and timing worker: measures candidates
PyCeler/Native_Optimizer.py - Candidate optimization → Backend selection policy: selects candidates
PyCeler/Native_Optimizer.py - Candidate optimization → Runtime guards and fallback: uses runtime resources
PyCeler/Native_Optimizer.py - Behavior and timing worker → Backend selection policy: applies selection rules
PyCeler/Selection_Worker.py - Runtime guards and fallback → Python developer: falls back to Python
README.md