Epic: DtRobust Standard Library Migration
Overview
Move DtRobust operators (integrate, decay, relax, etc.) from hardcoded compiler constructs to dynamically registered standard library functions while preserving SSA optimizations and vectorized performance.
Current Problem
- DtRobust operators are hardcoded in compiler with
DtRobustOperator enum
- Function names are hardcoded in parser/lowering phase
- Math implementations are hardcoded in vectorized executor
- Results in double implementation: hardcoded system AND stdlib functions
- Violates principle of configurable/extensible standard library
Goals
- ✅ Remove all hardcoded DtRobust logic from compiler
- ✅ Make DtRobust operators dynamically registered in standard library
- ✅ Preserve vectorized optimizations for performance
- ✅ Use consistent
kernel. prefix for all engine functions
- ✅ Maintain SSA pipeline benefits
Migration Strategy
Phase 1: Enhance Kernel Registry
- Extend kernel registry to support vectorized execution hints
- Add vectorized function registration macros
- Enable both scalar and vectorized implementations per function
Phase 2: Complete Standard Library
- Add missing DtRobust operators to kernels/functions
- Port vectorized implementations from compiler
- Ensure all operators work with kernel.* syntax
Phase 3: Remove Hardcoded System
- Remove DtRobustOperator enum and DtRobustCall variants
- Remove hardcoded parsing precedence in expr lowering
- Remove special SSA instruction handling
Phase 4: Update Execution Pipeline
- Enhance execution engine for vectorized kernel calls
- Update DSL syntax to require kernel. prefix
- Performance validation and testing
Breaking Changes
- DSL syntax changes:
integrate() → kernel.integrate()
- IR API changes: Remove DtRobust types from public API
Success Criteria
Timeline
4 weeks total, one phase per week
Related Issues
Will be created as sub-issues of this epic.
Epic: DtRobust Standard Library Migration
Overview
Move DtRobust operators (integrate, decay, relax, etc.) from hardcoded compiler constructs to dynamically registered standard library functions while preserving SSA optimizations and vectorized performance.
Current Problem
DtRobustOperatorenumGoals
kernel.prefix for all engine functionsMigration Strategy
Phase 1: Enhance Kernel Registry
Phase 2: Complete Standard Library
Phase 3: Remove Hardcoded System
Phase 4: Update Execution Pipeline
Breaking Changes
integrate()→kernel.integrate()Success Criteria
Timeline
4 weeks total, one phase per week
Related Issues
Will be created as sub-issues of this epic.