[E5-F7] Generate implementation issues for Warp/GPU Translation

[Gorkowski]

Dependencies:

#1115 [E5-F1] ──┐
#1116 [E5-F2] ──┤
#1117 [E5-F3] ──┼──► #THIS [E5-F7]
#1118 [E5-F4] ──┤
#1119 [E5-F5] ──┘

Plus: E3-F3 (Warp Integration) should be sufficiently advanced

Summary

Generate implementation issues for feature E5-F7: Warp/GPU Translation (Follow-on).

Translate the Python-native non-isothermal condensation implementation to NVIDIA Warp (wp.func / wp.kernel) for GPU acceleration. This is a follow-on feature that ships after the Python implementation (E5-F1 through E5-F5) is complete and validated. The GPU translation follows patterns established in E3-F3 (Warp Integration and GPU Kernels).

Feature Plan

Document: adw-docs/dev-plans/features/E5-F7-warp-gpu-translation.md

Parent Epic: E5: Non-Isothermal Condensation with Latent Heat (adw-docs/dev-plans/epics/E5-non-isothermal-condensation.md)

Phases to Generate

Phase	Description	Size
E5-F7-P1	Translate latent heat pure functions to wp.func kernels with tests	M (~80 LOC)
E5-F7-P2	Translate CondensationLatentHeat.step() to Warp kernel with tests	L (~120 LOC)
E5-F7-P3	Update development documentation for GPU feature	XS (~30 LOC)

Dependencies

• Blocked by: [E5-F1] [E5-F1] Generate implementation issues for Latent Heat Strategy Pattern #1115, [E5-F2] [E5-F2] Generate implementation issues for Non-Isothermal Mass Transfer Functions #1116, [E5-F3] [E5-F3] Generate implementation issues for CondensationLatentHeat Strategy Class #1117, [E5-F4] [E5-F4] Generate implementation issues for Builder, Factory, and Exports #1118, [E5-F5] [E5-F5] Generate implementation issues for Validation and Integration Tests #1119 — Python implementation is the reference
• Also leverages: E3-F3 (Warp Integration) for GPU kernel patterns; NVIDIA Warp package; CUDA-capable GPU
• Blocks: None (final follow-on feature)

ADW Instructions

When processing this issue:

1. Read the feature plan document: adw-docs/dev-plans/features/E5-F7-warp-gpu-translation.md
2. Read the parent epic for context: adw-docs/dev-plans/epics/E5-non-isothermal-condensation.md
3. For each phase in the Phase Checklist (E5-F7-P1 through E5-F7-P3), create an implementation issue with:
   • Full technical details from the feature plan
   • Specific file paths from the feature plan's scope section
   • Test file paths (co-located in module tests/ directories)
   • Co-located testing: tests ship with implementation in every phase
   • Coverage target: 80-85% per phase
4. Set dependency chain: E5-F7-P1 → E5-F7-P2 → E5-F7-P3
5. Label all phases with agent, blocked, type:complete, model:default

Files to Create/Modify

• Create: Warp function files (following E3-F3 patterns — exact paths TBD based on E3-F3 structure)
  • wp.func equivalents of get_thermal_resistance_factor and get_mass_transfer_rate_latent_heat
  • wp.kernel for non-isothermal condensation step
• Modify: docs/Examples/Dynamics/non_isothermal_condensation_example.py — add GPU usage examples
• Modify: adw-docs/dev-plans/features/index.md — GPU completion status
• Create: GPU parity and performance test files (following E3-F3 test patterns)

Testing Strategy

Unit Tests

Test File	Phase	Coverage Target
TBD (follows E3-F3 test patterns)	P1	Warp functions
TBD (follows E3-F3 test patterns)	P2	Warp kernels

Key Test Cases

1. Function parity: thermal_resistance_factor_wp matches get_thermal_resistance_factor for same inputs
2. Rate parity: mass_transfer_rate_latent_heat_wp matches Python version
3. Kernel parity: Full step kernel matches CondensationLatentHeat.step()
4. Energy tracking: GPU energy output matches CPU sum(dm * L)
5. Performance: GPU faster than CPU for >= 10000 particles (marked @pytest.mark.slow)

Numerical Tolerance

• Float32: < 1e-6 relative error (GPU default precision)
• Float64: < 1e-12 relative error
• Tests gracefully skip when Warp not installed (pytest.importorskip)

Risks and Mitigations

Risk	Mitigation
E3-F3 Warp patterns not yet finalized	Feature is explicitly deferred; wait for E3-F3 to stabilize
Float32 precision loss on GPU	Test both float32 and float64; document precision trade-offs
Warp not available in all environments	Skip GPU tests when Warp not installed
Performance regression	Benchmark tests with explicit particle count thresholds