particula.execution._condensation_reservoirs
Deterministic parallel reservoir reductions with bounded device scratch.
Functions
| Name | Description |
|---|---|
assemble_parts | Sum fixed strided slot groups without atomic floating-point additions. |
finish_gas | Finish in fixed lane order and write scaled gas derivatives. |
finish_schur | Add the diagonal/RHS once after deterministic partial reduction. |
gas_parts | Retain native slot weights and reduce independent fixed slot groups. |
assemble_partsfunction#
def assemble_parts(local: FloatTensor, response: FloatTensor, particular: FloatMatrix, counts: FloatMatrix, volume: FloatVector, scale: FloatVector, rhs: FloatVector, targets: IntVector, sources: IntVector, materials: int, gases: int, resolved: int, h: FloatVector, parts: FloatTensor)Sum fixed strided slot groups without atomic floating-point additions.
finish_gasfunction#
def finish_gas(parts: FloatTensor, mass_size: int, scale: FloatVector, derivative: FloatVector, invalid: IntVector)Finish in fixed lane order and write scaled gas derivatives.
finish_schurfunction#
def finish_schur(parts: FloatTensor, rhs: FloatVector, sources: IntVector, mass_size: int, gases: int, h: FloatVector, schur: FloatTensor)Add the diagonal/RHS once after deterministic partial reduction.
gas_partsfunction#
def gas_parts(rate: FloatTensor, counts: FloatMatrix, volume: FloatVector, mapping: IntVector, resolved: int, parts: FloatTensor)Retain native slot weights and reduce independent fixed slot groups.